Novel role for Grainy head in the regulation of cytoskeletal and junctional dynamics during epithelial repair

Epithelia are the first line of defense against the surrounding environment in multicellular organisms. As barrier maintenance is critical for survival, epithelia need to be able to self-repair. Multiple molecular mechanisms are used during wound response that differ depending on the species and developmental stage (Garcia-Fernandez et al., 2009). Injured simple epithelia efficiently close small holes by relying mostly on cytoskeleton dynamics, in particular through the formation of an actomyosin contractile cable at the wound edge. Such structure functions as a purse-string that brings cells together. This mechanism was first dissected in Drosophila (Wood et al., 2002) but is also conserved in vertebrate systems, such as humans, mice and zebrafish (Bement et al., 1993; Danjo and Gipson, 1998; Martin and Lewis, 1992; Mateus et al., 2012). The Drosophila pupal notum epithelium is an excellent system to study cellular dynamics (Antunes et al., 2013; Bosveld et al., 2012; Herszterg et al., 2013), being highly amenable to genetic manipulation and live imaging. We have previously shown that, in this system, the rapid formation of the actomyosin cable upon injury is preceded by a set of coordinated flows of Ca²⁺, actin and myosin, and apical cell constriction across the tissue (Antunes et al., 2013). The first signal detected upon tissue damage is an increase in intracellular Ca²⁺, which induces a flow of actomyosin polymerization and cell constriction from a few cell rows away toward the wound edge. Notably, rearrangement of junctional complexes takes place during the propagation of these flows (Antunes et al., 2013). E-Cadherin (E-Cad; also known as Shotgun, Shg, in flies), one of the major components of the adherens junctions (AJs), is highly regulated upon injury. E-Cad turnover and recycling are important in accommodating cell shape changes, as well as actomyosin flow formation and directionality (Antunes et al., 2013; Pinheiro et al., 2017; Rauzi et al., 2010). It has recently been shown in Drosophila embryos that inhibition of E-Cad turnover leads to an impairment in actin cable formation and wound closure (Carvalho et al., 2014; Hunter et al., 2015; Matsubayashi et al., 2015). However, it is still unknown how E-Cad is regulated during wound repair.

Grainy head (Grh) is an essential transcription factor (TF) expressed in virtually all epithelial tissues. This gene was first identified in Drosophila (Bray et al., 1989; Dynlacht et al., 1989) and is conserved in vertebrates within the grainyhead-like (Grhl) gene family (Janicke et al., 2010; Sueyoshi et al., 1995; Venkatesan et al., 2003; Wang and Samakovlis, 2012; Wilanowski et al., 2002). Interestingly, the evolution of this gene family seems to match the evolutionary origin of the epithelium itself (Traylor-Knowles et al., 2010), highlighting its importance as an epithelial master regulator. Grh TFs can act as activators or repressors according to the developmental context (Attardi and Tjian, 1993; Bray and Kafatos, 1991; Huang et al., 1995). In Drosophila, Grh is necessary during embryonic development, epidermal differentiation, central nervous system specification and epithelial repair (Attardi et al., 1993; Bray et al., 1989; Brody and Odenwald, 2000). During repair of the Drosophila embryonic epidermis, Grh induces the expression of genes associated with cuticle repair (Bray and Kafatos, 1991; Mace et al., 2005) as well as the activation of the receptor tyrosine kinase Stitcher (also known as Cadherin 96Ca), which in turn regulates actomyosin cable formation (Wang et al., 2009). In mammals, Grhl3 mutants fail to repair the epidermis upon injury (Ting et al., 2005) and cultured keratinocytes from these animals present actin polymerization defects upon wounding (Caddy et al., 2010). The conserved involvement of the grh/Grhl genes in tissue repair suggests that this gene family is essential for the regulation of cytoskeletal dynamics and junctional complexes during tissue repair.

In this study, we uncover a novel and central role for Grh in integrating AJ and actomyosin dynamics while maintaining epithelial identity during wound repair. We show that Grh is required both for the early cellular events after injury, such as actomyosin flow and cable formation, and for the maintenance of the epithelial characteristics in cells at the wound edge. Moreover, we demonstrate that Grh acts by tightly regulating E-Cad dynamics in a crucial manner for the proper repair of the injured epithelium.

