Modulation of junction tension by tumor suppressors and proto-oncogenes regulates cell-cell contacts

Tumor suppressors and oncogenes play fundamental functions in cell proliferation, growth and apoptosis. The analyses of these functions have led to important advances in our understanding of tissue development and homeostasis as well as pathologies, including tumorigenesis (for reviews, see Zhao et al., 2011; Patel and Edgar, 2014; Baillon and Basler, 2014). In Drosophila, analysis of the sizes and shapes of somatic clones affecting tumor suppressor and proto-oncogene activities is instrumental to understanding their contribution in tissue morphogenesis, organization and homeostasis (Resino et al., 2002; Baena-Lopez et al., 2005; Mao et al., 2011; Wartlick et al., 2011; Kuchen et al., 2012; Worley et al., 2013; Restrepo et al., 2014; Heemskerk et al., 2014). Accordingly, somatic mutant clones are essential for unveiling how tumor suppressor and proto-oncogene activities modulate tissue proliferation, growth, cell-cell interactions and cell competition (for a review, see Wagstaff et al., 2013). In particular, the functions of tumor suppressors and proto-oncogenes in tissue organization and morphogenesis have often been recognized as their respective loss and gain of function leads to the formation of a rounded group of mutant cells (somatic clones) having a smooth boundary with the surrounding wild-type (wt) cells and thus reducing their contacts with neighboring wt tissue (Justice et al., 1995; Prober and Edgar, 2000,, 2002; Baena-Lopez et al., 2005; Mao et al., 2006; Worley et al., 2013). This property is shared by the Ras and Myc proto-oncogenes as well as components of the Fat/Dachsous (Ft/Ds) and Hippo pathways (Justice et al., 1995; Adler et al., 1998; Johnston et al., 1999; Prober and Edgar, 2000,, 2002; Garoia et al., 2000; Baena-Lopez et al., 2005; Mao et al., 2011; Worley et al., 2013).

Experimental and modeling approaches converged to show that the cell-cell contacts between two cell populations can be modulated by cell-cell adhesion, cell cortical contractility and cell division rate and orientation; in particular, the analysis of tissue or compartment boundary formation has provided important insights into the mechanisms modulating cell-cell contacts between two cell populations in response to cell signaling (for a review, see Fagotto et al., 2013). An increase in cell junction tension (which is larger when cortical contractility increases or adhesion decreases) at the interface between two tissues is known to favor a straight boundary between these two tissues or two compartments within a tissue and is essential for tissue development (Graner, 1993; Brodland, 2002; Käfer et al., 2007; Krieg et al., 2008; Hilgenfeldt et al., 2008; Landsberg et al., 2009; Monier et al., 2010; Aliee et al., 2012; Röper, 2012; Fagotto et al., 2013; Calzolari et al., 2014; Umetsu et al., 2014). Theoretical analysis shows that randomly oriented cell divisions induce diffusive random cell displacements (Ranft et al., 2010), thereby adding a slight amount of disorder between two cell populations (Block et al., 2007; Radszuweit et al., 2009). Experimental analyses of cell division rate and orientation demonstrated that an increase in cell junction tension is necessary to prevent disorder introduced by cell divisions occurring near the anterior-posterior parasegment boundary of the Drosophila embryo (Monier et al., 2010). Although the role of cell division rate at the Drosophila dorsal-ventral boundary of the wing imaginal disc remains a matter of debate, cell division orientation has been shown to contribute to its shaping (Aliee et al., 2012). Cell divisions oriented perpendicular to the boundary can increase the number of cell junctions at the boundary, and hence its raggedness, whereas cell divisions parallel to the boundary can decrease its raggedness. Notably, these effects of cell divisions can be reinforced or conversely weakened according to the number and orientation of cell rearrangements following the divisions.

