Professional Documents
Culture Documents
1 Cell Biology and Biophysics Unit, European Molecular Biology Laboratory (EMBL), Heidelberg, Germany
2 Developmental Biology Unit, European Molecular Biology Laboratory (EMBL), Heidelberg, Germany
3 Department of Physics and Astronomy, Heidelberg University, Heidelberg, Germany
*Corresponding author. Tel: +49 6221 387 8367; E-mail: jan.rombouts@embl.de
**Corresponding author. Tel: +49 6221 387 8522; E-mail: erzberge@embl.de
†
These authors contributed equally to this work
ª 2023 The Authors. Published under the terms of the CC BY 4.0 license. EMBO reports 24: e57739 | 2023 1 of 24
EMBO reports Jan Rombouts et al
List of symbols
In the text, bold symbols indicate vector or tensor quantities. A subscript 0 indicates a steady-state quantity. A tilde indicates a small perturbation from
steady state.
be related to the microscopic parameters of cellular interactions molecules are subject to chemical reactions. Such processes lead to
(Oriola et al, 2022). reaction terms RðcÞ in the equation for the corresponding density or
If a system is made up of a small number of constituents, or if the concentration that describes how c changes locally. Such reaction–
focus is on individual interactions, a discrete framework might be a diffusion equations
more appropriate description. Discrete modeling frameworks can be ∂t c ¼ Dr2 c þ RðcÞ (2)
used to study, for example, the dynamics of individual cytoskeletal
filaments and associated proteins (Nedelec & Foethke, 2007) or inter- are classical pattern-forming systems. The seminal work of
actions between cells (Graner & Glazier, 1992; Alt et al, 2017; pre- Turing (1952) showed how biochemical feedback mechanisms com-
print: Belousov et al, 2023). Additionally, discrete models are also bined with diffusion can lead to the emergence of regular patterns.
essential to describe systems where the length scale of the pattern is Turing’s work was conceptually extended and described in the con-
similar to the length scale of the individual units (e.g., Collier text of activator–inhibitor systems (Gierer & Meinhardt, 1972; Mein-
et al, 1996; Manukyan et al, 2017; Fofonjka & Milinkovitch, 2021). In hardt, 2012). Such systems need at least two chemical species to
this paper, however, we focus on continuum approaches to pattern form patterns (see Box 1 for an example). Turing’s original two-
formation. Continuum theories have successfully predicted large- species example is mathematically simple but has often been consid-
scale flows, deformations, and patterns in many cellular and multi- ered biologically unrealistic due to, for example, its sensitivity to
cellular systems (Bergert et al, 2015; Streichan et al, 2018; Erzberger parameter values (Green & Sharpe, 2015). Recent theoretical and
et al, 2020; Saadaoui et al, 2020; Palmquist et al, 2022). computational work has shown, however, that a Turing mechanism
The dynamics in space and time of mesoscale fields, such as con- can robustly produce patterns in more complex reaction–diffusion
centrations, follow partial differential equations that are called conti- systems (Marcon et al, 2016; Haas & Goldstein, 2021).
nuity equations. When the fields correspond to densities of In most biological systems, patterns are not due to biochemical
conserved quantities such as mass or momentum, the governing or genetic interactions alone. Mechanical forces play an important
partial differential equations express conservation laws. For exam- role. The propagation of mechanical stresses occurs on different
ple, the concentration c of a conserved number of molecules at posi- temporal and spatial scales than molecular diffusion. For example,
tion x and time t evolves according to the spatial derivatives of the a small molecule requires seconds to diffuse over a 10-100 µm dis-
flux density j at that point: tance, whereas mechanical stress propagates over a similar dis-
tance within microseconds in the cytoskeleton (Wang et al, 2009).
∂t cðx; tÞ ¼ r j: (1) Mechanical or mechanochemical pattern formation thus broadens
the range of scales on which patterns can be formed (Howard
This equation states that the change in c depends on how many et al, 2011; Collinet & Lecuit, 2021). In fact, since all organisms
molecules enter and exit at a certain location. The r operator and cells are subject to physical laws, the effects of these laws are
can be expressed in terms of spatial derivatives, and in one dimen- likely exploited for the generation of patterns with biological
sion is equal to ∂x . The flux density j can arise, for example, functions.
from diffusion, which for molecules usually obeys Fick’s law: An aspect that is inherently linked to mechanics is the size and
jdiffusive ¼ Drc (Fick, 1855). Plugging the diffusive flux into equa- geometry of the domain on which these patterns form, such as the
tion (1) leads to the familiar diffusion equation ∂t c ¼ Dr2 c. The shape of cell membranes or tissue layers. Spatial derivative opera-
same equation can be used for collections of cells, rather than mol- tors, such as r, take different forms depending on the dimension of
ecules, that are moving about randomly (Berg, 1993). the system, its geometric properties, and the chosen coordinate sys-
In many biological contexts, the total number of units is not con- tem. This, in turn, influences the types of spatial modes that are
served. For example, cells and organisms reproduce and die, and used in a linear stability analysis (see Box 1) and the patterns that
10 μm 100 μm 1 mm 10 mm
B D O = 2π/qmax
System size L
exp(Zt) sin(qmaxx)
C q = π/L q =1.3π/L
x
L E
q = 2π/L exp(Zt) Z
q =3π/L F
Zmax
Re(Z)
q
qmax
can appear. Moreover, the shape of the domain can influence the contexts by neglecting the time-derivative term on the left-hand side
dynamics of the pattern: Nishide & Ishihara (2022) recently showed because, on the scale of cells and small organisms, forces due to
that Turing-type patterns that are static on a flat domain can propa- friction, or viscosity, dominate over inertial forces. Such over-
gate on curved shapes. damped motion appears very different from our everyday experi-
Furthermore, mechanical forces affect the flux term in equa- ences (Purcell, 1977; Lauga & Powers, 2009). In this limit, the
tion (1). For example, if there are fluid flows in the system that conservation law for momentum can be written as
carry along molecules or cells, there is a contribution of an advec-
tive flux jadvective ¼ cv, where v is the velocity field. The evolution of 0 ¼ r σ tot þ f; (3)
v is governed by a second equation that states the conservation of
momentum. This equation can be derived in the same way as equa- in which f denotes any external forces, such as those arising from
tion (1), but luckily, it can be simplified for many biological friction with the environment. Equation (3) is a continuity
∂t A ¼ DA ∂xx A þ k1 A2 B þ k2 k3 A;
( 18)
∂t B ¼ DB ∂xx Bk1 A2 B þ k4 :
The terms with the spatial derivatives describe diffusion with diffusion constants DA and DB and the remaining terms describe the chemical reactions.
