Article

Viscoelastic Dissipation Stabilizes Cell Shape

Changes during Tissue Morphogenesis
Highlights Authors
d Myosin-driven and optical forces reveal the dissipative ment, Benoı̂t Dehapiot,
Raphaël Cle
mechanics of cell contacts Claudio Collinet, Thomas Lecuit,
Pierre-François Lenne
d Deformations of cell contacts are stabilized by dissipation on
the minute timescale Correspondence
d Longer force application yields less reversible deformations [email protected] (R.C.),
[email protected]
d Dissipation and thus reversibility of deformations rely partly (P.-F.L.)
on actin turnover
In Brief
ment et al. investigate the mechanics
Cle
underlying the irreversibility of
morphogenetic deformations. They show
that deformations caused by contractile
pulses of myosin II or by optical tweezers
are stabilized by dissipation on the
minute timescale. Inhibiting actin
turnover reduces dissipation and yields
more reversible deformations.

ment et al., 2017, Current Biology 27, 3132–3142

Cle
October 23, 2017 ª 2017 Elsevier Ltd.
http://dx.doi.org/10.1016/j.cub.2017.09.005
 Current Biology

Article

Viscoelastic Dissipation Stabilizes

Cell Shape Changes during Tissue Morphogenesis

ment,1,3,* Benoı̂t Dehapiot,1 Claudio Collinet,1 Thomas Lecuit,1,2 and Pierre-François Lenne1,*
Raphaël Cle
1Aix-Marseille Univ., CNRS, IBDM, Marseille, France
2Collège de France, 11 place Marcelin Berthelot, Paris, France
3Lead Contact

*Correspondence: [email protected] (R.C.), [email protected] (P.-F.L.)

http://dx.doi.org/10.1016/j.cub.2017.09.005

SUMMARY elastic nor viscous but viscoelastic. Whereas elasticity provides

robustness to short-lived stresses, dissipation allows the system
Tissue morphogenesis relies on the production to change shape without dramatic stress accumulation, which
of active cellular forces. Understanding how such can threaten cellular integrity [9]. Thus, a likely mechanism to sta-
forces are mechanically converted into cell shape bilize deformation is the gradual dissipation of elastic energy dur-
changes is essential to our understanding of ing the morphogenetic process [9]. Dissipation should therefore
morphogenesis. Here, we use myosin II pulsatile ac- occur on timescales typically shorter, or commensurate with, but
definitely not much longer than the timescale of force generation.
tivity during Drosophila embryogenesis to study how
Hence, it is often assumed that tissues behave as fluids on devel-
transient forces generate irreversible cell shape
opmental timescales [10, 11]. Experimentally, the viscoelastic
changes. Analyzing the dynamics of junction short- nature of single cells or cell assemblies have been assessed
ening and elongation resulting from myosin II pulses, with a variety of techniques [12–15], and recent works using sus-
we find that long pulses yield less reversible defor- pended epithelial monolayers [16] confirm that dissipation oc-
mations, typically a signature of dissipative me- curs upon application of a constant deformation, typically on
chanics. This is consistent with a simple viscoelastic the minute timescale [17]. Yet very few studies involve direct me-
description, which we use to model individual short- chanical measurements in vivo in the context of morphogenesis.
ening and elongation events. The model predicts that In a previous study using the Drosophila embryonic epithelium as
dissipation typically occurs on the minute timescale, a model system, we delineated the mechanics of cell contacts at
a timescale commensurate with that of force genera- short timescales, typically below a few seconds, showing that
the response was mainly elastic and damped by fluid friction in
tion by myosin II pulses. We test this estimate by
the cytosol [18]. However, longer timescales remain largely un-
applying time-controlled forces on junctions with
explored, due to the technical challenge of actively probing cell
optical tweezers. Finally, we show that actin turnover mechanics in live embryos over long periods of time. Thus, a me-
participates in dissipation, as reducing it pharmaco- chanical analysis at morphogenetic timescales and an estimate
logically increases the reversibility of contractile of the relevant dissipation timescales in a morphogenetically
events. Our results argue that active junctional defor- active tissue are still lacking.
mation is stabilized by actin-dependent dissipation. Here, we take advantage of the irreversible shortening and
Hence, tissue morphogenesis requires coordination elongation of cell junctions of the Drosophila germband in
between force generation and dissipation. response to pulsatile myosin II activity to estimate the typical
timescale of dissipation at cell contacts. Thereby, we sought to
understand how transient forces generate irreversible deforma-
INTRODUCTION tion during this morphogenetic process.
Planar-polarized MyoII activity was shown to drive cell interca-
The course of animal development is a succession of morphoge- lation and participate in tissue extension in the germband [3, 4].
netic movements, which require the activity of force-generating Oscillations, or pulses, of MyoII activity along junctions aligned
cortical components that exert mechanical forces at the cellular with the dorso-ventral axis (Figure 1A) were shown to gradually
scale [1, 2]. Classic examples include cell intercalation, in which shorten junctions, acting as a ‘‘mechanical ratchet’’ [8], and
polarized activity of myosin II (MyoII) motors drives tissue elon- eventually resulting in the disappearance of vertical cell contacts
gation [3, 4], or apical constriction, in which MyoII drives tissue (Figures 1B, 1D, and 1F). Similar MyoII pulsations, located in
folding and invagination [5]. At the cell level, deformations rely adjacent anterior and posterior cells in the vicinity of a junction’s
on transient forces resulting from the pulsatile activity of MyoII vertices, were later shown to favor the formation and gradual
[5–8]. A key question is thus how these transient forces can result elongation of new junctions along the antero-posterior axis (Fig-
in persistent deformations. This requires a mechanical under- ures 1C, 1E, and 1G) [19]. Previous studies have shown that effi-
standing of how cells and tissues dissipate energy and thereby cient remodeling requires robust activation of MyoII in junctions
escape elastic recoil once contractile stresses (e.g., forces) are [8, 20, 21], and whereas it is clear that MyoII is required to
no longer applied. Indeed, cells and tissues are neither purely generate the force driving deformation, the mechanism by which

3132 Current Biology 27, 3132–3142, October 23, 2017 ª 2017 Elsevier Ltd.
 Figure 1. Junction Shortening and Elonga-
A B
tion by MyoII Polarized Activity
(A) Schematics of the germband of the Drosophila
embryo. A, P, D, and V symbolize the anterior,
posterior, dorsal, and ventral directions.
(B) Cartoon of junction shortening driven by MyoII
C pulsatile activity. Actin is in magenta and MyoII in
cyan.
(C) Cartoon of junction elongation driven by MyoII
pulsatile activity. Actin is in magenta and MyoII in
cyan.
D (D) Junction undergoing shortening in the germ-
F H
band (E-cadherin, red; MyoII, cyan). MyoII is
recruited at the junction. The scale bar repre-
sents 5 mm.
(E) Junction undergoing elongation in the germ-
band. MyoII is recruited in the anterior and pos-
terior cells, in the vicinity of vertices of newly
forming junctions.
(F) Shortening dynamics during a pulse of junc-
tional MyoII.
E G I
(G) Elongation dynamics during a pulse of MyoII
located in the vicinity of vertices.
(H) Region analyzed to measure MyoII activity
during shortening.
(I) Region analyzed to measure MyoII activity
during elongation.