To better understand the roles of Grh in cytoskeletal dynamics, cell-to-cell communication and tissue movement during epithelial repair, we made use of the Drosophila pupal notum epithelium and a laser ablation set-up and live imaging protocol previously established in our laboratory (Antunes et al., 2013). Since known grh mutations are embryonic lethal (Bray and Kafatos, 1991), we explored the possibility of inducing grh mutant clones in the notum. Unfortunately, grh^s2140 somatic clones could not be recovered. As an alternative, we manipulated Grh function using RNA interference (RNAi), expressing double-stranded RNA (dsRNA) through the Gal4/UAS system (Brand and Perrimon, 1993; Dietzl et al., 2007), to reduce Grh expression in a tissue-specific and time-controlled manner. We validated the grh knockdown (KD) by observing the reduced expression of the grh gene in a BAC-grh-GFP transgenic line expressing grh dsRNA under the control of the pupal notum driver stripe-Gal4 (sr-Gal4) (Usui et al., 2004) (Fig. S1). We started our analysis of grh phenotypes by comparing the localization of E-Cad (E-Cad::tomato or E-Cad::GFP) (Huang et al., 2009; Oda and Tsukita, 2001) and actomyosin (Utrophin::GFP, mCherry::Moesin and Spaghetti-squash::GFP) (Förster and Luschnig, 2012; Millard and Martin, 2008; Royou et al., 2002) in wild-type and Grh KD pupae before wounding. The effects on E-Cad were analyzed by expressing grh RNAi under the control sr-Gal4, which is expressed in a subset of cells in the notum (Usui et al., 2004), allowing us to compare wild-type and Grh KD cells in the same pupa. We observed a slight increase in E-Cad protein levels when grh was knocked down compared to control pupae, but the subcellular localization of E-Cad and the integrity of the epithelium were maintained (Fig. 1Ai,ii). This increase was more pronounced when normalizing the intensity levels to the junction length of each membrane (Fig. 1Aiii), probably due to the observed decreased apical area in the Grh KD cells compared to wild-type cells (Fig. S1B). In agreement with this E-Cad increase, we observed that E-Cad (shg) mRNA levels were also increased in grh mutant embryos when compared to wild-type embryos (Fig. 1Aiv). F-actin intensity levels were elevated and its subcellular distribution appeared normal when grh was knocked down (Fig. 1B,B′). Regarding myosin, although its intensity levels were also increased and its apical-basal distribution was maintained, the apical organization seemed to be affected, particularly at the medial (non-cortical) pool level (Fig. 1B,B′). While the mild changes in adhesion and cytoskeletal components observed did not appear to cause major changes in tissue organization, we observed that Grh KD had a striking effect after wounding (Fig. 2; Movie 1). We found that in the Grh KD pupae, E-Cad membrane levels were strongly reduced. This was observed in the first four to six cell rows around the wound, when compared to levels pre-wound and to cell rows further away from the wound site (Fig. 2Ai′–iv′,Bii). In contrast, wild-type pupae showed no drastic variation in E-Cad levels (Fig. 2Ai–iv,Bi). By quantifying E-Cad intensity levels, we observed a slight decrease in the wild type (Fig. 2Bi) that is likely due to photobleaching, since cells away and close to wound present the same level of decrease. Quantification of E-Cad::GFP levels in the Grh KD pupae showed the clear and rapid decrease in intensity in the first 10 min post wounding (mpw) in cells close to the wound, which becomes more gradual from this point onward (Fig. 2Aii′–iv′,Bii). The same phenotype was observed when using another reporter for E-Cad under control of its endogenous promoter (E-Cad::tomato) (Huang et al., 2009) (Fig. S2). Our analysis showed that, despite only having a modest effect in the expression of E-cad/Shg in notum epithelial cells before wounding, Grh is essential for correct E-Cad dynamics occurring at the wound periphery upon injury.

Having observed an alteration in E-Cad dynamics in Grh KD pupae, we wanted to understand whether this was associated with changes in actomyosin dynamics. In wild-type pupae, the typical actomyosin flow initiates shortly upon laser ablation, between four to six rows of cell rows away from the leading edge, and propagates towards the wound edge, leading to the formation of an actomyosin cable (Fig. 3Ai–iv). By contrast, in the Grh KD pupae, propagation of the actin flow was stalled and ectopic actin cables formed within the tissue surrounding the wound, a few cell rows away from the wound margin (Fig. 3Ai′–iv′, red arrowheads; Movie 2). This behavior was also confirmed using a different grh dsRNA line (Fig. S3B). To better understand the defects associated with this phenotype, we quantified the peaks of F-actin intensity in the four cell rows around the wound, which results in a clear profile of the sequential reaction of the cells. Whereas, in the wild-type, the intensity peaks are observed consecutively starting from the cells furthest away from the wound to the cells at the wound edge (Fig. 3Bi, asterisks), in Grh KD pupae, the F-actin intensity increased in the fourth cell row away from the wound, but not in the three cell rows closer to the wound margin (Fig. 3Bii, asterisk). The ectopic actin cable in Grh KD typically formed at the boundary between these two tissue regions that exhibited contrasting behaviors. Interestingly, the cells that lie between the ectopic cable and the wound margin are the same cells that show a strong reduction in E-Cad (Fig. S3A). To look in more detail at the actin flow in Grh KD pupae, we analyzed yz sections of the tissue surrounding the wound site during the first 60 mpw. In the wild-type, F-actin concentrated in the apical side of the cell during flow propagation, whereas in Grh KD this polarization was partially lost and F-actin localized not only apically but also along the whole apical-basal axis (Fig. 3A′i,i′, yellow arrowheads).

We detected similar dynamics for non-muscle myosin II (Royou et al., 2002); whereas the wild type presented a flow of apical myosin that propagated from cells away from the wound towards the wound edge (Fig. 3Ci–v; Movie 3), in Grh KD, this flow was stalled in the third cell row (Fig. 3Ci′–iv′,vi, Movie 3). Furthermore, similar to what was seen for F-actin pupae, myosin ectopically accumulated away from the wound in Grh KD, instead of forming the typical myosin cable at the wound edge observed in the wild type (Fig. 3Civ,iv′, arrowheads; Movie 3).

