Cytocortex-dependent dynamics of Drosophila Crumbs controls junctional stability and tension during germ band retraction

Epithelial cells are basic building blocks of all animals, lining the surface of their internal and external organs. As the first tissue specified in development, epithelial sheets extensively rearrange to drive morphogenesis in order to establish the basic body plan (Heisenberg and Bellaiche, 2013; McCaffrey and Macara, 2011; Quintin et al., 2008). Therefore, understanding the formation and remodelling of epithelia during development is crucial to deciphering the emergence of animal shape. Epithelial cells are intrinsically polarised, as manifested by an asymmetric distribution of proteins and organelles along the apico-basal axis. Furthermore, the plasma membrane is subdivided into the free apical surface, facing the outside or a lumen, and the basolateral side, contacting neighbouring cells and a basal lamina, respectively (Macara et al., 2014; Rodriguez-Boulan and Macara, 2014).

Epithelial morphogenesis depends on the tight coordination of cell shape changes, cell division and cell growth. In addition, it requires the spatially and temporally controlled action of several cellular components, such as the actomyosin cortex and cell–cell adhesion complexes. Forces generated by actomyosin drive morphogenesis by locally remodelling the plasma membrane and junctional connections (recently reviewed in Koenderink and Paluch, 2018; Pinheiro and Bellaïche, 2018; Sun and Toyama, 2018). Cortical and junctional functions strongly depend on the proper localisation of their components along the apico-basal axis of the cell, and in fact many of them are enriched in the apical and subapical domain, for example, components of adherens junctions (AJs) and of the actomyosin machinery (Guillot and Lecuit, 2013; Harris, 2018; Levayer and Lecuit, 2012, 2013; Martin and Goldstein, 2014; Qin et al., 2018). Therefore, morphogenesis is contingent on the maintenance of cell polarity (Macara and McCaffrey, 2013). Apico-basal polarity is controlled by a number of polarity protein complexes, loss of function of which can lead to impaired development and lethality (Campanale et al., 2017; Flores-Benitez and Knust, 2016; Macara et al., 2014; Tepass, 2012). However, despite extensive knowledge on the individual components important for apico-basal polarity, their temporal and spatial regulation, coordination and dynamics during morphogenesis are only partially understood.

Crumbs (Crb), initially identified in Drosophila (Tepass and Knust, 1990; Tepass et al., 1990), is an apically localised transmembrane protein, highly conserved from worms to mammals. Crb is necessary for the maintenance and organisation of the apical plasma membrane in many epithelia across different animal phyla. For example, Drosophila embryos lacking Crb die due to disorganisation of the epidermis, caused by loss of apico-basal polarity and failure to establish the zonula adherens (ZA), an adhesion belt encircling the apex of the cell (Grawe et al., 1996; Tepass, 1996; Tepass and Knust, 1990; Tepass et al., 1990). Similarly, mouse embryos mutant for Crb2, one of the three mammalian Crb orthologues, die during early gastrulation due to impaired cell ingression (Ramkumar et al., 2016; Xiao et al., 2011). In contrast, Crb overexpression can induce an increase of the apical domain and multilayering of the epithelium in Drosophila (Klebes and Knust, 2000; Letizia et al., 2011; Röper, 2012; Wodarz et al., 1995). These studies suggest that the amount of Crb at the apical cell surface must be tightly regulated in order to modulate apical domain size, which in turn is crucial for the maintenance of the integrity of developing epithelia. Although several mechanisms have been shown to modulate apical Crb trafficking and stability in different epithelia (Blankenship et al., 2007; Dong et al., 2014; Lin et al., 2015; Lu and Bilder, 2005; Pocha et al., 2011; Roeth et al., 2009; Zhou et al., 2011), the dynamics of Crb levels at the plasma membrane and its regulation during morphogenesis so far remain unknown.

Among all apical polarity proteins identified so far, Crb is the only transmembrane protein, and hence uniquely positioned to act as a structural and functional link between the plasma membrane and the apical cytocortex. Indeed, the short intracellular domain of Crb contains a FERM (protein 4.1/ezrin/radixin/moesin) domain-binding motif (FBM) (Klebes and Knust, 2000), which can directly interact with the FERM proteins Yurt (Laprise et al., 2006) and Moesin, a well-known organiser of the cytocortex (Fehon et al., 2010; Polesello et al., 2002; Tepass, 2009). A mutation in a conserved amino acid of the FBM abolishes the interaction with Moesin (Wei et al., 2015), resulting in actomyosin hyperactivation in the amnioserosa of Drosophila embryos and failure to undergo dorsal closure (Flores-Benitez and Knust, 2015). This points to an important role of the Crb–cytocortex interaction in morphogenesis. However, whether this interaction is required for regulating Crb amounts and behaviour at the apical membrane remains to be determined.

The embryonic epidermis of Drosophila is a powerful model to study epithelial morphogenesis, and the interplay between the apical plasma membrane and actomyosin-mediated contractile forces. Morphogenesis of the Drosophila embryo is characterised by a number of key events involving large-scale movements of epithelial sheets. One such event is germ band retraction (GBR). Within ∼2 h after egg laying, the germ band, encompassing the anlagen of the thoracic and abdominal segments, retracts so that the caudal end of the embryo comes to lie at the posterior end of the embryo. At the same time, the amnioserosa unfolds and stretches to cover the dorsal side of the embryo (Fig. 1A) (Campos-Ortega and Hartenstein, 1997; Gumbiner, 1992; Martinez Arias, 1993; Schöck and Perrimon, 2002). GBR involves the movement of the entire epithelial cell sheet and is mediated by changes in cell shape, cell organisation and cell packing. The latter includes shortening along the anterior–posterior axis and widening along the dorsal–ventral axis (Fig. 1A), processes that require controlled junctional remodelling (Clement et al., 2017; Yu and Fernandez-Gonzalez, 2016).