the resulting deformations are stabilized remains elusive. In the activation and is used as a proxy for motor activity. Whereas it
following, we introduce a minimal viscoelastic model of the so- is largely assumed that force increases with MyoII activity, the
called mechanical ratchet, which shows that active MyoII-driven precise scaling between MyoII and force production is un-
deformations can be stabilized by dissipation. Analyzing junc- known. For the sake of simplicity, we assume in the following
tional shortening and elongation in response to pulsatile forces that force is proportional to the excess of MyoII (with respect
generated by MyoII, we estimate the typical timescale of dissipa- to MyoII base level M0 observed in absence of pulses), so that
tion, which separates the elastic, reversible regime from the f = ± aðM  M0 Þ. The ± sign stands for shortening () or elonga-
viscous, irreversible regime. We find that it is typically one minute tion (+). M0 represents the amount of MyoII required to maintain
and confirm this estimate with optical tweezers experiments in constant junction length so that the junction is at rest when
which a controlled external force is applied to junctions. Finally, M = M0 . In practice, we assume that M0 is the minimum of
we show that pharmacological reduction of actin turnover limits MðtÞ, which is consistent with junctions being at rest in the
dissipation, so that contractile events in perturbed embryos are absence of pulses. Finally, a is an unknown scaling factor.
overall more reversible. From there, the dynamics of a viscoelastic junction submitted
to MyoII-driven shortening or elongation is given by the following
RESULTS reformulation of the Maxwell model of viscoelasticity (see STAR
Methods for details):
Our working hypothesis is that cell junctions are viscoelastic with m
a short-term elastic response and a long-term viscous response ± ka l_= m_ + : (Equation 1)
t
due to dissipation, allowing deformations generated by transient
MyoII activity to be stabilized. Note that cell junctions, whose In Equation 1, ka is the ratio between the elastic modulus k and
cortex consists of a complex and dynamic network of actin fila- the constant a, and m = M  M0 . Dots denote time derivatives,
ments that turn over and are crosslinked by a number of linkers so that l_ is the rate of elongation and m_ the rate of MyoII recruit-
[22], are likely to undergo dissipation on a distribution of time- ment. Note that, here, we neglect variations of tension in adja-
scales. For the sake of simplicity, we assume in the following cent junctions as well as the contribution of viscous drag in the
that dissipation occurs on a single typical timescale t. cytosol, which occurs on timescales much shorter than MyoII
Junction length dynamics during contractile pulses is corre- fluctuations [18]. The model has two free parameters: ka , which
lated with the fluctuation of MyoII intensity [5, 8], which is known controls how much MyoII can elastically deform a junction, and
to generate forces at junctions [23]. Therefore, we ultimately t, the typical dissipation timescale. In the case of shortening,
want the model to predict the associated dynamics of MyoII M is MyoII fluorescent intensity measured in the vicinity of
intensity MðtÞ and junction length lðtÞ. Because a mechanical the junction (Figure 1H), where it is geometrically the most
model relates force to deformation, we first need to relate MyoII effective to exert a shortening force. This view is supported by
intensity to force fðtÞ. MyoII recruitment at the cortex requires apical cell constriction events occurring prior to mesoderm

Current Biology 27, 3132–3142, October 23, 2017 3133

 A B

C D

E F

Figure 2. Viscoelastic Analysis of Contractile Pulses

(A) Shortening dynamics of a model viscoelastic junction in response to artificial Gaussian pulses. Black, t = 1h [ q (elastic regime); light gray, t = 1s  q
(viscous regime); dark gray, t = 1min  q (intermediate regime).
(B) Left: Cartoon of the irreversibility index I, defined as the ratio between the irreversible deformation and the maximal deformation. Right: Irreversibility index IðqÞ
increases with pulse duration q (n = 49 from 5 embryos). Center lines of boxes show the medians; boxes limits indicate the 25th and 75th percentiles; and whiskers
extend 1.5 times the interquartile range from the 25th and 75th percentiles.
(C) Time shift between the peak of deformation rate (here, shortening rate, dotted line) and the peak of MyoII activity (cyan). The former is ahead of the latter, both
in a model pulse (left panel) and in experiments (right panel).
(D) Boxplot of the time shift between the peak of deformation rate and the peak of MyoII activity, measured for shortening events (n = 40 from 4 embryos) and
elongation events (n = 35 from 3 embryos).
(E) Fit of junction shortening (left panel) and junction elongation (right panel) using the experimental MyoII signals depicted in Figures 1H and 1I as input for the
model.
(F) Estimates of t extracted from multiple fits of experiments, both for shortening (n = 16 from 3 embryos) and elongation (n = 15 from 3 embryos).

invagination, where MyoII accumulation in the medial part of the The model is, in essence, timescale dependent. It is expected
cell essentially shrinks cell area, with a mild effect on junction that the longer a force is applied, the less reversible the resulting
length. Importantly, this does not imply that MyoII from the deformation should be. Conversely, the shorter the dissipation
medial region plays no role in junction shortening, as MyoII is timescale is, the faster one will observe irreversible deforma-
known to flow toward junctions from the medial region [8]. In tions. This is illustrated in Figure 2A, where we show the pre-
the case of elongation, junctions have very low MyoII levels, dicted response of a junction to artificial MyoII pulses in three
and we measure MyoII intensity M in adjacent anterior and pos- extreme scenarios. As expected, when the timescale of dissipa-
terior cells in the vicinity of vertices (Figure 1I), where it was tion t is much longer than the pulse duration q (that is, when
shown to drive the growth of the new junction [19, 24]. t [ qÞ, the response is mainly elastic and deformations are

3134 Current Biology 27, 3132–3142, October 23, 2017

 A B E

F G

H I

Figure 3. Viscoelastic Analysis of Optical Tweezers Experiments

(A) Cartoon of the pull-release sequence and of the subsequent forces in the optical tweezers experiments. The trap is initially off (red), then switched on (green),
and switched off again (red). Forces displayed are the restoring force f, the drag force fd , and the trap force ft .
(B) Cartoon of deflection quantification. We monitor deflections along the x axis, perpendicular to the junction. xm is the junction position and xt the optical trap
position.
(C) Time-lapse sequence of a pull-release experiment, in which the junction is held for approximately 50 s.

(legend continued on next page)

Current Biology 27, 3132–3142, October 23, 2017 3135

essentially reversible, with complete recovery between pulses formed both for shortening events (Figure 2E, left panel) and
(black line). When t  q, the response is mainly viscous and elongation events (Figure 2E, right panel). Each fit yields an
deformations are essentially irreversible, with no recovery be- estimate of the timescale t. We find a value of 61 ± 9s for
tween pulses (light gray line). When t  q, the response is visco- junction shortening (mean ± SE) and a similar value of 56 ± 6s
elastic and a partial recovery is observed between pulses (gray for junction elongation (Figure 2F). Interestingly, this timescale
line). To test this experimentally, we defined the irreversibility in- is slightly shorter than the typical duration of MyoII pulses
dex I as the ratio between the irreversible part of the deformation (typically 80 – 100s), which, according to the model, allows
and the maximal deformation (Figure 2B, left panel) and dissipation during the pulse and efficient junction shortening
measured I as a function of the pulse duration q (Figure 2B, box- or elongation with little recovery (i.e., reversibility) between
plot). The resulting plot shows that shorter pulses indeed tend to consecutive pulses.
produce more reversible deformations, whereas longer pulses Our analysis of MyoII pulses partly relies on the supposed
tend to produce more irreversible deformations, a signature of scaling between MyoII activity and force generation. Therefore,
dissipative mechanics. we sought to confirm our mechanical analysis using better
Two terms contribute to the elongation rate l_ in Equation 1. known forces, applied in a controlled fashion. In a previous
One scales with MyoII activity m, whereas the other scales study, we introduced optical tweezers as a tool to deform junc-
with MyoII accumulation rate m. _ For a given pulse, the maximum tions in the embryonic epithelium of Drosophila and to study me-
of m_ precedes the maximum of m. Hence, a straightforward chanics on short timescales. Here, we use similar optical forces
consequence of viscoelasticity is that the maximal rate of elon- to perform pull-release experiments on the minute timescale to
gation should precede the maximum of MyoII activity (Figure 2C, study the irreversibility of deformations. Because of rapid cell
left panel). We observed that this is indeed the case in vivo (Fig- movements occurring during germband elongation (stage 7),
ure 2C, right panel). The model predicts a shift of about 25s be- we could not perform the tweezers experiments at that stage.
tween l_ and m, which is consistent with our data. Indeed, we Therefore, we performed the experiments at stage 6, typically
measured a shift of 29:5 ± 2s for shortening (mean ± SE) and 15 min prior to the onset of elongation (see STAR Methods for
28:5 ± 2s for elongation (Figure 2D). Whereas it was not inter- details).
preted in viscoelastic terms before, this shift was previously re- The trap is switched on at a distance xt of a few hundred
ported using temporal cross-correlation analysis during junction nanometers from the midpoint of a junction and then switched
shortening [8], junction elongation [19], and prior to mesoderm off after q seconds (Figure 3A). We monitor the deflection xm ðtÞ
invagination, when cells’ apical area undergoes MyoII-driven of the junction from its initial position, that is, xm ð0Þ = 0 (Figures
contractions [5]. We interpret this consistent shift between MyoII 3B–3D; Movies S2, S3, and S4). Assuming a viscoelastic
activity and deformation rate as another signature of viscoelastic behavior with dissipation on timescale t again yields a
mechanics. Maxwell-like constitutive equation relating the deflection xm ðtÞ
We have presented semiquantitative results showing that to the restoring force fðtÞ,
irreversible deformations generated by MyoII pulses bare the
f
signature of viscoelastic dissipative mechanics. To assess k x_m = f_ + ; (Equation 2)
more quantitatively the accuracy of the model, and hence to t
estimate the typical dissipation timescale, a direct comparison where k is the effective elastic stiffness of the junction (see STAR
between experimental and model deformation dynamics is Methods for details). Because the trap is instantly switched on
required. To that end, we extracted signals of MyoII activity or off, the junction undergoes rapid movements and the drag
from shortening and elongation movies (see, for example, force in the cytosol fd =  Ch x_m , where Ch is the drag coefficient,
Movie S1) and plugged them into our model. We then solved can no longer be neglected. The trap force simply writes
Equation 1 and predicted the resulting junction length dy- ft = kt ðxt  xm Þ, where kt is the optical trap stiffness [18].
namics lðtÞ. This prediction depends on the value of parameters Combining this constitutive equation to the force balance
ka and t, which we use as adjustable parameters to obtain the f + ft + fd = 0 yields the following equation of motion:
best possible fit between the predicted and the experimental  
1 1+b b b
lðtÞ. The best fit is determined by iterating integration of Equa- x€m + + x_m + xm = xt : (Equation 3)
t T tT tT
tion 1 using a gradient descent method. This analysis is per-