The chemical species A and B are produced at constant rates k2 and k4 , respectively. There is a conversion reaction with rate constant k1 , and A is
degraded with rate constant k3 . Appropriate scaling of the variables, space, and time yields equations in terms of the dimensionless variables u and v
∂t u~ ≈ ∂xx u
~ þ ∂u Ru ðu0 ; v0 Þ~
u þ ∂v Ru ðu0 ; v0 Þ~ v
( 20)
∂t ~v ≈ d∂xx ~ ~ þ ∂ v R v ð u 0 ; v0 Þ ~
v þ ∂u Rv ðu0 ; v0 Þu v;
in which we have kept terms in linear order in the small perturbations. In this linearized version, one can already see the feedback loops that are act-
ing in the system. For example, ∂v Ru ðu0 ; v0 Þ ¼ ðα þ βÞ2 is positive, indicating that there is positive feedback from B to A.
This linear system can be solved for any initial condition by decomposing the initial spatial profile into a sum of spatial modes and studying the
temporal evolution of these modes. The spatial modes that are appropriate depend on the geometry of the system. In this one-dimensional example,
suitable spatial modes are the functions cosðqxÞ and sinðqxÞ for different wavenumbers q. As known from Fourier analysis, a profile can be decom-
posed into combinations of these functions. These spatial modes are appropriate because their shape (but not their amplitude) is unchanged when
applying the diffusion operator ∂xx . This, together with the linearity of the equation, allows to study the evolution of equations (20) by solely consider-
ing what happens to the spatial modes. The temporal evolution of an initially sinusoidal profile is given by eωt sinðqxÞ and eωt cosðqxÞ. Each of the
modes’ amplitude thus evolves as eωt (Fig 1D and E). The value of ω depends on the wavenumber q. The relation ωðqÞ is called the dispersion relation
and can be computed from equations (20). This relation describes the growth rates of the different spatial modes (Fig 1F). Generally, ωðqÞ is a complex
number. Its real part describes the growth of the mode: perturbations with ReðωÞ < 0 decay in time, leaving the uniform solution intact. However, if
for a given q, one of the growth rates has a positive real part; the amplitude of the perturbation eωt grows in time. This implies that the uniform
steady state is unstable to perturbations with wavenumber q. The formation of patterns is associated with the existence of non-zero wavenumbers
that have ReðωÞ > 0. The fastest-growing mode—at the maximum of the dispersion relation—dominates, and thus sets the initial spatial wavelength
of the pattern. If ωðqÞ has an imaginary part, the solution is oscillatory over time. A mode that is growing in amplitude and is oscillatory leads to spa-
tiotemporal patterns such as waves, whereas a purely real ω is associated with stationary patterns. The dispersion relation of the Schnakenberg sys-
tem reveals the conditions on α; β, and d that lead to pattern formation, that is, for which values of these parameters unstable modes with q > 0
exist.
The geometry and size of the system impose further conditions on the solutions, mainly by constraining the form of the spatial modes and avail-
able wavenumbers. The sinusoidal form of the spatial modes is imposed by the fact that we study a one-dimensional system. The particular wave-
numbers that are possible are dictated by the domain size and boundaries: if the chemical reactions of the Schnakenberg system happen on a closed
interval ½0; L with either fixed concentrations or no-flux conditions at the boundaries, the only possible wavenumbers are of the form q ¼ nπ=L, with
an integer n. Patterns appear only when one of the permitted non-zero modes has a positive growth rate. The size of the domain thus restricts which
wavenumbers are possible. Analogously, other domain shapes and sizes dictate the spatial modes that can become unstable. This is illustrated in
Fig 1C.
Since we assumed small perturbations to the steady state in going from equations (19) to (20), the linear stability analysis only reveals the dynam-
ics of systems that are close to a steady state. Once the perturbation becomes too large, non-linear effects come into play and analytical results are
hard to obtain. Moreover, non-linear terms in the equations dictate at which amplitude the growing perturbations saturate, and also determine which
types of patterns appear; for example, whether spots or stripes appear in two-dimensional systems. Finally, we note that not all patterns exist close to
a homogeneous steady state—patterns may also exist “far from threshold” and depend crucially on the non-linearities in the system. We refer to the
book by Cross & Greenside (2009) for more details about these aspects. Even with these caveats, in many cases, the linear stability analysis already
provides good insight into the system’s pattern-forming properties and leads to an understanding of which biological parameters determine the exis-
tence and characteristic scale of patterns, as we illustrate throughout this review.
equation like equation (1), where the time derivative of the veloc- term. The counterpart of j in equation (1) is written as the negative
ity on the left-hand side is zero. In the same way in which we keep momentum flux density by convention and is called the total stress
track of how particles enter and leave a location in equation (1), tensor σ tot . It describes all forces per area that act on a small piece
we must account for the transport of momentum through a flux of the material for each of the possible force directions. Same as
for the molecular flux density, the components of the stress tensor one based on the synchronization of oscillators due to hydrody-
depend on the properties of the system. In an incompressible New- namic effects. The selected examples are ordered roughly along
tonian fluid, for example, a viscous stress arises between adjacent increasing spatial scales and levels of biological organization from
fluid layers that is proportional to the velocity difference between subcellular to organismal.
