Spatial organization of the extracellular matrix regulates cell–cell junction positioning Qingzong Tsenga, Eve Duchemin-Pelletierb, Alexandre Deshiereb, Martial Ballandc, Hervé Guilloud,e, Odile Filholb, and Manuel Thérya,1 a Laboratoire de Physiologie Cellulaire et Végétale, Unité Mixte de Recherche (UMR) 5168, Institut de Recherches en Technologies et Sciences pour le Vivant (iRTSV), Commissariat à l’Energie Atomiques et aux Energies Alternatives (CEA), Centre National de la Recherche Scientifique (CNRS), Institut National de la Recherche Agronomique, Université Joseph Fourier (UJF), 38054 Grenoble, France; bLaboratoire de Biologie du Cancer et de l’Infection, Unité 1036, iRTSV, Institut National de la Santé et de la Recherche Médicale, CEA, UJF, 38054 Grenoble, France; cLaboratoire Interdisciplinaire de Physique, UMR 5588, UJF, CNRS, 38402 Saint Martin d’Heres, France; dUnité Mixte International 2820, Laboratory for Integrated Micro Mechatronic Systems, CNRS, Institute of Industrial Science, University of Tokyo, Meguro-ku, Tokyo 153-8505, Japan; and eLaboratoire de Thermodynamique des Petits Systémes, Institut Néel, CNRS, UJF, BP166, 38042 Grenoble, France

Edited by Alexander D. Bershadsky, Weizmann Institute of Science, Rehovot, Israel, and accepted by the Editorial Board November 22, 2011 (received for review April 21, 2011)

tissue architecture crosstalk

| cytoskeleton | actin | traction force | integrin cadherin

E

pithelial sheets lie on a layer of extracellular matrix (ECM), the so-called basement membrane. In such epithelia, cells establish integrin-based adhesions on the basal part of the cell in contact with ECM, and cadherin-based intercellular adhesions on the apical part of contacting lateral domains, away from contact with ECM. The two adhesion systems display nonoverlapping spatial distributions. Both cell–cell and cell–ECM adhesions are required to establish proper epithelium morphology (1). They both participate in mechano-transduction of external physical cues into intracellular signaling (2). The biochemical nature of adhesion molecules engaged in intercellular adhesion, the energy of the interaction, as well as the mechanical tension developed along intercellular junctions have been shown to govern epithelial cell shape and orient intercellular junctions in various systems (3–6). However, whereas the contribution of cell–cell adhesion to epithelial topology has been the focus of many studies, much less attention has been paid to the role of ECM. ECM is a dynamic scaffold that is actively remodeled during morphogenesis, where it plays major roles in stimulating and guiding cell migration as well as orienting stem cell fate (7, 8). ECM is also known to impart morphoregulatory signals to epithelia, and thereby regulates tissue morphogenesis (8, 9). However, the mechanism by which ECM guides cell positioning at the single-cell scale is still not known. ECM geometry has been shown to regulate intracellular architecture (10) and provide

www.pnas.org/cgi/doi/10.1073/pnas.1106377109

spatial information for cell polarization (1, 11, 12), but how it regulates cell positioning and thereby spatially organizes multicellular architectures remained to be investigated. Results The effect of the spatial distribution of ECM on the localization of MCF10A intercellular junctions was investigated by controlling the location of ECM with fibronectin micropatterns (13, 14) (SI Methods). To provide a maximum degree of freedom to the intercellular junction and to minimize the number of parameters participating in its positioning, we focused our analysis on cell doublets, formed by daughter cells after mitosis. The positioning of cells, as revealed by the spatial coordinates of their nucleus, was recorded by time-lapse microscopy during a complete cell cycle and automatically quantified (Fig. S1, Movie S1, and SI Methods). We measured the angular distribution of the nucleus– nucleus axis as well as the proportion of time during which cells were moving. As expected from previous work (15), cells were almost randomly positioned and turned steadily around each other on [square]-shaped micropatterns (Fig. 1A and Movie S2). Under such conditions, ECM is present all along the contour of the cell doublet and provides a continuous peripheral track for cell movement. Because in epithelia, cadherins tend to flow from the ECM-rich basal pole to the ECM-free apical pole (16), we tested whether the absence of ECM could stabilize the intercellular junction and interfere with cell movement. [H]-shaped micropatterns were designed to provide two large regions devoid of ECM. In striking contrast to their behavior on [square], a large proportion of cell doublets on [H] did not move at all, resulting in highly stable configurations in which cells were positioned on each side of the gap (Fig. 1B and Movie S3). We further tested whether the nuclear axis was perpendicular to the intercellular junction by staining E-cadherin on fixed cells. We confirmed that the intercellular junctions were positioned over the gap where ECM was absent (Fig. 1 D and E). On [H], the absence of ECM was probably not the only factor contributing to cell-doublet stabilization. The presence of the large gap and the absence of a continuous peripheral track along

Author contributions: Q.T. and M.T. designed research; Q.T., E.D.-P., H.G., O.F., and M.T. performed research; A.D. and M.B. contributed new reagents/analytic tools; Q.T., O.F., and M.T. analyzed data; and M.T. wrote the paper. The authors declare some conflict of interest because M.T. is involved in the company CYTOO, which commercializes micropatterns. This article is a PNAS Direct Submission. A.D.B. is a guest editor invited by the Editorial Board. 1

To whom correspondence should be addressed. E-mail: [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1106377109/-/DCSupplemental.

PNAS Early Edition | 1 of 6

CELL BIOLOGY

Intercellular Junctions Are Stabilized in Regions Deprived of ECM.

APPLIED PHYSICAL SCIENCES

The organization of cells into epithelium depends on cell interaction with both the extracellular matrix (ECM) and adjacent cells. The role of cell–cell adhesion in the regulation of epithelial topology is well-described. ECM is better known to promote cell migration and provide a structural scaffold for cell anchoring, but its contribution to multicellular morphogenesis is less well-understood. We developed a minimal model system to investigate how ECM affects the spatial organization of intercellular junctions. Fibronectin micropatterns were used to constrain the location of cell–ECM adhesion. We found that ECM affects the degree of stability of intercellular junction positioning and the magnitude of intra- and intercellular forces. Intercellular junctions were permanently displaced, and experienced large perpendicular tensional forces as long as they were positioned close to ECM. They remained stable solely in regions deprived of ECM, where they were submitted to lower tensional forces. The heterogeneity of the spatial organization of ECM induced anisotropic distribution of mechanical constraints in cells, which seemed to adapt their position to minimize both intra- and intercellular forces. These results uncover a morphogenetic role for ECM in the mechanical regulation of cells and intercellular junction positioning.