(D) Kymographs of the deflection transverse to the junction, with q = 5s (left panel), q = 20s (middle panel), and q = 40s (right panel). Green and red arrows indicate
the trap being switched on and off. The scale bar for X represents 5 s and for Y represents 5mm.
(E) Graphs of junction deflection extracted from kymographs, with q = 3s (top panel) and q = 40s (bottom panel). The solid black line shows the junction position
xm ðtÞ. The horizontal line displays the optical trap position and its status (on, green; off, dotted red). The red line shows the fit obtained from our model.
(F) Estimates of t extracted from fits of experiments, both for optical tweezers (n = 19 from 6 embryos) and contractile pulses (shortening and extension; n = 31
from 6 embryos).
(G) Left: Kymograph (top) and quantification (bottom, a.u.) of the MyoII signal during a trapping experiment. Green arrow indicates the initial deflection. Right:
Boxplot of MyoII intensity at the onset of trapping versus after 40 s of trapping is shown (n = 20 from 5 embryos).
(H) Left: Cartoon of the irreversibility index I, defined as the ratio between the irreversible deformation and the maximal deformation. Right: Irreversibility index IðqÞ
increases with the pulling time q. Black dots show individual experiments (n = 31 from 6 embryos). The solid red line shows the analytical prediction, with no fit
parameters.
(I) Irreversibility index for q = 30s as a function of the maximum amplitude of deflection, for controls (n = 27 from 8 embryos) and for embryos treated with
cytochalasin-D (n = 35 from 11 embryos; see Figure 4 for Cyto-D experiments).

3136 Current Biology 27, 3132–3142, October 23, 2017

A B

D E

F G H

I K

(legend on next page)

Current Biology 27, 3132–3142, October 23, 2017 3137

 Note that inertia was neglected from the force balance versibility index as a function of the duration of force applica-
because the system is at very low Reynolds number. Equation 3 tion q (see STAR Methods for details). We measured IðqÞ for q
has three parameters. b = kt =k is dimensionless and compares ranging from 2 to 55 s (longer times were not experimentally
the trap stiffness to the junction stiffness. It typically controls accessible due to cell movements) and compared experimental
how far the junction elastically goes toward the optical trap. measurements to the theoretical prediction (Figure 3H, right
b = 0 when the trap is off; otherwise, our previous study showed panel). Note that this prediction has no adjustable parameters,
that b  1 [18]. T = Ch =k is the timescale associated with viscous because we directly used the values of b, T, and t obtained
damping in the cytosol, i.e., the typical time required for the junc- earlier. Finally, we controlled whether irreversibility was not
tion to elastically relax into and out of the trap. Our previous biased by the deflection amplitude, which might be larger for
study showed that T is in the order of one to a few seconds longer pulling experiments. To that end, we performed a series
[18]. Again, t is the typical dissipation timescale and underlies of experiments with a fixed pulling duration of q = 30s. Vari-
the extent of irreversibility. ability of junction stiffness or trapping stiffness generated
Consistent with our analysis of contractile pulses, we found a significant variability in deflection amplitude. We found no
that, if the duration of force application q is long enough, irrevers- correlation between the amplitude of deflection and the extent
ible deformations are observed. If q is small, typically a few sec- of irreversibility (Figure 3I). Overall, the good agreement be-
onds, deformation is mostly reversible (Figure 3E, top panel). For tween viscoelastic modeling and experiments further confirms
larger values of q, deformations become more and more irrevers- our estimate of t and the relevance of viscoelasticity to inter-
ible (Figure 3E, bottom panel). In addition, we used the solution pret junctional mechanics.
to Equation 3 to fit the dynamics of junction deflection in and The molecular origin of dissipation at cell contacts is a com-
out of the trap (Figure 3E). Each fit provides an estimate of plex issue. Turnover of the actomyosin cortex, believed to dissi-
the three independent parameters b, T, and t. Consistent with pate elastic energy stored in deformed actin filaments (F-actin)
our previous estimates [18], we found that b = 1:8 ± 0:7 and and of actin crosslinkers, might in part underlie irreversibility
T = 3 ± 1 s (mean ± SE). More importantly, we obtained [22, 27]. In an attempt to investigate this question, we tested
t = 50 ± 7 s, which is in agreement with the estimates obtained the possible role of F-actin turnover. We expect that increased
from MyoII pulses analysis (Figure 3F). We next controlled F-actin stability, by slowing down adaptation of the cortex to
whether the force applied by optical tweezers could trigger MyoII cell shape changes, might reduce dissipation and therefore
recruitment at junctions, as force application has been reported yield more reversible contractile events. To that end, we in-
to promote MyoII recruitment to the cortex [25, 26]. Analyzing jected embryos with low dosage of cytochalasin D (Cyto-D),
kymographs of MyoII signal (Figure 3G, left panel), we measured which is known in these conditions to reduce actin turnover
no increase of MyoII during trapping experiments (Figure 3G, by inhibiting both the polymerization and depolymerization of
right panel). This shows that stabilization is not due to MyoII filaments barbed ends [28–30]. Importantly, treated embryos
recruitment induced by force application. Note that this is not were still able to recruit MyoII at junctions and to undergo junc-
contradictory with observations by Fernandez-Gonzalez et al. tion shortening (Figure 4A; Movie S5). We also controlled that
in Drosophila [26], because we apply weaker forces during a MyoII and F-actin (using the actin binding domain of Utrophin
shorter period and at an earlier stage. fused to GFP as a proxy for F-actin) intensities were commen-
We observed that longer force application times yield defor- surate with controls and that the typical duration of contractile
mations that are more irreversible. We thus defined the irrevers- events was similar (Figure 4B). As reported for melanoma M2
ibility index IðqÞ as the ratio between the irreversible part of the cells [30], experiments of fluorescence recovery after photo-
deflection and the maximal deflection (Figure 3H, left panel). bleaching (FRAP) in actin-GFP-expressing embryos confirmed
From Equation 3, it is possible to calculate analytically the irre- that F-actin was more stable in embryos treated with low