The involvement of Crb in the regulation of apical cell surface size and AJ stability, on the one hand, and the requirement of cell surface area changes and cellular remodelling during GBR, on the other (Lecuit and Mahadevan, 2017; Martin and Goldstein, 2014; Rauzi et al., 2010) raises the question of whether both processes are connected. We hypothesise that modulation of apical surface area and AJ remodelling relies on continuous turnover and dynamic modulation of Crb localisation and stability at the plasma membrane. Therefore, we set out to investigate the regulation of Crb during GBR. A major challenge in understanding Crb behaviour so far has been the lack of live imaging studies to follow the dynamics of Crb in a developing organism. We overcame this limitation by optimising a fluorescence recovery after photobleaching (FRAP) assay to study Crb dynamics in the Drosophila embryonic epidermis, which allowed us to measure the mobile pool of Crb as well as its recovery kinetics and regulation. This enabled us to test Crb behaviour in different morphogenetic contexts, namely, during GBR and at later stages of embryogenesis, when morphogenesis is completed and cells are fully differentiated and have mature adherens and septate junctions (Huang et al., 2011). Genetic and pharmacological perturbations revealed that the mobile pool of Crb at the plasma membrane is maintained by endocytosis and the activity of the actomyosin cortex, while components of the actomyosin cortex are important to fine-tune Crb dynamics in a stage-dependent manner. Stabilisation of Crb at the plasma membrane depends on its FBM, loss of which results in increased junctional tension and higher DE-cadherin (DE-cad; also known as Shotgun) turnover, which eventually impairs junctional rearrangements. Our data reveal a mutual control between the apical cortex and Crb, in that Crb not only regulates the cytocortex as previously shown (Flores-Benitez and Knust, 2015; Salis et al., 2017; Sherrard and Fehon, 2015), but reciprocally the cytocortex modulates Crb dynamics at the plasma membrane.

During early stages of GBR (embryonic stage 12), the epidermal epithelium undergoes major morphogenetic transformation, during which the posterior cells are displaced over 50 µm along the anterior–posterior axis (Fig. 1A, black arrow; Fig. S1B) (Lynch et al., 2013; Schöck and Perrimon, 2002, 2003). Cells at this stage are not yet fully morphologically differentiated [manifested by incompletely formed septate junctions (SJs)] and are still reorganising their apical area and AJs, which is important to drive tissue morphogenesis (Huang et al., 2011; Tepass, 1997). To elucidate the role of Crb in this context, we studied the behaviour of Crb, fused to an EGFP tag in the extracellular domain, hereafter called Crb–EGFP. Crb–EGFP was expressed from a genomic transgene (foscrb_EGFP) in an otherwise crb-null mutant background. As previously shown, foscrb_EGFP fully rescues crb mutant phenotypes, and Crb–EGFP shows localisation and expression levels similar to that of the wild-type protein (Klose et al., 2013).

Evaluation of Crb–EGFP revealed a highly dynamic spatiotemporal localisation pattern at stage 12, particularly in the ventral epidermis (VE). In all epidermal cells, Crb–EGFP was localised in a belt in the subapical region (SAR) of the plasma membrane, outlining the perimeter of the cell (Fig. 1A, cartoon). Cells in the VE additionally showed Crb–EGFP on the free apical surface (apical free region; AFR), which previously has been only found in tracheal cells (Olivares-Castineira and Llimargas, 2017). However, no obvious pattern of Crb–EGFP distribution on the AFR was evident, except a decrease in dividing cells, which can be identified as they round up (Fig. 1B,B′,D; Fig. S1A, Movie 1). This dynamic behaviour is specific for Crb–EGFP and was not observed for DE-cad–mTomato, which marks the AJs (Fig. 1C,C′,D; Movie 2).

In order to quantitatively evaluate the dynamics of the Crb–EGFP pool at the SAR during GBR, we used the fluorescence recovery after photobleaching (FRAP) assay, a well-established technique to measure the behaviour of fluorescently labelled molecules (reviewed in Lippincott-Schwartz et al., 2018). We first performed FRAP experiments in the thoracic segments T1–T3 (defined as ‘anterior’) of the ventral epidermis of early stage 12 embryos. The mobile fraction of Crb–EGFP was compared with the mobile fraction of two other membrane proteins, junctional DE-cad–GFP, and the general membrane marker Resille–GFP (Res–GFP; also described as CG8668^117-2) (Figard and Sokac, 2011; Jayasinghe et al., 2013; Morin et al., 2001) (Fig. 1E). Crb–EGFP was much more mobile than DE-cad–GFP, but revealed a slower recovery kinetics than Resille–GFP (Fig. 1F,G). These data suggest that the observed Crb dynamics are subject to specific regulation and are not just a consequence of altered membrane turnover during GBR.

