The α-Catenin mechanosensing M region is required for cell adhesion during tissue morphogenesis

During morphogenesis, cell contacts are subjected to contractile forces to elicit coordinated cell shape changes and cell rearrangements. Maintaining cohesion and tissue integrity during cell contact changes is believed to require the response of adherens junction (AJ) components to force. AJs are both stable enough to resist actomyosin-generated forces, yet dynamic enough to be readily disassembled and rebuilt (Lecuit and Yap, 2015; Ladoux et al., 2015; Pinheiro and Bellaïche, 2018; Charras and Yap, 2018; Clarke and Martin, 2021). The cadherin–catenin complex (CCC) at AJs physically couples the cytoskeleton of neighboring cells and in response to force—generated often by actomyosin contraction—strengthens adhesive interactions (le Duc et al., 2010; Yonemura et al., 2010). α-catenin is thought to be the central mechanosensor of the CCC (Yonemura et al., 2010; Ishiyama and Ikura, 2012; Leckband and De Rooij, 2014; Ladoux et al., 2015; Angulo-Urarte et al., 2020). Although mutational analysis of α-catenin has demonstrated its essential role for cell adhesion similar to E-cadherin (Ecad) and β-catenin (Armadillo [Arm] in Drosophila) in several animal species (Torres et al., 1997; Kofron et al., 1997; Costa et al., 1998; Schepis et al., 2012; Sarpal et al., 2012; Nathaniel Clarke et al., 2019) the function of α-catenin mechanosensing in tissue morphogenesis remains largely unexplored.

Transmission of cytoskeletal forces to cadherins, and the integrity of AJs, requires the physical link between cadherins and actin provided by α-catenin (Rimm et al., 1995; Desai et al., 2013; Buckley et al., 2014). Force-induced conformational changes in cadherin extra-cellular domains increases the strength of trans-interactions between cadherins of neighboring cells (Leckband and De Rooij, 2014; Pinheiro and Bellaïche, 2018). Force-induced conformational changes also occur in two mechanosensory domains of α-catenin: the central M region and the C-terminal actin-binding domain (ABD; Fig. 1, A and B). α-catenin directly binds F-actin as a catch-bond, where binding strength increases with increasing force up to a threshold. Under tension, ABD conformational changes enhance direct F-actin binding (Buckley et al., 2014; Ishiyama et al., 2018; Xu et al., 2020; Wang et al., 2022). Moreover, reversible force-induced conformational changes in the M region cause recruitment of other F-actin binding proteins, believed to reinforce adhesion (Yonemura et al., 2010; Choi et al., 2012; Rangarajan and Izard, 2012; Twiss et al., 2012; Huveneers et al., 2012; Barry et al., 2014; Mège and Ishiyama, 2017; Ishiyama et al., 2018; Sarpal et al., 2019; Alégot et al., 2019).

Without tension, a dynamic electrostatic network of salt-bridges holds the α-helical bundles of M1, M2, and M3 in a closed conformation (Ishiyama et al., 2013; Li et al., 2015; Fig. 1, A and B). At forces of ∼5 pN, M1 unfurls and binds Vinculin (Vinc), recruiting Vinc to the cell membrane (Yonemura et al., 2010; Ishiyama et al., 2013; Barry et al., 2014; Yao et al., 2014; Kim et al., 2015; Maki et al., 2016; Seddiki et al., 2018). Junctional recruitment of Ajuba (Jub) is also regulated by mechanosensing through the M region, with M1 limiting Jub recruitment to α-catenin (Alégot et al., 2019; Sarpal et al., 2019). At higher forces of ∼10–15 pN, the M region unfolds further, partially or completely unfurling M2 and M3 (Yao et al., 2014), and the angle between M2 and M3 may increase (Ishiyama et al., 2013; Li et al., 2015). This is predicted to expose an Afadin interaction site (Pokutta et al., 2002; Ishiyama et al., 2013; Li et al., 2015; Matsuzawa et al., 2018; Sakakibara et al., 2020). A constitutively open M region α-catenin mutant increases Afadin and Vinc recruitment in cell culture (Matsuzawa et al., 2018). The α-catenin binding site on Afadin is also required for its mechanosensitive enrichment (Sakakibara et al., 2020). Hence, M region conformation states are thought to affect interaction between α-catenin and Afadin, Vinc, and Jub.

Here, we assessed the function of the M region of Drosophila α-Catenin (α-Cat) and its three domains (M1, M2, and M3) in embryonic morphogenesis. Contrary to expectation, α-Cat constructs lacking the entire M region can fully replace endogenous α-Cat in the wing disc epithelium (Sarpal et al., 2019). Deletion of individual M region domains also supports epithelial integrity of the wing disc and ovarian follicle (Desai et al., 2013; Sarpal et al., 2019). Partial M region deletion in zebrafish caused cell migration, but not adhesion defects, defects that were not phenocopied by Vinc mutants (Han et al., 2016; Han et al. 2017). Although several tissue culture studies argue that M1-dependent Vinc recruitment is essential for cadherin-based cell adhesion (Yonemura et al., 2010; le Duc et al., 2010; Huveneers et al., 2012; Twiss et al., 2012; Seddiki et al., 2018), a significant role for Vinc in AJ function is not supported by analysis of Vinc mutants in Caenorhabditis elegans (Barstead and Waterston, 1991), Drosophila (Alatortsev et al., 1997; Maartens et al., 2016), zebrafish (Han et al., 2017), or mouse (Xu et al., 1998). Similarly, whereas the role of M1 in limiting Jub recruitment regulates normal tissue growth (Rauskolb et al., 2014; Alégot et al., 2019; Sarpal et al., 2019), it is unclear whether this interaction supports adhesion. Jub null mutants in the mouse, zebrafish, and Drosophila complete embryogenesis with only subtle defects (Pratt et al., 2005; Witzel et al., 2012; Razzell et al., 2018) consistent with a minor role for Jub in cell adhesion, at best.

The only α-catenin M region interaction partner with substantive adhesion defects, though weaker than those resulting from α-catenin depletion, is Afadin (Canoe [Cno] in Drosophila). Loss of Cno causes mesoderm invagination defects (Sawyer et al., 2009) and tears in the ectodermal epithelium during convergent extension (Sawyer et al., 2009; Yu and Zallen, 2020). cno zygotic mutants have dorsal closure defects (Jürgens et al., 1984; Takahashi et al., 1998; Boettner et al., 2003; Choi et al., 2011) similar to some α-Cat zygotic mutants (Sarpal et al., 2012; Jurado et al., 2016). The comparatively mild-to-moderate defects produced by loss of Vinc, Jub, and Cno compared to the striking loss of epithelial integrity caused by loss of α-Cat led us to wonder (i) whether this is explained by redundancy between α-Cat interaction partners and (ii) whether, and if so how, the α-Cat M region contributes to adhesion and morphogenesis through recruitment of these interaction partners in response to force.

Vinc, Jub, and Cno show tension-sensitive distribution in the early Drosophila embryo, raising the possibility that their junctional recruitment is dependent on α-Cat mechanosensing. During germband extension, adhesion is challenged by rapid junction remodeling, and cell edges aligned with the dorsal–ventral (DV) axis (vertical edges) experience higher tension (Clarke and Martin, 2021; Fernandez-Gonzalez et al., 2009). Myosin becomes planar polarized, enriched at vertical edges, which contract during neighbor-exchange (Zallen and Wieschaus, 2004; Rauzi et al., 2010). Vinc, Jub, and Cno are also enriched at these high-tension, vertical edges, and at tricellular junctions (TCJs), another contact under high force. Cytoskeletal tension is required for normal Vinc, Jub, and Cno membrane localization, and laser ablation experiments revealed a direct correlation between Vinc enrichment and tension at cell edges (Kale et al., 2018; Razzell et al., 2018; Sawyer et al., 2009, 2011; Yu and Zallen, 2020). The natural variation in tension in the Drosophila germband thus provides a model for assessing M region mechanosensing.

Here, we report that the M region is required for cell adhesion during morphogenesis, in particular the M2 domain at contacts that experience higher tension. Our data suggest three distinct tension states read by α-Cat mechanosensing that cause the differential recruitment of Vinc, Jub, and Cno to enhance adhesion. Surprisingly, retention of the M1 domain was more deleterious than removal of the M region entirely. This effect is not due to the recruitment of Vinc, but likely to the M1-dependent regulation of Jub recruitment to AJs. Our findings also support the conclusion that Cno promotes cell adhesion in a parallel pathway to the CCC except at TCJs where M2 and M3 contribute to Cno enrichment. Our work provides evidence of a robust network of cooperative and redundant mechanosensitive interactions at AJs that support tissue morphogenesis.