Figure 4. Impact of Actin Dynamics on Dissipation

(A) Junctions undergoing shortening (yellow area) in both water and Cyto-D-injected embryos. White lines evidence the segmentation of surrounding cells.
(B) Averaged intensity of MyoII (left panel) and UtrophinABD (middle panel) at junctions in both water (n = 115 from 5 embryos) and Cyto-D embryos (n = 112 from
5 embryos). Right: Duration of contractile events in water (n = 50 from 3 embryos) and Cyto-D embryos (n = 55 from 3 embryos) is shown.
(C) Pre- and post-FRAP images showing the loss of actin-GFP signal caused by photobleaching. (Bottom panel) Time-lapse sequence of the FRAP area is shown.
Right: Recovery of the actin signal both in water (averaged from n = 25 junctions from 4 embryos) and Cyto-D embryos (averaged from n = 25 junctions from
8 embryos) is shown.
(D) Left: Cartoon of the maximal shortening rate of a contractile event (red line) and of the averaged shortening rate (green line). Middle: Maximal shortening rate in
water (n = 50 from 3 embryos) and Cyto-D embryos (n = 57 from 3 embryos) is shown. Right: Averaged shortening rate in water (n = 47 from 3 embryos) and Cyto-D
embryos (n = 52 from 3 embryos) is shown.
(E) Irreversibility index IðqÞ of contractile events as a function of their duration both in water (n = 51 from 3 embryos) and Cyto-D embryos (n = 48 from 3 embryos).
(F) Fit of junction shortening in both water (left panel) and Cyto-D embryos (right panel) using the experimental MyoII signal as input for the model.
(G) Estimates of t extracted from fits of shortening events for water (n = 26 from 3 embryos) and Cyto-D embryos (n = 38 from 3 embryos).
(H) Irreversibility ratio after 30 s pulling with optical tweezers in both water (n = 23 from 8 embryos) and Cyto-D embryos (n = 35 from 11 embryos).
(I) Sample kymographs of junction deflection in water and Cyto-D embryos.
(J) Segmented view of the germband in water-injected and Cyto-D-injected embryos. T1 transitions extracted from Movie S6 are depicted in green and red on the
image.
(K) Top: Cumulative sum of T1 transitions measured from Movie S6. Bottom: Averaged frequency of T1 transitions for water (n = 4 embryos) and Cyto-D (n = 4
embryos) is shown. Error bars represent the SE.

3138 Current Biology 27, 3132–3142, October 23, 2017

concentrations of Cyto-D than in embryos injected with water laser ablation [23, 26] or rapid MyoII inhibition through Rok
(Figure 4C). [31], still reveals the stress that was imposed prior to unloading
To assess the impact of Cyto-D on the dynamics of single junc- and therefore yields elastic recoil. Indeed, dissipation of elastic
tions, we monitored MyoII activity and junction length during energy does not imply that the system is devoid of mechanical
junction shortening. We first found that the maximal shortening stress, which in this case is continuously generated by MyoII.
rate during shortening events was similar in control and treated Dissipative mechanics thus provides a powerful explanation for
embryos, suggesting that the ability of MyoII to generate contrac- the so-called mechanical ratchet, suggesting that the stabiliza-
tile forces and deform junctions was not affected (Figure 4D, mid- tion of active junction shortening or elongation might result
dle panel). However, we found that the averaged shortening rate from dissipation without the need of specific stabilizing forces.
over longer periods, encompassing the contribution of several It is important to note that the observed shortening or elongation
consecutive contractile events, was slower in embryos treated of most junctions is quite far from the ‘‘canonical’’ dynamics with
with Cyto-D (Figure 4D, right panel). This suggests that elastic periodic pulses and stepwise deformations. Indeed, MyoII activity
recoil following shortening events is likely to be more important is usually far from being periodic, and its fluctuations are quite
in treated embryos. This was confirmed by the quantification noisy. Interestingly, this is not a limiting factor for our mechanical
of the irreversibility index (as defined earlier), which we found interpretation, as the model is independent of the shape of MyoII
consistently lower in embryos treated with Cyto-D (Figure 4E). signal. A consequence is that it provides no explanation as to why
We next used our viscoelastic model to fit the shortening dy- junctional remodeling occurs on a pulsatile basis. However, theo-
namics in both control and treated embryos (Figure 4F). The retical works have proposed generic mechanisms underlying the
good agreement obtained between predicted and observed emergence of pulsatile contractility, either as resulting from the
shortening suggests that the treatment preserves the visco- spontaneous instability of a turning-over material contracting
elastic nature of the system while increasing the typical timescale against elastic elements [32] or from the advection and diffusion
of dissipation (Figure 4G). To further confirm this result, we went of two antagonistic species that up- and downregulate stress [33].
back to optical tweezers and performed a series of 30-s pulling In this article, we sought to shed light on a series of biological
experiments in both water and Cyto-D-injected embryos (Fig- observations with simple physical concepts rather than to
ure 4I). We then measured the irreversibility index and found develop a realistic description of the system’s rheology. As
that deformations in Cyto-D-treated embryos were on average stated earlier, a consequence is that the viscoelastic model
almost twice as reversible than in the controls (Figure 4H). delineated here is an oversimplification, albeit a useful one,
Finally, we sought to determine whether the reduction of and is unlikely to reflect the complete rheological complexity of
F-actin turnover had an impact on cell intercalation at the scale the system. Indeed, junctions and their cortex are likely to dissi-
of the tissue, as could be expected from the increased revers- pate energy on a distribution of timescales and possibly involve
ibility of contractile events. To that end, we tracked cells by auto- non-linear viscoelasticity. Yet, our model is sufficient to quantita-
mated segmentation and counted the number of resolved T1 tively interpret the in vivo data accessible to our experiments.
transitions during germband extension (Figure 4J; Movie S6). More precise rheological measurements would be necessary
We found that Cyto-D treatment decreased the frequency of to develop more sophisticated rheological models. This remains
T1s (Figure 4K), consistent with recent reports showing that experimentally challenging, especially in vivo.
Cyto-D affects vertex resolution in this system [24] and suggest- At the tissue scale, morphogenetic deformations are likely to
ing that actin-dependent reversibility of contractile events af- involve a variety of additional contributors. It was recently shown
fects intercalation. that a subset of rosette formation events, which are not consid-
The analysis of Cyto-D-treated embryos thus indicates that ered in this study, result from basolateral protrusive activity [34]
F-actin turnover participates in dissipation at cell contacts. rather than from apical MyoII contractility. This suggests that
other mechanisms of cell deformations might be at play, pre-
DISCUSSION sumably involving different mechanical interactions. Similarly,
possible friction with the yolk and vitelline membrane are
The analysis of irreversible cellular deformations resulting from not considered here, nor topological transitions between cells.
both intrinsic and extrinsic forces reveals that dissipation at The formation of new antero-posterior junctions, because they
cell junctions occurs on the minute timescale in the germband modify tissue topology, might, for instance, be required for
of Drosophila. A direct consequence is that transient morphoge- long-term stabilization of large, tissue-scale elongation once My-
netic forces, such as those produced by pulsatile MyoII activity, oII polarity is lost. Such contributions are likely to involve effective
can locally generate cell shape changes that are irreversible if dissipation timescales at the tissue scale, possibly larger than a
they typically last a minute or more. This view is supported by minute, as suggested by recent rheological measurements per-
our results showing that longer pulses, or longer tweezing, yield formed using ferrofluid droplets [35]. Note that, in suspended
less reversible deformations. Unlike a purely elastic material, monolayers, which display no or very few topological transitions
junctions submitted to increased forces due to increased MyoII and have no substrate, stress relaxation at the tissue scale typi-
activity dissipate elastic energy. This results in a drift of the cally occurs on the minute timescale [17], consistent with our es-
reference configuration. Consequently, no or little elastic recoil timate. Although this is out of the scope of this study, continuum
is observed when MyoII activity drops down to its reference level mechanics models might provide a more adequate framework
after a pulse, because pulses typically span timescales to describe the full tissue dynamics [36, 37]. Bridging the
commensurate with the timescale of dissipation. On the con- gap between subcellular analysis, such as ours, and macro-
trary, sudden removal of mechanical loading, for instance, by scopic models of tissue flows, which can incorporate effective