Next, we set out to compare Crb–EGFP dynamics between the anterior and the posterior part of the embryo (abdominal segments A5–A7 were defined as ‘posterior’). The latter shows a higher degree of morphogenetic rearrangements, in that it is displaced on average by more than 50 µm during the first 40 min of GBR, resulting in a total displacement of >100 µm at the end of GBR (Fig. 1A, square; Fig. S1B). We photobleached spots of 2 µm diameter in the antero-posterior cell–cell boundaries (Fig. 2A) and extracted recovery curves (Fig. 2B), best-fitted by double exponential equations (Fig. S2, Table S1). This pointed to two components involved in the recovery of Crb–EGFP, with a fast (τ_fast) and a slow (τ_slow) halftime of recovery (10 s for τ_fast and 60–120 s for τ_slow), both in the anterior and the posterior part of stage 12 embryos (Fig. 2D,E). By comparing the amplitude values (C1 and C2) of each, we found that the slow process contributes more to the observed recovery than the fast process (Table S2). τ_slow was about twice as high in the anterior part than in the posterior region (Fig. 2E). Hence, the faster recovery kinetics of Crb–EGFP correlates with a more actively reorganising epithelium (largest cell displacement during GBR; Fig. 1A, Fig. S1). In contrast, the mobile fraction of Crb–EGFP did not differ between the anterior and the posterior region and was very high in both regions (∼80–90%) (Fig. 2F).

To corroborate the correlation between Crb–EGFP dynamics and the degree of tissue remodelling during GBR, we compared Crb–EGFP dynamics in the extensively remodelling epithelium (early stage 12) and the more static epithelium (end of GBR; stage 15) (Fig. 2A). At stage 15, when dorsal closure is completed, the epithelium is devoid of large-scale morphogenetic events and consists of fully differentiated cells connected by mature adhesive junctions (Huang et al., 2011). FRAP analyses revealed that at stage 15, the overall mobile fraction of Crb–EGFP was similar to the one at stage 12 (∼90%) (Fig. 2F). In contrast, the halftime of recovery τ increased as development proceeded (τ_fast from ∼10 s to 20 s; and τ_slow from ∼60–120 s to 230–300 s) (Fig. 2D,E), suggesting decreased turnover of Crb–EGFP at later stages. Interestingly, the difference between the anterior and posterior areas observed in stage 12 embryos was abolished at stage 15 (Fig. 2E). The behaviour of Crb–EGFP is specific, since Res–GFP showed no differences in recovery between anterior and posterior (∼56–89 s) nor between stages 12 and 15 (∼60 s) (Figs 1F and 2C, and data not shown). This demonstrates that Crb–EGFP mobility is under specific spatial and temporal control. To rule out that different Crb–EGFP dynamics merely reflect changes in membrane turnover rates, we analysed Crb–EGFP in the salivary glands of third-instar larvae (Fig. 2A, Fig. S3B,C). They represent a differentiated epithelium showing neither junctional remodelling nor proliferation, but high secretory activity (Andrew et al., 2000; Geron et al., 2013; Segal et al., 2018). Complete loss of crb results in rudimentary salivary glands in the embryo, but no later function has been described so far (Tepass and Knust, 1990). Crb–EGFP in the salivary gland epithelium displayed only negligible recovery after 25 min (compared to complete recovery after 7 min in the embryonic epithelium at stage 15) (Fig. 2B,F; Fig. S3A,D–D″), similar to what was seen for Res–GFP (purple in Fig. 2C), despite a very high degree of secretory activity of the glands.

Taken together, Crb–EGFP dynamics is spatially and temporally regulated in the developing epidermis. While the mobile fraction of Crb–EGFP stays high throughout GBR, the recovery kinetics is fast during early phases and decreases as development proceeds. Hence, Crb–EGFP dynamics correlate with the degree of rearrangements in the embryonic epidermis.

Next, we asked for the mechanism(s) that link Crb–EGFP dynamics and morphogenetic remodelling of the epithelium. To this end, we first explored the processes governing Crb–EGFP recovery. Diffusion coefficients measured at the plasma membrane by fluorescence techniques have previously been shown to be dependent on the size of the bleached area: increasing the analysed area resulted in a lower diffusion coefficient (Andrade et al., 2015; Bulgakova et al., 2013; Goehring et al., 2010). Therefore, if the fast recovery halftime τ_fast was due to lateral diffusion at the plasma membrane, we would expect it to be length-scale dependent and to decrease with increasing bleach spot size. Indeed, increasing the bleach spot size from 1 µm to 2 µm increased the τ_fast of Crb–EGFP recovery from 1 s to 10 s (Fig. 3A,B; Fig. S4A), but not τ_slow (∼92 s to ∼130 s) (Fig. 3A,B; Fig. S4B, Table S3). Hence, the component defined by τ_fast is likely due to diffusion.

In contrast, the recovery kinetics of τ_slow significantly slowed down when larger areas of up to 5 µm diameter were bleached (Fig. 3B), with τ_slow significantly increasing from ∼110 s to ∼139 s (Fig. S4B, Table S3). The mobile pool decreased as well (Fig. S4C, Table S3). In this case, the radius of the photobleached area (5 µm) was significantly wider than the size of a single-cell border and hence also photobleached molecules localised in the cytoplasm, including newly synthesised and recycled molecules on the way to and from the plasma membrane, respectively. The decrease in the observed recovery kinetics indicates that membrane–cytoplasmic exchange (Fig. 3B, red versus black) (Goehring et al., 2010) is involved in the slow (τ_slow) recovery component. Similarly, bleaching the intracellular pool of Crb–EGFP before performing FRAP experiments led to a decrease in fluorescence recovery (Fig. S4D,E). This further supports the assumption that the mobile pool of Crb–EGFP is maintained by cytoplasm–membrane exchange.

