Regulation of cell cycle progression by cell–cell and cell–matrix forces

It has long been proposed that the cell cycle is regulated by physical forces at the cell–cell and cell–extracellular matrix (ECM) interfaces1–12. However, the evolution of these forces during the cycle has never been measured in a tissue, and whether this evolution affects cell cycle progression is unknown. Here, we quantified cell–cell tension and cell–ECM traction throughout the complete cycle of a large cell population in a growing epithelium. These measurements unveil temporal mechanical patterns that span the entire cell cycle and regulate its duration, the G1–S transition and mitotic rounding. Cells subjected to higher intercellular tension exhibit a higher probability to transition from G1 to S, as well as shorter G1 and S–G2–M phases. Moreover, we show that tension and mechanical energy are better predictors of the duration of G1 than measured geometric properties. Tension increases during the cell cycle but decreases 3 hours before mitosis. Using optogenetic control of contractility, we show that this tension drop favours mitotic rounding. Our results establish that cell cycle progression is regulated cooperatively by forces between the dividing cell and its neighbours. Monitoring growing epithelial cells through the cell cycle, Uroz et al. find that cell–cell tension and cell–matrix traction forces differ across the cell cycle and affect cell cycle duration, the G1–S transition and mitotic rounding.

and from G2 to mitosis 10 . Later phases of the cell cycle, such as mitosis and cytokinesis have also been shown to be influenced by mechanical constrains 11 , by cell stretching 6 , and by the mechanical action of neighboring cells 9 . Altogether, these findings support the long-standing hypothesis that the cell reads out a force, a deformation, or their rates to decide whether it progresses through or exits the cell cycle 7 .
Despite the increasing evidence for mechanical regulation of cell proliferation, the evolution of cell-cell and cell-matrix forces during the cell cycle has never been reported in an epithelium. Without this key information, the question of whether progression through the various phases of the cell cycle is mechanically driven remains unresolved. Here we provide a systematic analysis of cell-ECM traction forces, cell-cell forces, as well as cell and nuclear shape throughout the cycle of a large number of cells in an expanding epithelial monolayer. We show that cellular forces impact various phases of the cell cycle, including its duration, the G1-S transition, and mitotic rounding.