We initially aimed to understand how Ft/Ds pathway, which modulates the activity of the myosin Dachs and of the Hippo/Yorkie pathway, promotes somatic clone rounding (Mao et al., 2006,, 2011; Cho et al., 2006). The tumor suppressors Ft and Ds encode large proto-cadherins that interact in a heterophilic manner and regulate proliferation via the Hippo/Yki pathway or mitochondrial metabolism, planar cell polarity and tissue morphogenesis by promoting oriented cell divisions and cell-cell rearrangements (Mahoney et al., 1991; Clark et al., 1995; Matakatsu and Blair, 2004; Baena-Lopez et al., 2005; Cho et al., 2006; Silva et al., 2006; Willecke et al., 2006; Harvey and Tapon, 2007; Mao et al., 2011; Donoughe and DiNardo, 2011; Bosveld et al., 2012; Brittle et al., 2012; Marcinkevicius and Zallen, 2013; Lawrence and Casal, 2013; Matis and Axelrod, 2013; Degoutin et al., 2013; Sing et al., 2014). Within tissues, opposing gradients of Ds and the Golgi resident kinase Four-jointed, which regulates Ft-Ds interaction, result in the planar polarization of Ft and Ds (Ishikawa et al., 2008; Brittle et al., 2010,, 2012; Simon et al., 2010; Bosveld et al., 2012; Ambegaonkar et al., 2012). One of the effectors of Ft and Ds planar polarization is Dachs, which regulates cell division orientation, cell rearrangements, cell affinity, proliferation rate and junction tension (Mao et al., 2006,, 2011; Bosveld et al., 2012; Brittle et al., 2012). Dachs membrane localization is promoted by the DHHC palmitoyltransferase Approximated (Matakatsu and Blair, 2008), and is inhibited by the interaction of the F-Box ubiquitin ligase FbxL7 with Ft (Rodrigues-Campos and Thompson, 2014; Bosch et al., 2014). Conversely, the Ds intracellular domain can form a complex with Dachs in vitro and Dachs is proposed to be planar polarized in response to Ds polarization on the same cell junction (Bosveld et al., 2012; Brittle et al., 2012). However, it remains to be shown whether the polarization of Ds intracellular domain would be sufficient to polarize Dachs in vivo. The loss of Ft or Ds function in mutant clones leads to the formation of rounded clones minimizing their contact with neighboring tissues (Adler et al., 1998; Garoia et al., 2000; Mao et al., 2006). Loss of Ft activity induces apical cell constriction and results in the membrane accumulation of Dachs, and ft clone rounding depends on Dachs (Mao et al., 2006,, 2011). Whether these changes in Dachs distribution are necessary for ft clone rounding remains to be determined.

Here, we investigated the role of loss of tumor suppressor and gain of proto-oncogene function in forming smooth clone boundaries in the Drosophila dorsal thorax (notum) epithelium where we could implement a time-lapse approach to follow the dynamics of clones. Initially focusing on the Ft/Ds pathway, we uncovered that ft clone rounding originates from two apparently opposed mechanical activities at the clone boundary and in the clone bulk. Extending the analysis to additional tumor suppressors and proto-oncogenes shows how distinct modulations of cell junction tension by tumor suppressors and proto-oncogenes could account for their common role in the regulation of cell-cell contacts.

To understand the mechanisms of clone rounding we implemented a time-lapse approach to follow the dynamics of ft^RNAi clones within the notum epithelium during pupal development (Fig. 1A). Using a combination of the flip-out and temperature-sensitive Gal4/Gal80^ts systems (Basler and Struhl, 1994; McGuire et al., 2001), the clones marked by the expression of membrane-CherryFP (UAS-PH:ChFP) were generated in second instar larvae that were then kept at 18°C to avoid the expression of the UAS-ft^RNAi. Following a temperature shift to 29°C at the end of larval development, we compared by time-lapse microscopy the proliferation and the circularity changes of a control group of cells and ft^RNAi clones (Fig. 1B-D; Movie 1). As a reporter of Ft activity we imaged Dachs:GFP (D:GFP), membrane levels of which increase upon loss of Ft function (Bosveld et al., 2012). At 10 hours after pupa formation (hAPF), the control and ft^RNAi clones had similar circularities and distributions of D:GFP. At 40 hAPF, whereas the control clones exhibited a reduced circularity and a D:GFP distribution similar to the surrounding cells, the ft^RNAi clones had a much higher circularity and an increased D:GFP level at cell junctions (Fig. 1B-D). This shows that our live-imaging approach is suitable for understanding how Ft/Ds signaling regulates tissue organization. To investigate whether clone rounding depends on cell division, we compared the respective timing of cell division and clone shape changes as well as the rates of proliferation in control and ft^RNAi clones (Fig. 1C). In both control and ft^RNAi clones, a wave of cell divisions occurred between 15 and 20 hAPF. During this wave neither the control nor the ft^RNAi clones rounded up. In fact, ft^RNAi clone rounding was observed after cell divisions had ceased and once D:GFP had started to accumulate (Fig. 1B). Furthermore, the proliferation rates were similar in control and ft^RNAi clones, showing that clone rounding can take place in the absence of over-proliferation (Fig. 1C). Together, these data show that rounding of ft^RNAi clones, and thus the reduction of their contacts with neighboring wt cells, is not concomitant with cell divisions and does not require a change in proliferation rate.