Finally, we assessed the contribution of biosynthetic recovery to the overall Crb–EGFP recovery by performing whole-cell FRAP (Huang et al., 2011). To this end, we photobleached 15-µm-diameter regions of the epidermis, resulting in bleaching of the entire Crb–EGFP pool present within the targeted cells, including molecules present at the plasma membrane and in the cytoplasm. The fluorescence recovery observed under this condition is dependent on the rate of biosynthesis and maturation of Crb–EGFP (Huang et al., 2011). We observed negligible recovery of fluorescence during the first 20 min after photobleaching both in the early and late developmental stages (Fig. 3C, Fig. S4F,G). This shows that the biosynthetic recovery of Crb–EGFP does not significantly contribute to the measured recovery using 2 µm spot FRAP, which takes only several minutes. Therefore, the observed spatiotemporal differences in Crb–EGFP recovery are independent of its biosynthetic rate.

Taken together, these results demonstrate that diffusion accounts for the fast recovery component, whereas cytoplasm–membrane exchange is the prevalent mechanism contributing to the recovery of Crb–EGFP at the plasma membrane at short time scales (up to 7 min).

Our finding that membrane-cytoplasm exchange accounts predominantly for Crb–EGFP recovery agrees with previous studies showing that endocytosis and recycling of Crb play an important role in Crb function, particularly in maintaining the apical domain of epithelial cells (Blankenship et al., 2007; Dong et al., 2014; Lin et al., 2015; Lu and Bilder, 2005; Pocha et al., 2011; Roeth et al., 2009; Zhou et al., 2011). To test whether endocytosis and recycling underlie Crb–EGFP exchange between the cytoplasm and the plasma membrane, we made use of shibire^ts1 (shi^ts1), a temperature-sensitive allele of Dynamin (Kosaka and Ikeda, 1983; van der Bliek and Meyerowitz, 1991), which blocks endocytic scission at restrictive temperatures (29°C and above) (Fig. S5). If endocytic recycling accounts for the recovery of Crb–EGFP, we would expect a decrease in Crb–EGFP recovery in a shi^ts1 mutant background at restrictive temperatures.

Indeed, blocking endocytosis in stage 12 shi^ts1 embryos by raising the temperature to 32°C for 10 min slowed down Crb–EGFP recovery (Fig. 4A–D). τ_slow increased from ∼50 s to ∼800 s in the anterior and from ∼50 s to ∼400 s in the posterior part of the embryo (Fig. 4E), without decreasing the mobile fraction (Fig. 4F; Table S4). τ_fast was not significantly affected under this condition (Table S4). An increase of τ_slow was also observed in shi^ts1 embryos at stage 15, from ∼60 s to ∼180 s in the anterior part (Fig. 4G,I). In contrast to stage 12, the mobile fraction was significantly decreased from ∼90% to ∼55% in shi^ts1 embryos at stage 15 (Fig. 4J). These results confirm that most of the membrane pool of Crb–EGFP is maintained by recycling at this short time scale. In addition, they point to an area- and stage-specific involvement of the endocytic machinery in regulating Crb–EGFP.

It has been shown that actomyosin activity is the driving force of tissue remodelling, and that cortical activity is linked to turnover of membrane proteins involved in morphogenesis (Braun et al., 2015; Jewett et al., 2017; Lee and Harris, 2014). The observed correlation between Crb–EGFP dynamics and epithelial remodelling suggested that actomyosin activity also plays a role in modulating Crb–EGFP behaviour. To test this hypothesis, we injected stage 12 embryos with Y-27632, an inhibitor of the Myosin II activator ROCK (Rho kinase) (Ishizaki et al., 2000). If epithelial rearrangements impacted on Crb–EGFP turnover, then blocking actomyosin contractility would be expected to decrease the recovery rate of Crb–EGFP. Injection of Y-27632 into the perivitelline space of stage 12 embryos indeed strongly decreased the mobile fraction from ∼90% to ∼50% in comparison to that of control embryos (Fig. 5A,B; Fig. S6A, Table S5). Such a decrease was not observed in stage 15 embryos injected with Y-27632 (Fig. 5C,D). Whereas the mobile fraction was significantly affected by Y-27632 at stage 12, recovery halftimes (τ_slow) were not affected by Y-27632 treatment at both stages (Fig. S6B,D). This suggests that in the early embryonic epidermis ROCK-mediated actomyosin activity is crucial for the maintenance of the mobile pool of Crb–EGFP. Strikingly, we observed a depletion of apical Crb–EGFP in cells of the VE at 15–20 min after Y-27632 injection (Fig. S6E–H, yellow dashed line). This was not due to overall loss of membrane integrity, as revealed by the presence of DE-cad-mTomato and by positive labelling of the plasma membrane with FM4-64 (Fig. S6E–H), demonstrating that cells were still intact.

Taken together, these results establish that actomyosin activity is required to maintain the apical Crb–EGFP pool and a high mobile fraction of Crb–EGFP at the plasma membrane in the ventral epidermis, specifically in stage 12 embryos.