We previously analyzed the ability of mutant α-Cat proteins to substitute for endogenous α-Cat in several different Drosophila tissues (Sarpal et al., 2012; Desai et al., 2013; Escobar et al., 2015; Ishiyama et al., 2018; Sarpal et al., 2019). Expression of an α-Cat construct that lacks the entire M region (α-Cat-ΔM) could fully support epithelial development of the wing disc epithelium (Sarpal et al., 2019). α-Cat-ΔM also provides strong rescue of α-Cat mutant cells of the follicular epithelium (Fig. S1), supports development of the head epidermis and prevents embryonic lethality of most zygotic α-Cat null mutant animals (Sarpal et al., 2019). In contrast to the widely considered model that M region–based mechanosensing enhances adhesion in response to mechanical force (Leckband and De Rooij, 2014; Mège and Ishiyama, 2017; Angulo-Urarte et al., 2020), our in vivo data at this point assigned only a minor or no essential role to the M region in cell adhesion and epithelial development. We therefore wondered whether M region function is most relevant during developmental periods of vigorous morphogenesis when cell contacts experience high levels of force. Drosophila gastrulation condenses several large-scale morphogenetic movements into a short ∼2-h time window. These movements entail the invagination of the mesoderm and endoderm (Martin, 2020), cell intercalations that drive germband extension (Paré and Zallen, 2020), the ingression of neural stem cells (Simões et al., 2017; An et al., 2017), and three rounds of cell division with a ∼40-min cell cycle time (Foe, 1989).

To assess M region function in early embryos, we removed the crucial maternal contribution of α-Cat (Sarpal et al., 2012) by RNAi (Fig. S2, A and B) and expressed shRNA-resistant constructs (Sarpal et al., 2019). Our constructs are recruited to AJs (Fig. S2 C) and are expressed at close to normal levels (Desai et al., 2013; Sarpal et al., 2019). Resistant constructs contain a mutated shRNA target site that preserves the amino acid sequence located in the M2 domain (denoted by “R”; e.g., α-CatR, with mutant α-Cat isoforms collectively referred to as α-CatX) or carry a deletion of M2 (Fig. 1 C). Maternally driven α-Cat-RNAi caused a dramatic phenotype like that reported for the loss of Arm (Cox et al., 1996) or Ecad (Tepass et al., 1996), leading to a highly fragmented epidermal/cuticle layer (Fig. 1 D). We quantified the cuticle defects, which are indicative of adhesion defects, by counting the number of intact abdominal denticle belts. α-Cat-RNAi embryos had 0 intact denticle belts. α-Cat-RNAi embryos expressing α-CatR had all eight denticle belts restored in >90% of animals. α-Cat-RNAi α-Cat-ΔM animals showed partial restoration of epidermal integrity compared to α-Cat-RNAi. Nearly all animals had denticle belt defects, with more than half of embryos producing no denticle belts (Fig. 1, C and D). These results demonstrate a substantive requirement of the M region in maintaining epithelial integrity in the Drosophila embryo.

To investigate the function of the M region in early morphogenesis, live embryos were observed during gastrulation with a focus on mesoderm invagination and ectodermal integrity during germband extension. Mesoderm invagination failed in α-Cat-RNAi embryos (Fig. 2 A, Fig. S2 A, and Video 1). As previously described (Martin et al., 2010), at the onset of mesoderm invagination actomyosin contraction overpowers residual adhesion in α-Cat-RNAi embryos. Cells detached with plasma membrane tethers formed between few remaining AJ puncta (Fig. S2 D). α-Cat-RNAi α-CatR embryos showed normal mesoderm invagination. In contrast, α-Cat-RNAi α-Cat-ΔM embryos displayed a range of defects. Approximately 40% of embryos showed normal mesoderm invagination where the ventral midline sealed along the entire anterior–posterior axis, whereas the remaining ∼60% of embryos showed either a partial closure or a complete failure to close the ventral midline (Fig. 2 A; and Videos 2, 3, and 4).

As the germband starts to extend, ectodermal cells in α-Cat-RNAi embryos lose adhesion and large gaps appear between the apical domain of cells (Fig. 2 B). Depletion of the maternal contribution of α-Cat eventually leads to cell dissociation and epithelial collapse (Cavey et al., 2008; Martin et al., 2010; Rauzi et al., 2010; Fernandez-Gonzalez and Zallen, 2011; Wang et al., 2013; Levayer and Lecuit, 2013; Eritano et al., 2020). Similar to the mesoderm, plasma membrane tethers are seen connecting cells across gaps (Fig. S2 E; Fernandez-Gonzalez and Zallen, 2011). α-Cat-RNAi α-CatR embryos displayed normal cell contacts, whereas α-Cat-RNAi α-Cat-ΔM embryos showed gaps in the epithelium with membrane tethers spanning the gaps (Fig. 2, B and C; and Videos 2, 3, and 4). Apical domains became separated, leading to gaps or tears that extend predominantly along the DV axis (Fig. 2 D). These defects were associated with a marked reduction in germband extension compared to α-Cat-RNAi α-CatR controls (Fig. 2 E). Thus, the M region makes an essential contribution to maintaining adhesion during mesoderm invagination and germband extension and its loss compromises these movements.

We next assessed the performance of α-Cat constructs that lacked individual M region domains (α-CatR-ΔM1, α-Cat-ΔM2, α-CatR-ΔM3) or both M2 and M3 (α-Cat-ΔM23). All constructs analyzed are robustly expressed in embryos, with no defects produced by overexpression in wild-type background (Desai et al., 2013; Sarpal et al., 2019). α-CatR-ΔM1 behaved like full-length α-CatR when we examined the cuticle (Fig. 1 C and Fig. S2 F), mesoderm invagination (stage 6, Fig. 2 A and Video 5), the ectodermal epithelium (Fig. 2, B–D, and Video 5), and germband extension (stages 7 and 8, Fig. 2 E), suggesting that the M1 domain makes no essential contribution to cell adhesion in these tissues. Surprisingly, expression of α-Cat-ΔM23 was much less capable of ameliorating the α-Cat-RNAi phenotype than α-Cat-ΔM. α-Cat-RNAi α-Cat-ΔM23 embryos showed little improvement of the cuticle defects seen with α-Cat-RNAi (Fig. 1 C and Fig. S2 F) and failed to rescue mesoderm invagination (Fig. 2 A and Video 6). Most embryos displayed prominent de-adhesion defects in the ectoderm, though these are not as severe as with α-Cat-RNAi alone (Fig. 2, B–D, and Video 6), and germband extension was substantially reduced (Fig. 2 E). Rescue by α-Cat-ΔM2 showed minor improvements when compared to α-Cat-ΔM23, whereas expression of α-CatR-ΔM3 showed a much better rescue than α-Cat-ΔM2 (Fig 1 C, Fig. 2, Fig. S2 F, and Videos 6, 7, and 8). This suggests that M2 is the most important domain within the M region for supporting cell adhesion, with M3 making a minor contribution. Notably, the poor rescue activity seen with α-Cat-ΔM23 compared to α-Cat-ΔM suggests that the M1 domain, which is predicted to remain unfurled in α-Cat-ΔM23, negatively regulates cell adhesion.

The ectoderm during germband extension displays an asymmetric distribution of non-muscle myosin II, which is enriched at vertical cell edges (oriented along the DV axis) and at TCJs. Laser ablation of junctions produces higher retraction velocities in myosin-rich vertical edges versus horizontal edges, and modeling predicts that TCJs show the highest tension (Bertet et al., 2004; Zallen and Wieschaus, 2004; Fernandez-Gonzalez et al., 2009; Tetley et al., 2016; Vanderleest et al., 2018; Higashi and Miller 2017; Trichas et al., 2012). For the purpose of our discussion and reflecting the distribution of myosin, we distinguish between low-tension horizontal bicellular junctions (BCJs), medium-tension vertical BCJs, and high-tension TCJs (vertices). Contraction of myosin at vertical edges either in multicellular rosettes or at bicellular contacts are important drivers of cell intercalation required for germband extension (Zallen and Wieschaus, 2004; Blankenship et al., 2006; Zallen and Blankenship, 2008; Fernandez-Gonzalez et al., 2009; Tetley et al., 2016; Vanderleest et al., 2018) and neuroblast ingression (Simões et al., 2017). We noticed that gaps or tears in the epithelium of embryos expressing M region deletions were commonly found at the centers of rosettes or at vertices, and were elongated along the DV axis (Fig. 2, B and D). These findings further suggest that the M region and its individual domains, except for M1, strengthen adhesion and that medium and high-tension cell contacts are more susceptible to loss of adhesion than other cell contacts.