Having found that D:GFP accumulates in ft^RNAi clones during rounding and knowing that Dachs is necessary for ft clone rounding (Mao et al., 2006,, 2011), we performed a detailed analysis of the distribution and the mechanisms of Dachs localization in ft clones. Ft signaling is known to prevent Dachs membrane accumulation (Mao et al., 2006,, 2011; Bosveld et al., 2012; Brittle et al., 2012). Accordingly, D:GFP levels were increased in ft cells and sometimes segregated in distinct regions of the cell junctions (Fig. 2A; Figs S1, S2). We also noticed that D:GFP was strongly enriched all around the circumference of ft clones, i.e. at the interface between wt and ft cells, but reduced at the transversal junctions (Fig. 1B; Fig. 2A, yellow arrowheads). Quantitative analyses of the D:GFP distribution at clone boundary and transversal junctions revealed that D:GFP within mutant cells neighboring wt cells specifically accumulated at the clone boundary (hereafter referred to as ‘Dachs polarization at clone boundary’) (Fig. 2J). We therefore investigated the mechanisms leading to the polarization of D:GFP at the ft clone boundary. In agreement with the facts that the Ds intracellular domain forms a complex with Dachs in vitro, that Ds polarization promotes Dachs polarization and that Ds:GFP is polarized at the boundary of ft clones (Fig. 2B, yellow arrowheads; Strutt and Strutt, 2002; Ma et al., 2003; Matakatsu and Blair, 2004; Bosveld et al., 2012), D:GFP polarization was reduced at the boundary between wt and ft ds double mutant cells (Fig. 2C, white arrowheads; Fig. 2J). Whereas the overexpression of a full-length Ds transgene in ft ds clones (ds ft ds^UP) restores D:GFP polarization at the clone boundary (Fig. 2D, yellow arrowheads; Fig. 2J), overexpression of the Ds intracellular domain (Ds:intra) did not (Fig. 2E, white arrowheads; Fig. 2J), suggesting that the extracellular heterophilic Ds-Ft interaction is required to polarize Dachs. Accordingly, D:GFP polarization was reduced at the clone interface of ds ft ds^UP clones, which abutted a ft hypomorphic mutant tissue (compare Fig. 2D and 2F; Fig. 2J). Because loss of Ft activity inhibits Dachs membrane levels, D:GFP is also elevated in ft cells surrounding ds ft ds^UP clones. Together, our results support the hypothesis that Dachs is not only enriched in ft mutant clones, but is also polarized at the clone boundary due to Ds polarization.

Two independent and complementary results strengthen the proposal that the polarization of the Ds intracellular domain is sufficient to polarize Dachs at the ft clone boundary. First, in vitro analyses of D:GFP polarization using an S2-induced polarity assay (Johnston et al., 2009; Ségalen et al., 2010), in which the extracellular and transmembrane domains of the homophilic adhesion molecule Echinoid are fused to the intracellular domain of Ds (Ed:mCh:Ds:intra), showed that the Ed:mCh:Ds:intra polarity domain was sufficient for D:GFP recruitment and polarization, which colocalized with Ed:mCh:Ds:intra (Fig. 2K). Second, within the tissue, we found that the circumference of clones that overexpressed a chimera of the Ds intracellular and Ft extracellular domains (Casal et al., 2006) in ds ft clones (ds ft ds:intra:ft:extra^UP) can harbor a D:GFP polarization at the clone interface (Fig. 2G, yellow arrowheads; Fig. 2J). Such polarization was lost in ds:intra:ft:extra^UP cells surrounded by ds cells (ds ft ds:intra:ft:extra^UP in ds; compare Fig. 2G and 2H; Fig. 2J), indicating that the polarization of the Ds intracellular domain appears to be sufficient to polarize Dachs independently of the orientation of the Ft-Ds extracellular heterophilic interaction. Altogether these results indicate that loss of Ft activity induces two distinct Dachs distributions within ft clones: an increase of Dachs levels within ft cells and a Dachs accumulation at clone boundary junctions in response to Ds intracellular domain polarization.