the layers. Additionally, in most biological systems, there are
active contributions to the stress, σ active (Kruse et al, 2005; Jülicher
et al, 2018), generated by processes such as the ATP-fuelled Geometry-dependent reaction–diffusion patterns
motion of molecular motors within the cytoskeleton (Bois et al,
2011; Dasanayake et al, 2011; Peleg et al, 2011), or the Pattern formation at the cell surface is coupled to shape and curva-
self-propelled movement of migrating cells or swimming ture and is influenced by the biophysical properties of the cell mem-
microorganisms (Yeomans, 2017; Alert & Trepat, 2020). Such brane and the cytoskeleton. In particular, reaction–diffusion
active processes are characteristic of biological systems and can dynamics in cells often involve an exchange of components between
give rise to behavior not seen in passive materials. This aspect of structures at the cell surface, that is, the plasma membrane or the
living systems motivated many successful developments within the actomyosin cortex, and the cytosol. These fluxes give rise to bio-
research field of active matter physics, with relevance beyond biol- chemical couplings between nearly two-dimensional structures that
A B
The dynamics in the bulk are coupled to those on the surface 2021), protein polymerization (Drew et al, 2005; Cytrynbaum &
through a boundary condition accounting for particle conservation: Marshall, 2007), interactions with particular lipids localized at the
the flux of particles leaving and entering the cytosol is balanced by cell poles (Renner & Weibel, 2012), and advective cytoplasmic flows
the rates of membrane binding and unbinding. Modeling details, (Vecchiarelli et al, 2014; Meindlhumer et al, 2023), have been exten-
including cooperative binding (Loose et al, 2008), further biochemi- sively studied and are proposed to play important roles in regulating
cal reactions (Meinhardt & Boer, 2001; Loose et al, 2011; Park Min patterns (Vecchiarelli et al, 2016; Wettmann & Kruse, 2018;
et al, 2011; Ayed et al, 2017; preprint: Carlquist & Cytrynbaum, Ramm et al, 2019; Takada et al, 2022).
The Min reaction system includes an ATPase, MinD, that enhances for theories on (biological) pattern formation, also beyond linear
the recruitment of itself to the membrane and a second protein, MinE, stability theory (Halatek & Frey, 2018; Brauns et al, 2020; Würthner
which drives MinD off the membrane again (Wettmann & Kruse, 2018; et al, 2022). This system is also an example of a “mass-conserving
Ramm et al, 2019). Under specific in vitro conditions, such systems reaction diffusion system,” where spatial redistribution, such as pro-
have been shown to exhibit stationary patterns (Glock et al, 2019), but tein depletion, drives the emergence of protein patterns. For a more
most research focuses on their emergent oscillatory dynamics. These detailed discussion of these systems, and their comparisons with
have been proposed to arise from a difference in the time MinD and non-mass conserving systems, see Halatek et al (2018).
MinE spend in the cytosol before re-binding to the membrane, intro- In elongated, rod-like cells such as E. coli, the standing waves
duced by the time needed for nucleotide replenishment of the ATPase produced by the Min system provide a mechanism to read out the
MinD (Huang et al, 2003; Glock et al, 2019). Whereas in vitro, this sys- mid-line and determine the cell division plane. A natural question is
tem admits different types of oscillating patterns, including both travel- then which Min patterns form in non-rod-like cells. Wu et al (2015)
ing waves and standing waves, in vivo cells robustly produce standing put E. coli cells in microchambers with different shapes, and
waves (Fig 2A; Hu & Lutkenhaus, 1999; Raskin & de Boer, 1999; Bonny observed that Min patterns oscillated along one of the symmetry
et al, 2013). Different groups have shown that the geometry of the cell axes. Subsequent simulations by Wettmann et al (2018) have also
may play a crucial role in the robust selection of standing wave patterns matched these results (Fig 2E). Walsh et al (2019) studied archaeal
geometry of the system. Many cellular patterning processes, how- functions h. Membrane shape dynamics arise when the system does
ever, rely on two-way interactions between the biochemical reac- not reside in this preferred configuration. Then, a restoring force
tions and the biophysical properties of the structures they take place that is proportional to δF=δh close to equilibrium drives the mem-
on (Schamberger et al, 2023). In this section, we discuss how mem- brane toward the preferred configuration. This term may compete
brane biophysics “by itself” can drive protein patterning. In particu- with other forces acting on the membrane. With the inclusion of
lar, the localization of proteins to and/or transport of proteins on frictional forces, a force balance equation (equation 3) governing
the plasma membrane can depend on the curvature of the mem- the membrane shape dynamics can be written as
brane (Heinrich et al, 2010; Sorre et al, 2012; Kwiecinski et al,
2017). Additionally, membrane–protein interactions modulate the δF
α∂t h ¼ : (7)
local biophysical properties of the membrane, giving rise to feed- δh
back effects that can drive pattern formation (Cooke & Deserno,
2006; Reynwar et al, 2007; Yue et al, 2010; Ramakrishnan Here, the left-hand side describes the frictional force on the membrane
et al, 2014; Sodt & Pastor, 2014; Hossein & Deserno, 2020). For due to movement, with friction coefficient α, and the right-hand side
example, curvature-mediated pattern formation plays a role in the corresponds to the forces arising from the free energy gradient. We
formation of cellular protrusions, such as those involved in cell continue to consider a simple theory for demonstration purposes,
A B C
n n
^ ^
x R1
R2 Curvature attracts
proteins
x
y
Membrane _( _ _ (
H= 1 1 + 1 _ _( _ _ (
H= 1 1 + 1
Membrane height
Proteins 2 R1 R2 2 R1 R2 Protein concentration
C p2
Cp
D
No
Growth rate ω
Instability
qc Control Parameter
10π/L 20π/L 30π/L
Wavenumber q 0
0 24/L
Average Spontaneous Curvature
C p c0
the free energy with respect to the protein distribution: spontaneous curvature by itself can drive patterning, we consider a
jcurv / rðδF=δcÞ (Mahapatra et al, 2021). system without chemical reactions, that is, R ¼ 0 (Agudo-Canalejo
The coupling between curvature and protein distribution can & Golestanian, 2017), and with a linear dependence of the spontane-
lead to the emergence of patterns. The conditions for pattern forma- ous curvature on protein concentration C p ðcðxÞÞ ¼ C p cðxÞ, where C p
tion in the coupled system given by equations (7) and (8), with the is positive. An increase in protein concentration thus induces an
free energy given by equation (6), can be obtained by linear stability increase in spontaneous curvature, which affects membrane shape,
analysis (see Box 1). To see how the concentration-dependent which in turn affects the protein distribution.