Current Biology 27, 3132–3142, October 23, 2017 3139

mechanical parameters as well as a wide variety of cell behaviors of tissue morphogenesis and the irreversibility of deformations
(topological transitions, growth, division, apoptosis, etc.), re- by an emphasis on viscoelastic, dissipative properties of cells
mains a challenge for the physics of tissue morphogenesis. beyond force generation per se.
The question as to how elastic energy is dissipated at the molec-
ular level remains rather open. A common view is that most of the STAR+METHODS
elastic energy is stored by deformed actin filaments and actin
crosslinkers and that their turnover is therefore likely to dissipate Detailed methods are provided in the online version of this paper
elastic energy [9, 27]. In that scenario, newly created filaments and include the following:
are indeed devoid of the elastic energy stored by the previous
ones, and the turnover rate critically affects the dissipation time- d KEY RESOURCES TABLE
scale. This view is supported by theoretical models [38] and by d CONTACT FOR REAGENTS AND RESOURCE SHARING
our Cyto-D experiments (Figure 4). We anticipate that further d EXPERIMENTAL MODEL AND SUBJECT DETAILS
work will be required to investigate other possible molecular sour- d METHOD DETAILS
ces of dissipation. MyoII in particular has, for instance, been shown B Imaging and quantification of MyoII pulses
to fluidize the cytoskeletal network [39, 40] and enhance actin B Optical tweezers
disassembly [41]. Irrespective of its contractile activity, MyoII is B Cytochalasin-D injections
also likely to contribute to dissipation as a turning-over crosslinker B Fluorescence recovery after photobleaching
of the cortical network. Anchorage of the actomyosin network to B Quantifications of T1 transitions
junctions, suggested to be affected by actin turnover inhibition B Model – MyoII pulses
[42], might affect cell shape changes [43] and possibly irrevers- B Model – optical tweezers
ibility. The redistribution of adhesion molecules at deformed junc- d QUANTIFICATION AND STATISTICAL ANALYSES
tions might also be important to stabilize deformations, and the d DATA AND SOFTWARE AVAILABILITY
rate of E-cadherin complex disassembly (e.g., via endocytosis)
might thus be a limiting factor for stabilization. Interestingly, such SUPPLEMENTAL INFORMATION
contributions of protein dynamics to dissipation allow cell-, tis-
Supplemental Information includes six movies and can be found with this
sue-, or organ-specific regulation of the dissipation timescales.
article online at http://dx.doi.org/10.1016/j.cub.2017.09.005.
This offers a versatile tool for morphogenesis: besides the ability
of cells to generate localized and/or polarized forces, local and/or
AUTHOR CONTRIBUTIONS
polarized control of the mechanical response to forces become
possible, not only in a quantitative manner (stiffer or softer), but R.C. and P.-F.L. conceived the project and discussed it with C.C. and T.L. B.D.
in a qualitative manner (more elastic or more viscous). Tissue-spe- and C.C. performed the myosin pulses experiments (Figure 2), and B.D., C.C.,
cific tuning of dissipative properties might also be important to and R.C. analyzed the data. R.C. performed the optical tweezers experiments
ensure that short-lived mechanical stresses do not produce dra- (Figures 3 and 4) and analyzed the data. B.D. performed the Cyto-D experi-
ments (Figure 4), and B.D. and R.C. analyzed the data. R.C. designed the
matic irreversible deformations. In the system investigated here,
model. All authors discussed the results. R.C. wrote the manuscript, and all au-
the timescale of force generation, that is, the timescale of MyoII thors commented on it.
pulses, is commensurate with the typical dissipation timescale,
which seems crucial to achieve efficient deformations. This obser- ACKNOWLEDGMENTS
vation might be relevant to other morphogenetic systems,
although further work will be required to confirm this hypothesis. We thank Claire Chardès for assistance with the optical tweezers setup and
In particular, other morphogenetic events displaying reversible members of the Lenne and Lecuit groups for stimulating and useful discus-
and irreversible contractions should be analyzed. For example, sions during the course of this project. This work was supported by an FRM
Equipe Grant FRM DEQ20130326509, Agence Nationale de la Recherche
actomyosin pulses during zebrafish optic cup morphogenesis
ANR-Blanc Grant Morfor ANR-11-BSV5-0008 (to P.-F.L.), the ERC Grant
drive essentially reversible shortening events of cells’ apico-basal BioMecaMorph (to T.L.), and LabEx INFORM (ANR-11-LABX-0054) of the
axis length [44]. Interestingly and in contrast with cells investigated A*MIDEX project (ANR-11-IDEX-0001-02), funded by the French Government
here, shortening and MyoII oscillations seem synchronous, which program ‘‘Investissements d’Avenir’’. We acknowledge France-BioImaging
is consistent with a reversible, elastic regime of deformation. Simi- infrastructure supported by the Agence Nationale de la Recherche (ANR-10-
larly, during the first phase of dorsal closure, pulses of MyoII in the INBS-04-01; ‘‘Investissements d’Avenir’’). B.D. was supported by the ERC
Grant BioMecaMorph (323027), and C.C. was supported by a Human Frontier
medio-apical cortex cause reversible contractions of amnioserosa
Science Program Long-Term Fellowship (LT000733/2011-L) and by a post-
cells’ apical area [45, 46]. This is followed by a second phase, dur- doctoral fellowship from the FRM (SPF20121226396).
ing which contractile events become more efficient, with net
shrinkage after each pulse. A possible mechanism is a gradual in- Received: May 11, 2017
crease of dissipation during the process, causing pulses to first Revised: July 27, 2017
probe an elastic regime, whereas, as time passes, they probe an Accepted: September 5, 2017
increasingly viscous regime, allowing more efficient deformations. Published: October 5, 2017
Altogether, our results provide an in vivo estimate of the typical
REFERENCES
dissipation timescale associated with junctional remodeling and
shed a new light on the mechanism of the so-called mechanical 1. Lecuit, T., Lenne, P.-F., and Munro, E. (2011). Force generation, transmis-
ratchet, highlighting the key contribution of dissipation in stabiliz- sion, and integration during cell and tissue morphogenesis. Annu. Rev.
ing deformations. It opens a new avenue for the understanding Cell Dev. Biol. 27, 157–184.