The identification of two distinct changes in Dachs distribution in ft clones prompted us to investigate their respective contribution to clone rounding. We previously found that D:GFP polarization at cell junctions correlates with a higher junction tension (Bosveld et al., 2012). To investigate further the role of Dachs polarization in junction tension regulation, we first confirmed that Dachs polarization promotes an increase in junction tension, i.e. the force exerted by the junction parallel to it, which is considered positive when it reduces junction length and which can be estimated by the initial recoil velocity of the junction vertices upon its severing by laser ablation (Hutson et al., 2003). As Ds planar polarization is independent of Dachs function (Fig. 3A-C; Bosveld et al., 2012; Brittle et al., 2012), we compared the relaxation velocity of junctions showing high or low Ds:GFP signal in a dachs tissue, the junctions being labeled by Baz:mChFP. The loss of Dachs function abolishes the difference in relaxation velocity between junctions bearing high and low Ds:GFP (Fig. 3D). Because Ds and Dachs are planar polarized at the same junctions, this indicates that Dachs activity is necessary to increase the tension of junctions harboring Ds and Dachs polarization.

Having confirmed that Dachs polarization is necessary to increase cell junction tension in wt tissue, we tested whether the Dachs polarization at the boundary of ft clones and the increase in cell junction tension (Fig. 4A,B; Bosveld et al., 2012) are sufficient to fully explain the higher circularity of ft clones relative to wt clones. The increased tension at ft clone boundaries was independent of the clone size and the orientation of the ablated junctions (Fig. S3). The loss of Ds activity in ft mutant clones is associated with a decreased D:GFP polarization at clone boundaries (Fig. 2C,J) and an average junction tension similar to that of wt (Fig. 4C,C′). Although the circularity of ds ft clones was lower than the circularity of ft clones, the circularity of ds ft clones remained higher than that of wt clones (Fig. 4A). As Dachs activity is essential to increase the ft clone boundary tension and the circularity of ft clones (Fig. 4A,B; Mao et al., 2006), this indicates that Dachs polarization at the clone boundary increases the circularity of ft clones and that an additional Dachs-dependent mechanism inside the clone contributes to ft clone rounding (Fig. 1; Fig. 4A).

To identify this additional mechanism, we probed the tension of junctions inside ft clones. Unexpectedly, we observed that inside ft clones the average cell junction tension was reduced (Fig. 4B). A lower tension inside the clone was also measured in ds ft and in ds clones, showing that the tension decrease is independent of Ds function (Fig. 4C,D). Cell junction tension inside ft dachs, ds dachs and ds ft dachs clones as well as their circularities were similar to the wt, indicating that Dachs activity is necessary inside ft clones to reduce junction tension (Fig. 4B-D). The changes in junction tension were not associated with changes in the levels of the Drosophila Myosin II light chain (Spaghetti Squash, Sqh) at the boundary of or inside ft or ds clones (Fig. 4E,F; Fig. S4).

A common feature of ft, ds and double ds ft clones is to exhibit a higher level of Dachs (Mao et al., 2006; Brittle et al., 2012; Fig. 2A,I; Fig. S1A,H), which is reported to inhibit Wts (Cho et al., 2006) via Zyxin-dependent or -independent mechanisms (Rauskolb et al., 2011; Gaspar et al., 2015). We uncovered that loss of Wts activity (wts^RNAi) leads to clone rounding and to a decrease of junction tension similar to that measured inside ft, ds or ds ft clones (Fig. 4G,H). Furthermore, overexpression of Wts in ft clones (ft wts^UP) partially rescued the rounding of ft clones by reducing the polarization of Dachs and the tension at clone boundary junctions (Fig. 4G-J). Finally, the lower tension associated with Wts loss of function might be mediated by the activation of Yki as clones overexpressing Yki (yki^UP) are characterized by a junction tension similar to that of wts^RNAi (Fig. 4G). yki^UP clones display a rounder shape than wts^RNAi clones (Fig. 4H), suggesting that overexpression of Yki promotes clone rounding by both tension-dependent and tension-independent mechanisms. Together, these experimental observations indicate that Dachs has two apparently opposed mechanical activities: the polarized Dachs distribution at the boundary of the clone increases junction tension and the Dachs accumulation within the clone bulk decreases junction tension.