To further characterize these adhesion defects, we analyzed the actomyosin cytoskeleton using the endogenously YFP-tagged myosin heavy chain. The expression of any M region deletion in an α-Cat-RNAi background led to increased total myosin signal (Fig. 3, A and E), while the enrichment of myosin to vertical edges was consistently observed (Fig. 3 A). As myosin is recruited by tension (Fernandez-Gonzalez et al., 2009) this suggests that the asymmetric distribution of tension in the ectoderm is not strongly affected in embryos expressing M region deletions. Although membrane association of myosin was preserved even with a poor rescue of adhesion as in α-Cat-RNAi α-Cat-ΔM23 embryos, separation of apical domains along the DV axis led to the apparent splitting of supracellular myosin cables (Fig. 3 B). In embryos with stronger defects, myosin accumulates in large puncta at the free apical cell edges of gaps after they had formed (Fig. 3 B).

Gaps in the ectoderm of α-Cat-RNAi α-CatX embryos could also be visualized by an enhanced signal of the active Rho1 probe Ani-RBD. Activation of Rho1 causes the phosphorylation of myosin regulatory light chain by Rho kinase, and hence myosin contraction, with enrichment of active Rho1 seen at sites of high actomyosin contractility in the germband (Munjal et al., 2015; Martino et al., 2018). Loss of apical cell contact was followed by the formation of large puncta of Ani-RBD (Fig. 3, C and D), which is enriched along the perimeter of epithelial gaps (Fig. 3 F). These observations are reminiscent of the accumulation of active Rho1 and myosin at wound margins (Abreu-Blanco et al., 2014; Rothenberg and Fernandez-Gonzalez, 2019), and Rho flares, which repair breaches in the epithelial barrier during Xenopus development (Stephenson et al., 2019). The appearance of membrane tethers (Fig. 2 B) and the accumulation of active Rho1 and myosin (Fig. 3) confirmed the loss of cell–cell contacts within the germband, particularly at medium- and high-tension cell edges or vertices, for embryos expressing deletions that remove the M2 domain (Fig. 2, B and C; and Fig. 3 C). Moreover, α-CatR-ΔM3 and α-Cat-ΔM23 expressing embryos showed comparatively lower levels than α-CatR in both cortical Ani-RBD signal and the proportion of junctional myosin (Fig. 3, E and F). In contrast, a higher fraction of junctional myosin was seen when M1 was deleted (Fig. 3 E). Together, these findings suggest that the M region contributes to the regulation of junctional actomyosin.

To further characterize the impact of the M region on cell adhesion, we examined the junctional distributions of Ecad, Vinc, and Jub using fluorescently tagged proteins under the control of their endogenous promoters (Huang et al., 2009; Sabino et al., 2011; Kale et al., 2018). In α-Cat-RNAi α-CatX embryos, the levels of Ecad correlated with the strength of adhesion defects observed (Fig. 4, A and C; for a summary of results, see Fig. 4 D), with lower levels of Ecad associated with more severe defects, as expected. Removal of M2 or M3 caused a significant reduction of cortical Ecad, suggesting that these domains predominantly contribute to the stability of the CCC at the junction (Fig. 4, A, C, and D). This concurs with a reduction of junctional Arm in wing disc cells lacking M2 or M3 (Sarpal et al., 2019). In contrast, the M1 domain has an inhibitory effect on Ecad. α-Cat-RNAi α-CatR-ΔM1 embryos showed elevated cortical enrichment of Ecad (Fig. 4, A, C, and D), consistent with an increase in junctional Arm in α-CatR-ΔM1 expressing wing disc epithelium (Sarpal et al., 2019; Alégot et al., 2019). An increase in Ecad levels as a result of the absence of M1 was also seen when comparing α-Cat-ΔM23 and α-Cat-ΔM embryos, and α-Cat-ΔM appears to be better maintained at the membrane and is less punctate than α-Cat-ΔM23 (Fig. 4 and Fig. S2 A). We conclude that the M region regulates junctional stability with M1 and M2/3 having opposing effects. Whereas M2/3 stabilizes the junction, M1 limits junctional Ecad.

Although direct binding between Vinc and Drosophila α-Cat is yet to be assessed in vitro, we found that M1 is required for all detectable junctional Vinc signal in gastrulating embryos (Fig. 4, A, C, and D). As removal of M2 and M3 also reduces Ecad, which is expected to reduce recruitment of AJ proteins, we wanted to know whether the reduction in Vinc in these conditions is explained by a reduction of the CCC. We therefore calculated the ratio of cortical Vinc to Ecad per cell (expressed within the same embryo). We found that the M2 and M3 domains contribute to Vinc junctional levels even when normalized to the amount of remaining Ecad (Fig. 4, A, C, and D). Removal of both M2 and M3 together significantly increases the Vinc/Ecad ratio compared to α-CatR control (Fig. 4, A, C, and D). Together, this suggests that both M2 and M3 individually inhibit M1, but also each other, such that removing M2 and M3 together relieves all inhibition of Vinc junctional recruitment (Fig. 4 E). Removing M2 alone leaves M3 free to inhibit the M1–Vinc interaction and vice versa. This agrees with work using cells in culture identifying a masking effect between domains (Yonemura et al., 2010; Matsuzawa et al., 2018; Sakakibara et al., 2020), and provides evidence within an animal model to support the existence of autoinhibition among M region domains.

Embryos where M1 is removed showed a dramatic increase in Jub at the junctions (Fig. 4, B–D). In contrast, Jub levels are reduced in α-Cat-RNAi embryos expressing α-Cat-ΔM2, α-CatR-ΔM3, or α-Cat-ΔM23, suggesting that the M2 and M3 domains support junctional Jub. Despite lacking M2 and M3, embryos expressing α-Cat-ΔM experience an enrichment of Jub (Fig. 4, B–D). Thus, as shown for the wing disc epithelium (Sarpal et al., 2019; Alégot et al., 2019), M1 also acts as an inhibitor of junctional Jub recruitment in the embryonic ectoderm. Retention of the M1 domain in α-Cat-RNAi α-Cat-ΔM23 embryos produces the largest reduction in Ecad, Jub (Fig. 4), and the strongest adhesion defects (Fig. 1 C, Fig. 2, and Fig. S2 F). As junctional Jub has a positive role in cell adhesion in the ectoderm (Razzell et al., 2018), we conclude that M1 is a negative regulator of cell adhesion, a role normally limited by the functions of M2 and M3, an inhibition resolved by the force-dependent unfurling of the M region.

We also found that manipulation of the M region significantly affected the mechanosensitive recruitment of Jub. Jub is recruited to the membrane in response to cytoskeletal tension (Rauskolb et al., 2019; Razzell et al., 2018; Alégot et al., 2019). In wing disc epithelium, removal of the M1 domain causes Jub to become hyper-recruited to junctions without an increase in tissue tension (Sarpal et al., 2019; Alégot et al., 2019). How this occurs is unclear as the Jub-binding site is not within the M region, but in the N2 domain (Marie et al., 2003; Alégot et al., 2019; Sarpal et al., 2019). A requirement for the α-Cat N-terminal domain for the junctional recruitment of Jub was confirmed in the embryo, as rescue with a construct that lacks this domain (DEcadΔβ::α-Cat-ABD) depleted Jub levels to the same degree as α-Cat knockdown (Fig. 5, A–C).