Together, our results show that Grh regulates the dynamics of actin and myosin during repair. Grh is required both for the correct propagation of actin and myosin flows and for the formation of a proper actomyosin cable at the leading edge of the wound.

E-Cad establishes a link between cell adhesion and cytoskeleton components that is essential for tissue-level responses (Martin et al., 2010). Moreover, the regulation of E-Cad dynamics is crucial for maintaining the directionality of intracellular actomyosin flows (Lecuit et al., 2011; Rauzi et al., 2010). To test the hypothesis that the observed E-Cad downregulation phenotype in the Grh KD pupae is linked to the defects in actin flow propagation, we knocked down E-Cad and analyzed F-actin dynamics upon wounding in notum epithelial cells. We reduced E-Cad levels by expressing E-Cad RNAi and found that the propagation of the actin flow was clearly impaired in E-Cad KD cells and the actin cable failed to assemble at the wound edge (Fig. 4A; Movie 4). In addition, the E-Cad KD epithelium was more disorganized and cells exhibited more irregular and rounded shapes than wild-type cells (Fig. 4Ai′, white arrowhead). We also found that, in these pupae, ectopic actin accumulations formed a few cell rows away from the wound edge, although not as strong as in the Grh KD (Fig. 4Aii′–iv′, red arrowheads). At 60 mpw, the E-Cad KD cells at the wound edge seemed to form longer and larger numbers of actin protrusions than in the wild type, which was also seen in other cell rows around the wound, a feature not observed in the wild type (Fig. 4Aiv′, white asterisk and inset). When quantifying the F-actin levels in several cell rows around the wound, it is clear that the actin flow was abolished, as actin intensity levels in the E-Cad KD did not show the characteristic intensity peaks observed in the wild type (Fig. 4B). We also tested the effects of elevated levels of E-Cad, to mimic the observations in grh-depleted tissue before wounding (Fig. 2Ai′), by overexpressing E-Cad in the tissue. The results showed that the overexpression did not affect the epithelial structure of the notum before wounding and did not prevent the actin flow or the formation of the actomyosin cable upon wounding (Fig. S4). These results suggest that, although pre-wound E-Cad intensity levels are higher in Grh KD that in wild type, the rapid loss of E-Cad upon wounding in Grh KD surrounding tissue might be responsible for the defects in F-Actin polymerization propagation during cable formation.

We next investigated how Grh loss-of-function impairs E-Cad dynamics during wounding and, thus, the repair process. It has been shown that, in Drosophila embryos, the downregulation and turnover of E-Cad at the AJs is required for actin polymerization and bundle formation, leading to the assembly of the contractile cable around the wound margin (Carvalho et al., 2014; Hunter et al., 2015; Matsubayashi et al., 2015). Moreover, the Toll-NF-κB signaling pathway is required for E-Cad turnover during wound closure. Namely, the NF-κB transcription factors Dif and Dorsal (Dl), which act as effectors of the pathway, the membrane receptor Toll and the inhibitor Cactus (Cact) have all been implicated in this process. In Dif dl and Toll mutants, as well as in Cact-overexpressing embryos, E-Cad is stabilized at the junctions in cells at the wound edge, thus impairing actin cable formation (Carvalho et al., 2014). This link led us to investigate whether the downregulation of E-Cad in the Grh KD pupae observed upon wounding is related to the misregulation of NF-κB pathway. First, we addressed this at the transcriptional level by quantifying Dif and dl mRNA levels in grh mutant embryos through quantitative real-time PCR. These data showed an increase of dl (but not Dif) expression in the mutants in comparison to wild type (Fig. 5A), suggesting that Grh might promote the downregulation of dl expression. Dl activation involves its translocation from the cytoplasm to the nucleus (Roth et al., 1989; Steward, 1989). Using a dl::GFP line as a reporter (DeLotto et al., 2007), we showed that in a wild-type pupal notum epithelium, its nuclear translocation begins soon after wounding (7 mpw) in some cells (Fig. 5Bii′), later spreading to the whole tissue around the wound (Fig. 5Biii′; Movie 5). This is similar to what has been shown for embryonic stages (Carvalho et al., 2014).

Therefore, taking into account the transcriptional upregulation of dl in grh mutant embryos and its activation upon wounding, it is possible that this factor is overactivated in the Grh KD notum epithelium. This could lead to the defective E-Cad downregulation and impaired F-actin dynamics. To test this, we overexpressed dl in the pupal notum epithelium (Fig. 5C,D; Movie 6). Upon wounding, F-actin dynamics and flow propagation were impaired and ectopic F-actin accumulated further away from the wound margin in these pupae (Fig. 5Cii–iv, red arrowheads), similar to the Grh KD (Fig. 3Ai′–iv′). Plotting the F-actin intensity for each cell row around the wound showed that the characteristic intensity peaks representing the wave propagation were missing (Fig. 5Cv). Regarding E-Cad, its intensity was quickly reduced upon wounding in the cells around the wound (Fig. 5Di–iii, yellow dashed line), between the ectopic actin accumulation spots (white arrowheads) and the wound margin (Fig. 5Div), similar to Grh KD (Fig. 2Ai′–iv′). Quantification of E-Cad intensity levels during wound closure showed a clear decrease in the cells near the wound site when compared to cells further away from the wound (Fig. 5Dv), especially during the first 2–3 mpw.