3140 Current Biology 27, 3132–3142, October 23, 2017

 2. Heisenberg, C.P., and Bellaı̈che, Y. (2013). Forces in tissue morphogen- 23. Rauzi, M., Verant, P., Lecuit, T., and Lenne, P.-F. (2008). Nature and
esis and patterning. Cell 153, 948–962. anisotropy of cortical forces orienting Drosophila tissue morphogenesis.
3. Bertet, C., Sulak, L., and Lecuit, T. (2004). Myosin-dependent junction Nat. Cell Biol. 10, 1401–1410.
remodelling controls planar cell intercalation and axis elongation. Nature 24. Yu, J.C., and Fernandez-Gonzalez, R. (2016). Local mechanical
429, 667–671. forces promote polarized junctional assembly and axis elongation in
4. Blankenship, J.T., Backovic, S.T., Sanny, J.S., Weitz, O., and Zallen, J.A. Drosophila. eLife 5, e10757.
(2006). Multicellular rosette formation links planar cell polarity to tissue 25. Effler, J.C., Kee, Y.S., Berk, J.M., Tran, M.N., Iglesias, P.A., and Robinson,
morphogenesis. Dev. Cell 11, 459–470. D.N. (2006). Mitosis-specific mechanosensing and contractile-protein
5. Martin, A.C., Kaschube, M., and Wieschaus, E.F. (2009). Pulsed contrac- redistribution control cell shape. Curr. Biol. 16, 1962–1967.
tions of an actin-myosin network drive apical constriction. Nature 457, 26. Fernandez-Gonzalez, R., Simoes, S. de M., Röper, J.C., Eaton, S., and
495–499. Zallen, J.A. (2009). Myosin II dynamics are regulated by tension in interca-
6. Munro, E., Nance, J., and Priess, J.R. (2004). Cortical flows powered by lating cells. Dev. Cell 17, 736–743.
asymmetrical contraction transport PAR proteins to establish and main- 27. Salbreux, G., Charras, G., and Paluch, E. (2012). Actin cortex mechanics
tain anterior-posterior polarity in the early C. elegans embryo. Dev. Cell and cellular morphogenesis. Trends Cell Biol. 22, 536–545.
7, 413–424. 28. Cooper, J.A. (1987). Effects of cytochalasin and phalloidin on actin. J. Cell
7. Skoglund, P., Rolo, A., Chen, X., Gumbiner, B.M., and Keller, R. (2008). Biol. 105, 1473–1478.
Convergence and extension at gastrulation require a myosin IIB-depen- 29. Shoji, K., Ohashi, K., Sampei, K., Oikawa, M., and Mizuno, K. (2012).
dent cortical actin network. Development 135, 2435–2444. Cytochalasin D acts as an inhibitor of the actin-cofilin interaction.
8. Rauzi, M., Lenne, P.-F., and Lecuit, T. (2010). Planar polarized actomyosin Biochem. Biophys. Res. Commun. 424, 52–57.
contractile flows control epithelial junction remodelling. Nature 468, 1110– 30. Fritzsche, M., Lewalle, A., Duke, T., Kruse, K., and Charras, G. (2013).
1114. Analysis of turnover dynamics of the submembranous actin cortex. Mol.
9. Wyatt, T., Baum, B., and Charras, G. (2016). A question of time: tissue Biol. Cell 24, 757–767.
adaptation to mechanical forces. Curr. Opin. Cell Biol. 38, 68–73. 31. Coravos, J.S., and Martin, A.C. (2016). Apical sarcomere-like actomy-
10. He, B., Doubrovinski, K., Polyakov, O., and Wieschaus, E. (2014). Apical osin contracts nonmuscle Drosophila epithelial cells. Dev. Cell 39,
constriction drives tissue-scale hydrodynamic flow to mediate cell elonga- 346–358.
tion. Nature 508, 392–396. 32. Dierkes, K., Sumi, A., Solon, J., and Salbreux, G. (2014). Spontaneous os-
11. Dicko, M., Saramito, P., Blanchard, G.B., Lye, C.M., Sanson, B., and cillations of elastic contractile materials with turnover. Phys. Rev. Lett.
Étienne, J. (2017). Geometry can provide long-range mechanical guidance 113, 148102.
for embryogenesis. PLoS Comput. Biol. 13, e1005443. 33. Kumar, K.V., Bois, J.S., Jülicher, F., and Grill, S.W. (2014). Pulsatory pat-
12. Forgacs, G., Foty, R.A., Shafrir, Y., and Steinberg, M.S. (1998). terns in active fluids. Phys. Rev. Lett. 112, 208101.
Viscoelastic properties of living embryonic tissues: a quantitative study. 34. Sun, Z., Amourda, C., Shagirov, M., Hara, Y., Saunders, T.E., and
Biophys. J. 74, 2227–2234. Toyama, Y. (2017). Basolateral protrusion and apical contraction
13. Desprat, N., Richert, A., Simeon, J., and Asnacios, A. (2005). Creep func- cooperatively drive Drosophila germ-band extension. Nat. Cell Biol.
tion of a single living cell. Biophys. J. 88, 2224–2233. 19, 375–383.
14. Plotnikov, S.V., Sabass, B., Schwarz, U.S., and Waterman, C.M. (2014). 35. Doubrovinski, K., Swan, M., Polyakov, O., and Wieschaus, E.F. (2017).
High-resolution traction force microscopy. Methods Cell Biol. 123, Measurement of cortical elasticity in Drosophila melanogaster embryos
367–394. using ferrofluids. Proc. Natl. Acad. Sci. USA 114, 1051–1056.
15. Bufi, N., Durand-Smet, P., and Asnacios, A. (2015). Single-cell mechanics: 36. Prost, J., Jülicher, F., and Joanny, J.-F. (2015). Active gel physics. Nat.
the parallel plates technique. Methods Cell Biol. 125, 187–209. Phys. 11, 111–117.
16. Harris, A.R., Bellis, J., Khalilgharibi, N., Wyatt, T., Baum, B., Kabla, A.J., 37. Tlili, S., Gay, C., Graner, F., Marcq, P., Molino, F., and Saramito, P. (2015).
and Charras, G.T. (2013). Generating suspended cell monolayers for me- Colloquium: mechanical formalisms for tissue dynamics. Eur. Phys. J. E
chanobiological studies. Nat. Protoc. 8, 2516–2530. 38, 33–63.
17. Wyatt, T.P.J., Harris, A.R., Lam, M., Cheng, Q., Bellis, J., Dimitracopoulos, 38. Kim, T., Gardel, M.L., and Munro, E. (2014). Determinants of fluidlike
A., Kabla, A.J., Charras, G.T., and Baum, B. (2015). Emergence of homeo- behavior and effective viscosity in cross-linked actin networks. Biophys.
static epithelial packing and stress dissipation through divisions oriented J. 106, 526–534.
along the long cell axis. Proc. Natl. Acad. Sci. USA 112, 5726–5731. 39. Le Goff, L., Amblard, F., and Furst, E.M. (2002). Motor-driven dynamics in
18. Bambardekar, K., Cle
ment, R., Blanc, O., Chardès, C., and Lenne, P.-F. actin-myosin networks. Phys. Rev. Lett. 88, 018101.
(2015). Direct laser manipulation reveals the mechanics of cell contacts €s, J. (2002). Active
40. Humphrey, D., Duggan, C., Saha, D., Smith, D., and Ka
in vivo. Proc. Natl. Acad. Sci. USA 112, 1416–1421. fluidization of polymer networks through molecular motors. Nature 416,
19. Collinet, C., Rauzi, M., Lenne, P.F., and Lecuit, T. (2015). Local and tissue- 413–416.
scale forces drive oriented junction growth during tissue extension. Nat. 41. Wilson, C.A., Tsuchida, M.A., Allen, G.M., Barnhart, E.L., Applegate, K.T.,
Cell Biol. 17, 1247–1258. Yam, P.T., Ji, L., Keren, K., Danuser, G., and Theriot, J.A. (2010). Myosin II
20. Munjal, A., Philippe, J.-M., Munro, E., and Lecuit, T. (2015). A self- contributes to cell-scale actin network treadmilling through network
organized biomechanical network drives shape changes during tissue disassembly. Nature 465, 373–377.
morphogenesis. Nature 524, 351–355. 42. Jodoin, J.N., Coravos, J.S., Chanet, S., Vasquez, C.G., Tworoger, M.,
21. Kerridge, S., Munjal, A., Philippe, J.-M., Jha, A., de las Bayonas, A.G., Kingston, E.R., Perkins, L.A., Perrimon, N., and Martin, A.C. (2015).
Saurin, A.J., and Lecuit, T. (2016). Modular activation of Rho1 by GPCR Stable force balance between epithelial cells arises from F-actin turnover.
signalling imparts polarized myosin II activation during morphogenesis. Dev. Cell 35, 685–697.
Nat. Cell Biol. 18, 261–270. 43. Roh-Johnson, M., Shemer, G., Higgins, C.D., McClellan, J.H., Werts, A.D.,
22. Blanchoin, L., Boujemaa-Paterski, R., Sykes, C., and Plastino, J. (2014). Tulu, U.S., Gao, L., Betzig, E., Kiehart, D.P., and Goldstein, B. (2012).
Actin dynamics, architecture, and mechanics in cell motility. Physiol. Triggering a cell shape change by exploiting preexisting actomyosin con-
Rev. 94, 235–263. tractions. Science 335, 1232–1235.

Current Biology 27, 3132–3142, October 23, 2017 3141


rez, M., Kuchling, F., Letelier, J., Polvillo, R., Wittbrodt, J., and
44. Nicolás-Pe 47. Guillot, C., and Lecuit, T. (2013). Adhesion disengagement uncouples
Martı́nez-Morales, J.R. (2016). Analysis of cellular behavior and cytoskel- intrinsic and extrinsic forces to drive cytokinesis in epithelial tissues.
etal dynamics reveal a constriction mechanism driving optic cup morpho- Dev. Cell 24, 227–241.
genesis. eLife 5, 1–24. 48. Aigouy, B., Farhadifar, R., Staple, D.B., Sagner, A., Röper, J.C., Jülicher,
45. Solon, J., Kaya-Çopur, A., Colombelli, J., and Brunner, D. (2009). Pulsed F., and Eaton, S. (2010). Cell flow reorients the axis of planar polarity in
forces timed by a ratchet-like mechanism drive directed tissue movement the wing epithelium of Drosophila. Cell 142, 773–786.
during dorsal closure. Cell 137, 1331–1342. 49. Cavey, M., and Lecuit, T. (2008). Imaging cellular and molecular dynamics in
46. Machado, P.F., Duque, J., Étienne, J., Martinez-Arias, A., Blanchard, G.B., live embryos using fluorescent proteins. Methods Mol. Biol. 420, 219–238.
and Gorfinkiel, N. (2015). Emergent material properties of developing 50. Spitzer, M., Wildenhain, J., Rappsilber, J., and Tyers, M. (2014). BoxPlotR:
epithelial tissues. BMC Biol. 13, 98. a web tool for generation of box plots. Nat. Methods 11, 121–122.