During germband extension, vertical edges are under higher tension than horizontal contacts (Fernandez-Gonzalez and Zallen, 2011). As Jub signal increases with increasing tension, this leads to the planar polarized enrichment of Jub at vertical edges (Razzell et al., 2018). Similarly, Vinc and Cno localization depend on cytoskeletal tension and are also planar polarized (Hara et al., 2016; Kale et al., 2018; Sawyer et al., 2011; Yu and Zallen, 2020). The mean fluorescent intensity (FI) of Vinc, Jub, and Cno increases in accordance with expected relative tension in the embryo, lowest at horizontal BCJs, higher at vertical BCJs and highest at TCJs (Fig. 5 D). This provided us with a system in which we could test Jub response to tension. We determined the planar polarity of Jub in α-Cat-RNAi α-CatX embryos by measuring the average FI of Jub along vertical edges (Fig. 5 C), divided by that of horizontal edges (Fig. 5 B) within the same embryo (Fig. 5 E). This analysis revealed that the M2 domain is in fact required for the specific enrichment of Jub to vertical cell contacts. Expression of α-Cat-ΔM, α-Cat-ΔM23, and α-Cat-ΔM2 failed to rescue the enrichment of Jub to edges approaching 90° (where the anterior–posterior axis = 0°), and in some instances a reversal of Jub planar polarity was seen with higher Jub levels at horizontal versus vertical edges (Fig. 5 E and Fig. S3 A). Furthermore, in all experimental conditions, Jub levels at TCJs versus BCJs were reduced as compared to control (Fig. 5 F). The M region is thus required for the tension-dependent distribution of Jub.

In α-Cat-RNAi α-Cat-ΔM1 embryos, Jub planar polarity and enrichment at TCJs is significantly reduced, but in this case is due to a constitutive enrichment of Jub to all cell contacts, regardless of the state of tension (Fig. 5, A–C and E; and Fig. S3 A). Similarly, in wing disc epithelium, loss of M1 desensitizes junctional Jub recruitment to a reduction of myosin activity (Alégot et al., 2019). The planar polarity of Jub thus requires M2 for the mechanosensitive recruitment to higher-tension edges, while M1 limits Jub at lower-tension edges. Only at the highest tension sites, at TCJs, was a role for M3 alone seen in supporting Jub enrichment (Fig. 5 F). Although Jub is required for normal germband extension, its planar polarity is not essential for adhesion or germband extension (Razzell et al., 2018). M2 and M3 may therefore function by supporting the necessary amount of Jub at higher-tension edges, and hence support Ecad membrane stability against enhanced actomyosin contractility. As junctional Jub is more enriched in α-Cat-ΔM than α-Cat-ΔM23 embryos, we wondered whether the lack of M1-mediated inhibition of Jub helps to explain the difference in rescue of cell adhesion by these two constructs. Indeed, depletion of Jub by RNAi worsened epithelial integrity of α-Cat-RNAi α-Cat-ΔM embryos (Fig. 5 H). These results suggest that Jub and the M2/3 domain cooperate with each other to support AJs under morphogenetic stress and underline a requirement for M2/3 in the mechanosensitive recruitment of Jub to junctions under higher tension (Fig. 5 I).

In contrast to the enrichment of the α-Cat interaction partners Jub, Vinc, and Cno at vertical, high-tension edges during germband extension (Blankenship et al., 2006; Sawyer et al., 2011; Kale et al., 2018; Razzell et al., 2018), AJ components such as Baz, Arm, α-Cat, and Ecad become planar polarized and enriched at horizontal edges (Paré and Zallen, 2020). Within an Ecad::GFP, GAP43::mCherry background, embryos expressing α-Cat-RNAi α-Cat-ΔM and α-Cat-RNAi α-Cat-ΔM23 showed significantly enhanced planar polarized localization of Ecad. Ecad was reduced specifically from the actomyosin-enriched vertical edges compared to α-CatR controls (Fig. 6 A), suggesting that the M2 and M3 domains support Ecad at high-tension edges. Interestingly, within an Ecad::GFP, Vinc::mCherry background, a reversal of Ecad planar polarity was seen in α-Cat-RNAi α-Cat-ΔM23 embryos (in which Vinc is ectopically enriched per CCC), while there was no change to Ecad planar polarity in embryos rescued by α-Cat-ΔM or α-CatR-ΔM1 (which Vinc is not predicted to bind; Fig. 6 A). Expression of Vinc::mCherry to assess Vinc distribution provided an additional genomic copy of Vinc, representing an over-expression condition. This additional copy of Vinc also improves adhesion in the ectoderm of α-Cat-RNAi α-Cat-ΔM23 embryos compared to controls without Vinc::mCherry (Fig. 6 B). Total gap areas in α-Cat-RNAi α-Cat-ΔM embryos are not significantly changed by additional Vinc (Fig. 6 B), suggesting this role for Vinc is mediated through its interaction with the CCC. As another method to enhance Vinc activity, we expressed an active, constitutively open form of Vinc (Vinc-CO; Maartens et al., 2016). This caused a strong enrichment of Ecad (Fig. 6 C), and a reversal of Ecad planar polarity such that Ecad is enriched at vertical edges (Fig. 6 D). These results suggest that the M1–Vinc interaction can reinforce Ecad recruitment or stability in response to actomyosin contraction and, consequently, cell adhesion during Drosophila gastrulation (Fig. 6 I). However, as the M1 domain is dispensable for adhesion and embryonic development, it is unlikely that the recruitment of Vinc is the primary mode of junctional reinforcement through the M region in wild type. Instead, our data suggests that M1 functions as an inhibitor of AJs.

Since M1 has an inhibitory effect on Ecad and Jub, and retention of M1 both enriches Vinc and causes more severe defects than removal of the whole M region, we wondered whether Vinc has an inhibitory effect on Ecad or Jub, and thus adhesion. A Vinc null mutant allele was incorporated into the rescue set-up to see if this would ameliorate epithelial integrity defects in α-Cat-RNAi α-Cat-ΔM23 embryos. However, removal of Vinc in fact worsened the defects in α-Cat-ΔM23 embryos, suggesting that Vinc cooperates with the M2 and M3 domains to support adhesion (Fig. 6 E). Furthermore, we confirmed that Vinc null mutants have no discernable epithelial defects during gastrulation, as previously reported (Alatortsev et al., 1997; Maartens et al., 2016), but found subtly yet significantly reduced α-Cat and Jub levels (Fig. 6, F and H). Although M1 is similarly dispensable for embryonic development, abnormally large clusters of Ecad, Jub, and α-Cat-ΔM1 were seen in α-Cat-RNAi α-Cat-ΔM1 embryos along AJs compared to α-CatR expressing controls (Fig. 5 A, Fig. 6 G, and Fig. S4 A). We used the normalized SD of Ecad to estimate the fragmentation of junctional signal, and this was significantly increased in the absence of M1 (Fig. 6 G). Likewise, junctional α-Cat signal is more fragmented in Vinc embryos (Fig. 6 H). M1 recruitment of Vinc therefore does support a more uniform distribution of the CCC at AJs, but the impact of its loss has no apparent phenotypic consequences. Overall, our observations suggest that the inhibitory effect the M1 domain has on adhesion is not due to Vinc recruitment, and that the M1 domain plays a dual inhibitory and supportive role at AJs.

Loss of either Vinc or Jub alone did not produce significant adhesion defects in Drosophila embryos (Maartens et al., 2016; Razzell et al., 2018), and Jub knockdown in a Vinc mutant background did not enhance embryonic lethality. However, loss of Vinc or Jub enhanced epithelial defects in M region deletion construct expressing embryos (Fig. 5 E and Fig. 6 E), raising the possibility that they work in parallel with an additional interaction partner of the M region, contributing to its mechanosensory output. As Afadin binds the mammalian α-catenin M2/3 domains (Pokutta et al., 2002; Sakakibara et al., 2020), we were wondering whether force-dependent junctional localization of Cno could help explain M region function in cell adhesion.

A direct M3–Afadin interaction was reported in MDCK cells, which is inhibited by M1 (Sakakibara et al., 2020). Although direct interactions between Drosophila Cno and α-Cat have not been reported to date, we found that removal of either M3 or M1 led to an increase in cortical Cno (Fig. 7, A and B). In fact, cortical Cno, representing largely the bicellular Cno pool, increased modestly with the expression of all M region deletions except for α-Cat-ΔM23, which showed a reduction but not a loss of Cno at AJs (Fig. 7, A and B). At gastrulation, Cno is significantly reduced by the knockdown of α-Cat (Fig. 7, A and B), when few cell junctions remain. Since Cno levels are increased in α-Cat-ΔM expressing embryos, the reduction of Cno in α-Cat-RNAi α-Cat-ΔM23 embryos may be a consequence of the strong adhesion defects in this condition. These observations suggest that Cno has alternate means of junctional recruitment, in particular at BCJs, apart from M region interactions. Deletion of M1 is predicted to unmask M2 and M3, but bicellular α-Cat-ΔM1 signal does not colocalize better with Cno than α-CatR control when expressed in α-Cat-RNAi embryos (Fig. S4, A and B), further supporting the conclusion that Cno can localize to BCJs independently of the M region. Echinoid (Ed) is thought to cooperate with Ecad to support Cno recruitment in the wing disc (Wei et al., 2005; Sawyer et al., 2009). However, even in a tissue that does not express Ed, the amnioserosa (Lin et al., 2017), junctional recruitment of Cno was not depleted in α-Cat-RNAi α-CatR-ΔM3 or α-Cat-RNAi α-Cat-ΔM embryos (Fig. S4 C). Knockdown of Ed in α-Cat-RNAi α-Cat-ΔM embryos also did not produce noticeable changes in phenotype (Fig. S4 D). Furthermore, α-Cat constructs were consistently located more basally than Cno in germband BCJs (Fig. S4 E). We also found that Cno enrichment at medium-tension vertical BCJs was not reduced in α-Cat-RNAi α-Cat-ΔM and α-Cat-RNAi α-Cat-ΔM23 embryos (Fig. 7 E and Fig. S3 A). Together, these findings suggest that pathways independent of Ed and α-Cat regulate the association of Cno with BCJs in the embryonic ectoderm.