To further address the impact of dl overexpression and activation in wound closure, we knocked down its known inhibitor, Cact. This IκB homolog inhibits the Toll signaling pathway by binding to Dl, thereby impairing its dimerization and consequently its activation (Belvin et al., 1995; Geisler et al., 1992; Roth et al., 1991). Upon knocking down cact through RNAi expression (Fig. 5E,F), we observed that the actin flow propagation and actin cable formation were impaired (Fig. 5Ei–iv). This was also evident after quantifying F-actin intensities in the cells around the wound along time (Fig. 5Ev). E-Cad dynamics were also affected around the wound (Fig. 5Fi–iii), in a similar trend to what was observed for dl overexpression (compare Fig. 5Fv and Fig. 5Dv), giving rise to two distinct phenotypes between cells closer to and away from the wound.

These data show that the deregulation of Dl affects F-actin and E-Cad dynamics during the wound closure process in a similar manner to the reduction of Grh. This points to a possible regulatory role for Grh in NF-κB signaling during wound healing.

To investigate the long-term consequences of the impairment of E-Cad and actomyosin dynamics on epithelial repair in Grh KD, we tracked wound recovery until 12 h post wounding. Wound closure was not achieved (Fig. S5A), which was expected as the cable plays a critical role in this process (Martin and Lewis, 1992; Wood et al., 2002). Surprisingly, after the ectopic cable is fully established in Grh KD (at ∼60 mpw), cells localized between the cable and the wound margin remained in the tissue but their phenotypic characteristics changed dramatically regarding AJs and cytoskeletal dynamics. One known component that acts as a bridge between AJs and the cytoskeleton is α-catenin (α-cat) (Kobielak and Fuchs, 2004; Rimm et al., 1995), therefore we analyzed α-cat::GFP dynamics in the cells located between the ectopic cable and the wound margin. While in wild type the integrity of the epithelium around the wound was maintained during the entire closure process (Fig. 6A), in the Grh KD cortical α-cat quickly collapsed to the apical center of the cells (Fig. 6B, white arrowheads; Movie 7). This extremely fast behavior was also observed for myosin (Fig. 6C,D; Movie 8) and F-actin (Fig. 6E,F; Movie 9), although for the latter this was not as evident due to the several F-actin populations that exist in each cell. All these events occurred simultaneously between 1 and 3 h post wounding. After the collapse of the cortical components, the basal side of the cells rounded up. Moreover, large sets of filopodia formed at the apical side (Fig. 6Fi–iii, red arrowheads), including in the cells not directly at the wound margin (Fig. 6Fi′–iii′; Movie 9), a feature never observed in the wild type (Fig. 6E). When labeling myosin and F-actin simultaneously, we observed that these filopodia only started to form after the cortical collapse (Fig. S5B; Movie 10). We tested whether these observed phenotypes were recapitulated by Dl overexpression, since in this condition E-Cad dynamics are also compromised around the wound. These experiments revealed that the cortical collapse and subsequent filopodia formation in the cells not directly at the wound edge did not occur upon Dl overexpression (Fig. S5C). Although this experiment does not allow us to measure the degree of overactivation of the NF-κB pathway, these results suggest the Grh phenotypes shown in Fig. 6 are independent of Dl deregulation.

The above data suggest that in Grh KD the cells remaining between the ectopic cable and the wound margin lose the typical characteristics of an epithelial tissue. Epithelial cells, in certain developmental or disease contexts, can partially or fully transition into mesenchymal cells (Baum et al., 2008; Lim and Thiery, 2012). This conserved process, called epithelial-to-mesenchymal transition (EMT), is characterized by the loss of tight epithelial adhesion and polarity, the reorganization of the cytoskeleton and the expression of new regulatory pathways, which can lead to increased cell motility and invasiveness (Greenburg and Hay, 1982; Hay, 1995; Lamouille et al., 2014). Several conserved genes linked to the mesenchymal phenotype have been classically used as markers to detect EMT, not only during developmental processes but also in disease contexts. Many of these genes belong to the conserved Snail family of transcription factors, such as snail (sna), twist (twi), escargot (esg) and worniu (wor). In addition, the upregulation of Matrix metalloproteinase 1 (Mmp1), an extracellular matrix regulator, and N-Cadherin (CadN), an AJ protein, has also been associated with the EMT phenotype in Drosophila (Oda et al., 1998; Stevens and Page-McCaw, 2012; Vicidomini et al., 2015).