Main
As a model system for epithelial growth we used the expansion of a micropatterned colony of MDCK cells. We placed a polydimethylsiloxane (PDMS) membrane with a 300 µm-wide rectangular opening on top of a collagen-I-coated polyacrylamide gel (11 kPa in stiffness) 26,27 . To monitor the cell cycle during growth of the colony, we seeded MDCK-Fucci cells on the pattern and allowed them to adhere and form a confluent monolayer. MDCK-Fucci cells express Ctd1-RFP during G1 and S phases and Geminin-GFP during the S-G2-M phases, which allowed us to monitor the state of each cell in the cycle 6 (Fig. 1a,f). Four hours after seeding, the PDMS membrane was removed and cells migrated uni-dimensionally towards the newly available surface 26,27 (Fig. 1a, Supplementary Video 1). During the first hours of expansion, the monolayer flattened but no differences in height were observed between leading edge and bulk ( Supplementary Fig. 1a-f). We used Traction Force Microscopy (TFM) to map traction forces at the cell substrate interface 28 (Fig. 1b), and Monolayer Stress Microscopy (MSM) to map in-plane tension between and within cells (Fig. 1c) 29 . Cell traction and tension increased during the first hours of expansion, as previously reported 27 , but then decreased towards a plateau (Figure 1d-e, Supplementary Fig. 2a-c, Supplementary Video 1). In parallel with collective migration, cells divided frequently across the monolayer (Supplementary Video 1). During the first ~12h of expansion, the average cell area increased smoothly. The cell area then remained constant until the end of the experiment (Fig. 1g-i). By contrast, the average cell area in monolayers in which cell cycle progression was arrested with thymidine and aphidicolin showed a continuous growth, reaching a 4-fold increase from the initial area ( Fig. 1g-k, Supplementary Video 2). This result points to a mechanism by which MDCK monolayers coordinate their growth and division cycle to maintain a constant cell density.
To investigate whether the duration of the cell cycle was mechanically regulated we followed 120 cells from three independent monolayers throughout a complete cycle. The average duration of the cell cycle was 21.4 ± 0.4 h, of which 12.6 ± 0.3h corresponded to G1 and 8.8 ± 0.1h corresponded to S-G2-M (Fig. 2a). On average, cell area exhibited a linear 5-fold increase during the cycle, which arises from both cell growth and spreading (Fig. 2b,c). We then asked whether a specific mechanical property promoted the transition from G1 to S. We reasoned that, if this was case, this mechanical property should be significantly different in cells that had transitioned to S than in cells of the same age (defined as time elapsed since anaphase) that had remained in G1 30 . To test this rationale, we focused on the time points at which only a small fraction of the cells had transitioned to S, and plotted the average cell area, nuclear area, traction force, and cellular tension for cells of the same age in G1 and S. Cells that had transitioned to S exhibited significantly higher cell and nuclear areas than those that had stayed in G1 (Fig.  2d,e). Tension was also higher in cells in S than in identically-aged cells in G1 (Fig. 2f). By contrast, traction forces did not show significant differences in G1 and S (Fig. 2g). This analysis is consistent with previous studies showing that cellular and nuclear spreading favor the G1-S transition 6,8 , but it also raises the possibility that tension favors progression from G1 to S.
Since geometric and mechanical properties might be linked, we sought to determine the extent to which each one of them predicts cell cycle progression. As potential geometric predictors we considered the cell and nuclear areas shortly after the beginning of the cycle (A 0 c and A 0 n , measured 60 min after cell birth), as well as their average over the full length of G1 (A G1 c and A G1 n ), and over the first  hours of the cycle (A  c and  ), where =6h is the latest time point at which all cells remained in G1. We also considered the average rates of change of cell and nuclear area over G1 (̇G 1 c and ̇1 ) and over  (̇ c and ̇ n ). Note that volume and mass were not experimentally accessible in our system and, therefore, growth rates reported in our study are changes in 2D projected areas of cells and nuclei. Given that cells are simultaneously growing and spreading, the relationship between area and volume is not straightforward 31, 32 . As mechanical predictors, we considered traction force and tension averaged over G1 (T G1 c and  G1 c ) and over  (T  c and   c ).
We then tested whether these properties correlated with the duration of G1 ( 1 ) (Fig. 2h, see Supplementary Fig. 3a-b for a complete correlation matrix between all measured properties).
Cell and nuclear area at the beginning of the cycle, as well as their time averages over  and over the duration of G1 did not correlate with 1 (Fig. 2h, Supplementary Fig. 4a-d). By contrast, the area growth rates of the cell and the nucleus correlated with 1 with high statistical significance (p=2×10 -5 and p=5×10 -13 , respectively) ( Fig. 2l). This result shows that cells that grow faster in area divide earlier independently of their initial area. Importantly, tension also exhibited a high correlation with 1 (p=2×10 -6 ) (Fig. 2m). This was not the case for traction forces (Fig. 2h). Similar results were obtained by averaging mechanical properties between cell birth and the peak of Ctd1 signal of each cell (Supplementary Figure 3c,d). Tension and nuclear area growth rate also correlated negatively with 1 on softer substrates (2.4 kPa) and, to a lesser extent, on substrates coated with fibronectin, thereby supporting the generality of our findings (Fig. 3). Overall, this correlation analysis shows that both mechanical and geometric properties are excellent candidates to explain 1 .
To establish which of these properties is more predictive of 1 , we carried out a modelselection analysis 33, 34 . This analysis compares a set of statistical models and selects those that are more plausible according to the Bayesian Information Criterion (BIC, see methods). We first compared all models that are linear in one of the properties measured either at the beginning of the cycle or averaged over . We excluded properties averaged over the full length of G1 from this analysis, as these averages can include implicit 1 and lead to trivial predictions. This analysis showed that tension was the most predictive property of 1 , followed by nuclear area growth rate and nuclear area (Fig. 2j). Thus, if one has experimental access to all 2D geometric and mechanical properties over the first 6h of the cycle, then the tension average of this period will be the best predictor of the duration of G1.
We also asked whether combination of cellular properties could better predict 1 than single properties. To do so we tested all possible models involving the product of any pair of properties, irrespective of any a priori consideration with regard to their physical or biological meaning (Fig. 2p, Supplementary Fig. 5). Our analysis concluded that the product between cell area and tension outperforms the predictive power of any single property by more than one order of magnitude (Fig. 2p,q). The product of tension by area has units of energy and is related to the mechanical energy associated with the growth of the cell during the first  hours of G1 (  ). A direct consequence of the predictive power of  is that cells with the same area should enter G1 earlier if they are subjected to higher tension. To test this prediction, we ranked the 120 cells in area deciles and, for each decile, we computed the correlation between tension at 6h and 1 (Fig. 2r). With the exception of the first decile (smallest cells), the correlation was always negative, confirming that within a population of cells with same area, those subjected to higher tension will progress faster in their cycle. This relationship cannot be explained through changes in cell height, as height and tension did not correlate for cells of the same area ( Supplementary Fig. 1g-i). Finally, we explored all models that are multilinear on any pairwise combination of single properties and property products. None of these models had more predictive power than the best linear models of a single product of two properties (Supplementary Table 4).