Cno is normally elevated both at vertical BCJs during gastrulation and at TCJs (=vertices; Sawyer et al., 2009; Sawyer et al., 2011; Fig. 5 D). Interestingly, enrichment of Cno at TCJs, but not BCJs, is responsive to cytoskeletal tension (Yu and Zallen, 2020). We assessed Cno levels at vertices as a TCJ/BCJ ratio within the same embryo. Although the minimal Afadin binding site of αE-catenin is located within M3 (Pokutta et al., 2002; Sakakibara et al., 2020), removing M3 alone had no effect on the TCJ enrichment of Cno (Fig. 7, A and D). The enrichment of Cno at vertices was significantly reduced in embryos lacking M2, and this is worsened when M3 is also removed (Fig. 7, A, D, and E). However, some enrichment under M2/3 deletion is not abolished (Fig. 7, C and D) suggesting that M2/3 is not solely responsible for Cno recruitment to TCJs. Our results suggest that M3 along with M2 support normal levels of Cno recruitment to TCJ but that the M region is not involved in the localization of Cno to BCJ (Fig. 7 I).

Some vertices experience actomyosin forces from more than three junctions during the cell intercalations associated with germband extension. Each junction shows actomyosin generated tension (Fernandez-Gonzalez et al., 2009), predicting that connecting additional junctions increases the force at the vertex. We took advantage of this and compared vertices with three junctions to vertices with four or five junctions that are frequently observed in rosettes (Blankenship et al., 2006). In control α-Cat-RNAi α-CatR embryos, we found that Ecad is reduced as the number of junctions connected to a vertex increases (Fig. 7 H), suggesting that an increased number of edges linked to a vertex increasingly challenges the AJ. Interestingly, Jub enrichment increased only slightly with the number of junctions connected to a vertex (Fig. 7 G), whereas Cno remained unchanged when three, four, and five junction vertices were compared in α-Cat-RNAi α-CatR embryos (Fig. 7 F). In contrast, vertex enrichment of Cno and Jub dropped significantly in embryos expressing α-Cat-ΔM or α-Cat-ΔM23 (Fig. 7, F and G). Embryos lacking M2 and M3 showed a reduction in Cno, Jub, Vinc, and Ecad enrichment at all vertices, with a loss of Jub, Vinc, and Ecad enrichment at vertices joining four and five junctions (Fig. 7, F–H and Fig. S3 C). Vertices joining five edges showed Cno levels lower than BCJs in α-Cat-RNAi α-Cat-ΔM23 embryos (Fig. 7 F). Taken together, these findings suggest that although Cno membrane recruitment occurs independently of the α-Cat M region, the enrichment of Cno at vertices requires an interaction with the α-Cat M2 and M3 domains, and these domains support AJ stability at vertices to resist the increasing forces vertices experience during cell intercalation (Fig. 7 I). Moreover, the lack of progressive enrichment of Jub and Cno when three, four, and five junction vertices are compared would be consistent with a model posing that virtually all ⍺-Cat molecules at a three-junction vertex are in an open conformation, maximizing the mechanosensitive recruitment at these sites that cannot further be enhanced by increasing force.

To further test the hypothesis that the M region supports the mechanosensitive recruitment of Jub and Cno, we manipulated tension through a wound-healing response. Laser ablation of the cell cortex to cut the medial actomyosin network initially led to an expansion in cell area and reduction of junctional tension. Subsequently, actomyosin accumulates and contracts, increasing force at junctions of the wounded cell and its neighbors (Yu and Fernandez-Gonzalez 2016; Fig. 8, A–C). Following laser ablation, we measured the responses of Jub and Cno at TCJs of the contracting cell, and the adjacent BCJs in a neighboring cell that are being pulled at. In α-Cat-RNAi α-CatR embryos, a consistent peak in Jub intensity was observed following cell contraction—at around 10 min of analysis at BCJs and 3 min at TCJs (Fig. 8, D and E, and Video 9). On average this resulted in a 49 ± 18 and 43 ± 5% increase in Jub intensity at BCJs and TCJs, respectively (Fig. 8, F and G). In α-Cat-RNAi α-Cat-ΔM embryos, changes in Jub intensity at both BCJs and TCJs were significantly reduced, especially at BCJs where Jub intensity showed no increase in response to tension (Fig. 8, D–G). Jub signal at TCJs in α-Cat-RNAi α-Cat-ΔM embryos also did not reduce in response to cell expansion, when junctional tension is decreased, in contrast to α-Cat-RNAi α-CatR controls. Thus, without the M region Jub sensitivity to both increases and decreases in tension is reduced. In α-Cat-RNAi α-CatR embryos, bicellular Cno was not impacted by increased tension (Fig. 8, H and J, and Video 10), similar to findings by Yu and Zallen (2020) where Cno at BCJs was unaffected by conditions reducing tension. At TCJs, however, we did observe an increase in Cno intensity by 33 ± 6% following cell contraction (Fig. 8, I and K). Cno intensity in TCJs is already depleted in α-Cat-RNAi α-Cat-ΔM embryos, in which we see a smaller increase in Cno intensity at vertices (22 ± 5%, P = 0.054; Fig. 8, I and K). Together, these data further support the conclusion that the M region supports the normal recruitment of Jub and Cno in response to cytoskeletal tension.

Our findings support both ⍺-Cat–dependent and independent mechanisms for Cno junctional recruitment. To further ask whether Cno and ⍺-Cat act in parallel we examined embryos expressing ⍺-Cat M region deletions in a Cno compromised background. Indeed, Cno knockdown enhanced the epithelial defects of all α-Cat deletion constructs (Fig. 9 A). Furthermore, overexpression of Cno improves epithelial integrity in α-Cat-RNAi, α-Cat-RNAi α-Cat-ΔM23, and α-Cat-RNAi α-Cat-ΔM embryos (Fig. 9, B and C). In particular, overexpression of Cno significantly ameliorated the loss of ventral epithelium in α-Cat-RNAi α-Cat-ΔM23 and α-Cat-RNAi α-Cat-ΔM embryos, which are predicted to abolish any binding to Cno, as well as α-Cat-ΔM2 (Fig. 9, C and D). Interestingly, while Cno is significantly depleted at TCJs in α-Cat-RNAi α-Cat-ΔM23 embryos, Cno overexpression rescues adhesion defects and enhanced tricellular Cno to support adhesion during morphogenesis. Together with the M region–independent association of Cno with AJs, these observations suggest that Cno supports adhesion at TCJs through both a mechanism depending on ⍺-Cat mechanosensing and an α-Cat–independent parallel mechanism. In further support of a parallel pathway we found that knockdown of the Rap1 GTPase, which contributes to Cno membrane recruitment and activation (Sawyer et al., 2011; Bonello et al., 2018; Perez-Vale et al., 2021), enhances the defects in α-Cat-RNAi α-CatR-ΔM embryos (Fig. 9 E). During cellularization, Cno recruitment is unaffected by the loss of Arm, and instead requires the small GTPase Rap1 (Sawyer et al., 2009). Our results suggest that Rap1 interaction with Cno shares some parallel function with the M region to support adhesion. These findings suggest that Cno has a role in adhesion that is at least in part redundant to the M region, and to the M1 domain in particular as knockdown of Cno enhances loss of junctional Ecad and cell adhesion defects in α-CatR-ΔM1 embryos compared to α-CatR controls (Fig. 9, A and D).

