Src kinases and ERK activate distinct responses to Stitcher receptor tyrosine kinase signaling during wound healing in Drosophila

Epidermal injury induces a coordinated array of reactions involving epithelial, mesenchymal, endothelial and blood cells that reconstruct the damaged site. An understanding of wound responses in each cell type, and of the pathways that orchestrate the activities of different cells, is expected to advance the development of new treatments for the multiple pathologies caused by defective wound healing (Gurtner et al., 2008). Drosophila offers a tractable model for the molecular dissection of wound responses in vivo (Galko and Krasnow, 2004; Razzell et al., 2011). Epidermal cells in wounded embryos rapidly migrate and assemble an actin-rich ring around the wound that constricts and closes the gap. Wounding also initiates inflammation and activates ERK to induce the expression of barrier-repair genes in epithelial cells at the wound site (Mace et al., 2005; Pearson et al., 2009; Wood et al., 2006). The evolutionarily conserved transcription factor Grainy head (Grh) is phosphorylated by ERK and lies at the heart of wound healing (Kim and McGinnis, 2011). It activates the barrier-repair genes Dopa decarboxylase (Ddc) and pale (ple, encoding tyrosine hydroxylase) in Drosophila and transglutaminase-1 in mice (Ting et al., 2005; Wang and Samakovlis, 2012). Stitcher (stit, also known as Cad96Ca) encodes a Ret-family receptor tyrosine kinase (RTK), which is also activated by Grh following wounding. Live-imaging analysis of wounded stit embryos reveals that Stit promotes actin-cable assembly during re-epithelialization after wounding. Stit also triggers ERK activation at the injury site, leading to a Grh-dependent upregulation of wound-repair genes and of its own transcription. Thus, Stit activation is crucial for the activation of epidermal wound responses in embryos (Wang et al., 2009). Additionally, embryos with mutations in the genes encoding either the Drosophila epidermal growth factor receptor (EGFR^t1) or the ERK homolog rolled^10a show defective wound closure (Geiger et al., 2011). In Drosophila larvae, the Platelet-derived growth factor and vascular endothelial growth factor (PDGF and VEFG) receptor (Pvr) is required for wound closure. This RTK becomes activated by its ligand Pvf1 upon epithelial disruption and induces actin polymerization and wound closure. Thus, genetic analysis in flies has identified three different RTKs that are required for epithelial-wound closure at different developmental stages. Numerous signal transducers including Ca²⁺ (Razzell et al., 2013), the Rho family of GTPases (Wood et al., 2002), ERK (Geiger et al., 2011; Mace et al., 2005) and JNK (Galko and Krasnow, 2004; Rämet et al., 2002) orchestrate the activation of transcriptional and cytoskeletal wound responses. Recent genetic screens for mutants with defective wound closure have rapidly expanded the catalog of new signaling molecules and effectors of wound healing in flies (Campos et al., 2010; Juarez et al., 2011; Lesch et al., 2010). However, a mechanistic view of the links and regulatory relationships between the different components of wound-healing responses is still missing.

RTK signaling plays prominent roles during wound re-epithelialization in vertebrates as well as in Drosophila. Several RTKs become activated after injury to control epithelial migration and wound closure (Müller et al., 2012). Conditional inactivation of hepatocyte growth factor (HGF, also known as scatter factor) receptor (c-Met) in keratinocytes abolishes the migratory capacity of these cells during wound re-epithelialization. This is accompanied by a decrease in the phosphorylation of ERK and other kinases, and a failure in cytoskeletal rearrangement in response to HGF (Chmielowiec et al., 2007). Simultaneous inactivation of two other RTKs, the fibroblast growth factor receptors FGFR1 (isoform IIIb) and FGFR2 (isoform IIIb) also delays wound healing and keratinocyte migration. In this case, cell adhesion and the expression of focal adhesion kinase and paxillin were reduced in the mutants (Meyer et al., 2012). Recent studies in zebrafish revealed a direct role for Fynb [an intracellular transducer of the Src family kinases (SFKs)] and Ca²⁺ signaling in the early events of regeneration of amputated fins. In this system, ERK activation is required in parallel with Fynb for regeneration, and the two signal transducers might share downstream effectors (Yoo et al., 2012). Collectively, research in several model organisms has identified several RTKs and signal transducers that control epithelial wound responses. To understand the mechanisms of RTK functions in epidermal wound healing, we dissected Stit signaling and its signal transducers after wounding in Drosophila embryos.