180°

proportion of moving frames (%) 100 60

0° 8

N cell = 146 N angle = 8561

16 24%

20

max

averaged E-cadherin

orientation of nucleus-nucleus axis 90°

E-cadherin actin

D

A

180° n = 16

proportion of moving frames (%) 100 60

180°

0° 8 20

24 N cell = 117 N angle = 5442

n = 23

min

40%

0° 8 16 24% 270°

E-cadherin actin

proportion of moving frames (%) 100 60 20

max

averaged E-cadherin

180°

0° 8

16 24%

0° 8 16 24% 270°

180°

16 24%

270°

270°

N angle = 6769

N angle = 5406

N angle = 5171

proportion of moving frames (%)

proportion of moving frames (%)

proportion of moving frames (%)

100

100

100

60

60

60

20

20

20

N cell = 118

N cell = 90

N cell = 86

F

270° orientation of nucleus-nucleus axis 90°

180°

orientation of nucleus-nucleus axis 90°

max

averaged E-cadherin

orientation of nucleus-nucleus axis 90°

E-cadherin actin

E

N cell = 69 N angle = 3991

0°

orientation of nucleus-nucleus axis 90°

8

min

270°

B

C

orientation of nucleus-nucleus axis 90°

n = 18

min

Fig. 1. The cell–cell junction is stabilized in regions deprived of ECM. (A– C) Time-lapse acquisition in phase contrast of an MCF10A cell doublet on a [square]- (A) or [H]-shaped micropattern (B). Time frame is 15 min. Automated movie analysis of Hoechst-stained cells (Fig. S1) provided the angular distribution of the nucleus–nucleus axis orientation and quantification of cell-doublet movements on [square]- (A), [H]- (B), and [hourglass]-shaped micropatterns (C). (D–F) Examples of immunofluorescent stainings of E-cadherin on cell doublets (Upper) and averaged staining over several images (Lower) on [square] (A), [H] (B), and [hourglass] (C). Micropattern width is 35 μm.

the cell-doublet contour may also have physically separated cells and prevented cell movement. [Hourglass]-shaped micropatterns were designed to provide both a continuous peripheral track for cell movement and two regions along the cell-doublet contour in which ECM was absent (Fig. 1C). Thereby we could test the specific effect of ECM on intercellular junction positioning under conditions where migration was not impaired. Analysis of both nucleus position and junction position confirmed that cell doublets tend to stabilize their positions in configurations where their intercellular junction lies above regions deprived of ECM (Fig. 1 C and F). Under these conditions, the central part of the junction lies above ECM. However, cadherins tend to concentrate at the extremity of the junction (17), where most of the intercellular force is applied (18), and we could show that it is indeed the absence of ECM at the extremity and not in the central part of the junction that stabilizes junction displacements (Fig. S2). Notably, these orientations of cell doublets did not merely result from the orientation of the mother cell division but instead implied an active repositioning mechanism of the intercellular junction in response to ECM geometry (Fig. S3). 2 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1106377109

Fig. 2. The distance from ECM regulates the stability of intercellular junction position. Analyses of cell-doublet positioning from time-lapse acquisition (Fig. S1) provided angular distributions of nucleus–nucleus axis orientations as well as the distributions of the proportion of moving frames in [gapped-square] shapes with gap sizes of 7, 14, and 20 μm (from left to right). Micropattern width is 35 μm.

If epithelial cells actively tend to form their intercellular junction away from ECM, the geometry of ECM and the size of ECM-deprived regions should directly regulate the degree of stability of the intercellular junction positioning. We tested this hypothesis by varying the size of the gap in [C]-shaped micropatterns (Fig. 2). As the gap size increased from 7 to 20 μm, the proportion of cells with the intercellular junction positioned over the gap increased progressively, whereas the proportion of cell movement gradually decreased (Fig. 2). Intercellular Junctions Are Destabilized by the Presence of ECM. To further confirm whether ECM promotes intercellular junction displacement, we tested a complementary approach in which ECM was added along a stable intercellular junction position. To that end, we first designed a micropattern geometry in which cell doublets could adopt two stable configurations. On [cross], the intercellular junction was positioned along the midlines, either horizontally or vertically, within the two large regions devoid of ECM (Fig. 3A). Cell-doublet angular speed was high when the intercellular junction passed over ECM-rich diagonals. It slowed down when the junction passed over ECM-free midlines (Fig. 3C). When ECM was introduced along the horizontal midline, the cell–cell junction was no longer slowed down along this axis, and cell doublets could not adopt the corresponding orientation anymore (Fig. 3 B and C). From these experiments, we concluded that the proximity of cell–cell adhesion sites with ECM promotes the displacement of the intercellular junction, whereas the distance between cell adhesion sites and ECM stabilizes it. Regulation of Tensional Forces Is Implicated in Intercellular Junction Guidance by ECM. To further investigate the molecular mechanism

governing intercellular junction positioning in response to ECM geometry, we inactivated some of the signaling pathways specifically associated with cell adhesion and modified the formation Tseng et al.

A

orientation of nucleus-nucleus axis

A

N cell = 131 N angle = 5991

averaged E-cadherin

90°

180°

max

junction extremities coordinates

0°

junction orientation

8% 16% 270°

n = 38

min

B FAK inhibition

orientation of nucleus-nucleus axis

B

N cell = 149 N angle = 7619

90°

180°

averaged E-cadherin

max

n = 31

min

ROCK inhibition

0° 8% 16% 270°

θ2,t 2

angular speed =

θ2 - θ1 t - t 2 1

2.0 1.5 1.0 0.5

n = 129

n = 121

n = 96

D

C

laminin

siRNA GFP

siRNA P120-catenin

0.0

0 30 60 90 120 150 180 nucleus-nucleus axis orientation (degree)

APPLIED PHYSICAL SCIENCES

Fig. 3. Contact with ECM destabilizes the cell–cell junction. (A and B) Analyses of cell-doublet positioning from time-lapse acquisition (Center) (Fig. S1) and averaged staining of E-cadherin over several cells (Right) on [cross]- (A) and [cross+bar]-shaped micropatterns (B). Micropattern width is 35 μm. (C) Quantification of averaged angular rotation speed of cell doublets with respect to axis orientation. Arrows indicate cell acceleration when the junction passes over ECM along diagonals on [cross] (red arrows), and along diagonals and horizontal bars on [cross+bar] (blue arrow).

Tseng et al.

E

n = 97

n = 116

30

***

40

***

FAK inhibition

laminin

10 0

control (fibronectin) ROCK inhibition

20

**

cell proportion (%)

50

ns

n = 122

or maturation of cell–cell and cell–ECM adhesion constituents. We tested the effects of these treatments on cell doublets plated on the [hourglass], because this geometry guides intercellular junction positioning without impairing cell migration. We quantified the orientation of the intercellular junction and the precise positions of its extremities on fixed cells stained for α-catenin (Fig. 4A). Under control conditions, most intercellular junction extremities were positioned in the region deprived of ECM, on both sides of the narrow part of the [hourglass] (Fig. 4B and Fig. S4). This phenotype was not significantly perturbed by the inhibition of FAK, ERK, JNK, Rac, or Src, which are the main kinases conveying biochemical signaling from cell adhesions (19) (Fig. 4B and Fig. S4). However, intercellular junction position and orientation were strongly perturbed by the inactivation of cell contraction using either Rho kinase inhibition or myosin II inhibition (Fig. 4 B and E and Fig. S4). Down-regulation of focal adhesion proteins, on which traction forces are applied, impaired cell spreading and precluded the analysis of their specific contribution to intercellular junction positioning. However, cell spreading on laminin, which is known to engage a distinct subset of integrins than fibronectin (20), significantly perturbed junction positioning (Fig. 4 C and E and Fig. S4) and thereby showed that this positioning was directly regulated by the nature of the cell–ECM interactions. We then tried to perturb the assembly of junctional complexes that are able to transfer the forces applied on ECM to the adjacent cells. Down-regulation of p120-catenin by siRNA treatment is known to affect intercellular junction turnover and actin dynamics (21). It significantly perturbed junction positioning (Fig. 4 D and E and Fig. S4). These results

CELL BIOLOGY

θ1,t 1

angular speed (degree/min)

C

0 10 20 30 40 50 60 70 80 90 junction orientation (degree)

siRNA GFP siRNA p120-catenin

Fig. 4. Cell contractility regulates intercellular junction positioning. (A) Immunostaining of α-catenin on cell doublets plated on [hourglass] (Left) (α-catenin in green, DNA in blue) allowed the detection and measurement of intercellular junction positions (Right). (B–D) The positions of junction extremities (black dots) were measured on cells treated for 6 h with PF-573228 (1 μM) to inactivate FAK or Y27632 (5 μM) to inactivate ROCK (B), cells plated on laminin-coated micropatterns (C), or cells pretransfected with siRNA against GFP or p120-catenin (D). Images are examples of treated cells (α-catenin in green, DNA in blue). More representative stainings are shown in Fig. S4. Micropattern width is 35 μm. (E) Curves indicate the proportion of junctions for each angular sector with the same set of data as in B and C. Differences between the two curves were compared using the Kolmogorov– Smirnov test; **, 0.5% and ***, 0.1% error probability in the rejection of the hypothesis that the two distributions are identical; ns indicates that the probability would be higher than 5%.