The equations for the steady state are solved by a uniform pro- Thus, an increase in local curvature will lead to an increase in the
tein concentration c0 and a corresponding flat equilibrium shape h0 . flux of curvature-inducing proteins into the region, providing a posi-
Linearizing around this steady state in the small gradient limit— tive feedback loop that clusters curvature-inducing proteins, accom-
corresponding to weak bending of the membrane—results in the panied by a further increase in the local curvature. Further
equations for small perturbations c~ðxÞ and h~ðxÞ: constraints on the membrane, such as incompressible enclosed vol-
umes, may play a role in mode selection and the formation of stable
2
!
κC p c20 2 ~ protein patterns (Ramaswamy et al, 2000; Tozzi et al, 2019). Mem-
α∂t h~ ¼ κr4 þ r h κC p r2 c~; (9a)
2 brane tension, in particular, acts in opposition to increases in
membrane curvature and provides a biophysical means to tune the
wavelength of emergent patterns (Agudo-Canalejo & Golesta-
2 ~
τ∂t c~ ¼ τD þ C p r2 c~ þ C p r4 h: (9b) nian, 2017). Many of the results discussed in this section are valid
in the limit of small membrane deformations, and it remains an
Here, τ is a timescale dependent on friction, and inversely propor- interesting problem for future research to model curvature-mediated
tional to the bending rigidity. The last term in each of these equa- patterning in the large deformation limit (see Box 2).
tions describes the protein–membrane feedback loop, whose In the biological context, curvature-coupled proteins do not oper-
αv ¼ ∂x σ; (11)
units that contribute to a larger, tissue-scale pattern. Adherent cells
typically exert active contractile stresses on their environment
(Schwarz et al, 2002; Schwarz & Safran, 2013; Tanimoto & where the left-hand side corresponds to an external frictional stress
Sano, 2014). Similar to the way in which myosin molecules inside with friction coefficient α as before. For a contractile Newtonian
the actin network act as contractile elements, the cells themselves fluid, the stress consists of two parts: σ ¼ η∂x v þ σ A . Here, η is the
can constitute contractile units within a meshwork of extracellular viscosity, and the active stress σ A ðcÞ depends on the concentration
matrix (ECM) filaments. As such, they can induce active rearrange- of the regulator. For contractile regulators, an increase in c leads to
ments of the cell–ECM system, leading to contractile patterning at an increase in stress. Many studies assume a saturating relation
the multicellular level (Fig 4B, Harris et al, 1984; Shyer et al, 2017; between the active stress and the concentration of contractile
Palmquist et al, 2022). Systems consisting of contractile cells particles (e.g., Bois et al, 2011; Mietke et al, 2019b). However, in
embedded in extracellular matrix were, in fact, one of the earliest the simplest case, valid for low concentrations, each unit (e.g.,
systems where mechanical forces were studied as a biological myosin molecule or contractile cell) imparts the same average
pattern-forming mechanism (Oster et al, 1983; Harris et al, 1984). stress σ A to the material. Then, the active stress contribution
To study contractile patterning theoretically, we write the conti- depends linearly on the concentration of contractile particles,
nuity equation (equation 1) for the concentration of contractile par- which we write relative to the average concentration c0 :
convective flow 0
10/R
B 0
0 10/R 20/R 30/R 0 10 20
20 min
Perturbation wavenumber q Péclet number
D 150
1 μm
Péclet number
100 Unstable
20 h
50
Stable
0
1000 μm 0 0.25 0.5 0.75 1
Hydrodynamic length Lh/R
!
σA
ω ¼ Dq2 1 αD
: (12) vivo, the depletion of material in between aggregates eventually
1 þ αη q2 appears to prevent sustained flows.
Another experimental system in which a contractile instability
The equation reveals that the existence of an instability only has been studied is the Drosophila tracheal tube. For this system,
depends on two quantities. The ratio σ A =ðαDÞ is called the Peclet Hannezo et al (2015) included a reaction term of the form
number and indicates the relative strength of the advective trans-
pffiffiffiffiffiffiffiffi RðcÞ ¼ ðc0 cÞ=τ to describe turnover at a timescale τ. The linear
port compared with the diffusive transport. The ratio η=α is a stability analysis shows that here too, patterns emerge when advec-
length scale characterizing the spatial range of flows called the tion dominates over other effects. Like diffusion, the turnover coun-
hydrodynamic length. There are solutions ω > 0, corresponding to teracts the self-amplifying feedback loop of accumulation. Genetic
the emergence of patterns, only when the Peclet number is larger and pharmacological perturbations to actin patterns in these cells
than 1 (Fig 4C). This shows that patterns form if the hydrodynamic also yield wavelength changes that match theoretical predictions.
flow induced by active stress is strong enough to overcome the sta- Kumar et al (2014) extended the framework of contractile insta-
bilizing effect of random diffusive motion. bilities to two chemical species. One species upregulates the active
The fundamental pattern-forming process is a self-amplifying stress and another downregulates it. In addition to stationary pat-
feedback loop where more regulator means more inward flow, terns, such a system also shows an oscillatory instability which
velocities of migrating tumor (Hawkins et al, 2011; Bergert a thin object is subject to compressive forces, which lead to its bend-
et al, 2015), immune (Liu et al, 2015), and embryonic progenitor ing. Buckling instabilities are also relevant for pattern formation in
cells (Ruprecht et al, 2015). biological systems, where they can lead to regular wrinkles or folds.