Given the aforementioned findings, we wondered whether the recruitment of Vinc by M1 is redundant with Cno function. Epithelial integrity was significantly worsened when Vinc is removed from a cno-RNAi background, consistent with functional redundancy between Cno and the M1–Vinc interaction (Fig. 9 F). Functional redundancy between Cno and the M1 domain could explain why M1, and similarly, Vinc are dispensable in the embryo. Furthermore, double knockdown of Jub and Cno significantly enhanced adhesion defects compared to Cno knockdown alone (Fig. 9 G). Such findings argue that multiple mechanisms build a robust partially redundant network of interactions to foster AJ stability. Our evidence supports a model where the α-Cat M region itself, and the α-Cat interaction partners Vinc and Jub cooperate with Cno to promote AJ integrity and function (Fig. 9 H).

α-Catenin acts as a physical linker between the cadherin/β-catenin complex and the actin cytoskeleton (Desai et al., 2013; Buckley et al., 2014), and through its mechanosensory properties has the potential to modify interactions between cadherins and the cytoskeleton in response to external or internal forces (Yonemura et al., 2010; le Duc et al., 2010; Borghi et al., 2012). Two mechanosensory regions within α-catenin have been identified: the C-terminal actin-binding domain of α-catenin increases binding strength to F-actin as force is applied (Buckley et al., 2014; Ishiyama et al., 2018; Xu et al., 2020; Wang et al., 2022), and the central M region of α-catenin that acts principally by modifying interactions between α-catenin and interaction partners in response to force (Yonemura et al., 2010; Ishiyama et al., 2013; Thomas et al., 2013; Yao et al., 2014; Li et al., 2015; Maki et al., 2016; Seddiki et al., 2018; Barrick et al., 2018; Matsuzawa et al., 2018; Terekhova et al., 2019; Sarpal et al., 2019; Alégot et al., 2019; Boutillon et al., 2022). How α-catenin mechanosensing cooperates with multiple interaction partners that can also bind to F-actin to dynamically adjust adhesion during morphogenesis has remained largely unexamined.

We determined that the M region of Drosophila α-Cat plays a key role in maintaining cell adhesion during active junctional remodeling that drives morphogenesis in the early embryo. Taking advantage of the well-described tension landscape in the ectoderm as it undergoes cell intercalation during germband extension (Bertet et al., 2004; Fernandez-Gonzalez et al., 2009; Tetley et al., 2016), we determined that the M region supports the junctional recruitment of Vinc, Jub, and Cno in response to cytoskeletal tension, which function redundantly to support adhesion. While direct physical interaction between Drosophila ⍺-Cat and Vinc, Jub, or Cno have not been reported, our data and work of others (Alégot et al., 2019) suggest that recruitment of Vinc and Jub to AJs critically depends on direct or indirect interaction between Vinc and the M1 domain and Jub and the N2 domain of ⍺-Cat. In contrast, Cno localizes at AJs through mechanisms that are independent of α-Cat and supports cell adhesion in a parallel pathway. Only the mechanosensitive enrichment of Cno at TCJs (Yu and Zallen, 2020) requires its interaction with the M2 and M3 domains of α-Cat.

Characterization of the M region in both mammalian αE-catenin and C. elegans HMP1 by single molecule manipulation suggest that unfolding of M region domains occurs within physiological ranges of force (<20 pN) in a stepwise manner (Yao et al., 2014; Maki et al., 2016; Le et al., 2021,Preprint). Without force, the M region of α-catenin is stabilized in a closed conformation by a dynamic network of salt-bridges, which is largely conserved in Drosophila (Ishiyama et al., 2013; Li et al., 2015). Under application of force (5 pN αE-catenin, 10–15 pN HMP1), M1 is unfurled, allowing the binding of Vinc (Yao et al., 2014; Maki et al., 2016; Le et al., 2021,Preprint). This force activated Vinc binding is conserved even in the more distantly related HMP1 (Le et al., 2021,Preprint). Next, at a higher range of forces (10–15 pN αE-catenin, 15–20 pN HMP1), a further conformational change, likely affecting the M2 and M3 domains takes place (Yao et al., 2014; Le et al., 2021,Preprint). Moreover, a higher fraction of molecules in an open conformation found at vertices/TCJs that could be further enhanced through the application of force has been reported for αE-catenin (Yonemura et al., 2010). Projecting these findings onto our results, we propose a model where the frequency of three different tension states of α-Cat may correspond to the differential junctional recruitment (through direct or indirect binding) of interaction partners (Fig. 10). (i) Moderate amounts of Vinc and Jub are recruited to AJs at low-tension BCJs (=horizontal junctions in the extending germband). (ii) At medium-tension BCJs (=vertical junctions in the extending germband), recruitment of Vinc and Jub is significantly increased. We suggest this difference between horizontal and vertical edges could be explained by a higher range of forces at the vertical edge, which allow more α-Cat molecules to adopt an unfurled M1 conformation, facilitating both Vinc binding to M1 and Jub recruitment by N2. In contrast, at horizontal edges, which are under lower tension, a smaller fraction of α-Cat molecules may be found in the unfurled M1 conformation. (iii) Finally, at TCJs, which experience the highest levels of tension, Vinc and Jub are further elevated, and Cno is enriched in response to α-Cat mechanosensing. Here, the highest range of forces may lead to the highest fraction of α-Cat molecules with unfolded M2 and\or M3, allowing α-Cat to engage not only high levels of Vinc and Jub but also to interact with Cno to further stabilize AJs (Fig. 10, A–C).

The M1, M2, and M3 domains of α-Cat make distinct contributions to cell adhesion. M2 has a major role in supporting surface stability of the CCC and is required for the mechanosensitive AJ enrichment of Jub, which further stabilizes the junction. Together with M3, M2 supports Cno recruitment to TCJs to further reinforce AJs (Fig. 10 A). Recently, mechanical coupling between the M2 and N2 domains has been proposed (Terekhova et al., 2019), as well as direct, transient interaction between the disordered N-terminal region of β-catenin and the M region (Bush et al., 2019), suggesting that M2 could participate in strengthening interactions at the N-terminal domain with Jub and Arm. In embryos lacking M2, cortical levels of both Jub and Vinc, as well as Cno enrichment at TCJs, are reduced. Effects on multiple interaction partners at once could explain the strong adhesion defects observed in these embryos.

The M2/3 and M1 domains appear to have opposing roles in both adhesion and regulation of tissue growth (this work, Sarpal et al., 2019). M1 is required for growth regulation, and dispensable for adhesion, while the opposite is largely true for the M2 and M3 domains. This suggests that M region mechanosensing acts similarly in cell adhesion and tissue growth and that the tissue-specific consequences are due to differences in downstream pathways that impact on growth versus adhesion. Contrary to current models, we find that M1 negatively regulates CCC-dependent adhesion through inhibition of Jub recruitment, which may be released by M1 unfurling in response to force. While M1 recruitment of Vinc supports AJs stability, this contribution is subtle and dispensable for both adhesion and growth (Sarpal et al., 2019, this work). This is consistent with dispensability for the α-catenin Vinc binding domain and Vinc for barrier function in zebrafish (Han et al., 2016) and worsening of the zygotic α-Cat mutant phenotype by combination with a Vinc mutation (Jurado et al., 2016). M region mechanosensing and Vinc were recently implicated in coordinating the migration of non-epithelial Polster cells in the zebrafish gastrula (Boutillon et al., 2022). In Drosophila, we noticed that the loss of M1 or Vinc let to a somewhat enhanced clustering of the CCC, suggesting that the M1–Vinc interaction supports the uniform distribution of the CCC at cell contacts. Genetic analysis suggests that the M1–Vinc interaction is redundant with Cno and that Cno and the M region act in parallel to support AJs. Additional α-Cat M region interaction partners such as α-Actinin (Nieset et al., 1997) or srGAP (Serre et al., 2022) remain to be investigated in our system, and could contribute to the redundant, multivalent interactions at the CCC–actin interface.

Recent analysis of Cno function in the Drosophila ectoderm highlighted the need for multiple, partially redundant interactions between Cno, the plasma membrane, and the cytoskeleton in support of AJ stability during tissue morphogenesis (Manning et al., 2019; Perez-Vale et al., 2021). Thus, both Cno and α-Cat stabilize AJs through multiple interactions. It is unclear, however, whether Cno is a mechanosensor or only responds to mechanosensory inputs by α-Cat as we show here or by other mechanisms as Cno enrichment at TCJs is also enhanced by phosphorylation of Cno by the Abl non-receptor tyrosine kinase (Yu and Zallen, 2020). Whether phosphorylation of Cno modifies the interactions with α-Cat remains to be explored. Cno at TCJs needs to be dynamic to enable normal vertex resolution during the cell rearrangements required for cell intercalation (Yu and Zallen, 2020).