suggested that the production of mechanical forces on intercellular junctions was responsible for junction positioning away from ECM. Intercellular Tension Is Reduced in Regions Deprived of ECM. To measure intercellular tensional forces, we grafted ECM microPNAS Early Edition | 3 of 6

A

n = 30

500 400 300

500 400 300

200 100

200 100

n = 33 Cell-cell force (nN)

B

Fcell-cell = −ΣF i

C

1500 pN

n = 30

Force (nN) 300

D

200

**

1

100

0.5

0

0

1250 1000 750

1250 1000 750

500 250

500 250

1500 pN

n = 33

Force (nN) 300

ns

200

200

100

100

0

E

Ratio cell-cell/total force

***

***

0

intra inter adhesive adhesive

intra inter adhesive non-adhesive INTER-cellular force (nN)

INTRA-cellular force (nN)

***

***

150

200 100 100

50

0

0

with ECM

without ECM

with ECM

without ECM

F cell-cell inter

H = 2. (Jintra.Lintra + Jinter.Linter) + Jcell-cell.Lcell-cell inter

ECM dependent Energy (a.u.)

intra

on ECM: Jintra = Jinter = 1 away from ECM: Jintra = 1.3 , Jinter = 0.8 Jcell-cell = 0.1

2 1.5 1

ECM independent

0.5 0 -45

Jintra = Jinter = 1 Jcell-cell = 0.1 0 45 90 135 angle (degrees)

patterns on soft polyacrylamide gels (22) (SI Methods). Deformations of polyacrylamide gels were used to measure the forces exerted by cell doublets on the substrate (23–25) (Fig. 5A) and indirectly derive the forces they exert on each other (26) (SI Methods). The analysis of force balance in cell doublets on [square] and [H] revealed that intercellular forces were significantly reduced when intercellular junction extremities lay above ECM gaps (Fig. 5B). Because intercellular tension reduction on 4 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1106377109

Fig. 5. Junction positioning away from ECM is associated with relaxation of intercellular forces. (A) Fibronectin micropatterns with [square] and [H] shapes on polyacrylamide gels and traction maps, averaged over several cells, on the corresponding geometries. The color code indicates the local traction in Pascal. (B) Cell–cell force measurement. The mechanical balance in each cell imposes that the force exerted between cells counterbalances the sum of traction forces exerted on the substrate. Cell–cell force was measured and compared between cells plated on [square] and [H]. Each dot corresponds to a measure on a single cell. The ratio between the intercellular force and the total traction force (sum of all force magnitude over the micropattern) was calculated and plotted to confirm the specific reduction of cell–cell forces on [H]. (C) Traction force fields averaged over several cells plated on [square] (Left) and [H] (Right). Magnifications correspond to the white square regions on global maps. Arrows indicate force orientation; color and length both represent local force magnitude in pN. (D) Decomposition of traction forces into “intra” forces oriented toward intracellular space and “inter” forces oriented toward the intercellular junction. These forces were noted “adhesive” or “nonadhesive” whether they were oriented along a cell edge in contact or not with ECM. The absence of ECM is associated with relaxed intercellular forces. (E) Force measurements along cell edges on various micropattern shapes ([square], [X], and [H]; Fig. S6) were combined depending on their orientation (intra or inter) and the presence of ECM along their length (with or without ECM). These graphs represent three separate experiments and 100 cells per condition. The presence of ECM showed opposite effects on intra- and intercellular tension. All statistical comparisons were Student’s t tests, **P < 0.01, ***P < 0.001, ns, P > 0.05 (F) Physical modeling. The various orientations of the intercellular junction can be described by a simple energy function H taking into account the length L and the tension J along all edges of the cell doublets. The two curves correspond to numerical simulations of the energy function for various junction orientations when the tension depends (red) or not (blue) on the local presence of ECM. A favored orientation (junction orientation at 0°), corresponding to experimental observations, only occurs when tension depends on ECM.

[H] could result from a decrease in global traction force rather than a specific decrease of the intercellular force, we calculated the ratio between the intercellular force and the total traction force. Like the intercellular force, this ratio was lower when the junction was stabilized over ECM gaps (Fig. 5B). To further explain how large traction forces on sites flanking ECM gaps could be associated with reduced levels of intercellular forces over such gaps, we analyzed force orientations depending on the Tseng et al.

unstable

low inter-cellular forces low intra-cellular forces non moving inter-cellular junction

stable

Fig. 6. Summary and illustration. On micropatterns, cells develop elevated intra- and intercellular tension when the intercellular junction is in contact with ECM (Left) and reduced tension when it is far from ECM (Right). In situ, as shown here in the case of tubulogenesis, this mechanism would promote the relocalization of junctions away from ECM and the formation of multilayered structures of opposed polarities. ECM is shown in red, intercellular junction complexes are in green, and nuclei are in blue.

presence or absence of ECM at the extremities of the junction (Fig. 5 C and D). Traction forces at cell apices were decomposed into a component perpendicular to the junction, considered as contributing mainly to intercellular forces, and a component parallel to the junction, considered as contributing mainly to intracellular forces (Fig. 5D). This component of the traction perpendicular to the junction was a good approximation of the global cell–cell force (Fig. S5). On [square], intercellular forces along ECM-rich edges were equivalent to intracellular forces. However, on [H], intercellular forces along ECM-free edges were almost half the strength of intracellular forces (Fig. 5D). We further compared the intra- and intercellular forces on various micropattern shapes (Fig. S6) and found that the presence of ECM had an opposite effect on each: It reduced intracellular forces but increased intercellular forces (Fig. 5E). Overall Tension Minimization Accounts for Cell Positioning. To test whether the regulation of intra- and intercellular forces in response to the presence of ECM could account for the favored cell positioning we observed, we performed numerical simulations to take into account all cortical forces and compare the energetic costs of all cell positions. We used a well-established physical model of the energy associated with the development of cortical forces (4, 27) and applied it to the simplified case of cells having a fixed area (28) (Fig. 5F and SI Methods). Numerical simulations showed that when tension was considered to be independent of the presence of ECM, all cell orientations had approximately the same energetic cost. However, when we incorporated into the model the relative differences of tension we measured in response to the presence of ECM, the tension along cell edges became anisotropic, and the configurations associated with the lower total energy actually corresponded to those we observed experimentally (Fig. 5F and Fig. S7). This suggested that cell positioning could actually result from minimization of global tension in response to ECM geometry.

Discussion Here we described the development of an experimental system to study multicellular morphogenesis. Although this system partly Tseng et al.