We next carried out an analogous analysis asking whether mechanical properties during S-G2-M (labelled hereafter with the subscript SG2M) correlate with the combined duration of these phases of the cycle ( 2 ). Unlike the case of G1, none of the mechanical properties measured during S-G2-M correlated with 2 ( Fig. 2i, left panel). By contrast, cell and nuclear area growth rates averaged over G1 correlated with 2 (p=4×10 -4 and p=7×10 -7 , respectively) and so did cellular tension (p=7×10 -7 ) ( Fig. 2i right panel, Figs. 2n,o). This result suggests that the S-G2-M phases have mechanical memory of G1, in the sense that they are influenced by the mechanical state of the cell during that earlier phase ( Supplementary Fig. 3e). Analysis of linear models involving a single mechanical property showed that nuclear area growth rate and tension averaged over G1 were the best predictors of 2 (Fig. 2k). Conversely, analysis of models involving a product between pairs of properties showed that the product between nuclear area growth rate and tension during G1 was the most predictive ( Supplementary Fig.  5). This product can also be interpreted in terms of energy because the average area growth rate during G1 and the area at the G1-S transition are highly correlated (Supplementary Fig.  4e-g). In summary, our statistical analysis shows that mechanical properties such as tension and mechanical energy are powerful predictors of the duration of the different phases of the cell cycle, outperforming geometric features such as cell and nuclear areas and their rates of change.
To study in further detail the mechanical regulation of cell division, we focused on the time evolution of local tension during the full length of the cell cycle. For each time point, we averaged tension in three concentric regions around each cell of interest (Fig. 4a,b). The first region covers the area of the cell of interest (green area in Fig 4a,b), and the second and third regions are the two annuli consecutively concentric to that cell (red and blue areas in Fig. 4a,b, respectively). To average out intercellular variability and to isolate variations associated with the cell cycle from global mechanical trends of monolayer expansion, we computed ratios of tension between pairs of regions (Fig. 4c). A tension ratio between regions i and j was then labelled as ij, where indices i and j run from 1 to 3 (Supplementary Video 3). Throughout the cycle, 23 was close to unity, indicating the absence of systematic long-ranged tensional trend in the neighborhood of the dividing cell (Fig. 4d). By contrast, 12 and 13 showed systematic departures from unity. At the beginning of the cycle, 12 and 13 decreased slightly for ~1 hour. Afterwards, they increased steadily through most of the cycle. Three hours before mitosis, 12 and 13 begun a progressive decrease and attained their minimum value at mitosis (Fig. 4d).
13 dropped further than 12 indicating that tension in the immediate neighbors of the dividing cell decreased more than in distant cells.
This slow decline in tension suggests a regulatory mechanism that precedes mitosis. To investigate such mechanism we studied how mitosis is affected by tensional differences in the local environment of the dividing cell. To do so, we resorted to optogenetics to selectively increase or decrease tension in the neighbors of the dividing cell. We generated mosaic monolayers in which 10% of the cells expressed Fucci and the remaining 90% were engineered to either increase (MDCK OptoGEF-Contract) or decrease (MDCK OptoGEF-Relax) tension upon illumination with low doses of blue light 35 (Fig. 4e,g). When tension in the neighbors of a dividing cell was optogenetically increased, mitotic rounding slowed down (Fig. 4e,f). By contrast, when tension in the neighbors was decreased, rounding time accelerated (Fig. 4g,h). Previous work at the single cell level showed that preventing mitotic rounding by placing an object in contact with a dividing cell delays mitosis and causes mitotic defects such as spindle misassembly and pole splitting 11 . This delay was attributed to the need of the mitotic cell to generate an additional force to round up against the object. Our work suggests that increased tension in the neighbors of the mitotic cell also delays mitosis by preventing rounding. The drop in intercellular tension observed well before division might thus be a regulated process to ensure proper rounding and the absence of mitotic defects.
We next focused on the time evolution of traction ratios Tij throughout the full length of the cell cycle (Fig. 5a,b). Unlike tension ratios, traction ratios T12 and T13 were constant during most of G1 and S-G2. This result is consistent with recent findings in single cells showing that tractions plateau between late G1 and S phases 12 (Fig. 5b). As cells rounded up for division, T12 and T13 exhibited a peak flanked by two periods of low traction (Fig. 5b,d). By contrast, single MDCK cells fully relaxed their tractions during rounding until respreading of daughter cells (Fig. 5c,d). To explore how tractions can develop under a rounding cell, we turned to mosaic monolayers in which 80% of the cells expressed Lifeact-GFP and the remaining 20% expressed Lifeact-Ruby. Confocal stacks revealed that neighbors of the dividing cell wrapped around it to maintain a largely continuous interface throughout mitosis (Fig 5e-g) Fig. 5h,i). Together, these results raise the possibility that coordinated protrusion and retraction of neighboring cells assists rounding of the mother cell and re-spreading of its daughters.
Cell mechanics has long been implicated in the regulation of cell proliferation 1-12, 22, 38 , but how cellular forces evolve through the cell cycle, and whether this evolution is associated with the duration of each of its phases has been unknown thus far. Here we showed that cell-cell forces impact various phases of the cell cycle, including its duration, the G1-S transition, and mitotic rounding. We showed, further, that tension of the dividing cell relative to that of its surrounding neighbors increases smoothly through most of G1, S and G2. After an initial spreading phase, MDCK monolayers maintained a largely constant density, which suggests coordination between cell growth and division machineries during tissue expansion. This type of coordination has been observed in a variety of tissues during development and homeostasis 19, 31, 39 , which has led to the idea that cells read out their size to control progression through the cycle 6,24,30 . In the specific case of epithelial monolayers, previous experiments provided evidence that the G1-S and G2-M transitions are regulated by exogenous control of the cell area 6,8,10 . Here we confirmed that cell and nuclear areas, as well as their rates of change, predict progression through the cell cycle. More importantly, we found that tension and mechanical energy have a higher predictive power than geometrical properties; for a given cell area, cells under higher tension display a shorter G1. Together these results point to distinct mechanisms by which cells probe area and tension to progress through their cycle. The nature of these mechanisms and how they might be integrated to control the duration of G1, S and G2 are major questions that our study raises for future investigations.