The dimension and shape of the domain on which the actin cor- The formation of folds is important in many systems, including the
tex is modeled are important for the resulting pattern, by gut (Savin et al, 2011; Shyer et al, 2013), the wrinkles of the brain
constraining the spatial modes that can become unstable (see also (Llinares-Benadero & Borrell, 2019), the development of the Dro-
Box 1, Fig 1C and section “Geometry-dependent reaction-diffusion sophila wing (Tozluoğlu et al, 2019), and others (see Nelson, 2016,
patterns”). In real biological systems, contractile stresses can lead to for a review). Buckling due to differential growth is a key mecha-
deformations of this shape. Some studies have addressed how non- nism in the determination of leaf shapes (reviewed in Guo
uniform contractile stresses and flows change the system’s shape et al, 2022). We do not discuss this in this review, but mechanical
(Ruprecht et al, 2015; Callan-Jones et al, 2016), and how this in turn forces in general play an important role in plant morphogenesis (see
may feed back onto the flow pattern (Mietke et al, 2019a, 2019b). Sampathkumar, 2020, for a review).
Ruprecht et al (2015) and Callan-Jones et al (2016) study shape Other mechanisms besides buckling, such as localized contrac-
deformations due to cortical flows. As in the paper by Hawkins tion or non-uniform growth, can also lead to folding (see Tozluoğlu
et al (2011), these studies were inspired by cell migration in three & Mao, 2020, for an overview). In biological systems, it is not easy
A B
C D E
only elastic penalties to changes in the height. As before, one can this problem, see the paper by Almet et al (2020) and references
do an analysis of the stability of a flat epithelial layer, which therein.
shows that a pattern-forming instability is possible when the Buckling instabilities may be followed by changes in signaling,
growth of the layer overcomes the resistance to bending and the as has been shown for example in the formation of villi in chick
resistance to deformation of the substrate. The wavelength of the embryos (Shyer et al, 2015). There, the curved geometry induced by
pattern depends on the mechanical properties of the tissues the folding leads to a patterned distribution of morphogens, which
involved (Brau et al, 2013). The equation above is a simplified impacts the localization of stem cells in this system. There is thus
illustration; for an overview of different modeling approaches to an interplay among mechanics, geometry, and signaling.
Buckling due to constrained growth is not only seen in epithelial et al (2017), Petridou et al (2017), and Kindberg et al (2020), and
tissues but also in multicellular systems of another kind: bacterial references therein, for more information.
biofilms. While a biofilm’s size increases by the division and growth In this section, we discuss one particular example where the ability
of bacteria, its substrate does not grow. This leads to the accumula- of cells to respond to mechanical changes gives rise to pattern forma-
tion of stresses that result in a buckling instability and the appear- tion, involving an interplay between extracellular signal-regulated
ance of wrinkles in the biofilm (Figs 5B and 1A). In a study by Yan kinase (ERK) activity and cellular shape. The pattern discussed in this
et al (2019), these wrinkles develop into a striking star-like pattern section is varying not only in space but also in time, in the form of
with a well-defined wavelength. Classical theory for film–substrate waves. Waves of ERK activity and contractility appear in systems
systems predicts that the wrinkle wavelength should scale as of collectively moving cells (Aoki et al, 2017; Hino et al, 2020;
ðGf =Gs Þ1=3 , where Gf and Gs are the shear moduli of biofilm and sub- Boocock et al, 2021). Collective cell motion is important for many bio-
strate. These moduli, which characterize a material’s response to logical processes in the development and maintenance of an organ-
shear stresses, can be measured using a shear rheometer. The scal- ism, and a wide range of physical modeling approaches exists (see
ing relation is seen to hold true in a range of stiffnesses, but the Alert & Trepat, 2020, for a review). Collectives of moving cells in epi-
two-layer model is not an accurate description of the data for small thelia show varied dynamics, including oscillatory motion and the
Gf . Yan et al (2019) show that a model including a third layer is a propagation of mechanical waves (see, e.g., Dierkes et al, 2014;
In the preceding sections, we mainly discussed mechanochemical in which spatial scales have been normalized to the reference
systems in which feedback arises directly due to geometrical or length of an individual cell. The time scale τr depends on the fric-
mechanical effects on the distribution of constituents, such as tion of the cells with the substrate and the strength of the spring-
curvature-dependent localization (section “Pattern formation by like coupling.