To understand whether Grh is required to maintain the epithelial characteristics of the cells surrounding the wound, we investigated whether Grh depletion affects the expression of the EMT-related genes previously mentioned. For this, we quantified their expression in wild-type and grh mutant embryos through real-time quantitative PCR. Although the levels of sna and esg were not significantly altered, CadN, Mmp1, twi and wor showed significantly increased expression in grh mutants when compared to that in wild-type embryos (Fig. 6G). These data show that cells lacking Grh upregulate the expression of genes typical of a mesenchymal phenotype.

Our results indicate that knocking down Grh leads to cortical instability and loss of epithelial identity upon wounding. Cells close to the wound margin lose E-Cad and α-cat at their AJs, which might affect the attachment of the cortical cytoskeleton and induce a mesenchymal-like phenotype. Furthermore, the upregulation of mesenchymal markers in grh embryos supports our hypothesis that Grh is needed to maintain the epithelial phenotype of cells upon injury.

The Grh family of transcription factors has been linked to epithelial repair not only in Drosophila but also in vertebrates (Ting et al., 2005). In Drosophila, Grh function has been mainly linked to the activation of wound response genes necessary for cuticle repair (Mace et al., 2005). Here, we propose a novel function for Grh in the early stages of wound closure, namely, that it regulates cytoskeletal and junctional dynamics. Moreover, we show the importance of this factor in the maintenance of epithelial identity during the repair response.

By knocking down Grh using tissue-specific RNAi, we uncovered the importance of this factor in regulating the cellular dynamics necessary to attain tissue closure in the pupa notum epithelium. Upon wounding, Grh-depleted tissue is incapable of forming the characteristic contractile actomyosin cable at the wound edge and achieve wound closure. Instead, the actomyosin flow does not propagate and leads to formation of cable-like structures at ectopic locations. The location where these ectopic cables form depends on where the actomyosin flow starts. In turn, the place where the actin flow begins depends on the size of the wound, consequent tissue displacement and tension release that it entails (Antunes et al., 2013). Notably, knocking down E-Cad in the notum epithelium leads to a similar phenotype to Grh loss of function. This leads us to hypothesize that the Grh KD phenotype is related to the impairment of E-Cad dynamics in the cells around the wound.

E-Cad transmembrane clusters are central components of AJs that are essential for the communication of forces and signals that enable cells to work collectively. AJs respond to mechanical cues that are transmitted to the interior of the cell through actin-binding proteins, thereby leading to cytoskeletal rearrangements (Martin et al., 2010; Michael and Yap, 2013; Sawyer et al., 2009). These cellular events have consequences at the tissue level, resulting in coordinated cell rearrangements that are at the basis of epithelial morphogenesis, including epithelial repair (Gustafson and Wolpert, 1962; Holtfreter, 1943). In these processes, AJ localization and turnover are very dynamic and their regulation is crucial (Gumbiner, 2005; Lecuit and Lenne, 2007). In fact, it has recently been described that regulated removal of junctional E-Cad from the wound margin is necessary for the formation of the actomyosin cable in the Drosophila embryonic epidermis (Carvalho et al., 2014; Hunter et al., 2015). Moreover, tissue displacement caused by wounding leads to the redistribution of E-Cad in cells around the wound (Antunes et al., 2013). Here, we have confirmed and extended these findings by showing that Grh plays a key role in the regulation of tissue dynamics through E-Cad. Upon wounding, Grh KD cells close to the wound fail to regulate E-Cad levels, leading to an extremely rapid and irreversible loss of this cortical component. Without E-Cad, affected cells possibly lose an important communication channel, becoming unable to transduce the tension response cues necessary for the propagation of the actomyosin flow, formation of the contractile cable and, eventually, wound closure.