Using numerical simulations, we challenged more quantitatively this modeling by simulating the growth and proliferation of mutant clones using the cellular Potts model directly based on the ratios of experimental tension (Fig. 5C,D; Fig. S5; Movie 3). They showed that: (1) the experimental tension values in wt, ft, ds and ds ft both at clone boundary and inside the clone agree with the rounding of ft clones and the lower circularity of ds and double ds ft clones (Fig. 5D); (2) the combination of an increase of junction tension at the clone boundary and a decrease of junction tension inside the clone leads to a higher circularity than an increase of junction tension solely at the clone boundary (compare ft, simulated using tensions γ_b and γ_c with ft_b, simulated using γ_b and γ_c=γ; Fig. S5B); and (3) the dimensionless clone tension is relevant, and the observed clone circularity increases with σ similarly in experiments and in simulations (Fig. 5D; Fig. S5B). Our experimental observations and modeling of the ft clones uncovers how two distinct activities of the myosin Dachs associated with two distinct changes in its localization could induce clone rounding (reducing their contact with neighboring wt cells) by both increasing junction tension at the clone boundary and reducing it inside the clone.

Having found that the modulation of junction mechanical properties by the Ft/Ds and Hippo/Yki pathways promote clone rounding, we investigated whether the overexpression of proto-oncogenes known to be associated with clone rounding also modulates junction tension. As previously reported (Prober and Edgar, 2000,, 2002), we observed that the overexpression of Myc (myc^UP) as well as the overexpression of a gain-of-function allele of Ras (rasV12^UP) promote clone rounding (Fig. 6A). Probing the junction tension at the boundary and inside myc^UP and rasV12^UP clones unveiled distinct regulations of bulk or clone boundary junction tension as observed in the case of Ft/Ds or Hippo/Yki pathways (Fig. 6B). Whereas junction tension was reduced inside myc^UP clones and unchanged at the boundary, the junction tension was unchanged inside rasV12^UP clones but greatly increased at the boundary. The measured changes in junction tension are compatible with the rounding of somatic clones shown by our computer simulations (Fig. S5B). As the overexpression of Myc and RasV12 are both known to be associated with over-proliferation (Prober and Edgar, 2000,, 2002), their distinct impacts on junction tension strongly suggest that over-proliferation cannot be the only cause of their clone rounding. Accordingly, and as observed for the development of Gal4/Gal80^ts ft^RNAi clones, the proliferation rate in Gal4/Gal80^ts rasV12^UP clones was similar to that of control clones, and rasV12^UP clone rounding was observed after the main cell division wave (Fig. 6C-E). We therefore investigated whether the observed changes in junction tensions and in clone circularities were associated with altered myosin distributions.

In myc^UP clones, D:GFP was not polarized at the interface of the clone. Furthermore, a heterogeneous D:GFP enrichment was observed at only a few rasV12^UP clone boundaries, and this enrichment was weak (Fig. 6I,J). Such a weak increase is unlikely to be responsible for the rasV12^UP clone circularity, which is similar to that of ft clones. Moreover, in all experimental conditions, the D:GFP levels were normal inside the clones. We therefore turned our attention to the distribution of MyoII (Sqh:GFP). rasV12^UP clones were associated with an increase in MyoII localization at the clone boundaries (Fig. 6F,H). Such an increased level of MyoII at the boundary of rasV12^UP clone is in full agreement with the observed higher tension also reported for the MyoII distribution along the anterior-posterior compartment boundaries or during cell intercalation (Rauzi et al., 2008; Fernandez-Gonzalez et al., 2009; Monier et al., 2010; Umetsu et al., 2014). In contrast to rasV12^UP, MyoII was globally reduced within myc^UP clones (Fig. 6F,G). To determine whether the reduction of MyoII levels can account for the rounding of myc^UP clones, we tested whether mimicking a reduction of MyoII levels using sqh^RNAi was sufficient to decrease junction tension and to induce a rounded clone shape. Indeed, reducing MyoII levels led to a reduced tension inside sqh^RNAi clones and promoted clone rounding (Fig. 6A,B).