Methods Cell Culture. The culture of mammary epithelial cells MCF10A was described previously (38). Cells were seeded on patterned substrates at a density of 8 × 104/cm2. Cells not attaching to the adhesive region on the substrate were

PNAS Early Edition | 5 of 6

CELL BIOLOGY

high inter-cellular forces high intra-cellular forces moving inter-cellular junction

inter-cellular junctions away from the ECM

APPLIED PHYSICAL SCIENCES

inter-cellular junctions in contact with the ECM

lacks some of the physiological characteristics found in situ in animal systems (3) and in vitro in 3D cyst formation (29, 30), it offers the possibility of quantifying cell movement and accurately measuring the spatial distribution of forces in a large number of reproducible assays. In addition, the fine manipulation of ECM geometry allows precise control of the degree of freedom for cell movements. Thus, although spatially confined, multiple cells can reveal their natural self-assembly process. ECM has been shown to guide tissue morphogenesis by modulating cell–cell interactions (31, 32). The biochemical, structural, and mechanical interactions between cell–ECM and cell–cell adhesions have been well-characterized (2, 33, 34). However, how these interactions translate into spatial organization of multicellular arrangements remained to be clarified. Our results suggest that intercellular junction positioning away from ECM did not result from the action of a single ECM signal. Indeed, most of the main signaling pathways associated with cell–ECM adhesion could be impaired individually without affecting junction position, at least in the range of inhibitor concentration and the time window we tested. Neither did it result from oriented cell divisions. Instead, it appeared to stem from ECM-regulated production of intra- and intercellular forces and overall minimization of global tension. We found that the contact of intercellular junctions with ECM caused the displacement of the cell–cell junction and the production of large perpendicular tensional forces and constant motion of the cells. In support of these findings, work from others has shown that cell adhesion to ECM alters cell–cell adhesion and promotes the development of large tensional forces responsible for junction disruption and subsequent cell migration (35, 36). In addition, we found that when cells place their intercellular junction on ECM-deprived regions, intra- and intercellular tension gets reduced and cell position is stabilized (Fig. 6). How cell interaction with ECM regulates inter- and intracellular forces in an opposed way remains to be elucidated. We suspect this involves complex interplays between the growth and dynamics of several types of actin-based structures formed in proximity to cell–ECM and cell–cell contacts. Anisotropic distribution of forces along the cell periphery has been shown to regulate cell shape, junction remodeling, and cell spatial repositioning within tissues (3). Our results suggest that ECM could participate in similar morphogenetic processes. In response to heterogeneous distribution of ECM, cell doublets developed anisotropic force fields and adopted stable positions along the axis of low tension. According to this mechanism, cells tend to stabilize the position of their intercellular junctions away from ECM. This could account for many important processes during epithelium morphogenesis (Fig. 6). For example, during tubulogenesis, cells first form a single-file chain of cells. This stage, during which cell–cell junctions are close to ECM, is only transient. Cells reorganize and move their junctions away from ECM to form a multilayered cord with a central lumen (30, 37). During this transformation, intracellular forces that were developed across the ECM gap get relocalized along ECM fibers, whereas intercellular forces undergo the opposite reorientation. According to the mechanism we describe, this transformation would be favored by the relaxation of both intra- and intercellular tensions (Fig. 6). Although much attention has been paid to the role of cell– cell interaction in the shaping of epithelia, our results suggest that we should also consider the organizing role of ECM, which, far from being a mere supporting scaffold, plays an instructive role in regulating mechanical forces and orienting multicellular assembly.

washed away 1 h after seeding. After cell spreading on micropatterns, Hoechst 33342 was added at 5 ng/mL to label the nucleus during timelapse acquisition. Chemical Inhibitors. Chemical inhibitors were added 24 h after cell plating on micropatterns at the following concentrations: PF573228, 1 μM (FAK inhibition); PD98059, 2 μM (ERK1 inhibition); SP600125, 1.8 μM (JNK inhibition); SU6656, 5 μM (Src inhibition); NCS23766, 5 μM (Rac inhibition); Y27632, 5 μM (ROCK inhibition); blebbistatin, 15 μM (myosin II inactivation); ML7, 5 μM (MLCK inhibition). Cells were fixed 6 h later. Micropatterning. Glass coverslip micropatterning has been described elsewhere (14). Micropatterned polyacrylamide gels were made as previously described (22) (SI Methods). Immunofluorescent Staining. Thirty hours after plating cells on a micropatterned coverslip, cells were either extracted in cytoskeleton buffer containing 0.5% Triton X-100 and fixed in 4% paraformaldehyde or fixed in methanol at −20 °C. Fixed cells were incubated with a 1:200 dilution of antiα-catenin (B52975; Calbiochem) or a 1:50 dilution of anti-E-cadherin (sc8426; Santa Cruz Biotechnology) for 1 h, and then incubated with corresponding secondary antibodies and FITC-phalloidin (Invitrogen) at 1 μg/mL for 30 min.

was calculated by the normalized correlation coefficient algorithm, so that an individual interrogation window was compared with a larger searching window. The next iteration takes into account the displacement field measured previously, so that a false correlation peak due to insufficient image features is avoided. The normalized cross-correlation also allowed us to define an arbitrary threshold to filter out low correlation values due to insufficient beads presented in the window. The resulting final grid size for the displacement field was 1.63 × 1.63 μm, with six beads per interrogation window on average. The erroneous displacement vectors due to insufficient beads present in the window were filtered out by their low correlation value and replaced by the median value from the neighboring vectors. With the displacement field obtained from the PIV analysis, the traction force field was reconstructed by the Fourier transform traction cytometry (FTTC) method with regularized scheme (25) on the same grid (1.63 × 1.63 μm) without further interpolation or remapping. The regularization parameter was set at 9 × 10−10 for all traction force reconstructions. The FTTC code was also written in Java as an ImageJ plugin, so that the whole traction force microscopy procedure from PIV to force calculation could be performed with ImageJ. The entire package of traction force microscopy software is available at https://sites.google.com/site/qingzongtseng/tfm.

Traction Force Microscopy. Images of fluorescent beads with and without cell doublets were first aligned to correct experimental drift using the ImageJ plugin “align slices in stack.” The displacement field was subsequently calculated by a custom-written particle image velocimetry (PIV) program implemented as an ImageJ (http://rsb.info.nih.gov/ij) plugin. The PIV was performed through an iterative scheme. In each iteration, the displacement

ACKNOWLEDGMENTS. We thank Laurent Blanchoin and Thomas Lecuit for interesting discussions about this work, Benedikt Sabass for help on the Java FTTC code, as well as Matthieu Piel, Alexandra Fuchs, and Susana Godinho for critical reading of the manuscript. Kevin Berton, Philippe Huber, and Julien Verove kindly provided reagents. This work was supported by Agence National pour la Recherche Grants ANR-PCV08-322457 (to H.G., O.F., and M.T.) and ANR-08-JC-0103 (to M.T.), the Ligue Nationale contre le Cancer (O.F.), and a PhD fellowship from the Irtelis Program of the Commissariat à l’Energie Atomique et aux Energies Alternatives (to Q.T.).