Blocking the contractility of the actomyosin cytocortex can impair various processes, including intracellular membrane trafficking, assembly and scission of endocytic vesicles and transport of clathrin-coated pits (Collins et al., 2011; Georgiou et al., 2008; Grassart et al., 2014; Harris and Tepass, 2008; Leibfried et al., 2008). To further define the role of the cytocortex in stage-dependent Crb–EGFP dynamics, we genetically modified the organisation of the cytocortex in a more controlled way by depleting either β_H-Spectrin [encoded by Drosophila karst (kst)] or Moesin (Moe), using zygotic mutants. β_H-Spectrin is a structural component of the­ cytocortex, implicated in regulating membrane rigidity and endocytosis (Lee and Thomas, 2011; Tjota et al., 2011), while Moe is a FERM-domain protein that connects the cytocortex via F-actin interaction with the plasma membrane (Fehon et al., 2010; Kunda et al., 2008; Polesello et al., 2002). Moe and β_H-Spectrin can interact with Crb directly and indirectly, respectively (Médina et al., 2002; Wei et al., 2015). Lack of zygotic Moe had no or only minor effects on Crb–EGFP recovery (Fig. 6A–C; Fig. S7A–C, Tables S7, S8), but specifically decreased the mobile fraction of Crb–EGFP at both stages (Fig. 6C,F), suggesting a role in maintaining the mobile pool of Crb–EGFP. In contrast, loss of zygotic β_H-Spectrin differentially affected τ_slow of Crb–EGFP at early and late stages. While τ_slow was increased ∼3-fold (from ∼56 s to ∼143 s) at stage 12, it was decreased from ∼230 s to ∼85 s at stage 15 (compare Fig. 6A,B with D,E). These results revealed a distinct, stage-dependent role of β_H-Spectrin in regulating Crb–EGFP mobility and recovery kinetics.

The observation that there was barely any effect on Crb–EGFP dynamics in Moe mutant embryos was puzzling, since the FBM of Crb can directly interact with Moe (Wei et al., 2015). To better understand the functional importance of this interaction, we generated an EGFP-tagged version of foscrb_EGFP that additionally carries a point mutation in the FBM motif (Y10A; hereafter called foscrb_{EGFP_Y10A}) (Fig. 7A). The τ_slow of Crb_{EGFP_Y10A} was lower than that of wild-type Crb–EGFP, both at stage 12 (Fig. 7B,C, Fig. S8A), and, even more pronounced, at stage 15, when τ_slow decreased by ∼3-fold, from >300 s in wild-type to ∼100 s in the mutant (Fig. 7D,E, Fig. S8B, Table S6). This result shows that the stability of Crb–EGFP depends on an intact FBM in its intracellular domain.

In agreement with previous reports, homozygous mutant embryos carrying only the foscrb_Y10A allele (and no functional crb allele) develop a continuous epidermis and thus completely rescue the loss of apico-basal polarity in the epidermis observed upon loss of crb (Flores-Benitez and Knust, 2015; Klose et al., 2013). However, closer inspection of the epithelium revealed small gaps and breaks in junctional DE-cad–GFP foscrb_Y10A embryos, both at stage 12 and stage 15 (Fig. 8A; Movie 3). To unravel the basis for these junctional defects, we analysed the cell rearrangements that occur during stage 12. Similar to what is seen during germ band extension (Bertet et al., 2004), epithelial cells of the retracting germ band undergo defined transitions, during which they change their neighbours, divide and extrude (Movies 3, 5 and 6, wild-type control). In foscrb control embryos, the DE-cad–GFP belt was continuous during junctional remodelling (Fig. 8C, top; Fig. S8C,D, top). In contrast, foscrb_Y10A mutant cells failed to maintain the integrity of DE-cad–GFP staining at the vertices (Fig. 8C, bottom; Fig. S8C,D, bottom), resulting in gaps. Live-imaging revealed that these gaps were transient, lasting ∼10 s, after which DE-cad–GFP staining was again continuous (Fig. 8C, coloured arrows; Movie 4). Similar breaks were observed in foscrb_Y10A mutants undergoing more complex rearrangements, which were also accompanied by discontinuities of DE-cad–GFP at cell–cell borders (Fig. 8D; Movie 5). These observations demonstrate that an intact FBM in Crb is required for proper junctional remodelling.

Junctional remodelling depends on the tight control between actomyosin contractility and turnover of AJ components (Baum and Georgiou, 2011; Pinheiro and Bellaïche, 2018). The defects in junctional remodelling in foscrb_Y10A mutant embryos prompted us to test for DE-cad–GFP turnover through FRAP experiments in mutant and wild-type embryos (Movie 7). DE-cad–GFP was much more mobile in foscrb_Y10A mutant embryos than in the foscrb background in the ventral epidermis of stage 12 embryos (Fig. 8B; Fig. S8E; Movies 7, 8; Table S9).

Proper regulation of junctional tension is important during cell rearrangements to modulate junctional stability and cell shape by deforming the plasma membrane. Increased turnover of DE-cad–GFP has been linked to increased tensile forces during junctional rearrangements (de Beco et al., 2015; Iyer et al., 2019; Kale et al., 2018). Therefore, to test whether the observed gaps in DE-cad–GFP continuity in foscrb_Y10A embryos are associated with aberrant tension, we performed laser ablations in the embryonic epidermis. This assay enables the evaluation of mechanical tension by measuring the velocity of vertex displacement (Farhadifar et al., 2007). These experiments showed that DV tension in foscrb_Y10A mutants was significantly higher in comparison to that observed in control foscrb embryos (Fig. 8F). The increase in cortical tension in foscrb_Y10A mutants further coincides with aberrant organisation of F-actin (Fig. 8E; Fig. S8C) and Myosin II (Spaghetti squash, sqh; Movies 9, 10) in the ventral epidermis. F-actin revealed diffuse or punctate distributions, distinct from that in foscrb embryos (Fig. S8C).