The segregation of tissues or of tissue compartments plays fundamental roles in the regulation of growth, patterning and morphogenesis (for a review, see Fagotto, 2014). Numerous tumor suppressors and proto-oncogenes share the common feature of promoting cell segregation apparent by the rounding of somatic clones upon their loss or gain of functions, respectively (Adler et al., 1998; Johnston et al., 1999; Prober and Edgar, 2000,, 2002; Garoia et al., 2000; Baena-Lopez et al., 2005; Mao et al., 2006,, 2011; Worley et al., 2013). Here, we show that loss of Ft/Ds and Hippo tumor-suppressor pathways as well as gain of function of the proto-oncogenes Myc, Ras and Yki lead to changes in cell junction tension. Although the changes are distinct and depend on distinct molecular mechanisms, taking into account both clone boundary and bulk tensions is sufficient to account for their segregation from surrounding wt tissue. Although previous works demonstrated that cell cortical contractility or cell adhesion at the interface between two compartments, tissues or clones can promote their separation or contribute to tissue invagination (Graner, 1993; Le Borgne et al., 2002; Brodland, 2002; Wei et al., 2005; Laplante and Nilson, 2006; Landsberg et al., 2009; Monier et al., 2010; Chang et al., 2011; Aliee et al., 2012; Röper, 2012; Fagotto et al., 2013; Calzolari et al., 2014; Umetsu et al., 2014), our genetic and modeling findings show that both boundary and bulk cell junction tensions need to be considered to fully account for segregation. This point is illustrated by the analysis of the mechanisms of clone rounding in the case of the Ft/Ds pathway where loss of Ft activity leads to change in cell junction tension both at the clone boundary and within the clone bulk.

We provide evidence that the rounding of ft clones can result from two apparently opposed mechanical inputs that require Dachs myosin activity. On one hand, the polarization of Ds at clone boundaries promotes Dachs polarity and increases junction tension. On the other hand, loss of Ft or Ds activity in the bulk of the clone leads to elevated Dachs levels and a decrease in junction tension. Our genetic and modeling findings show how these two distinct mechanical inputs both contribute to ft clone rounding. Importantly, increased levels of Dachs lead to an increased junction tension only at cell junctions where both Ds and Dachs are polarized, providing the first example of polarization and changes in the levels of a myosin leading to opposite changes in junction tension. The mechanism by which Dachs increases junction tension upon Ds-dependent polarization within tissues might be distinct from the one proposed for MyoII, because in vitro recombinant Dachs does not bind F-Actin in an ATP-dependent manner, but rather modulates F-Actin organization by promoting the binding of Zyxin to F-Actin (Cao et al., 2014). The mechanisms by which Dachs reduces tension inside ft and ds clones are independent of Ft and Ds activity and correlate with an increase of Dachs level. Dachs is reported to inhibit Wts by both Zyxin-dependent and -independent mechanisms (Rauskolb et al., 2011; Gaspar et al., 2015) and loss of Wts activity is associated with a similar decrease of cell junction tension. Wts is localized at cell-cell junction and its activity correlates with its binding to distinct partners and with distinct apical-basal distributions (Rauskolb et al., 2014; Sun et al., 2015). Whether these changes in Wts binding partner at cell-cell junctions can account for the regulation of tension by Dachs remains to be analyzed. Finally, the Ft/Ds pathway controls the shape of mammalian tissue (Saburi et al., 2008), therefore it will be relevant to examine whether Ft and Ds regulate tissue shape in vertebrate systems through modulation of cell junction tension.

Our findings on the mechanisms of clone rounding provide information on cell competition and on a possible negative feedback between proliferation and tension regulation. In the context of cell competition, it has been recently reported that a clonal decrease of Myc levels, or a clonal increase of Myc levels in conjunction with an inhibition of apoptosis, promotes clone fragmentation and cell mixing; such cell-cell mixing increases cell-cell contacts and depends on reduced levels of F-actin, independently of changes in MyoII levels (Levayer et al., 2015). Our experimental results and previous findings show that Myc overexpression leads to clone rounding and a reduction of cell-cell contacts with neighboring wt cells (Prober and Edgar, 2002). Although we did not study Myc-mediated cell competition and our experimental setup was different from that of Levayer et al. (2015), it would be relevant to analyze whether Myc-mediated cell competition also depends on additional mechanisms associated with reduced cell-cell contact formation owing to a decrease of MyoII level. Lastly, our findings on Yki suggest the existence of a possible negative feedback between Yki activation and cell junction tension. Both experimental and modeling approaches converge to support the idea that mechanical tension is an important regulator of growth and proliferation (Shraiman, 2005; Hufnagel et al., 2007; Aegerter-Wilmsen et al., 2007,, 2012; Mammoto and Ingber, 2010; Dupont et al., 2011; Sansores-Garcia et al., 2011; Aragona et al., 2013; Schluck et al., 2013; Rauskolb et al., 2014; Benham-Pyle et al., 2015). In particular, an increase in mechanical tensile stress leads to the activation of the Hippo/Yki (YAP/TAZ) pathway (Dupont et al., 2011; Sansores-Garcia et al., 2011; Rauskolb et al., 2014; Codelia et al., 2014). We have provided evidence that ectopic activation of the Hippo/Yki pathway in the notum epithelial tissue decreases junction tension in the bulk of the clone. The decrease of junction tension upon activation of Yki might unveil the possible existence of negative feedback from Hippo and Yki signaling in response to tensile stress in Drosophila tissues. Although the exact molecular mechanisms remain to be better characterized, we propose that such negative feedback could be instrumental to prevent prolonged activation of the Hippo and Yki pathways in response to an increase of mechanical tissue tension associated with tissue development or with external stress. Finally, by demonstrating that other signaling pathways regulate clone bulk or boundary tensions, we foresee that junction tension regulation might support cross-talk between tumor suppressor/proto-oncogene pathways and the Hippo/Yki pathways to define the size and shape of tissues and organs.