3142 Current Biology 27, 3132–3142, October 23, 2017

STAR+METHODS

KEY RESOURCES TABLE

REAGENT or RESOURCE SOURCE IDENTIFIER

Chemicals, Peptides, and Recombinant Proteins
Cytochalasin D Sigma C8273
Deposited Data
Mendeley dataset This paper http://dx.doi.org/10.17632/my2dcbrf8g.1
Experimental Models: Organisms/Strains
D. melanogaster / ; Sqh::GFP; Gap43::mCherry [47] N/A
D. melanogaster / ; E-cad::GFP, Sqh::mCherry; [47] N/A
D. melanogaster / ; ; GFP::UtrophinABD, Sqh::mCherry This paper N/A
D. melanogaster / ; UAS-GFP::Actin5C ; Bloomington Drosophila Stock Center 9258
Software and Algorithms
Tissue Analyzer [48] https://grr.gred-clermont.fr/labmirouse/
software/WebPA/index.html
Other
Invitrogen FluoSpheres carboxylate-modifies Thermo Fisher F8812
0.5mm red fluorescent

CONTACT FOR REAGENTS AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Raphaël

ment ([email protected]).
Cle

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Flies were maintained under standard lab conditions at 22 C with yeast food. In order to obtain embryos, flies were put in standard
cages. Cages feature an agar plate with apple juice at the bottom, supplemented with fresh yeast paste. Flies lay eggs on these
plates. To collect embryos we filtered the yeast paste supplemented with H20. Embryos are covered by a non-transparent chorion
which needs to be removed for live imaging. To do so we treated the embryos with commercial bleach for 45 s, and then washed them
abundantly with H20. Embryos were then staged and aligned on coverslips using binoculars. Coverslips were then taken to fluores-
cence microscopes for live imaging. For more details about embryo preparation, see [49]. Please refer to method details for strains
used for each experiment.

METHOD DETAILS

Imaging and quantification of MyoII pulses

Fly embryos were imaged during germband elongation (stage 7) in the ventro-lateral region for about 30 min. For Figures 1 and 2,
embryos labeled with E-cadherin::GFP (shg in Drosophila) and MyoII regulatory light chain (MRLC)::mCherry (sqh in Drosophila)
were from a fly stock sqhAX3; E-cad::GFPKIn, sqh-Sqh::mCherry. For Figure 4, embryos labeled with GFP::UtrophinABD and
Sqh::mCherry were used. The fly line ; ; GFP::UtrophinABD, Sqh::mCherry was generated using a sqh-GFP::UtrophinABD trans-
gene [8] recombined with the sqh-Sqh::mCherry transgene [5], both inserted on the third chromosome.
Time-lapse imaging was performed with a spinning disk confocal microscope (Eclipse TE 2000-E with an Ultraview ERS, Perkin
Elmer spinning disk) using a 100X/N.A. 1.4 oil-immersion objective. Z series of 6-10 planes spanning 4-6mm from the cell apex
were acquired with a frame rate of 3 to 4.5 s/frame to collect both the medio-apical and the junctional MyoII pools. Maximum pro-
jections of confocal Z stack were generated for the image analysis, and cytoplasmic background was subtracted as previously
described [19].
MyoII signal at the junctions was measured by measuring the mean fluorescence intensity of the MyoII channel in a 0.8mm wide
straight line connecting the junction vertices (Figure 1H). Fluorescence intensity at the vertices was excluded by shrinking the line
of 5 pixel (0.4mm) at both ends.
MyoII signal close to adjacent vertices was measured in adjacent anterior and posterior cells as the mean fluorescence intensity of
the MyoII channel in two elliptical regions of interest (ROIs) close to the junction vertices. Ellipses have an aspect ratio of 0.7 and a
major axis of 3mm aligned along the line of the junction (Figure 1I).

Current Biology 27, 3132–3142.e1–e4, October 23, 2017 e1

Optical tweezers
For optical tweezers experiments, fly embryos double-labeled with GFP::UtrophinABD and Sqh::mCherry were used, except Fig-
ure 3G, for which flies labeled with Sqh::GFP and Gap43::mCherry were used. Tweezers experiments were performed typically
15 min before germband elongation (stage 6). Rapid cell movements occurring throughout the extension process (stage 7) prevent
‘‘long’’ trapping experiments (more than 10-20 s). At stage 6 the tissue is static enough to allow longer trapping, although one-minute-
long pulling experiments are already an upper limit considering cell movements. Imaging was done in a custom light sheet micro-
scope coupled with a single-beam gradient trap (wavelength 1070nm, ytterbium fiber laser; IPG photonics). A 100X water immersion
lens (N.A. 1.1, Nikon) was used for imaging as well as introducing the optical trap in the imaging plane. Before every experiment, we
calibrated the relationship between the galvanometers voltages (V1,V2) and the optical trap position (x,y), using the following proced-
ure: 500nm-diamater fluorescent polystyrene beads were trapped in water and moved slowly in circles by imposing galvanometers
voltages. Images were acquired simultaneously to voltage commands, and successive (x,y) positions were localized by 2D Gaussian
fits, providing the mapping between (V1, V2) and (x,y). This information was used to infer the trap position during in vivo experiments.
Kymographs of interface deflection were produced from movies using Fiji, and analyzed with a custom MATLAB script described
below. Note that we neglected possible Z displacements likely to be caused by radiation pressure, which in a few experiments can
move junctions slightly off focus while the trap is on (see for instance Figure 3D, right panel). To obtain the position of the cell-cell
interface in the XY plane, the custom script performs a Gaussian fit of the fluorescence intensity of UtrophinABD or Gap43 along
the kymograph line, perpendicular to the contact line. This is repeated for each frame, and allows subpixel detection of the deflection
xm ðtÞ, with an error < 100nm [18].

Cytochalasin-D injections
For injections experiments shown Figure 4, Cyto-D suspended in DMSO at a concentration of 10mg/ml was diluted 20 times in water,
and injected in the yolk at the onset of gastrulation (stage 5-6). The final dilution factor in the embryo is around 1/50, so that we expect
a final concentration of approximately 0.01mg/ml. Water supplemented with the same quantity of DMSO (1/20) was injected for control
experiments.

Fluorescence recovery after photobleaching

For FRAP experiments, F2 progeny of females ; 67Gal4 ; crossed with males ; UAS-GFP::Actin5C ; were used. Photobleaching was
performed using the targeted laser of the Nikon Eclipse Ti spinning disc (iLAS2). A circular junctional region of typically 2.5mm diam-
eter was photobleached. Laser power and integration settings were adjusted to photobleach up to 80% of the initial intensity. The
photo-bleached region was tracked using a custom macro in ImageJ. The fluorescence intensity of the photo-bleached region was
measured at each time point, and normalized as Inorm = ðI  Imin Þ=ðImax  Imin Þ (Figure 4C).

Quantifications of T1 transitions
To measure the number of T1 transitions (Figures 4J and 4K), we used the Tissue Analyzer plugin for ImageJ developed by Benoı̂t
Aigouy [48]. Segmentation was automatically performed by the plugin and corrected by the experimenter. T1 events were automat-
ically detected by the plugin and then checked individually by the experimenter to prevent false detections.

Model – MyoII pulses

Let us first consider an elastic cell-cell junction with modulus k submitted to an elongation/shortening force f. The length of the junc-
tion is given by:
f
l = l0 + : Eq. 4
k
Now let us assume that the reference configuration l0 drifts due to dissipation. Since the system has a number of dissipation sources,
possibly including the turnover of Actin and its many crosslinkers, it is likely to dissipate energy on a distribution of timescales. For the
sake of simplicity we will assume that it dissipates energy on a single typical timescale t and therefore that the reference configuration
accordingly drifts over timescale t. We thus have:
l  l0
l_0 = : Eq. 5
t
Dots denote time derivatives. Derivating Equation 4 with respect to time, using that l_0 = ðl  l0 Þ=t = f=ðktÞ, and multiplying by k yields
the well-known Maxwell model of viscoelasticity:
f
kl_= f_+ : Eq. 6
t
We now need to relate the force f to MyoII activity. We assume that f results from a balance between MyoII at the junction which
tends to shorten junctions (minus sign), and MyoII in the vicinity of the junctions vertices (and possibly in the neighboring junctions),
which tend to elongate it (plus sign), so that f =  fj + fv . Note that with this convention both fj and fv are positive.