Acknowledgments
We thank Natalia Castro for technical assistance, Alberto Elosegui-Artola for assistance with data analysis, and Lars Hufnagel for generously providing MDCK-Fucci cells. We are grateful to Buzz Baum and members of our groups for insightful comments and discussion.

Preparation of polyacrylamide gels
Glass-bottom dishes were activated by using a 1:1:14 solution of acetic acid/bind-silane (M6514, Sigma)/ethanol. The dishes were washed twice with ethanol and air-dried for 5 min. For 11kPa (2.4kPa) gels, a 500µl stock solution containing HEPES 10mM, 93.75µl (68.75µl) acrylamide 40% (161-0140, BioRad), 25µl (11µl) bisacrylamide 2% (161-0140, BioRad), 2.5µl 10% ammonium persulfate diluted in water (161-0700, BioRad), 0.25µl TEMED and 12µl of 200-nm-diameter far red fluorescent carboxylate-modified beads (F8807, ThermoFisher) was prepared. A drop of 18 µl was added to the centre of the glass-bottom dishes and the solution was covered with 18-mm-diameter GelBond film (Lonza) coverslips (hydrophobic side down) that were custom cut by an electronic cutting tool (Silhouette Cameo). After 40 min polymerization, the coverslip was removed and gels were functionalized using sulfo-sanpah. Briefly, a 80µl drop of sulfo-sanpah (22589, Thermo-Scientific) was placed on the top of the polyacrylamide gel and activated by UV light for 3 min. Sulfo-sanpah was diluted in miliQ water to a final concentration of 2mg/ml from an initial dilution 50mg/ml kept at -80ºC. Gels were then washed twice with miliQ water and once with PBS for 5min each. Afterwards, gels were incubated with 200µl of a collagen I or fibronectin solution (0.1mg ml −1 ) overnight at 4ºC.