Taken together, our results show that Crb–EGFP dynamics decreases as development proceeds. Furthermore, the pool of dynamic Crb–EGFP is maintained and modulated by direct and indirect interactions with the actomyosin cortex. A functional FBM in Crb is important to allow for proper rearrangement of epithelial cells by controlling junctional tension and DE-cad turnover during GBR.

Here, we provide the first study of the behaviour of the important polarity regulator Crb in different contexts of epithelial remodelling during embryogenesis. Live imaging and FRAP analysis in Drosophila embryos revealed a previously unknown, temporally and spatially regulated dynamics of Crb during GBR. These dynamics correlate with the degree of epithelial reorganisation, which is manifested by changes in cell and tissue shape, cell rearrangements and junctional remodelling. We identified the endocytic machinery and the actomyosin cortex as important regulators of Crb dynamics. Abolishing the link between Crb and the cytoskeleton results in increased junctional tension and enhanced DE-cad turnover, which impairs junctional remodelling. These results provide novel insights into how epithelial cells can adjust to changes occurring during dynamic morphogenetic processes in the Drosophila embryo, which may also have implications for epithelia in other species.

It has recently been hypothesised that during germ band extension, Crb may act as a sensor for ‘morphogenetic stress’ (as defined in Letizia et al., 2018) and may be involved in controlling cell packing by modulating the apical surface area depending on the mechanical stress. A similar role has been suggested in pupal wing development, where Crb is important to stabilise the actomyosin cortex and DE-cad, and loss of Crb results in fragmentation of junctional DE-cad and impaired epithelial cell packing (Salis et al., 2017). These authors speculated that an altered membrane tension induced by loss of crb in pupal epithelial cells may account for junctional defects at the end of tissue rearrangement. We now show that, in fact, a mutation in the FBM of Crb (foscrb_Y10A) results in increased tension in the epidermis during GBR, which is associated with increased mobility of DE-cad and junctional defects. Whether Crb exerts a similar role during pupal wing development remains to be assessed. Our findings are in agreement with earlier results showing that Crb can act as a regulator of actomyosin activity (Das and Knust, 2018; Flores-Benitez and Knust, 2015; Médina et al., 2002; Röper, 2012; Salis et al., 2017; Sherrard and Fehon, 2015; Wodarz et al., 1995). But how can Crb perform this function?

Structural studies have not only shown a direct interaction between the FBM of Crb and Moe, but additionally revealed an unexpected interaction between the PDZ-domain binding motif (PBM) at the C-terminus of Crb and Moe. The PBM interaction site in Moe overlaps with the site at which it binds to the phosphatidylinositol 4,5-bisphosphate (PIP₂) head group, which is required for recruitment of Moe to the plasma membrane. Therefore, interaction between Moe and Crb could enhance membrane localisation and activation of Moe, and thereby modulate the interaction between the actin cytocortex and plasma membrane (Wei et al., 2015). This underscores an important function of Crb in linking the plasma membrane to the cytocortex, thereby modulating the stability of the adherens junctions.

Unexpectedly, we found that the apical cortex conversely influences the Crb recovery kinetics and mobile fraction. Crb recovery is spatially and temporally controlled during GBR: at early stages, recovery is faster in the posterior part of the germ band, a region which undergoes extensive rearrangements. At stage 15, when the epidermis shows overall reduced morphogenetic remodelling, Crb recovery slows down and no difference between anterior and posterior was observed, suggesting a functional link between Crb dynamics and tissue rearrangements. β_H-Spectrin regulates Crb mobility in a stage-specific manner: it sustains high Crb mobility at early stages but ensures a decrease in Crb mobility at later stages, thus promoting stabilisation of Crb. Moe, on the other hand, is required to maintain a high mobile fraction of Crb but does not play a major role in Crb recovery. This result conflicts with those obtained for Crb_Y10A, a Crb protein lacking a functional FBM and hence unable to bind Moe (Wei et al., 2015). Crb_{EGFP_Y10A} shows faster recovery than the wild-type protein at all stages analysed. This discrepancy could be explained by assuming that the maternal component of Moe is sufficient to maintain Crb recovery kinetics. However, this assumption cannot be tested experimentally, since removal of maternal Moe by inducing germ line clones leads to defects in oocyte development (Polesello et al., 2002). Interactions with Yurt, a negative regulator of Crb at late stages of embryogenesis (Laprise et al., 2006) could also contribute to the regulation of Crb recovery. In the follicle epithelium, other FERM-domain-containing proteins (Moe, Merlin and Expanded) participate in maintaining the apical pool of Crb (Aguilar-Aragon et al., 2019 preprint), possibly by changing actomyosin cortex organisation (Sherrard and Fehon, 2015).