1. Yu W, et al. (2005) β1-integrin orients epithelial polarity via Rac1 and laminin. Mol Biol Cell 16:433–445. 2. Papusheva E, Heisenberg CP (2010) Spatial organization of adhesion: Force-dependent regulation and function in tissue morphogenesis. EMBO J 29:2753–2768. 3. Rauzi M, Verant P, Lecuit T, Lenne PF (2008) Nature and anisotropy of cortical forces orienting Drosophila tissue morphogenesis. Nat Cell Biol 10:1401–1410. 4. Käfer J, Hayashi T, Marée AF, Carthew RW, Graner F (2007) Cell adhesion and cortex contractility determine cell patterning in the Drosophila retina. Proc Natl Acad Sci USA 104:18549–18554. 5. Krieg M, et al. (2008) Tensile forces govern germ-layer organization in zebrafish. Nat Cell Biol 10:429–436. 6. Foty RA, Steinberg MS (2005) The differential adhesion hypothesis: A direct evaluation. Dev Biol 278:255–263. 7. Guilak F, et al. (2009) Control of stem cell fate by physical interactions with the extracellular matrix. Cell Stem Cell 5(1):17–26. 8. Rozario T, DeSimone DW (2010) The extracellular matrix in development and morphogenesis: A dynamic view. Dev Biol 341(1):126–140. 9. Fata JE, Werb Z, Bissell MJ (2004) Regulation of mammary gland branching morphogenesis by the extracellular matrix and its remodeling enzymes. Breast Cancer Res 6(1):1–11. 10. Théry M, Pépin A, Dressaire E, Chen Y, Bornens M (2006) Cell distribution of stress fibres in response to the geometry of the adhesive environment. Cell Motil Cytoskeleton 63:341–355. 11. Wang AZ, Ojakian GK, Nelson WJ (1990) Steps in the morphogenesis of a polarized epithelium. I. Uncoupling the roles of cell-cell and cell-substratum contact in establishing plasma membrane polarity in multicellular epithelial (MDCK) cysts. J Cell Sci 95(Pt 1):137–151. 12. Théry M, et al. (2006) Anisotropy of cell adhesive microenvironment governs cell internal organization and orientation of polarity. Proc Natl Acad Sci USA 103: 19771–19776. 13. Théry M (2010) Micropatterning as a tool to decipher cell morphogenesis and functions. J Cell Sci 123:4201–4213. 14. Azioune A, Carpi N, Tseng Q, Théry M, Piel M (2010) Protein micropatterns: A direct printing protocol using deep UVs. Methods Cell Biol 97:133–146. 15. Huang S, Brangwynne CP, Parker KK, Ingber DE (2005) Symmetry-breaking in mammalian cell cohort migration during tissue pattern formation: Role of random-walk persistence. Cell Motil Cytoskeleton 61:201–213. 16. Kametani Y, Takeichi M (2007) Basal-to-apical cadherin flow at cell junctions. Nat Cell Biol 9(1):92–98. 17. Yamada S, Nelson WJ (2007) Localized zones of Rho and Rac activities drive initiation and expansion of epithelial cell-cell adhesion. J Cell Biol 178:517–527. 18. Maruthamuthu V, Sabass B, Schwarz US, Gardel ML (2011) Cell-ECM traction force modulates endogenous tension at cell-cell contacts. Proc Natl Acad Sci USA 108:4708–4713. 19. Playford MP, Schaller MD (2004) The interplay between Src and integrins in normal and tumor biology. Oncogene 23:7928–7946.

20. Hynes RO, Naba A (2011) Overview of the matrisome—An inventory of extracellular matrix constituents and functions. Cold Spring Harb Perspect Biol, 10.1101/cshperspect. a004903. 21. Reynolds AB, Roczniak-Ferguson A (2004) Emerging roles for p120-catenin in cell adhesion and cancer. Oncogene 23:7947–7956. 22. Tseng Q, et al. (2011) A new micropatterning method of soft substrates reveals that different tumorigenic signals can promote or reduce cell contraction levels. Lab Chip 11:2231–2240. 23. Butler JP, Tolic-Nųrrelykke IM, Fabry B, Fredberg JJ (2002) Traction fields, moments, and strain energy that cells exert on their surroundings. Am J Physiol Cell Physiol 282:C595–C605. 24. Dembo M, Wang YL (1999) Stresses at the cell-to-substrate interface during locomotion of fibroblasts. Biophys J 76:2307–2316. 25. Sabass B, Gardel ML, Waterman CM, Schwarz US (2008) High resolution traction force microscopy based on experimental and computational advances. Biophys J 94:207–220. 26. Liu Z, et al. (2010) Mechanical tugging force regulates the size of cell-cell junctions. Proc Natl Acad Sci USA 107:9944–9949. 27. Farhadifar R, Röper JC, Aigouy B, Eaton S, Jülicher F (2007) The influence of cell mechanics, cell-cell interactions, and proliferation on epithelial packing. Curr Biol 17:2095–2104. 28. Vianay B, et al. (2010) Single cells spreading on a protein lattice adopt an energy minimizing shape. Phys Rev Lett 105:128101. 29. Debnath J, Brugge JS (2005) Modelling glandular epithelial cancers in three-dimensional cultures. Nat Rev Cancer 5:675–688. 30. O’Brien LE, Zegers MMP, Mostov KE (2002) Building epithelial architecture: Insights from three-dimensional culture models. Nat Rev Mol Cell Biol 3:531–537. 31. Sakai T, Larsen M, Yamada KM (2003) Fibronectin requirement in branching morphogenesis. Nature 423:876–881. 32. Kurpios NA, et al. (2008) The direction of gut looping is established by changes in the extracellular matrix and in cell:cell adhesion. Proc Natl Acad Sci USA 105:8499–8506. 33. Weber GF, Bjerke MA, DeSimone DW (2011) Integrins and cadherins join forces to form adhesive networks. J Cell Sci 124:1183–1193. 34. Borghi N, Lowndes M, Maruthamuthu V, Gardel ML, Nelson WJ (2010) Regulation of cell motile behavior by crosstalk between cadherin- and integrin-mediated adhesions. Proc Natl Acad Sci USA 107:13324–13329. 35. de Rooij J, Kerstens A, Danuser G, Schwartz MA, Waterman-Storer CM (2005) Integrindependent actomyosin contraction regulates epithelial cell scattering. J Cell Biol 171 (1):153–164. 36. Marsden M, DeSimone DW (2003) Integrin-ECM interactions regulate cadherin-dependent cell adhesion and are required for convergent extension in Xenopus. Curr Biol 13:1182–1191. 37. Pollack AL, Runyan RB, Mostov KE (1998) Morphogenetic mechanisms of epithelial tubulogenesis: MDCK cell polarity is transiently rearranged without loss of cell-cell contact during scatter factor/hepatocyte growth factor-induced tubulogenesis. Dev Biol 204(1):64–79. 38. Deshière A, Theis-Febvre N, Martel V, Cochet C, Filhol O (2008) Protein kinase CK2 and cell polarity. Mol Cell Biochem 316(1-2):107–113.

6 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1106377109

Tseng et al.

Supporting Information Tseng et al. 10.1073/pnas.1106377109 SI Methods siRNA Treatments. Transient silencing was performed using 15 nM

siRNA P120-catenin: 5′-AACGAGGUUAUCGCUGAGAAC3′; GFP: 5′-GACGUAAACGGCCACAAGUUC-3′) and retrotransfection with Lipofectamine 2000 according to the manufacturer’s instructions (Invitrogen). Lysates of cells were prepared in RIPA lysis buffer [10 mM Tris·HCl (pH 7.4), 150 mM NaCl, 1% (vol/vol) Triton X-100, 0.1% (vol/vol) SDS, 0.5% (vol/vol) sodium deoxycholate, and 1 mM EDTA] containing a protease and phosphatase inhibitor mixtures (P8340, P2850, P5726; SigmaAldrich). Equivalent amounts of protein were processed by SDS/ PAGE and transferred to PVDF membrane (Hybond-P; Amersham Biosciences). Immunoblotting was performed using the following primary antibodies: Integrin-β1 and P120-catenin (BD Biosciences) and β-actin (Abcam). Video Microscopy and Image Acquisition. Time-lapse acquisitions were taken with an inverted microscope (Axiovert 200M; Carl Zeiss). Temperature, CO2, and humidity control was performed using a Box and Brick System (Life Imaging Services). Multiple positions were recorded using an XY motorized stage (Marzhauser). Nucleus movements were followed with a 15-min time frame and a 10× dry phase-contrast objective over 48 h. Fluorescent image acquisitions of fixed cells were taken using an upright microscope (BX61; Olympus) with 100× (N.A. 1.4) oilimmersion objectives mounted on a piezo motor (Physics Instruments). Both microscopes were controlled with MetaMorph software (MDS Analytical Technologies). Averaged images were obtained by aligning multiple images taken on distinct cells and projecting them onto a single image by calculating the average fluorescence intensity of each pixel. Automated Nucleus Tracking. Micropattern images (taken with a 10× objective) from each stage position were first segmented to 36 subregions corresponding to a 6 × 6 array of micropatterns. Subregions were screened for the existence of a nucleus by standard deviation of pixel intensities. Nucleus images were then convolved by a 9 × 9 “Mexican hat” kernel. Binarized nucleus images were obtained by automatic thresholding and a median filter to remove isolated noise pixels. After “watershed” segmentation, the number of nuclei was determined considering their size and shape factor as well. Only subregions containing only one nucleus at the beginning were selected for subsequent mitosis detection. Based on the nucleus shape factor, intensity variation, and the number of nuclei detected, division from one cell to two cells could be detected. At these positions only, orientations of the nucleus–nucleus axis were recorded after cell division. The measurements continued until the end of the timelapse series or the detection of the second mitosis event. Glass Slide Micropatterning. Glass coverslip micropatterning has