Microfabrication of the PDMS membranes
SU8-50 masters containing rectangles of 300×2500 μm were fabricated using conventional photolithography. Uncured PDMS was spin-coated on the masters to a thickness lower than the height of the SU8 feature (35 μm) and cured for 4h at 60 °C. A thicker border of PDMS was applied at the edges of the membranes for handling purposes. PDMS was then peeled off from the master and kept in ethanol at 4 °C until use.

Cell patterning
Before seeding the cells, PDMS membranes were incubated in a solution of 2% Pluronic F-127 (Sigma-Aldrich) in PBS to prevent damage of the gel coating caused by the PDMS membranes. At the same time, the gels coated with collagen were washed twice with PBS, covered with cell media and kept in the incubator.
One hour after incubation of PDMS membranes in 2% Pluronic solution, the membranes were washed twice with PBS and air dried for 20 min. After removing the media, the gels were air dried for 4min. The PDMS membranes were then deposited on the surface of the gels and air dried for 2min. A small volume (8 μl) containing 40,000 cells was placed on the exposed region of the polyacrylamide gel defined by the PDMS membrane. Aftert 30 min, the unattached cells were washed off and 200 μl of medium were added. Four hours after seeding the cells (overnight for the gels coated with fibronectin), 2 ml of medium were added and the PDMS membranes were carefully removed with tweezers before the beginning of the experiment. Time-lapse recording started approximately 1h after removing the PDMS membrane. The interval between image acquisition was 10 min and a typical experiment lasted for 40h. A row of 9 images was acquired for each pattern with a ×40 objective and then stitched with a MATLAB script.

G1/S cell cycle arrest
To arrest cells at the beginning of S-phase, a cocktail of 2mM thymidine (T9250-1G, Sigma) and 5µg/ml aphidicolin (A0781-1MG, Sigma) was added to the media of MDCK Fucci cells 24h before the beginning of the experiment (in the cell culture flask). The patterning procedure was identical to that of control cells but cells were seeded with double the density to compensate for the lack of proliferation during 4h seeding. To quantify the number of nuclei we used a custom-made MATLAB software based on a sequential thresholding of the image to capture nuclei with distinct levels of intensity. Cell area was computed by dividing the monolayer area by the number of nuclei.