Taken together, our work suggests that the α-Cat M region acts in cell adhesion to support embryonic morphogenesis through mechanosensitive modulation of interactions with Vinc, Jub, and Cno. The M1–Vinc interaction, dispensable on its own, shows redundancy with Cno function. While Vinc and Jub association with AJs is completely dependent on α-Cat mechanosensing, the association of Cno with AJs depend on α-Cat mechanosensing only for its enrichment at TCJs. Cno supports adhesion at bicellular contacts in a parallel pathway to α-Cat. We suggest that these redundant and cooperative multivalent interactions are the molecular basis of the mechanoresponsive dynamic stabilization of AJs that maintain tissue integrity as the epithelium undergoes cell contacts changes.

Flies were raised on standard media at 25°C for all experiments. For rescue experiments, males carrying mat-GAL4 (P{matα4-GAL-VP16}67; Häcker and Perrimon, 1998) and the α-Cat constructs (inserted at attP2) were crossed to females carrying fluorescently tagged proteins of interest and UAS-α-Cat-RNAi (TRiP HMS00317, Transgenic RNAi project [TRiP]). F1 virgin females carrying mat-GAL4, fluorescent protein, and α-Cat construct were then crossed to OregonR wild-type males (except jub::GFP males were used for rescue experiments measuring Jub::GFP) and the progeny of this cross were analyzed. Similarly, for double knockdown analysis, F1 virgin females carrying both mat-GAL4 and transgenes were crossed to OregonR wild-type males and their progeny assessed. In Vinc mutant analysis, a complete genomic deletion of Vinc, Vinc^102.1 was used (Klapholz et al., 2015). See Table S1 for details of crosses used for each figure. The following fly lines were employed:

α-CatR, α-CatR-ΔM1, α-Cat-ΔM (aka: α-CatR-ΔM), α-CatR-ΔM3 (Sarpal et al., 2019); α-Cat-ΔM2 (aka: αCatΔVH2-N), α-Cat-ΔM23 (aka: αCatΔVH2), DE-cadΔβ::α-CatABD (aka: DE-cadΔβ::VH3-CTD); UAS-GFP (Desai et al., 2013); Vinc-CO (Maartens et al. 2016); UAS-Cno (aka: CanoeFL::GFP; Bonello et al., 2018).

α-Cat-RNAi (TRiP HMS00317); jub-RNAi (TRiP HMS00714); canoe-RNAi (TRiP HMS00239 and GL00633); Rap1-RNAi (TRiP HMJ21898) were produced by the TRiP at Harvard (Zirin et al., 2020). GFP-RNAi (BL41552, BDSC) and Ed-RNAi (v938, VDRC) were soured from the Bloomington Drosophila Stock Center (BDSC) and the Vienna Drosophila Resource Center (VDRC).

α-Cat::YFP (α-CatCPTI002516), Canoe::YFP (cnoCPTI000590), and Zipper::YFP (zipCC01626) are derived from the Cambridge Protein Trap Insertion project, Kyoto Stock Centre (Lye et al., 2014; Lowe et al., 2014). Other fluorescently tagged proteins are insertions under the control of the respective endogenous promoter: Ecad::GFP (aka: DEcad::GFP, shg>DEcad::GFP; Huang et al., 2009); Vinc::mCherry (recombined with Ecad::GFP; Kale et al., 2018); Jub::GFP (Sabino et al., 2011). Fluorescently tagged proteins ubiquitously expressed in the early embryo: Ani-RBD (aka: ubi>Anillin-RBD::GFP; Munjal et al., 2015) and sqh>GAP43::mCherry (Martin et al., 2010).

To analyze the terminal cuticle phenotype of embryos, eggs collected overnight at 25°C were aged for 2 d or 24 h if viable, and then washed in water and dechorionated for 5 min in 2.15% sodium hypochlorite. After a second wash, embryos were mounted on a slide into a 3:2 mixture of Hoyer’s medium and lactic acid and incubated at 85°C overnight. To determine the percentage of embryonic lethality, flies were allowed to lay eggs at 25°C for an 8-h period, and between 100 and 300 eggs were counted and arranged into rows on a new agar plate. This plate was then examined 24 and 48 h later to count hatched larvae and dead embryos.

Drosophila embryos were dechorionated in 2.15% sodium hypochlorite and then either heat-fixed in a salt solution (Miller et al., 1989; Figs. 9 and S4) or fixed for 20 min in 4% formaldehyde in a 1:1 PBS:heptane mixture (Fig. S2). Primary antibodies used under heat fixation were anti-HA (rat monoclonal, 3F10; 1:500; Roche), anti-Cno (rabbit 1:1,000, a gift from Mark Peifer, University of North Carolina, Chapel Hill, NC, USA), anti-DE-cadherin (rat monoclonal DCAD2 1:25; Developmental Studies Hybridoma Bank). Under heptane fixation: anti-HA (mouse 16B12 1:100; Abcam). For immunostaining in Fig. S1, ovaries were fixed in 5% formaldehyde in phosphate buffer for 10–12  min. Primary antibodies used for immunostaining were: anti-HA (rat monoclonal 3F10, 1:600; Roche); anti-Arm (mouse monoclonal N2 7A1, 1:100; DSHB); anti-α-spectrin (rabbit polyclonal, 1:1,000, gift from Daniel Branton, Harvard University, Cambridge, MA, USA); anti-GFP (mouse monoclonal JL-8, 1:400; Clontech). Secondary antibodies were used at a concentration of 1:400 (Invitrogen).

A heat-shock inducible MARCM system was employed as previously described (Sarpal et al., 2012; Desai et al., 2013). For cell clones produced in the follicular epithelium, the percentage of rescued cells was calculated. The following recombinant lines were used for MARCM analysis: hs–Flp FRT40A; da–Gal4 UAS–mCD8::GFP α-Cat¹/TM6B; tub–Gal80 ubi–α-Cat FRT40A; act–Gal4 α-CatX α-Cat¹/TM6B (Sarpal et al., 2012; Desai et al., 2013; Sarpal et al., 2019).

Live imaging was performed on dechorionated embryos mounted in Halocarbon Oil 27 (Sigma-Aldrich) between an oxygen permeable membrane (YSI; Xylem Inc.) and coverslip, and short time-lapse movies were acquired, similar to other previously described methods (Blankenship et al., 2006). Imaging was performed at room temperature using a Leica TCS SP8 scanning confocal microscope with 40× or 63× objectives (HC PL APO CS2 with NAs of 1.30 and 1.40, respectively) and using Leica LASX Software (Leica Microsystems). A Carl Zeiss Axiophot2 microscope using a phase-contrast 20× lens (NA 0.5) connected to a Canon Rebel XSi camera was used to capture images of embryonic cuticles, and time-lapse movies were captured using this scope and a 10× lens (NA 0.14) to determine rates of germband extension. The number of embryos per genotype is listed on each figure.

Z-stacks were collected using a z-step size of 0.35–0.4 µm, starting from above the vitelline membrane and moving well through all visible apical junctional structures for 15–20 steps. Planes containing autofluorescence of the vitelline membrane were removed in post-processing and maximum projections were produced using Fiji (Schindelin et al., 2012). All images shown are such maximum projections unless otherwise indicated. GAP43::mCherry signal in supplemental videos is displayed as a median projection of three z-slices. Adobe Illustrator was used to assemble confocal images. Curves were adjusted in Adobe Photoshop for Vinc::mCherry confocal images (Fig. 4), antibody staining of Ecad (Fig. 9), and images of cuticle preparations, and this was performed equally across conditions within the same experiment. For images shown using a colormap legend, the “gem” lookup table in Fiji was used. The same settings on confocal equipment and image processing were applied for all images within the same experiment.

α-Cat-RNAi α-CatX expressing embryos were live imaged during stages 6, 7, and 8 and phenotypic categories were determined for embryos within a Jub::GFP, Ecad::GFP, GAP43::mCherry, or myosin::YFP background. For quantification of germband extension, the change in position of the proctodeal invagination after 40 min was normalized to the total length of the embryo.