Crb mobility is not only affected in the absence of individual cytocortex components (β_H-Spectrin and Moe), but also upon blocking actomyosin activity, pointing to a more general impact of the cortex on cell polarity. This is corroborated by previous observations showing that treatment with Y-27632 results in a similar loss of proteins of the apical Par complex from the apical plasma membrane (Atwood and Prehoda, 2009; Simões et al., 2010; Yu and Fernandez-Gonzalez, 2016). Finally, the apical cytocortex can contribute to the regulation of endocytic trafficking, including proper assembly and scission of endocytic vesicles, transport of clathrin-coated pits (Kaksonen et al., 2006; Yarar et al., 2005) or targeting of cargo from the ER to the Golgi network (Devarajan et al., 1997; Salcedo-Sicilia et al., 2013). Therefore, it will be interesting in the future to more specifically unravel the individual roles of actomyosin organisation and activity on polarity, junctional remodelling and epithelial morphogenesis in the developing Drosophila epidermis.

To conclude, the ability of Crb to bind FERM-domain proteins makes it an important player at the cross-road of actomyosin activity, intracellular trafficking and plasma membrane remodelling during morphogenesis. Additionally, Crb can recruit the Myosin II regulator PATJ, which could provide an additional way of modifying actomyosin (Sen et al., 2012). It remains to be explored to what extent elevated dynamics of Crb facilitates tissue rearrangement, or, conversely, whether remodelling of the tissue induces differential dynamics of Crb. Both ways would allow the integration between polarity and junctional regulation to achieve cell shape changes in response to the changes in the actomyosin activity. Answering this question will provide a deeper understanding of the function of the Crb protein in maintaining epithelial tissue integrity.

All flies were kept as described (Ashburner et al., 2011) and maintained on standard food. Crosses were performed at room temperature, unless otherwise stated. Experiments using the temperature-sensitive allele shi^ts were performed at 32°C (FRAP experiments) and 34°C (live imaging). For all experiments, mutant alleles were balanced over balancers carrying GFP-encoding transgenes to distinguish the homozygous mutant embryos. The protein null allele crb^GX24 (Huang et al., 2009) was used in combination with various crb fosmids to ensure that the different Crb variants are expressed from the fosmid only. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study. Fly stocks used in this study are listed in Table S10, more detailed information on fly genotypes can be found at Flybase (http://flybase.org/; release FB2018_06) (Thurmond et al., 2019).

A counter-selection strategy was used to generate the foscrb variants in this study (Klose et al., 2013). A list containing the sequences of the modification rpsl-neo cassette and the counter-selection oligonucleotides is provided in Tables S11, S13, S14. All exons in the newly created transgenic constructs were sequenced (Table S12).

Transgenic flies carrying the fosmid were generated in y, v, P(nos-phiC31\int.NLS)X; P(CaryP)attP40 (Bloomington #25709), using the phiC31 integrase-mediated site-specific integration into attP landing-site attP40. Correct transformants were screened for red fluorescent eyes (from the 3×P3-DsRed marker in the fosmid backbone) (Ejsmont et al., 2009). Injection and establishment of transgenic lines were performed according to standard protocols (Bachmann and Knust, 2008).

Embryos were collected in fly cages on apple juice agar plates with yeast paste at 25°C for 1 h and collected for imaging after 7 and 11 h (stage 12 and 15, respectively). Staging of embryos was according to Campos-Ortega and Hartenstein (1997). Mutant embryos were collected under a dissecting microscope based on the absence of GFP fluorescence expressed from the balancer chromosomes.

Embryos were dechorionated by hand and aligned with their ventral side up on an apple juice agar plate. They were transferred to a bottom glass Petri dish (MatTek P35G-1.5.14-C, Ashland, MA), coated with heptane glue, so that the ventral side of each embryo was facing the coverslip. Embryos were covered with a 1:1 mixture of halocarbon oil 27 and 700 and imaged either at 25°C or at restrictive 32°C (FRAP) and 34°C (live imaging). Embryos were imaged using an Andor Revolution Spinning Disk confocal microscope (Andor Technology) with Andor iXon 888 Ultra with fringe suppression (Andor Technology), a 60× silicon oil-immersion lens (NA 1.30; Olympus). 16-bit z-stacks were acquired at 0.25-µm steps every 200 ms (25 slices/stack) at a magnification of 0.196 µm/pixel. Image manipulation was fully compliant with the image guidelines for proper digital image handling (Rossner and Yamada, 2004).

The following antibodies were used at the indicated concentrations for immunofluorescence: mouse anti-tubulin (1:500; Sigma-Aldrich), rabbit-anti-Crb 2.8 (1:500; (Richard et al., 2006)), rabbit anti-GFP (1:1000; Invitrogen), and mouse anti-GFP (1:1000; Sigma-Aldrich) antibodies, secondary antibodies were conjugated to Alexa Fluor 488, 568 or 647 (1:500; Invitrogen). Phalloidin-568 (1:200; Invitrogen) was used to label F-actin.

Embryos were collected in fly cages on apple juice agar plates at 25°C. For all experiments, 1 h collections were used. Embryos of different genotypes used in the same experiment were collected and aged for the same time and fixed and stained in parallel. Embryos were dechorionated with 3% sodium hypochlorite (3 min), fixed in 4% formaldehyde in phosphate-buffered saline (PBS)/heptane (v/v of 1:1) on a rotating shaker (40 min). Embryos were devitellinized by hand. Embryos were incubated in blocking solution with 5% normal horse serum (NHS; Sigma-Aldrich H1270, St. Louis, MO) in PBT (0.3% Triton X-100) for 2 hours, followed by an overnight incubation at 4°C with primary antibodies diluted in 5% NHS-containing PBT (0.3% Triton X-100). The embryos were washed in PBT (0.3% Triton X-100) four times for 15 min each and then incubated with the appropriate secondary antibodies (Alexa conjugated), diluted in 5% NHS-containing PBT (0.3% Triton X-100), for 1 h at room temperature. Embryos were washed in PBS four times for 15 min each and mounted on glass slides using VectaShield (Vector Laboratories). Homozygous mutant embryos were identified by selecting against the GFP signal from the fluorescent balancers. Images were acquired using ZEISS LSM 880 with Airy scan detector (ZEISS Microscopy) with a 3×/1.3 LCI Plan-Neofluar, W/Glyc, DIC, Zeiss objective.