Drosophila stocks used were: ds⁰⁵¹⁴², ft^G-rv, ft⁸, ft¹, d^GC13, d²¹⁰, UAS-ft^RNAi, UAS-sqh^RNAi, UAS-wts^RNAi, UAS-yki, UAS-myc, UAS-Ras^V12 (see Table S1 for full genotypes and references). Clones were generated using FLP/FRT, flip-out or mosaic analysis with a repressible cell marker (MARCM) techniques (Xu and Rubin, 1993; Basler and Struhl, 1994; Lee and Luo, 1999). Clones were induced in second instar larvae (20 min heat-shock at 37°C) and analyzed 3-4 days later in 18-22 hAPF pupa. Live-imaging was carried out on ft^RNAi (hs-flp; ft^[G-rv], tub-GAL80^[ts]/AyGAL4; D:GFP, Baz:mChFP/UAS-ft^{[RNAi, JF03245]}, UAS-PH:ChFP), rasV12 (hs-flp; UAS-ras^[V12]/E-Cad:GFP, tub-GAL80^[ts]; tub-FRT-GAL80-FRT-GAL4, UAS-mRFP); clone induction and temperature changes are listed in the legend of Fig. 1 with further details in supplementary materials and methods. Their behaviors were compared with control clones: groups of 25-50 cells, with a circularity similar to that of ft^RNAi or rasV12^UP clones at the onset of pupariation (10 hAPF), were tracked in a region of the epithelial tissue devoid of ft^RNAi clones (n=4) or in E-Cad:GFP pupae filmed at 29°C (n=4). Clones were analyzed in the anterior notum where the Ds and Fj gradients are absent and where cells undergo only one division (Bosveld et al., 2012).

Pupae were mounted and imaged at 25°C or 29°C using an inverted confocal spinning disk microscope (Nikon) (David et al., 2005). Time-lapse images were acquired every 10 min (0.5 µm/slice, 20-30 slices/z-stack). Images of live or fixed pupae were captured in 18-22 hAPF pupae.

Laser ablations were performed in 18-22 hAPF pupae using a Ti:Sapphire laser (Mai Tai DeepSee, Spectra Physics) and images were acquired using a two-photon laser-scanning microscope (LSM710 NLO, Carl Zeiss) in single-photon mode with bidirectional scan lasting 156 ms in a region of 18 µm x 18 µm (100×100 pixels) as previously described (Bosveld et al., 2012). Ablations were performed in tissues expressing D:GFP and Baz:mChFP (wt, ft, ds, ds ft), Ds:GFP and Baz:mChFP (wt, dachs), Baz:mChFP (ft dachs, dachs) or E-Cad:GFP (ft, dachs, ds dachs, ds ft dachs, myc^UP, yki^UP, rasV12^UP, sqh^RNAi, wts^RNAi, wts^UP, ft wts^UP). To determine the initial relaxation velocity, the vertex-vertex distance of the pre-cut and post-cut (two frames after ablation) junction was measured manually using ImageJ. For each given ablation, the relative uncertainty on the velocity measurement was of order of 5%, much lower than the biological variability between different ablation experiments.