membrane curvature”) or convective flows (section “Contractile The equation for r is complemented by one for the preferred cell
fluid patterning”). At the multicellular level, however, the capacity length, ‘0 and a third variable E that represents the local ERK activ-
of cells to sense mechanical properties and forces through special ity. Hino et al (2020) found that ERK activity induces cell contrac-
mechanosensory machineries and signaling pathways can become tion. This can be modeled by stating that ERK levels tend to
relevant for pattern formation. Mechanical forces sensed by the cell decrease ‘0 . Experiments have also shown that cell extension
can lead, among others, to changes in gene expression, subcellular induces an increase in ERK levels, indicating that a mechanosensory
localization, or post-translational modifications of proteins. In turn, process gives rise to feedback between the cell’s mechanical and
these biochemical changes can affect the mechanical properties of a biochemical states (Fig 5D). However, how exactly the mechanical
cell. In this way, mechanosensors can be part of mechanochemical changes lead to a change in ERK activity is not fully understood
feedback loops. Mechanosensing and mechanotransduction are (Hirashima et al, 2023). How this feedback is implemented thus
important processes, but a more detailed discussion lies outside the becomes a modeling choice: Is ERK activation induced by absolute
scope of the current paper. We refer to the reviews by Chan cell size, the cell strain, or strain rate? Boocock et al (2021), in an
extended discussion of the model’s assumptions, show that only the The directed propagation of ERK waves, which corresponds to cell
first choice leads to patterns in the model. This could provide a the- migration in the direction opposite to the waves, can be explained by
oretical pointer toward the biological mechanism. coupling the ERK–cell extension dynamics to cell polarity. Boocock
The equations for the model, with preferred cell length and ERK et al (2021) investigate a model where cell polarity is induced by
levels rewritten as a deviation from a basal level, are given by stress gradients. Cell polarity, in turn, affects the evolution of r. Simu-
lations of this extended model show that the system can indeed pro-
τr ∂t r ¼ ∂xx r∂x ‘0 ; duce collective cell migration. In this model, ERK waves that travel
τ‘ ∂t ‘0 ¼ ‘0 aE; (15) away from a free edge correspond to a persistent polarization of the
τE ∂t E ¼ E þ b∂x r: follower cells, thereby guiding them to migrate collectively in the
direction of the edge. Moreover, the model predicts that there is an
The three constants τr ; τ‘ and τE are the timescales of evolution of optimal wavelength and frequency of ERK waves that produce the
the three variables. They are, like the Peclet number in the section strongest average polarization. The ERK waves that have been
“Contractile fluid patterning”, combinations of physical parameters observed in experiments have frequency and wavelength which
such as friction constants and elastic coefficients. Importantly, roughly correspond to these optimal values, which suggests that the
these timescales correspond to directly measurable quantities: if system is tuned for an optimal migration speed.
Equation (16) is an equation for discrete oscillators, which is the Importantly, the range of spatial coupling determines the stability of
form mostly used for studying metachronal waves. Some studies, metachronal waves (Wollin & Stark, 2011; Brumley et al, 2016).
however, use continuous phase equations, where the phase is given The mechanical aspect relevant to metachronal waves is mainly
by a field φðx; tÞ (Fig 5G, Chakrabarti et al (2022) and Quillen (2023)). the hydrodynamic coupling. However, the shape of the domain on
Interestingly, continuous phase equations with non-local coupling which the ciliary arrays are present can also be important. In the
also appear in models of collections of neurons, where long-range natural world, ciliary arrays often appear on non-flat domains such
interactions are due to the spatial extent of axons (e.g., Crook et al, as spheres. The question of how different domain shapes affect the
1997). A continuous version of equation (16) reads synchronization of coupled cilia has recently been addressed theo-
retically (e.g., Westwood & Keaveny, 2021, and references therein).
Z
Ciliary arrays are not the only biological systems that show meta-
∂t φðx; tÞ ¼ ωðxÞ þ dyGðjxyjÞH ðφðx; tÞ; φðy; tÞÞ: (17)
chronal waves. They also appear on the level of multiple organisms,
as has been described by Peshkov et al (2022). In this study, the
A uniformly oscillating system is a solution to this equation with authors observe swimming nematodes (T. aceti or vinegar eels, see
φðx; tÞ ¼ ωt. It is, similar to the examples in the other sections, also Fig 1A). These small worms, when confined in liquid droplets,
possible to do a linear stability analysis of this state. Moreover, synchronize their undulating motion when their density is high. The
close to a steady state, it proves to be remarkably useful in under- for example (equations 19, Box 1), the non-linearity arises from a
standing pattern formation in general settings. Yet, a more complete trimolecular reaction term with a single rate constant. For mass
understanding of patterns, including their long-time behavior, can action kinetics, this is unrealistic since it requires the simultaneous
require additional theoretical tools or computer simulations (Box 2). collision of three molecules with a non-negligible probability. Many
In many living systems, processes are coupled across different spatio- biochemical pattern-forming models make assumptions of—often
temporal scales and levels of biological organization. For example, in unknown—reactions that are summarized by a few non-linear
some cases, pattern formation in a tissue can be described by contin- terms. Taking the—typically known—effects of mechanics and
uum equations, treating the cells as microscopic units whose dynam- geometry into account can make invoking additional biochemical
ics can be described by fields, but sometimes the cell’s individual complexity unnecessary.
features matter too, in which case a model with cells as discrete units The inevitability of geometry and mechanical interactions raises
may be more appropriate. In particular, when microscopic units the question of how patterns that rely on mechanics have evolved,
adapt and respond to changes at larger scales, behaviors can arise and whether mechanical pattern formation is more robust com-
that are not seen in passive physical systems, motivating theoretical pared to genetic or biochemical patterns (Box 2). It will be inter-
approaches that not only explain how large-scale phenomena arise esting to investigate—for specific cases and across different
from microscopic dynamics but which can also capture the effects of contexts—whether patterns or their functions become less vulnera-
Alt S, Ganguly P, Salbreux G (2017) Vertex models: from cell mechanics to Boocock D, Hino N, Ruzickova N, Hirashima T, Hannezo E (2021) Theory of
tissue morphogenesis. Philos Trans R Soc Lond B Biol Sci 372: 20150520 mechanochemical patterning and optimal migration in cell monolayers.