been described elsewhere (1). Coverslips were first spin-coated with adhesion promoter Ti Prime (MicroChemicals) and then with 0.5% polystyrene in toluene at 3,000 rpm. Polystyrenecoated coverslips were oxidized through oxygen plasma (FEMTO; Diener Electronic) for 10 s at 30 W before incubation with 0.1 mg/mL poly-L-lysine(20)-g[3.5]-poly-ethyleneglycol (PLLPEG) (Surface Solutions) in 10 mM Hepes (pH 7.4) for 15 min. After drying, coverslips were exposed to deep UV (UVO cleaner; Jelight) through a photomask (Toppan) for 2 min. Right after UV activation, coverslips were incubated with 20 μg/mL fibronectin (Sigma), another selected ECM protein when speTseng et al. www.pnas.org/cgi/content/short/1106377109

cifically mentioned (type 1 collagen from rat tail; Sigma), or laminin (Sigma), and 10 μg/mL of fluorescent fibrinogen conjugate (Invitrogen) solution in PBS for 30 min. Coverslips were washed three times with sterile PBS before plating cells. Polyacrylamide Micropatterning. Micropatterned polyacrylamide gels were made as previously described (2). Briefly, fluorescent beads (dark red 200 nm; F-8807; Invitrogen) were first passivated by PLL-PEG (2) (Surface Solutions). Acrylamide (6.67%) and bis-acrylamide (0.167%) solution, containing the passivated beads, was polymerized between the photomask (Toppan) and acryl-silanized coverslips. The resulting polyacrylamide gel had a thickness of 70 μm and rigidity of 7 kPa. After deep UV activation through the photomask, the gel was incubated with the crosslinker N-hydroxysuccinimide (Fluka) and 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (Pierce) before coating with fibronectin (Sigma) and fluorescent conjugated fibrinogen (Invitrogen). Physical Modeling and Numerical Simulations. When cells spread on a patterned substrate, they optimize their shape by minimizing energy. This simple argument has been demonstrated in the case of single cells and a lattice of adhesive proteins. The extension of this argument to the case of two cells is the purpose of this material. We suppose that the energy of the cell doublet can be separated into three parts as in the equation

H ¼ H1 þ H2 þ H12 ; where H1 and H2 are the self-energy, respectively, of cell 1 and of cell 2 within the doublet. The coupling term H12 accounts for the interaction energy between cell 1 and cell 2. In our model, cells are considered as 2D homogeneous material. The self-energy terms include (i) an adhesion term, corresponding to the energy gain upon binding of integrins with ECM, (ii) a line energy term mimicking the energy cost of the tension generated by the actomyosin contractility, and (iii) a compressibility term that fixes the mean cell area and allocates an energy cost to variations of cell area. These terms have been thoroughly discussed in the framework of the cellular Potts model in refs. 3–6. To further simplify the analysis, we suppose that first, cell shapes adopt the convex envelope of the micropattern, second, cells have equal areas in the doublet, third, the interfaces are straight, and fourth, this interface passes through the center of the pattern. With these simplifications, terms (i) and (iii) in the self-energy of cells can be dropped. The interaction energy H12 is written as H12 ¼ Jcellcell :Lcellcell ; where Jcell-cell measures the strength of the cadherin–cadherin interaction and Lcell-cell is the length of the interface between cell 1 and cell 2. Dropping all of the constant terms, the self-energy of cell i = 1,2 is expressed as X JK :Lk ; Hi ¼ k

where k is the interface type, Lk is the interface length, and Jk is the line tension of the interface. According to the experimental measurements of the forces (Fig. 5 and Fig. S7), the interfaces have been classified into four types: 1 of 4

k = intra-noECM; corresponds to the type of interface linking integrins to integrins and standing above nonadhesive regions; k = inter-noECM; corresponds to the type of interface linking integrins and cadherin above nonadhesive regions; k = intra-ECM; corresponds to the type of interface linking integrins to integrins and standing above adhesive regions; k = inter-ECM; corresponds to the type of interface linking integrins and cadherin above adhesive regions. The total energy of the doublet depends then on a set of parameters describing the cell states and on a single variable that defines the orientation of the interface. Our computation will consider various ECM patterns, and we will identify the energy minima as a function of the doublet configuration described by a unique variable. We compare the computations on various patterns with experimental observations. At first we ignore the role of ECM and set all of the cortical line energy equal, that is,

orientation for various types of patterns. Because the cortical line energies are all equal, the self-energy is a constant and is proportional to the perimeter of the doublet. Thus, the variation of the doublet energy is determined only by the changes in intercellular junction length. The energy profile is also independent of the pattern, because they have by construction all of the same perimeters. Considering isotropic line tension is not sufficient to find all of the equilibrium positions of the doublets on the patterns, except in the case of the [square]-shaped micropattern (Fig. S7). Next, we consider anisotropic cortical energy in response to the presence of ECM and fix the values of the line energies in the same ratio as the force measurements: JintranoECM ¼ 1:3; JinternoECM ¼ 0:8; and JintraECM ¼ JinterECM ¼ 1:

We also set a value for the cadherin–cadherin interface line energy. Because this interface fluctuates more than the others, the line energy must be smaller and we arbitrarily fixed it to 10% of Jintra-ECM. The energy was plotted as a function of the doublet

We kept Jcell-cell = 0.1. The energy plots are represented in Fig. S7. For clarity, all of the profiles have been shifted by a constant value. In the case of anisotropic cortical tensions, the positions of the energy minima vary from pattern to pattern and systematically correspond to the orientations observed experimentally.

1. Azioune A, Carpi N, Tseng Q, Théry M, Piel M (2010) Protein micropatterns: A direct printing protocol using deep UVs. Methods Cell Biol 97:133e146. 2. Tseng Q, et al. (2011) A new micropatterning method of soft substrates reveals that different tumorigenic signals can promote or reduce cell contraction levels. Lab Chip 11:2231e2240. 3. Käfer J, Hayashi T, Marée AF, Carthew RW, Graner F (2007) Cell adhesion and cortex contractility determine cell patterning in the Drosophila retina. Proc Natl Acad Sci USA 104:18549e18554.

4. Vianay B, et al. (2010) Single cells spreading on a protein lattice adopt an energy minimizing shape. Phys Rev Lett 105:128101. 5. Farhadifar R, Röper JC, Aigouy B, Eaton S, Jülicher F (2007) The influence of cell mechanics, cell-cell interactions, and proliferation on epithelial packing. Curr Biol 17: 2095e2104. 6. Rauzi M, Verant P, Lecuit T, Lenne PF (2008) Nature and anisotropy of cortical forces orienting Drosophila tissue morphogenesis. Nat Cell Biol 10:1401e 1410.