The FRAP procedure was adapted from Bulgakova et al. (2013). FRAP experiments were performed in the ventral embryonic epidermis in the thoracic segments T1–T3, referred to as ‘anterior’, and in abdominal segments A5–A7, termed ‘posterior’. Experiments were performed either at the beginning of stage 12, when GBR is initiated, or during stage 15 of embryogenesis, when dorsal closure is completed. Image acquisition of protein accumulation and photobleaching was performed using an inverted confocal microscope (Andor Spinning Disc FRAPPA) equipped with a 60×/1.35 NA oil UPlan-SApochromat objective lens. 16-bit depth images were taken at a magnification of 0.111 µm/pixel (FRAP). The camera used for acquisition was Andor iXon EM+ DU-897 BV back illuminated EMCCD; dexel size of EMCCD chip: 16 µm. FRAP experiments were performed using z-stack image acquisition of 2.1 µm. In each embryo, several point regions were photobleached at junctions so that there was only one bleaching event per cell. Photobleaching was performed using 60% of laser power, with the dwell time of 5 µs resulting in the reduction of EGFP signal. Before each FRAP experiment, five frames were acquired (pre-bleaching). Pre-bleaching and photobleaching were performed in the same plane, using spot sizes of 1,2, 5 or 15 µm. Following bleaching, 500 frames were taken, with an exposure time of 300 ms/frame rate. Images were registered for movement of the embryo during the acquisition if necessary (StackReg plugin) (Thevenaz et al., 1998).

An F-test was used to choose the equation and compare datasets. Recovery data were fitted in Graph Pad using a nonlinear regression model.

Laser ablation experiments were performed on a Zeiss microscope stand, equipped with a spinning disc module (CSU-X1; Yokogawa), EMCCD camera (Andor) and a custom-built laser ablation system, using a 355 nm, 1000 Hz, pulsed laser. Ablations were performed in stage 12 embryos along a 10 µm line oriented either along the dorsoventral (DV) axis or antero-posterior (AP) axis of the embryo. The ablation was only performed in one single plane where the entire ROI could be acquired in one focal plane. To capture the rapid recoil of the ablated front, single plane images were acquired with 50 ms exposure and the interval between frames was 85 ms. The initial recoil velocity of the ablated region was computed for the estimation of mechanical stress in the tissue. The images obtained from the laser ablations were analysed using Fiji and MATLAB (Mathworks). Kymographs were generated on both sides of the ablated line using the FIJI plugin Multi Kymograph (http://imagej.net/Multi_Kymograph). Both recoil velocities were calculated using a custom written routine in MATLAB and the average of the two recoil velocities is presented as the recoil velocity for each ablation.

Embryos were collected for 1 h and incubated at 25°C for 6 h 30 min or 10 h 30 min in order to have stage 12 and 15 (since the preparation for injection plus injections varied by up to 20 min), respectively. Staged embryos were glued onto a coverslip covered with heptane glue, dehydrated for 10 min, and covered with 1:1 mixture of halocarbon oils 27 and 700 before injection. Pharmacological inhibitors were injected anteriorly into the perivitelline space of embryos using an Eppendorf FemtoJet micromanipulator. Using this condition, drug solutions are predicted to be diluted 50-fold in the embryo (Foe and Alberts, 1983). FM4-64 (Thermo Fisher Scientific) was injected at a concentration of 5–10 mM in H₂O and Y-27632 (Tocris) was used at a concentration of 5mM in H₂O to inhibit ROCK activity. H₂O was used as control. Embryos were imaged as described for FRAP or using an Andor Revolution Spinning Disk confocal microscope (Andor Technology) with Andor iXon 888 Ultra with fringe suppression (Andor Technology), a 60× silicon oil-immersion lens (NA 1.30; Olympus). 16-bit z-stacks were acquired at 0.25-µm steps every 200 ms (15 slices/stack).

Graphs were plotted and statistical analyses were performed using GraphPad Prism 6. Results are expressed as means±95% confidence intervals. Statistical significance was calculated by an independent-samples two-tailed t-test or one-way ANOVA followed by Dunnett's or Tukey's post hoc multiple comparison test. For FRAP data analysis and fitting, we performed a statistical test assessing which model is preferred. We assessed significance by applying an extra sum-of-squares F test (α=0.05).

We would like to acknowledge the support by FlyBase (NIH #U41 HG000739 and MRC #MR/N030117/1) and the Bloomington Drosophila Stock Center (NIH P40OD018537), which was instrumental for this work. We are indebted to the Light Microscopy Facility (Jan Peychl) and the Scientific Computing Facility (Benoit Lombardot) at MPI-CBG for the outstanding technical assistance, and N. Bulgakova for constructive technical advice. We also thank E. Martìn-Blanco, David Flores-Benitez, A. Thommen and C. Dahmann for helpful comments on the manuscript, and the Knust lab for scientific discussions.