Time lapse imaging
Multidimensional acquisition routines were performed on an automated inverted microscope (Nikon Eclipse Ti) equipped with thermal control, CO2 and humidity control using MetaMorph/NIS Elements imaging software.

Spinning-Disk imaging
An inverted Nikon microscope with a spinning disk confocal unit (CSU-WD, Yokogawa) and Zyla sCMOS camera (Andor) was used for high-resolution image acquisition.

Traction microscopy
Traction forces were computed using Fourier transform based traction microscopy with a finite gel thickness 28 . Gel displacements between any experimental time point and a reference image obtained after monolayer trypsinization were computed using home-made particle imaging velocimetry software 28 .

Monolayer Stress microscopy
Monolayer stresses were computed using monolayer stress microscopy 29 . Monolayer stress microscopy uses traction forces and force balance demanded by Newton's laws to map the twodimensional stress tensor σ in the monolayer. By rotating these stress components at each point in the cell sheet, we computed the magnitude of the two principal stress components σmax and σmin and their corresponding, mutually perpendicular, principal orientations. For each point in the monolayer, we then computed the average normal stress within and between cells defined as = (σmax + σmin)/2. This is the value reported in the paper as tension. Boundary conditions during migration were those described previously 27 .

Monolayer height measurements
To measure monolayer height while monitoring the state of the cell in the cycle we used a MDCK cell line expressing Fucci and CIBN-CAAX-GFP. z-stacks were acquired every 30 min using a ×60 oil objective. To image the full width of the monolayer, we tiled 6 fields of view in a row at every time point. To compute monolayer height, we first divided the monolayer image in adjacent 14x14µm (xy) square regions. We then averaged intensity values in each region for each plane, thereby obtaining a profile of intensity in the z-direction. We computed the monolayer height as the width at half-maximum of the z profile (that is, half of the distance between the maximum intensity and the background level). To report the xz-profile shown in Supplementary Fig. 1f we averaged monolayer height in the y direction. The monolayer volume was calculated by integrating monolayer height along x and y.

Nucleus and cell shape tracking
To track individual cells within the monolayer a custom-made MATLAB software was used. First, the position of the nucleus of interest was acquired manually in the frame before division, defined as the last frame in which one single nucleus could be distinguished. Then, proceeding backwards in time until the beginning of the cell cycle, each new position of the nucleus was obtained by 1) binarizing the images with a threshold (in a small ROI centered at the nucleus), 2) labeling the binarized nucleus and, 3) overlapping the labeled image with a dilatation of the nucleus of interest in the previous analyzed timepoint. The nucleus overlapping with the dilatation was defined as the nucleus of interest and its position and mask were recorded. Finally, using the same method but this time proceeding forward in time, and starting again in the division frame, the tracks of the two daughter nuclei during the first hour after division were obtained. In this case, each new region was centered between the two nuclei. After obtaining the nuclei tracks, a custom-made MATLAB software was used to manually draw the shapes of the cells during the cell cycle.
To define the G1-S transition timepoint, the mean fluorescence intensity in the nucleus was computed. Red and green channels were normalized separately (by the mean of top 15% intensity values) and the transition timepoint was defined as the timepoint in which the two curves intersected. Using this procedure, 40 complete cycles (20 for the soft gels and fibronectin experiments) were analyzed per experiment. All analyzed cells started the cycle in the first 6h of experiment.
The nuclear and cell shape masks were used to compute the area and area growth rate of the nucleus and the cell.