Ambrosi D, Ben Amar M, Cyron CJ, DeSimone A, Goriely A, Humphrey JD, Kuhl Nat Phys 17: 267 – 274
E (2019) Growth and remodelling of living tissues: perspectives, challenges Bowick MJ, Fakhri N, Marchetti MC, Ramaswamy S (2022) Symmetry,
and opportunities. J R Soc Interface 16: 20190233 thermodynamics, and topology in active matter. Phys Rev X
Aoki K, Kondo Y, Naoki H, Hiratsuka T, Itoh RE, Matsuda M (2017) 12: 010501
Propagating wave of ERK activation orients collective cell migration. Dev Brangwynne CP, MacKintosh FC, Kumar S, Geisse NA, Talbot J, Mahadevan L,
Cell 43: 305 – 317 Parker KK, Ingber DE, Weitz DA (2006) Microtubules can bear enhanced
Argudo D, Bethel NP, Marcoline FV, Grabe M (2016) Continuum descriptions compressive loads in living cells because of lateral reinforcement. J Cell
of membranes and their interaction with proteins: towards chemically Biol 173: 733 – 741
accurate models. Biochim Biophys Acta 1858: 1619 – 1634 Brau F, Damman P, Diamant H, Witten TA (2013) Wrinkle to fold transition:
Ayed SH, Cloutier AD, McLeod LJ, Foo ACY, Damry AM, Goto NK (2017) influence of the substrate response. Soft Matter 9: 8177 – 8186
Dissecting the role of conformational change and membrane binding by Brauns F, Halatek J, Frey E (2020) Phase-space geometry of mass-conserving
the bacterial cell division regulator MinE in the stimulation of MinD reaction-diffusion dynamics. Phys Rev X 10: 041036
ATPase activity. J Biol Chem 292: 20732 – 20743 Brauns F, Pawlik G, Halatek J, Kerssemakers J, Frey E, Dekker C (2021)
Cytrynbaum EN, Marshall BDL (2007) A multistranded polymer model Gordon VD, Davis-Fields M, Kovach K, Rodesney CA (2017) Biofilms and
explains MinDE dynamics in E. coli cell division. Biophys J 93: 1134 – 1150 mechanics: a review of experimental techniques and findings. J Phys D
Dasanayake NL, Michalski PJ, Carlsson AE (2011) General mechanism of Appl Phys 50: 223002
actomyosin contractility. Phys Rev Lett 107: 118101 Goriely A (2017) The mathematics and mechanics of biological growth, Vol. 45,
De Gennes PG, Taupin C (1982) Microemulsions and the flexibility of oil/water Interdisciplinary applied mathematics. New York, NY: Springer
interfaces. J Phys Chem 86: 2294 – 2304 Gov NS (2018) Guided by curvature: shaping cells by coupling curved
Denk J, Kretschmer S, Halatek J, Hartl C, Schwille P, Frey E (2018) MinE membrane proteins and cytoskeletal forces. Philos Trans R Soc Lond B Biol
conformational switching confers robustness on self-organized Min Sci 373: 20170115
protein patterns. Proc Natl Acad Sci U S A 115: 4553 – 4558 Graner F, Glazier JA (1992) Simulation of biological cell sorting using a two-
Deserno M (2015) Fluid lipid membranes: from differential geometry to dimensional extended Potts model. Phys Rev Lett 69: 2013 – 2016
curvature stresses. Chem Phys Lipids 185: 11 – 45 Green JBA, Sharpe J (2015) Positional information and reaction-diffusion:
Deserno M, Kremer K, Paulsen H, Peter C, Schmid F (2014) Computational two big ideas in developmental biology combine. Development 142:
studies of biomembrane systems: theoretical considerations, 1203 – 1211
simulation models, and applications. In From Single Molecules to Gross P, Kumar KV, Grill SW (2017) How active mechanics and regulatory
Nanoscopically Structured Materials, Advances in Polymer Science, Basche biochemistry combine to form patterns in development. Annu Rev Biophys
Haussman RC, Deserno M (2014) Effective field theory of thermal Casimir Kuramoto Y, Battogtokh D (2002) Coexistence of coherence and incoherence
interactions between anisotropic particles. Phys Rev E Stat Nonlin Soft in nonlocally coupled phase oscillators. Nonlinear Phenom Complex Syst 5:
Matter Phys 89: 062102 380
Hawkins RJ, Poincloux R, Benichou O, Piel M, Chavrier P, Voituriez R (2011) Kurtisovski E, Taulier N, Ober R, Waks M, Urbach W (2007) Molecular origin
Spontaneous contractility-mediated cortical flow generates cell migration of model membrane bending rigidity. Phys Rev Lett 98: 258103
in three-dimensional environments. Biophys J 101: 1041 – 1045 Kwiecinski J, Chapman SJ, Goriely A (2017) Self-assembly of a filament by
Heinrich MC, Capraro BR, Tian A, Isas JM, Langen R, Baumgart T (2010) curvature-inducing proteins. Phys D Nonlinear Phenom 344: 68 – 80
Quantifying membrane curvature generation of drosophila amphiphysin Lauga E, Powers TR (2009) The hydrodynamics of swimming microorganisms.