Analysis of cortical levels of fluorescence was performed using the Matlab script SIESTA (scientific image segmentation and analysis; Fernandez-Gonzalez and Zallen, 2011). For a number of embryos, a polyline 3 pixels in width was drawn along the perimeter of each cell, and the mean FI along this line, minus the mean FI within the center of the cell, was plotted as cortical FI. Total junctional myosin is given as the sum of the mean fluorescence along the perimeter and mean fluorescence within the perimeter for each cell (total FI). Junctional/total myosin fraction is given as the average FI of myosin at the cell perimeter (junctional) out of total myosin (fraction of total FI). For comparison of Ani-RBD at the edge of gaps versus cells not in contact with a gap, the average FI along the perimeter (perimeter FI) was plotted for polylines drawn along cells (intact) versus along the gap (gap).

Using SIESTA, trajectories 3 pixels in width were drawn along individual cell edges, avoiding overlap or vertices, and these edges were grouped into 15° bins, reflected about the DV axis. For each bin, the cytoplasmic fluorescence of the image was subtracted from the average FI of cell edges. This value was then divided by that of the 0–15 (horizontal) bin. Enrichment to vertical edges is plotted from this calculation for the 75–90° bin. To show the FI of Jub or Ecad at vertical or horizontal edges, for each embryo cytoplasmic fluorescence is subtracted from the average FI of edges in the vertical (75–90°) and the horizontal (0–15°) bins.

Fluorescence within cell vertices was measured using the ellipse tool in Fiji for 30–45 vertices per embryo. For enrichment at vertices, the mean intensity of vertices was divided by the mean intensity of bicellular cell edges per embryo. Bicellular cell junctions were measured through trajectories drawn in SIESTA, or lines drawn in Fiji.

In Fiji, on an image representing a field of view of 2,136.4 μm², the brush tool was used to mask each gap, and then the wand tool was used to measure area of gap and fit ellipse.

To determine the degree of colocalization of Cno versus α-Cat construct signal, regions of interest of junctions were compared for a number of embryos using the coloc2 plugin in ImageJ.

Live embryos were prepared as described above and imaged using a Revolution XD spinning-disk confocal microscope equipped with an iXon Ultra 897 camera (Andor), a 60x oil immersion lens (NA 1.35; Olympus), and MetaMorph Software (Molecular Devices). 16-bit Z-stacks were acquired at 0.5 μm steps every 10s (20 slices per stack) for 15 min, and maximum intensity projections were used for analysis. Jub::GFP was illuminated using a 488 laser with 30% intensity and 100 gain. Cut cells were selected based on position and stage (ventral ectoderm, stage 7), and were centered within the field of view. A pulsed Micropoint N2 laser (Andor) tuned to 365 nm, with an attenuation plate setting of 20, was applied to the cell cortex for 10 pulses. Image analysis was performed using PYJAMAS (Fernandez-Gonzalez et al., 2021). In order to segment cells, polylines were drawn using the Livewire function for each cell/frame individually. TCJs were segmented by placing fiducials at each junction. Background intensity was removed by subtracting the mode of the image and photobleaching correction was performed by dividing by the mean pixel intensity of the frame. Pixel width for analysis was 3 pixels. Cell outlines were used to calculate cell area at each time point. Cell area and vertex intensity traces over time were smoothed using a Gaussian filter with a width of 1 min (six time points). Since wounded cells expanded prior to rapidly contracting, the maximum cell area was identified as the start of cell contraction and cell area was tracked for the next 4 min at the end of which cell area had stabilized. To measure the change in tricellular vertex intensity in response to the change in cell area, we compared the minimum vertex intensity in the 1 min after maximum area to the maximum intensity reached by the vertex in the 4 min following initiation of cell contraction. To measure the change in bicellular junction intensity in the cell neighboring the contracting cell, we compared the minimum junction intensity at the time the junction fully extended to the maximum intensity at the junction 10 min later.

Prism v9 (GraphPad) was used for statistical analysis and plots, except for the balloon plot in Fig. 4, which was made using Microsoft Excel, and plots in Fig. 8, which were generated using MATLAB. In violin plots, bold lines indicate the median, thinner lines represent interquartile ranges. For bar graphs, height of bar indicates mean, and error bars indicate SD (except Fig. S1, error bars give SEM). In box-and-whisker plots (Fig. 8), gray line indicates the mean, edges of box mark SEM, and whiskers represent SD. Significance was calculated on measurements from each cell, junction, or vertex for a number of embryos, denoted on each plot. In experiments with only two conditions, the Mann-Whitney unpaired two-tailed t test was used. When comparing more than two conditions, and output consisted of single measurements per n, ordinary one-way ANOVA with Tukey’s multiple comparisons test was used unless distribution was non-normal, where Kruskal–Wallis test with Dunn’s multiple comparisons test was used to determine significance. Normality was assessed using D’Agostino & Pearson test, Shapiro-Wilk test and normal QQ plot. To compare enrichment of AJ proteins at bicellular junctions within the same bin of cell edge trajectories under multiple conditions, ordinary two-way ANOVA with uncorrected Fisher’s Least Significant Difference (LSD) was performed. Two-way ANOVA used to determine significance when comparing enrichment of AJ proteins in vertices joining three, four, or five junctions, with Tukey’s multiple comparisons test to compare across conditions, and a mixed effects model with Geisser-Greenhouse correction and Tukey’s multiple comparison test on vertices within the same embryo. To compare terminal cuticle phenotypes, Chi-square analysis was performed on pairs of conditions.

Fig. S1 shows the analysis of α-Cat-ΔM within ovarian follicular epithelium. Fig. S2 shows additional data on the effectiveness of knockdown by α-Cat-RNAi, cellular distribution of constructs, and examples of cuticle phenotypes. Fig. S3 provides a detailed breakdown of the analysis of FI of Ecad, Vinc, Jub, and Cno by edge angle or vertex type. Fig. S4 shows data illustrating the relationship between α-Cat and Cno. Fig. S5 demonstrates the phenotypic categories used to score cuticle defects and contains a legend for graphics used in figures. Table S1 contains a summary of genotypes used and breakpoints of constructs used. Video 1 shows a gastrulating embryo expressing α-Cat-RNAi. Videos 2, 3, 4, 5, 6, 7, and 8 show examples of gastrulation in embryos expressing α-Cat-RNAi and M region deletion constructs. The constructs expressed in these embryos are as follows: Video 2: α-CatR; Video 3: α-Cat-ΔM (an embryo with strong defects); Video 4: α-Cat-ΔM (an embryo with moderate defects); Video 5: α-CatR-ΔM1; Video 6: α-Cat-ΔM23; Video 7: α-CatR-ΔM3 (an embryo with stronger defects); and Video 8: α-Cat-ΔM2. Video 9 shows examples of movies analyzed in the laser ablation experiment to measure Jub mechanosensitive recruitment, and examples of movies analyzed to investigate Cno mechanosensitive recruitment are shown in Video 10.

All data are available in the manuscript or supplementary materials. Transgenic Drosophila stocks from Desai et al. (2013) or Sarpal et al. (2019) are available upon request.

We thank Ritu Sarpal and Limin Wang for providing the data shown in Fig. S1. We thank Mark Peifer for discussion of unpublished data and comments on the manuscript. We thank Dorothea Godt for critical reading of the manuscript. We thank Nick Brown, Thomas Lecuit, Mark Peifer, the BDSC, the Drosophila RNAi Screening Center at Harvard Medical School, and the Developmental Studies Hybridoma Bank for reagents. We like to thank the Imaging Facility of the Department of Cell and Systems Biology, University of Toronto, for support.

This work was funded by a project grant from the Canadian Institutes of Health Research (CIHR; to U. Tepass). R. Fernandez-Gonzalez is the Canada Research Chair in Quantitative Cell Biology and Morphogenesis. U. Tepass is a Canada Research Chair for Epithelial Polarity and Development. K.E. Rothenberg was supported by a postdoctoral fellowship from the CIHR, and by a grant from the CIHR to R. Fernandez-Gonzalez.

Author contributions: The project was conceived and experiments were designed by L. Sheppard and U. Tepass. L. Sheppard carried out experiments and data analysis. Laser ablation experiments were performed and analyzed by D.G. Green and G. Lerchbaumer, with formal analysis performed by K.E. Rothenberg, under the supervision of R. Fernandez-Gonzalez. U. Tepass provided supervision and raised funds. The paper was written by L. Sheppard and U. Tepass.