Averaging cell properties
All properties at each timepoint were obtained from the nuclear and cell shape masks. The initial areas 0 and 0 were defined as the area 60 min after the beginning of the cycle. For the prediction of 2 initial areas were defined as the area at the G1-S transition.
We defined the area growth rate at each time point i as: Traction and tension at each time point were defined as the mean of these properties over the cell area. For fibronectin coating and soft gel experiments, cell properties were averaged on a circle of 24µm radius centered at the cell nucleus.

Linear model-selection analysis
To establish which properties are more predictive of the duration of the G1 and the S-G2-M phases, we carried out a systematic model comparison. In particular, we considered all models that are linear in one of the measured properties or in a product between two properties. We also explored all models that are multilinear on any pairwise combination of single properties and property products. To estimate the plausibility of each model the Bayesian information criterion was used (BIC) 34 . All models were compared to the most plausible one (the model with lowest BIC) using the Bayes factor, which is given by 33 : where p(D|A) is the probability of the observed data given model A. Therefore, the Bayes factor gives the ratio between model plausibility (when all models are considered a priori equally plausible). For example, if the Bayes factor of a model B with respect to the most plausible model A is 10, this means that A is 10 times more likely than B.

Tension and Traction ratios
To average forces in the three ROIs for each timepoint and cell, a custom-made MATLAB software was used. To average across the cell population, the time axis of each cell was linearly compressed or expanded so that the duration of G1 and S-G2-M was the average of all cells (G1 duration of 760min and S-G2-M duration of 530min). The ratios were then calculated for each cell and averaged afterwards.

Optogenetic experiments
The optogenetic system used here was described previously 35 . Briefly, the system is based on overexpressing a RhoA activator (DHPH domain of ARHGEF11) fused to a light-sensitive protein CRY2-mcherry. The resulting protein is called optoGEF-RhoA. Upon illumination, CRY2 changes conformation and binds its optogenetic partner CIBN.
To increase contractility, optoGEF-RhoA was forced to localize at the cell surface, where RhoA is located, by targeting CIBN-GFP to the plasma membrane. To decrease contractility, optoGEF-RhoA was forced to localize at the mitochondria, by targeting CIBN-GFP to the mitochondrial membrane. Optogenetic experiments were performed using a stable MDCK cell line 35 .
For experiments, the gels were air dried for 15 minutes after the media was removed. A drop of 10µl containing a mixture of 90% optogenetic cells and 10% Fucci cells was deposited on top of the gel. After 30 min, 2ml of media were added and the cells were left in the incubator overnight. The following day, dividing Fucci cells surrounded by optogenetic cells were imaged, both in activation and in control conditions. A laser of 488 nm was used to activate optogenetic cells. In the control case, time lapse images were acquired using only brightfield (green light filtered) and 561 nm laser. Cells were imaged for eight hours with time intervals of 4 min. Experiments were performed using a spinning disk microscope using a ×100 objective. Rounding time was defined as time from nuclear envelope breakdown to anaphase.

Single Cell experiments
Single cell imaging was acquired with ×40 objective with a timeframe of 6min.

Lifeact imaging experiments
LifeAct experiments were performed micropatterning the cells with the same protocol than the Fucci expansions but with a mixture of 80% LifeAct-GFP and 20% LifeAct-RFP cells instead. The interval between image acquisition was 5 min and a typical experiment lasted 15 h. Images were acquired with ×40 objective. Shapes of dividing cells were acquired manually with a custom-made MATLAB software. Radial traction and green fluorescence were then averaged over time by using these masks. High resolution LifeAct images were acquired with a Spinning Disk microscope and a ×60 objective.

Cadherin imaging experiments
Cadherin experiments were performed in glass, with Spinning Disk and ×60 objective.

Statistics and Reproducibility
Statistical comparisons were performed by using non-parametric Mann-Whitney's test or Bayesian analysis as indicated in each figure caption. All observations were reproducible and the number of experimental repeats is indicated in each figure caption.

Data availability
The data that support the findings of this study are available from the corresponding authors on reasonable request.

Code availability
Code used in this article can be made available upon request to the corresponding author.