N-BAR domains. J Phys Chem Lett 1: 3401 – 3406 Rep Prog Phys 72: 096601
Helfrich W (1973) Elastic properties of lipid bilayers: theory and possible Leibler S (1986) Curvature instability in membranes. J Phys 47: 507 – 516
experiments. Z Naturforsch C 28: 693 – 703 Lin S, Hirayama D, Maryu G, Matsuda K, Hino N, Deguchi E, Aoki K, Iwamoto
Hino N, Rossetti L, Marin-Llaurado A, Aoki K, Trepat X, Matsuda M, Hirashima R, Terai K, Matsuda M (2022) Redundant roles of EGFR ligands in the ERK
T (2020) ERK-mediated mechanochemical waves direct collective cell activation waves during collective cell migration. Life Sci Alliance 5:
polarization. Dev Cell 53: 646 – 660 e202101206
Hirashima T, Hino N, Aoki K, Matsuda M (2023) Stretching the limits of Lipowsky R, Döbereiner H-G, Hiergeist C, Indrani V (1998) Membrane
Mim C, Unger VM (2012) Membrane curvature and its generation by BAR Quaas J, Wylie C (2002) Surface contraction waves (SCWs) in the Xenopus
proteins. Trends Biochem Sci 37: 526 – 533 egg are required for the localization of the germ plasm and are
Müller MM, Deserno M, Guven J (2005) Interface-mediated interactions dependent upon maternal stores of the kinesin-like protein Xklp1. Dev Biol
between particles: a geometrical approach. Phys Rev E 72: 061407 243: 272 – 280
Murray JD (1982) Parameter space for Turing instability in reaction diffusion Quillen AC (2023) Robust formation of metachronal waves in directional
mechanisms: a comparison of models. J Theor Biol 98: 143 – 163 chains of phase oscillators. Phys Rev E 107: 034401
Murray JD (2003) Mathematical biology: II: spatial models and biomedical Quillen AC, Peshkov A, Wright E, McGaffigan S (2021a) Metachronal waves in
applications, Vol. 18. Interdisciplinary applied mathematics. New York, NY: concentrations of swimming Turbatrix aceti nematodes and an oscillator
Springer chain model for their coordinated motions. arXiv https://arxiv.org/abs/
Naganathan SR, Oates AC (2017) Mechanochemical coupling and 2101.06809 [PREPRINT]
developmental pattern formation. Curr Opin Syst Biol 5: 104 – 111 Quillen AC, Peshkov A, Wright E, McGaffigan S (2021b) Metachronal waves in
National Academies of Sciences, Engineering, and Medicine (2022) Physics of concentrations of swimming Turbatrix aceti nematodes and an oscillator
life. Washington, DC: The National Academies Press chain model for their coordinated motions. Phys Rev E 104: 014412
Nedelec F, Foethke D (2007) Collective Langevin dynamics of flexible Ramakrishnan N, Sunil Kumar PB, Radhakrishnan R (2014) Mesoscale
cytoskeletal fibers. New J Phys 9: 427 computational studies of membrane bilayer remodeling by curvature-
Sawai T (1982) Wavelike propagation of stretching and shrinkage in the Trushko A, Di Meglio I, Merzouki A, Blanch-Mercader C, Abuhattum S, Guck J,
surface of the newt’s egg before the first cleavage. J Exp Zool 222: Alessandri K, Nassoy P, Kruse K, Chopard B et al (2020) Buckling of an
59 – 68 epithelium growing under spherical confinement. Dev Cell 54: 655 – 668
Schamberger B, Ziege R, Anselme K, Ben Amar M, Bykowski M, Castro APG, Tu ZC, Ou-Yang ZC (2014) Recent theoretical advances in elasticity of
Cipitria A, Coles RA, Dimova R, Eder M et al (2023) Curvature in biological membranes following Helfrich’s spontaneous curvature model. Adv Colloid
systems: its quantification, emergence, and implications across the scales. Interface Sci 208: 66 – 75
Adv Mater 35: 2206110 Turing AM (1952) The chemical basis of morphogenesis. Philos Trans R Soc
Schnakenberg J (1979) Simple chemical reaction systems with limit cycle Lond B Biol Sci 237: 37 – 72
behaviour. J Theor Biol 81: 389 – 400 Uchida N, Golestanian R (2010) Synchronization and collective dynamics in a
Schwarz US, Safran SA (2013) Physics of adherent cells. Rev Mod Phys 85: carpet of microfluidic rotors. Phys Rev Lett 104: 178103
1327 – 1381 Ul Islam T, Wang Y, Aggarwal I, Cui Z, Eslami Amirabadi H, Garg H, Kooi R,
Schwarz US, Balaban NQ, Riveline D, Bershadsky A, Geiger B, Safran SA (2002) Venkataramanachar BB, Wang T, Zhang S et al (2022) Microscopic
Calculation of forces at focal adhesions from elastic substrate data: the artificial cilia – a review. Lab Chip 22: 1650 – 1679
effect of localized force and the need for regularization. Biophys J 83: Valet M, Siggia ED, Brivanlou AH (2022) Mechanical regulation of early
1380 – 1394 vertebrate embryogenesis. Nat Rev Mol Cell Biol 23: 169 – 184
Würthner L, Brauns F, Pawlik G, Halatek J, Kerssemakers J, Dekker C, Frey E Zaritsky A, Kaplan D, Hecht I, Natan S, Wolf L, Gov NS, Ben-Jacob E,
(2022) Bridging scales in a multiscale pattern-forming system. Proc Natl Tsarfaty I (2014) Propagating waves of directionality and
Acad Sci U S A 119: e2206888119 coordination orchestrate collective cell migration. PLoS Comput Biol 10:
Yan J, Fei C, Mao S, Moreau A, Wingreen NS, Kosmrlj A, Stone HA, Bassler BL e1003747
(2019) Mechanical instability and interfacial energy drive biofilm Zhou W, Hao Z, Gravish N (2021) Collective synchronization of undulatory
morphogenesis. Elife 8: e43920 movement through contact. Phys Rev X 11: 031051
Yang S, Miao X, Arnold S, Li B, Ly AT, Wang H, Wang M, Guo X, Pathak MM, Zieske K, Schwille P (2013) Reconstitution of pole-to-pole oscillations of Min
Zhao W et al (2022) Membrane curvature governs the distribution of proteins in microengineered polydimethylsiloxane compartments. Angew
Piezo1 in live cells. Nat Commun 13: 7467 Chem Int Ed 52: 459 – 462
Yeomans JM (2017) Nature’s engines: active matter. Europhys News 48: Zieske K, Schwille P (2014) Reconstitution of self-organizing protein gradients
21 – 25 as spatial cues in cell-free systems. Elife 3: e03949
Yolcu C, Haussman RC, Deserno M (2014) The Effective Field Theory Zimmerberg J, Kozlov MM (2006) How proteins produce cellular membrane
approach towards membrane-mediated interactions between particles. curvature. Nat Rev Mol Cell Bio 7: 9 – 19
Adv Colloid Interface Sci 208: 89 – 109
Yoneda M, Kobayakawa Y, Kubota HY, Sakai M (1982) Surface contraction License: This is an open access article under the