JintranoECM ¼ JintraECM ¼ JinternoECM ¼ JinterECM ¼ 1:

Fig. S1. Automated analysis of cell positions and movements. (A) Automated detection of mitosis events from fluorescent images of nucleus staining with Hoechst in 10× video-microscopy acquisition and a 15-min time frame. Fluorescence intensity increase during metaphase and object separation during anaphase were both used to detect mitosis. Detections of first and subsequent mitosis were then used to determine the part of the movie in which daughter cell positions could be analyzed. (B) Example of image thresholding and segmentation to detect nucleus positions and automatically record nucleus–nucleus orientation. (C) The measured angles of nucleus–nucleus axis orientation over time (Upper Left) from all movies were pooled and plotted together in a circular histogram showing the angular distribution of cell doublets (Upper Right). The histogram has been made circular for clarity but is identical modulo 180°. In addition, for each movie, the instantaneous angular speed was calculated at every time point (Lower Left). Comparison between measured value and personal visual appreciation of movement led to the determination of an arbitrary threshold of 0.3°/min above which cells were considered as moving. For each movie, this threshold was used to calculate the proportion of moving versus nonmoving frames. All these percentages (one per movie) are graphically represented in a scatterplot graphic (Lower Right). Fig. S1

Fig. S2. Intercellular junctions are stabilized by the absence of ECM at their extremities. Cell positions and movement analysis (Fig. S1) of cells on five different geometries of fibronectin-coated micropatterns (images). The presence of anisotropic adhesive or nonadhesive regions in the central part of the geometry had no effect on junction orientation (first to third rows). The absence of ECM at the extremities of the junction stabilizes the junction (fourth and fifth rows). Fig. S2

Fig. S3. Cell division orientation is not sufficient to determine daughter cell positioning. The automated detection of mitosis (Fig. S1A) was used to measure cell division orientation (nucleus–nucleus orientation during anaphase) on [H] (Upper) and [hourglass] (Lower). Angular distributions of cell division orientation were only slightly biased by ECM geometry (Left) because micropattern size was not adapted to single cells that were moving on the micropatterns (Movie S3). Daughter cells refine their positioning after cell division and finally adopted much more biased orientations (Right). These data were duplicated from Fig. 1 to be easily compared with cell division orientation. Cell-doublet orientation appeared to result from an active positioning process of daughter cells rather than oriented cell division. Fig. S3

Tseng et al. www.pnas.org/cgi/content/short/1106377109

2 of 4

Fig. S4. Molecular players implicated in intercellular junction positioning. Panels 1–9: Wild-type daughter cell doublets were plated on fibronectin micropatterns, treated with chemical inhibitors for 6 h, and fixed. Panel 10: Wild-type daughter cell doublets were plated on laminin micropatterns. Panels 11 and 12: siRNA-treated cells were plated on micropatterns and fixed 30 h later. Panel 13: Overlay of all angular distributions of cell–cell junctions to help comparison. Cells were fixed in methanol and immunostained against α-catenin (green) and DNA (Hoechst; blue). Each panel shows the spatial distribution of intercellular junction extremities and the angular distribution of junction orientations (measured as described in Fig. 4A) as well as a representative set of images that have been used to perform these measurements. For chemical inhibitors, the panel title indicates the drug target and drug concentration (panels 1–9). For siRNA, the panel title indicates the siRNA target, and bottom images show Western blot quantifications of RNA content with control siRNA (directed against GFP) and the tested siRNA. Fig. S4

Fig. S5. Comparison between cell–cell forces and inter oriented forces. (A) Global traction maps were used to list all traction forces and their positions in the cell. (B) Cell–cell force corresponds to the force that opposes traction forces on the substrate (Left), whereas inter force represents the sum of all force components that were oriented perpendicular to the intercellular junction (Right). (C) Inter force, that is, the total traction force component oriented toward the intercellular region, is plotted against the genuine cell–cell force calculated without any consideration of force orientation and spatial distribution. In all cases, the inter oriented forces match the genuine cell–cell force. This was particularly true on [H], where all traction force components toward the intercellular junction contributed to the total cell–cell force. Fig. S5

Fig. S6. ECM induces opposite effects on intra- and intercellular forces. (A) Decomposition of traction forces into intra forces oriented toward intracellular space and inter forces oriented toward the intercellular junction. The role of ECM was tested by comparing intra- and intercellular forces along cell edges with ECM (blue arrows in scheme, blue dots in graphs) or without ECM (red arrows in scheme, red dots in graphs). Intracellular forces were compared between [H] and [X], which provide similar intercellular forces along edges without ECM (Left). Intercellular forces were compared between [square] and [H], which provide similar intracellular forces along edges with ECM (Left). The experiments were repeated three times. Because gel rigidities vary from one slide to the other, the absolute force value changed slightly but the overall tendencies were reproducible. All these data were combined and are shown in Fig. 6E. (B) Averaged traction force field on a [C]-shaped micropattern. Magnifications correspond to the white square regions in global maps. Arrows indicate force orientation; color and length both represent local force magnitude in pN. Forces were decomposed into intra- and intercellular forces. Intercellular forces along the adhesive edge were compared with those on the nonadhesive edge. The reduction of intercellular forces due to the absence of ECM, previously observed in distinct cells plated on different micropatterns, was even confirmed within individual cells. All statistical comparisons used Student’s t tests, **P < 0.005, ***P < 0.001. Fig. S6

Fig. S7. Modeling of the energetic costs of intercellular junction orientations. Numerical simulations of the model described in Fig. 6 and SI Methods were performed for various ECM micropattern geometries (Center and Left, respectively). Simulations were performed in the case of constant tension (regardless of the presence of ECM) (blue curves) and regulated intra- and intercellular tension in response to the presence or absence of ECM (red curves) as described in Fig. 6. Energies were calculated for all intercellular junction orientations (0° corresponds to horizontal junction orientation). Experimental data are reported (Right). These data are reported from Figs. 1–3 where they are shown as rose diagrams except for the data shown in the second row, which were not shown in these figures. In all cases, the most frequent intercellular junction orientations observed experimentally corresponded to the energetic wells obtained with the simulations performed in the case of ECM-dependent tensions. This correlation suggests that the regulation of intra- and intercellular forces by ECM could account for the preferred intercellular junction orientation observed experimentally. Fig. S7

Movie S1. Cell movement on [ring]-shaped micropatterns. Montage of time-lapse movies acquired through a 10× dry objective. Nuclei were stained in living cells with Hoechst 33342 and visualized with epifluorescence microscopy. Images were taken every 15 min. Movie S1

Tseng et al. www.pnas.org/cgi/content/short/1106377109

3 of 4

Movie S2. Cell movement on a [square]-shaped micropattern. A [square]-shaped fibronectin micropattern is shown in the left panel. Phase-contrast images during time-lapse acquisition are shown on the right. Daughter cells moved regularly around each other. Images were taken every 15 min. Movie S2

Movie S3. Cell movement on an [H]-shaped micropattern. An [H]-shaped fibronectin micropattern is shown in the left panel. Phase-contrast images during time-lapse acquisition are shown on the right. At the top, cell division is oriented with respect to micropattern geometry, and daughter cells remained positioned according to this orientation. On the bottom, cell division was misoriented but daughter cells corrected their position and adapted it to the micropattern geometry. In both cases, cell positions were maintained steadily until the next division. Images were taken every 15 min. Movie S3

Tseng et al. www.pnas.org/cgi/content/short/1106377109

4 of 4

Figure S1: automated analysis of cell position and movements t=0

A

1h30

1h45

MITOSIS DETECTION B

3h30

4h45

12h15

14h45

15h

15h30

17h15

120min

135min

MITOSIS DETECTION

selected frames for nuclei movement analysis

raw data t=0

15min

30min

45min

60min

75min

90min

105min

thresholded and segmented pictures

nucleus-nucleus axis measurement

C

angular distribution of nucleus-nucleus axis orientation

90°

180

angle (degree)

150

hundreds of movies

120 90

30 to 40 angle values per analyzed movie

60

180°

0°

30 0

0

200

400

600

8%

800

16%

time (min)