e2 Current Biology 27, 3132–3142.e1–e4, October 23, 2017

 Let us first consider shortening events, powered by pulses of MyoII at junctions ðMj Þ. We assume that the shortening force is pro-
portional to MyoII level, so that fj = aMj . During these events, MyoII in the vicinity of vertices ðMv Þ is low and does not display coherent
fluctuations. Hence we assume that the opposing external force fv is constant, so that fv = fv0 . Thus we have f =  aMj + fv0 .
At rest (in the absence of pulses), the length l is constant and f = 0, and MyoII at the junction Mj is at a base level M0j .
Hence aM0j + fv0 = 0, so that fv0 =  aM0j . Therefore during shortening, f can be approximated as f =  aðMj  M0j Þ. One can interpret
the base level M0j as the amount of MyoII required at the junction to withstand the ‘‘constant’’ external force fv0 exerted by neighboring
cells. Only the excess of MyoII compared to that base level will be able to shorten the junction. Practically, we assume that M0j is the
minimum of Mj , which is consistent with shortening junctions being at rest in the absence of pulses at the junction. Defining
m = Mj  M0j , and ka = k=a, we obtain for shortening:
m
ka l_= m_ + : Eq. 7
t
Let us now consider elongation events, powered by pulses of MyoII in the vicinity of vertices ðMv Þ. We assume that the elongation force
is proportional to Mv , so that fv = aMv . During elongation events, MyoII at the junction ðMj Þ is low and does not display coherent fluc-
tuations. Hence in that case we assume that fj is constant, so that fj = fj0 . With similar arguments on equilibrium, we obtain that fj0 = aM0v ,
so that f = + aðMv  M0v Þ. M0v is the base level of MyoII at vertices, and corresponds to the level of MyoII required to withstand the ‘‘con-
stant’’ junction tension fj0 . Again, practically we assume that M0v is the minimum of Mv , consistent with elongating junctions being at rest
in the absence of pulses in the vicinity of vertices. Again, defining m = Mv  M0v , and ka = k=a, we obtain for elongation:
m
+ ka l_= m_ + : Eq. 8
t
Equations 7 and 8 have two parameters. ka tells us to which extent MyoII can deform junctions, and hence contains both the elastic
modulus of the cortex and the conversion constant from MyoII to force. t is the typical dissipation timescale. Note that since MyoII
pulses are slow compared to the timescale associated to viscous drag in the cytosol (one to a few seconds [18]), we neglected drag in
the context of MyoII pulses.
To generate artificial data of shortening during pulses, we integrated Equations 7 and 8 using Gaussian pulses of MyoII activity as
input for m. To fit experimental data and estimate t, we integrated Equations 7 and 8 using the experimental MyoII signals as input for
m. For each fit, the integration was iterated using a gradient descent method to obtain the best fit values for ka and t. Note that MyoII
intensity being in arbitrary units, ka also has arbitrary units.

Model – optical tweezers

Let us now consider a cell-cell junction submitted to an external normal force exerted by optical tweezers. The contact line is de-
flected from the straight reference configuration with an angle 4, hence the restoring force f is:
f =  2T sin 4 Eq. 9
where T is the tension in the junction. For small deflection angles, sin 4x2xm =h, where h is the junction length, xm the deflection, and
TxT0 = cst [18]. Hence,
4T0
f=  ðxm  x0 Þ =  kðxm  x0 Þ Eq. 10
h
where k = 4T0 =h is the effective stiffness of the junction in response to perpendicular deflection, and x0 the abscissa of the junction in
the absence of trapping, with x0 ðt = 0Þ = 0. Again, let us now assume that the reference configuration drifts on a typical timescale t due
to dissipation:
xm  x0
x_0 = : Eq. 11
t
Combining Equations 10 and 11 again yields a Maxwell-like viscoelastic equation:
f
k x_m = f_+ : Eq. 12
t
The restoring force f is balanced by viscous drag in the cytosol and by optical trapping, which we can assume is linear [18]. Note that
since we switch the trap rapidly from on/off to off/on, junctions are submitted to rapid movements, and unlike the case of MyoII
pulses, here drag cannot be neglected. Finally we neglect the inertial term, since the system is at very low Reynolds number. The
force balance reads:
f + kt ðxt  xm Þ  Ch x_m = 0 Eq. 13
where kt is the stiffness of the optical trap, and Ch the damping coefficient due to drag in the cytosol. Combining Equations 12 and 13
yields the equation of motion for the junction submitted to trapping:
 
1 1+b b b
x€m + + x_m + xm = xt : Eq. 14
t T tT tT

Current Biology 27, 3132–3142.e1–e4, October 23, 2017 e3

 This equation has three parameters. b = kt =k is the dimensionless ratio between the trap stiffness and the junction stiffness. Note
that b = 0 if the trap is off. Otherwise b  1 [18]. T = Ch =k is the relaxation timescale associated to drag in the cytosol. In practice, we
have T  3s. Finally t is the timescale of dissipation, which we are primarily interested in.
Let us now introduce a and b such that a = 1=t + ð1 + bÞ=T and b = b=tT. Equation 14 now reads:
x€m + ax_m + bxm = bxt : Eq. 15
Let D be the discriminant of the characteristic polynomial associated to Equation 15, that is D = a  4b. Using that xt is a particular
2

solution, the general solution to Equation 15 reads:

xm ðtÞ = xt 1 + Aet=tA + Bet=tB Eq. 16

where tA = 2=ða + D1=2 Þ and t B = 2=ða  D1=2 Þ. The first initial condition is: xm ðt = 0Þ = 0. In addition, at the onset of deflection ðt = 0 + Þ,
the force balance reads: Ch x_m ðt = 0Þ = kt xt , hence x_m ðt = 0Þ = kt xt =Ch . Plugging these conditions in Equation 16 yields the constants A
and B. We find that A = t A ðT  btB Þ=ðTðt B  t A ÞÞ and that B = t B ðbtA  TÞ=ðTðtB  t A ÞÞ. We use Equation 16 to fit pull-release exper-
iments and to estimate the three parameters of the model b, T, and t.
Let us now consider the relaxation process after a pulling experiment of duration q (hence for t > q). At time t = q, the trap is switched
off, so that for t > q, kt = 0 and b = 0, and Equation 15 simplifies into:
x€m =  ax_m Eq. 17
where a = 1=t + 1=T. Let vm = x_m be the velocity of deflection. Equation 17 yields:
vm ðt > qÞ = vm ðqÞeaðtqÞ Eq. 18
+ +
where vm ðqÞ = vm ðt = q Þ is the initial velocity of relaxation. At the onset of relaxation ðt = q Þ, the force balance reads
fðqÞ =  Ch vm ðqÞ. Hence:
fðqÞ aðtqÞ
vm ðt > qÞ =  e : Eq. 19
Ch
Integrating Equation 19 yields:
fðqÞ aðtqÞ
xm ðt > qÞ = e + C: Eq. 20
aCh
At the onset of relaxation, xm = xm ðqÞ. Plugging this initial condition in Equation 20 yields:
fðqÞ aðtqÞ

xm ðt > qÞ = xm ðqÞ + e 1 : Eq. 21
aCh
Let the irreversibility index I be the ratio of irreversible deformation over maximal deformation after a pulling time q. IðqÞ simply reads:

xm ðt/NÞ xm ðqÞ  fðqÞ ðaCh Þ t fðqÞ
IðqÞ = = =1  : Eq. 22
xm ðqÞ xm ðqÞ t + T kxm ðqÞ
Equation 16 provides the expression for xm ðqÞ. Combining Equations 13 and 16 also provides the expression of fðqÞ. Plugging the
explicit expressions of xm ðqÞ and fðqÞ in Equation 22 leads to the irreversibility index IðqÞ:
t AðT=tA  bÞeq=tA + BðT=t B  bÞeq=tB
IðqÞ = 1  : Eq. 23
t+T 1 + Aeq=tA + Beq=tB
Since we obtained estimates of b, T, and t (and therefore A, B, tA and tB ) from individual pulling experiments, this expression pro-
vides a prediction with no free parameter for IðqÞ (see Figure 3H).

QUANTIFICATION AND STATISTICAL ANALYSES

All information concerning the statistical details are provided in the main text and in figure legends. This includes the means and stan-
dard errors, the nature of error bars, as well as the number of samples analyzed for each experiment (number of junctions, number of
animals). All boxplots use the following standards: center lines of boxes show the medians; boxes limits indicate the 25th and 75th
percentiles; whiskers extend 1.5 times the interquartile range from the 25th and 75th percentiles. Boxplots were generated using
BoxPlotR, developed by the Tyers and Rappsilber labs [50].

DATA AND SOFTWARE AVAILABILITY

Data and software presented in this paper have been deposited at Mendeley Data and are available at http://dx.doi.org/10.17632/
my2dcbrf8g.1.

e4 Current Biology 27, 3132–3142.e1–e4, October 23, 2017