Although E-Cad levels severely diminish upon wounding, we should not forget that in the Grh KD, E-Cad appears to be upregulated in unwounded tissue, suggesting that during homeostasis Grh acts as a negative regulator of this protein (Fig. 7Ai). Such regulation is likely to be at the expression level, as grh mutant embryos exhibit an increase of shg mRNA levels. This role of Grh does not come as a surprise as both Drosophila Grh and at least one of its vertebrate homologs (Grhl2) regulate the expression of E-Cad in neuroblasts and different types of epithelia (Almeida and Bray, 2005; Pyrgaki et al., 2011; Werth et al., 2010). Nevertheless, the repression effects that Grh exerts on E-Cad in undamaged epithelia are dramatically reversed during the wound closure process, where Grh becomes essential for the maintenance of E-Cad levels at the junctions. Our current model is that, upon wounding, Grh acts, at least in part, through Dl, the effector of the Toll-NF-κB signaling pathway (Fig. 7Aii). This pathway has been extensively studied in Drosophila and is conserved in vertebrates, with important roles in the initiation of the inflammatory response, morphogenesis and early embryo patterning (Matova and Anderson, 2006; Valanne et al., 2011). More recently, we showed that dl activation (in combination with Dif) is responsible for downregulating E-Cad at the wound margin in Drosophila embryos during the closure process (Carvalho et al., 2014). This seems to occur via a two-tiered mechanism: in the early stages of closure, Dl regulates E-Cad at the post-transcriptional level; in the later stages, Dl represses the transcription of shg, thereby contributing to the sustained E-Cad downregulation at the wound edge. Here, we show that dl expression is increased in grh mutant embryos, suggesting that Grh might repress dl. This is supported by other studies showing that the transcription starting site of dl contains a presumptive binding domain for Grh (Potier et al., 2014), and that the expression of Drosomycin, a bona fide target of Dl, is increased in grh mutant embryos in comparison to wild type (Nevil et al., 2017; Paré et al., 2012). The regulation of Dl by Grh might be especially relevant during wound healing. In agreement with those results, we show that overexpressing Dl or knocking down its inhibitor Cact leads to phenotypes that are similar to those seen upon Grh loss of function regarding both F-actin and E-Cad dynamics. It is possible that the increase of dl transcripts in the Grh KD leads to the hyperactivation of this pathway, upon activation of its membrane receptor Toll upon injury, resulting in the uncontrolled depletion of junctional E-Cad in the cells surrounding the wound. The mechanism by which this depletion occurs is still unknown, but recent studies point to the importance of E-Cad endocytosis during wound healing (Hunter et al., 2015; Matsubayashi et al., 2015). We also show that, both upon Grh loss of function and Dl overexpression, a specific range of cells around the wound is affected in terms of F-actin dynamics and loss of AJs. This might be related to the gradient of tension disruption sensed by the tissue, conveyed by the size of the wound (Antunes et al., 2013). Although our Dl reporter analysis shows that this factor translocates to the nucleus in the tissue surrounding the wound, the effect on junctional E-Cad could obey a signaling threshold that we cannot measure with the currently available tools.

Grh has already been shown to act through specific pathways in a stress response context. Upon injury, Grh is phosphorylated by ERK1/2 and this modification is required for the activation of the wound response genes but not for the homeostatic processes where Grh participates (Kim and McGinnis, 2011). Therefore, it is possible that this phosphorylation is important for the inhibitory interaction between Grh and Dl in the nucleus of the cells affected by a wound, which could explain the rapid control of the E-Cad depletion both in the Grh KD and Dl overexpression. More experiments are necessary to test this link, especially regarding protein–protein interactions, but altogether this work points to a novel regulatory function for Grh in epithelial stress response.

In the absence of Grh, cells located between the ectopic cable and the wound margin remained alive and active, as evident from the formation of filopodia in these cells (Fig. 6F). Moreover, we did not observe effector caspase activation when using the fluorescent biosensor Apoliner (Bardet et al., 2008) (data not shown). These cells lose junctional stability, with the loss of E-cad and α-cat. The latter plays an important role in mechanotransduction and tension sensing through its link to the F-actin cytoskeleton (Buckley et al., 2014; Kobielak and Fuchs, 2004; Rimm et al., 1995; Yonemura et al., 2010). The subsequent collapse of the cortical actomyosin cytoskeleton ring in these cells may be a response to the exacerbated cortical tension promoted by the surrounding tissue. Interestingly, this resembles the phenotype seen in β-catenin (also known as armadillo) mutant embryos during high-tension morphogenetic movements (Martin et al., 2010). Adding to this, the formation of numerous filopodia in these cells after the collapse occurs, as well as the rounding of their basal side, suggests that they could be losing their epithelial identity.

As a master regulator of epithelial determination and maintenance, Grh not only activates the expression of characteristic epithelial components but also transcriptionally represses typical mesenchymal genes. Our results support this idea by showing that, in grh mutant embryos, the expression of some of these factors (twi, wor, Mmp1 and CadN) is upregulated. Notably, these mesenchymal factors are involved in EMT events, both during development and in cancer (Barrallo-Gimeno and Nieto, 2005; Leptin, 1991). Thus, our data suggest that, in the absence of Grh, cells might be more prone to undergo a loss of epithelial identity. A question that remains is how these genes behave in Grh-depleted cells in the notum upon wounding. So far, we have not been able to address this in the notum owing to technical hurdles.

In vertebrates, TFs of the Grhl family, homologs of Grh, have been shown to be important in preventing EMT (Cieply et al., 2013; Farris et al., 2016; Ray and Niswander, 2016) and their misregulation has been observed in different types of tumors (Cieply et al., 2012; Mlacki et al., 2015). It is still not well understood how Grhl TFs regulate these events, but a link to AJs is likely.

Although the link to an EMT-like event in our observed phenotype is still weak, the relevance of EMT to wound healing has already been reported, with the inflammatory response as a key mediator (Lee and Nelson, 2012). In more complex systems, such as vertebrates, prompt inflammation responses are essential for proper wound closure. While Drosophila tissue repair is achieved in a scar-free manner, the inflammatory response in vertebrate systems leads to the onset of fibrosis, which in extreme cases can lead to cancer (Coussens and Werb, 2002; López-Novoa and Nieto, 2009). Interestingly, this process has been linked to NF-κB overactivation, and inhibiting this pathway can diminish the fibrotic reaction (Kuwabara et al., 2006), demonstrating the importance of regulating these pathways.