270° ncell = 164 speed =

nangle= 7669

angle(t2) - angle (t1) t2 - t1

angular speed (degree/min)

proportion of moving frames (%) 1.5

100

1.0

moving frames

0.5

non-moving frames 0.0

0

200

400

600

one percentage of moving frame per analyzed movie

hundreds of movies

60

800

time (min)

20

nmovies = 164

Figure S2: Intercellular junctions are stabilised by absence of ECM at their extremities

proportion of moving frames (%)

orientation of nucleus-nucleus axis ncell = 164

100

90°

nangle= 7729 60 180°

0° 8%

20 16%

270° ncell = 156

100

90°

nangle= 7759

cell doublet periphery is homogeneously adhesive

60 180°

0° 8%

20 16%

270° ncell = 205

90°

100

nangle= 9078

60 180°

0° 8% 20

16% 270°

ncell = 136

90°

100

nangle= 6124

60 180°

0° 8% 20 16%

cell doublet periphery contains non-adhesive regions

270° ncell = 211

90°

100

nangle= 7812

60 180°

0° 8%

20 16%

270°

Figure S3: Cell division orientation is not sufficient to determine daughter cell positioning

CELL DIVISION ORIENTATION

DAUGHTER CELL DOUBLET ORIENTATION

90°

90°

180°

0°

8

180°

0°

8 24

24

40%

40% 270°

270°

180°

N cell = 117 N angle = 117

N cell = 117 N angle = 5442

90°

90°

0°

8

180°

0°

8 24

24

40%

40% 270° N cell = 69 N angle = 69

270° N cell = 69 N angle = 3991

Figure S4-1

CONTROL

cell proportion (%)

40 30 20 10 0

50 40 30 20 10 0

0

10

20

30

40

0 10 20 30 40 50 60 70 80 90 angle (degree)

Figure S4-2

FAK inactivation with PF573528, 1µM

cell proportion (%)

40 30 20 10 0

50 40 30 20 10 0

0

10

20

30

40

FAK inhibition control

0 10 20 30 40 50 60 70 80 90 angle (degree)

Figure S4-3

ERK1 inactivation with PD98059, 2µM

cell proportion (%)

40 30 20 10 0

50 40 30 20 10 0

0

10

20

30

40

ERK1 inhibition control

0 10 20 30 40 50 60 70 80 90 angle (degree)

Figure S4-4

Src inactivation with SU6656, 5µM

cell proportion (%)

40 30 20 10 0

50 40 30 20 10 0

0

10

20

30

40

Src inhibition control

0 10 20 30 40 50 60 70 80 90 angle (degree)

Figure S4-5

JNK inactivation with SP600125, 1,8µM

cell proportion (%)

40 30 20 10 0

50 40 30 20 10 0

0

10

20

30

40

JNK inhibition control

0 10 20 30 40 50 60 70 80 90 angle (degree)

Figure S4-6

ROCK inactivation with Y27632, 5µM

cell proportion (%)

40 30 20 10 0

50 40 30 20 10 0

0

10

20

30

40

ROCK inhibition control

0 10 20 30 40 50 60 70 80 90 angle (degree)

Figure S4-7

Myosin II inactivation with Blebbistatin, 15µM

cell proportion (%)

40 30 20 10 0

50 40 30 20 10 0

0

10

20

30

40

MyosinII inhibition control

0 10 20 30 40 50 60 70 80 90 angle (degree)

Figure S4-8

MLCK inactivation with ML7, 5µM

cell proportion (%)

40 30 20 10 0

50 40 30 20 10 0

0

10

20

30

40

MLCK inhibition control

0 10 20 30 40 50 60 70 80 90 angle (degree)

Figure S4-9

Rac inactivation with NCS23766, 5µM

cell proportion (%)

40 30 20 10 0

50 40 30 20 10 0

0

10

20

30

40

Rac inhibition control

0 10 20 30 40 50 60 70 80 90 angle (degree)

Figure S4-10

laminin on micropattern

cell proportion (%)

40 30 20 10 0

50 40 30 20 10 0

0

10

20

30

40

laminin fibronectin

0 10 20 30 40 50 60 70 80 90 angle (degree)

Figure S4-11

siRNA against GFP (si RNA control)

cell proportion (%)

40 30 20 10 0

50 40 30 20 10 0

0

10

20

30

40

si GFP control

0 10 20 30 40 50 60 70 80 90 angle (degree)

Figure S4-12

siRNA against p120-catenin

cell proportion (%)

40 30 20 10 0

50 40 30 20 10 0

0

10

20

30

40

si p120-catenin si GFP

0 10 20 30 40 50 60 70 80 90 angle (degree)

si GFP si p120catenin anti-p120catenin anti-actin

Figure S4-13

50

cell proportion (%)

40

PF573528 (FAK) SP 600125 (JNK) si GFP SU6656 (Src) collagen NCS 23766 (Rac) CONTROL (fibronectin)

30

20

PD 98059 (ERK1) ML7 (MLCK) si P120 catenin laminin Y27632 (ROCK) Blebbistatin (Myosin II)

10

0 0

10

20

30

40

50

angle (degree)

60

70

80

90

Figure S5: Comparison between genuine cell-cell force and «inter» force

A measurement of all traction forces

Fi

B

decomposition of force coordinate

mechanical balance in each cell

Fi

Fi

Fi = F i + F i Fi

F = −ΣF i

F = −ΣF i

inter

cell-cell

C 200

slope = 1.13

50

2

(nN)

200

150

inter

100

slope = 1.04

50

2

0

50

100

150

Fcell-cell

150 100

slope = 1.06

50

2

R = 0.93

R = 0.83

R = 0.4 0

F

(nN) inter

100

F

(nN)

150

F inter

200

200

(nN)

0

0

50

100

Fcell-cell

150

200

(nN)

0

0

50

100

Fcell-cell

150

(nN)

200

Figure S6: ECM induces opposite effect on intra- and inter-cellular forces

A

INTER-cellular forces

***

**

300

Force (nN)

Force (nN)

INTRA-cellular forces

experiment 1

200 100

***

300 200

experiment 2

100 0

100 50

200

***

150 100 50 0

**

300

Force (nN)

Force (nN)

150

0

Force (nN)

Force (nN)

0

200

200

experiment 3

100

** 200 150 100 50 0

0

WITH WITHOUT ECM ECM

WITH WITHOUT ECM ECM

B 1250

1250

1000

1000

750

750

500

500

250

n = 25

Averaged traction map Force (nN)

Fibronectin

1500 pN

100

***

50 0

inter inter adhesive non-adhesive

250

Figure S7: Modeling of the energetic cost of intercellular junction positioning

Energy (a.u.)

cell proportion (%)

2 1.5 1 0.5 0 -45

0

45

90 135

2 1.5 1 0.5 0 -45

0

45

90 135

2 1.5 1

Energy (a.u.)

0

45

90 135

2 1.5 1 0.5 0

45

90 135

cell proportion (%)

0 -45

Energy (a.u.)

cell proportion (%)

2.5

cell proportion (%)

Energy (a.u.)

3

0.5 -45

EXPERIMENT

cell proportion (%)

Energy (a.u.)

MODEL

2 1.5 1 0.5

0 -45 0 45 90 135 intercellular junction orientation (degrees)

ECM dependent ECM independent

20

Ncell = 146 Nangle = 8561

15 10 5 0 -45

0

45

20

90

135

Ncell = 121 Nangle = 4931

15 10 5 0 -45

0

45

30

90

135

Ncell = 131 Nangle = 5991

20 10 0 -45

0

45

40

90

135

Ncell = 117 Nangle = 5442

30 20 10 0 -45 40 30

0

45

90

135

Ncell = 86 Nangle = 5171

20 10 0 -45 0 45 90 135 intercellular junction orientation (degrees)