For immunohistochemistry, pupae were dissected, fixed and stained as described by Ségalen et al. (2010). Primary antibodies used were rabbit anti-GFP (1:2000; A-11122, Molecular Probes) and mouse anti-FasIII (1:50; DSHB 7G10). Secondary antibodies were Cy3-conjugated donkey-anti-mouse (Jackson ImmunoResearch) and Alexa-488-conjugated goat-anti-rabbit (Molecular Probes). Images were collected with a confocal microscope (LSM710 NLO, Carl Zeiss).

In the literature, the terms ‘circularity’, ‘compactness’, ‘segregation’ or ‘separation’ are used to express a size reduction of the contact with neighboring cells and can be quantified by measuring the clone area to square perimeter ratio (C=4π×area/perimeter²). C=1 for a circle. Circularity quantifications were performed on fixed and live tissues using membrane markers (mRFP, PH:ChFP, FasIII, Baz:mChFP, E-Cad:GFP) to delineate clone outlines. Clones were identified using loss (ubi-nlsGFP, ubi-H2B:RFP) or gain (UAS-GFP, UAS-mRFP, UAS-PH:ChFP) of expression. Clone outlines were manually drawn and measured using ImageJ.

D:GFP polarity was quantified by measuring on average projected images the D:GFP intensity at clone boundary junctions and transversal junctions within clones. The intensity was quantified manually (∼10-40 transversal, ∼20-60 boundary junctions per clone) and corrected by subtraction of the cytoplasmic signal. For each clone, the ratio between the average intensity at boundary junctions (<I_boundary>) and average intensity at transversal junctions (<I_transversal>) was determined. D:GFP is polarized when the ratio is >1. Junctional Sqh:ChFP or Sqh:GFP levels were quantified by manually measuring on average projected images the intensity on junctions inside clones, at clone boundaries as well as in wt junctions surrounding the clone. In each clone, ∼15-50 boundary, bulk and wt junctions were measured. The intensities were corrected by subtraction of the cytoplasmic signals. Boundary and bulk junctional intensities were normalized to the wt junctional intensity to determine the average boundary and bulk intensities per clone.

The Ubi-Dachs:mRFP transgene was generated by cloning a full length dachs cDNA (D isoform) into pUWR (DGRC) by Gateway cloning. Transgenesis was performed by Bestgene Inc. For the S2 cell polarity assay, the Ds intracellular domain was fused to the extracellular domain of echinoid fused to mCherry (Ed:mCh:Ds:intra). The Ds intracellular fragment used in the S2 cell polarity assay was cloned into pMT:Ed:mCherry with a five-amino acid linker (GGGGS) between the mCherry and the Ds intracellular domain. Dachs was cloned into pUWG (DRGC). Induced cell polarity assays were carried out as described previously (Johnston et al., 2009; Ségalen et al., 2010). The polarization of the GFP-tagged Dachs construct relative to the induced Ed:mCh:Ds:intra polarity interface was quantified from single confocal scans as follows. The mean GFP intensities all around the cell cortex I_cell-cortex and at the Ed:mCh:Ds:intra interface I_{polarity-domain} were quantified blindly using ImageJ. The D:GFP was scored as polarized when the I_{polarity-domain}/I_cell-cortex ratio was >1.1.

See supplementary materials and methods for theoretical analysis and numerical simulations.

All error bars represent the standard deviation (s.d.). Statistical significance between experimental conditions was assessed using Student's t-tests: distribution normalities were checked using Kolmogorov–Smirnov test. In cases where the variances were different (F-test), significance was assessed using the unequal variance t-test. The one-way ANOVA Tukey's test was used to assess significance between genotypes within the same dataset (circularity, D:GFP polarity). Significance between either clone boundary or clone bulk tensions was assessed using one-way ANOVA Dunnett's test. By reporting significances based on one-way ANOVA, statistical differences between two different experimental conditions might have been missed.

The code used for simulations based on the cellular Potts model is provided as a zip file in supplementary information along with the procedure to install and run it on MacOS.

We thank J. Axelrod, A. J. Bardin, S. Blair, J. Casal, Y. Hong, K. Irvine, E. Martin-Blanco, D. Pan, M. Simon, D. Strutt, N. Tapon, S. Tsukita, the Bloomington Drosophila Stock Center, Transgenic RNAi Project at Harvard Medical School, Vienna Drosophila RNAi Center, Developmental Studies Hybridoma Bank for reagents; PICT-IBiSA of the UMR3215 and U934 for assistance with microscopy; F. Elias, M. Durand, V. Leroy, C. Derec for discussions; and D. Lubensky, D. Pinheiro for comments.