According to our model, Grh is a negative regulator of E-Cad in notum epithelia in homeostatic conditions (Fig. 7Ai). Upon wounding, Dl is activated during closure and leads to the fast downregulation of E-Cad at the AJs in an early stage of the process (Fig. 7B, full line). The weakening of the epithelial junctions is likely to be necessary to facilitate tissue repair movements and resembles the EMT processes that have been associated with vertebrate skin wound healing. However, this regulation needs to be tuned down quickly as repair proceeds, so that the tissue can regain epithelial integrity. The activation of Grh plays a key role in this process by limiting Dl action at the proper time, acting as a stop signal (Fig. 7Aii). In the absence of the Grh brake, the downregulation of E-Cad by Dl is exacerbated leading to the impairment of the tissue repair process (Fig. 7B, dashed line). The interaction between Grh and Dl may be of general importance for tissue homeostasis. While Dl promotes the tissue movements necessary for repair, Grh prevents an excessive response that might otherwise result in malignant processes. Alternatively, Grh and/or Dl might also regulate factors involved in E-cad dynamics that are already altered upon Grh KD during homeostasis, which could in turn influence the outcome of the wound response, especially taking into account that both Grh and Dl have many described target genes.

In conclusion, our results show that Grh participates in both early and late stages of wound response, regarding cellular components and transcriptional identity. Future studies analyzing the functional interactions between the vertebrate homologs of these TFs (Grhl and NF-κB) might lead to better understanding of how wound healing occurs at the cellular level in more complex epithelia. Grh factors might thus be central regulators of homeostasis in epithelial tissues, not only by controlling cytoskeletal and cell–cell interactions but also by maintaining epithelial identity.

Flies were cultured and maintained on standard conditions. Actin dynamics in notum epithelia were assessed by expressing UAS-mcherry::moesin (Millard and Martin, 2008) under the control of the pnr-GAL4 driver (Calleja et al., 1996) or sqh-utrophin::GFP (Förster and Luschnig, 2012). sqh-sqh::GFP (Royou et al., 2002) was used to visualize myosin levels. To analyze E-Cadherin dynamics, DE-cadherin::mtomato (Huang et al., 2009) or ubi-e-cadherin::GFP flies (Oda and Tsukita, 2001) were crossed with sr-GAL4 (Usui et al., 2004) recombined with UAS-mcherry::moesin.

To knockdown Grh expression, UAS-grh-RNAi KK (Dietzl et al., 2007; Vienna Drosophila Resource Center, Vienna, Austria) and UAS-grh-RNAi VALIUM10 TRiP (Transgenic RNAi Project, Bloomington Drosophila Stock Center, Bloomington, IN) were used. Downregulation of endogenous Grh expression was assessed through the use of a BAC-grh::GFP line (Spokony insertions, Bloomington Drosophila Stock Center, Bloomington, IN, USA).

UAS-shg-RNAi VALIUM 20 TRiP (Transgenic RNAi Project, Bloomington Drosophila Stock Center, Bloomington, IN) was crossed with pnr-GAL4, UAS-mcherry::moesin to analyze actin flow dynamics in the shg/e-cad knockdown background. UAS-shg (Sarpal et al., 2012) was crossed with pnr-GAL4,UAS-mcherry::moesin to analyze actin flow dynamics in a shg/e-cad overexpression background.

dl reporter analysis was performed using the transgenic line dl::GFP (DeLotto et al., 2007) on a Dif dl/+ background (Matova and Anderson, 2006) crossed with pnr-GAL4,UAS-mcherry::moesin. For overexpression of Dl and Cact knockdown, UAS-dl (Matova and Anderson, 2006) and UAS-cact-RNAi VALIUM20 TRiP (Transgenic RNAi Project, Bloomington Drosophila Stock Center, Bloomington, IN), respectively, were crossed with pnr-GAL4, UAS-mcherry::moesin for F-actin flow analysis and w; ubi-e-cad::GFP; sr-GAL4, UAS-mcherry::moesin for E-Cad dynamics analysis.

Transheterozygote embryos resulting from two P-element insertions, grh⁰⁶⁸⁵⁰ and grh^s2140 (Hemphälä et al., 2003; Spradling et al., 1999) were used for gene expression analysis in grh mutants.

Due to lethality of phenotypes in some experiments, pnr-GAL4 was replaced by sr-GAL4, which shows a smaller expression domain and weaker expression levels, allowing the survival of the organism.

Crosses were performed at 22°C (crosses with UAS-shg-RNAi and dl::GFP on a Dif dl/+ background), 25°C (crosses with UAS-grh-RNAi) or at 29°C (crosses with UAS-shg, UAS-dl and UAS-cact-RNAi) to maximize RNAi expression and viability using the UAS/GAL4 system (Brand and Perrimon, 1993).

Pupae were staged at 13 h after puparium formation (APF) (Bainbridge and Bownes, 1981), mounted and prepared for imaging as previously described (Antunes et al., 2013). Laser wounding and live imaging were performed as previously described (Antunes et al., 2013). Briefly, wounds were performed using a MicroPoint (Andor Technology, Belfast, UK) UV nitrogen-pumped dye laser (435 nm). Live imaging was performed using a Nikon/Andor Revolution spinning disk system with an EMCCD Camera (iXon 897; Andor Technology, Belfast, UK) at 25°C, using the iQ software (Andor Technology, Belfast, UK). Z-stacks were acquired with a Plan Apo VC PFS 60× objective (NA 1.40; Nikon) every: 10 s for F-actin and Myosin movies; 30 s for E-Cad and F-actin, and F-actin and Myosin movies; or every 60 s for dl::GFP movies.

Raw data was processed using Fiji (Schindelin et al., 2012) to obtain maximum intensity projections and orthogonal views. Kymographs were made using a 1-pixel width line. To quantify F-actin and Myosin fluorescence intensity over time for actomyosin flow analysis, four cell rows close to the leading edge were analyzed. The average intensity measurements from all cells in each cell row were obtained, according to their distance from the wound margin (four sequential circumferential areas), along the different time points, normalized to average values before wounding and then averaged between movies. To measure E-Cad fluorescence intensity along time, 10 cells from each row (for each movie) were selected and cortical intensity plots were obtained. For cortical levels analysis, a segmented line defined as the outline of the cell with a 3-pixel width was used to extract average intensity levels along this line. The same was performed for quantification of cortical myosin levels. For myosin medial intensity levels, an ellipsoid region of ∼70 pixels in area was used to extract the average intensity levels at the center of each cell. Relative intensity values along time were obtained by normalizing each time point to average values before wounding. All quantifications were performed manually using Fiji.

Wild-type (w1118) and grh embryos at stage 16 were collected and dechorionated. RNA extraction was performed with the RNeasy Plus Micro Kit (Qiagen, Hilden, Germany). Total RNA (1 μg) was reverse-transcribed using the Transcriptor High Fidelity cDNA Synthesis kit (Roche, Basel, Switzerland). Quantitative real-time PCR (qPCR) was performed with the Fast Green Master Mix (Roche, Basel, Switzerland) and the Roche LightCycler 480 (Roche, Basel, Switzerland). Normalization of gene expression values was done regarding the βTub56 housekeeping gene expression levels and the fold change was calculated through the ΔΔCT method (Livak and Schmittgen, 2001). Dopa decarboxilase (Ddc), a known transcriptional target of Grh (Bray et al., 1989), and grh were used as negative controls for the experiment.

Four biological replicates (50 embryos each) and two technical replicates were analyzed for each condition. Primers were as follows: βTub56 Fwd-5′-CGGCCAACTGAACGCTGAT-3′; βTub56 Rev-5′-GCGGCCATCATGTTCTTGG-3′; shg Fwd-5′-CGACCTTGGCAGGCGACTA-3′; shg Rev-5′-GGTCGACGGCCTCTACGG-3′; grh Fwd-5′-CCTCACCGTTCCTGCTCCA-3′; grh Rev-5′-GTCATCCTGCCGCGCTTC-3′; ddc Fwd-5′-GCCCACATCCGCAGACAC-3′; ddc Rev-5′-GGCCGCGTCCATTGAT TC-3′; dl Fwd-5′-CGGCAGGAAAGGCGCTAA-3′; dl Rev-5′-GCAGCCCTCCTTGCCAAC-3′; dif Fwd-5′-CCCGAGCCAAGTGCAAGA-3′; dif Rev-5′-GCTCGTCGCTCGTCACACA-3′; sna Fwd-5′-CAGCAGCGGTGCCAGTGTA-3′; sna Rev-5′-ACAGGCCCATCGAGGTGG-3′; esg Fwd-5′-CATGGCTGCTGCAAGGACA-3′; esg Rev-5′-TCCGCCCTCTGAATGCTTG-3′; CadN Fwd-5′-TCGTTCGCGATCGGACTG-3′; CadN Rev-5′-GCCGCCCACGATTGAAAG-3′; Mmp1 Fwd-5′-CACCCGTTTCCACCACCAC-3′; Mmp1 Rev-5′-AGTCCCGCGAAGCTTTGG-3′; Twi Fwd-5′-TGTCGCCCAAAGTGCTGC-3′; Twi Rev-5′-GCGTACTGGGCATGCTGC-3′; wor Fwd-5′-GGGAGTGCCGCGATACCA-3′; wor Rev-5′-AGGAGCCATGGTCGCGA-3′.

Statistical analysis was performed using Prism (GraphPad Software, La Jolla, CA). Error bars in graphs correspond to the standard deviation (s.d.) or standard error of the mean (s.e.m.), as stated in figure legends.

We would like to thank A. Bensimon-Brito for advice on qPCR analysis and critical manuscript review; Y. Bellaiche for helpful discussion and manuscript review; A. Bensimon-Brito, M. Antunes, A. Brandão and F. Vasconcelos for helpful discussions; T. Pereira for advice on data analysis; J. Lima for advice on statistical analysis; R. DeLotto for the dl::GFP transgene; Y. Bellaiche for the DE-Cad::tomato transgene; TRiP at Harvard Medical School (NIH/NIGMS R01-GM084947) and VDRC for transgenic RNAi fly stocks and the Bloomington Stock Center for other Drosophila lines used in this study.