RTKs become auto-phosphorylated upon ligand-induced dimerization, and these phosphorylation events are necessary for the activation of downstream effectors (Lemmon and Schlessinger, 2010). We searched for putative tyrosine (Y) phosphorylation sites in the intracellular domain of Stitcher (http://www.cbs.dtu.dk/services/NetPhos/). Four of the six predicted phospho-tyrosine (pY) residues resided in the kinase domain and two were in the C-terminal tail domain (Fig. 1A). Sequence alignments of Stit-like proteins from multiple species revealed the conservation of five of these six residues across a range of insect species (supplementary material Fig. S1A). To investigate whether these residues become auto-phosphorylated upon Stit overexpression, we generated V5-tagged Stit constructs by either replacing each single tyrosine residue with a non-phosphorylatable phenylalanine (F), by altering the four residues in the kinase domain (Stit^{Kinase Quad}) or by replacing all six tyrosine residues (Stit^Hex) (Fig. 1B). We overexpressed these constructs along with a wild-type and a kinase-dead form of Stit [Stit^KD, in which lysine 504 is replaced with alanine (K504A)] (Wang et al., 2009) in S2 cells. Following immunoprecipitation of the Stit proteins with anti-V5 antibodies, we probed western blots with an anti-pY antibody to assess the phosphorylation levels of each protein construct. As expected, wild-type Stit became heavily phosphorylated when it was overexpressed, and the pY signal was abolished in the Stit^KD mutant form, indicating that Stit kinase activity is essential for auto-phosphorylation of the tyrosine residues of the intracellular domain. Whereas single tyrosine substitutions did not affect pY signal intensity (data not shown), we detected a severe reduction in the signal in Stit^{Kinase Quad}. This suggested that tyrosine residues in both the kinase domain and in the cytoplasmic tail are phosphorylated. The phosphorylation level of Stit^Hex was much reduced compared with the Stit^{Kinase Quad} control, and it was comparable to the kinase-dead control (Fig. 1C). This indicates that the Y751 and Y762 residues in the tail domain are phosphorylated upon Stit activation by overexpression. In summary, this analysis of the constructs indicates that the six predicted pYs account for most, if not all, Stit phosphorylation sites.

To identify Stit signal transducers, we performed a yeast two-hybrid screen using the intracellular domain of Stit as bait. By screening a prey library of 72 million clones, we identified 185 interacting clones. We focused on four unique binding partners; Drk (Downstream of receptor kinase), Src42A, Src64B and Btk29A. Each of these possessed a Src homology 2 (SH2) domain that specifically recognizes pY (Sadowski et al., 1986), and each was represented by multiple interacting clones (supplementary material Fig. S1B). Drk, a homolog of mammalian GRB2, contains an SH2 and two SH3 domains that couple activated RTKs to the Ras guanidine exchange factor Sos (Olivier et al., 1993), which further activates the Ras1–MAPK pathway in flies. Src42A and Src64B are the closest Drosophila relatives of the Src oncogene product, the defining member of the SH2 and SH3 family of proteins (Takahashi et al., 1996). The gene btk29A (also known as tec29) also encodes non-receptor tyrosine kinases (Lu et al., 2004) (supplementary material Fig. S1B). Members of the Src family and Btk29A often act together to regulate epithelial proliferation and morphogenesis (Shindo et al., 2008; Tateno et al., 2000; Vidal et al., 2007; Wouda et al., 2008).

In another assay to investigate whether the putative Stit effectors bind to phospho-Stit, we produced GST-tagged versions of the SH2 domains of Drk, Src42A and Src64B in bacteria, and tested their binding to V5-tagged Stit versions produced in S2 cells. All three GST–SH2 fusions bound to wild-type Stit but only showed very weak association with the inactive Stit^KD form (Fig. 2A–C). This indicates that the three effectors selectively bind to phosphorylated Stit.

We further tested which tyrosine residue mediates Stit binding to the SH2 domains of the three binding partners. We produced six V5-tagged Stit versions, each replacing one of the six predicted pYs, and one construct substituting phenylalanine residues for the two C-terminal tail pYs. All Stit versions apart from the ones lacking kinase activity (Stit^KD) or replacing Y⁷⁶² (Stit^Y762F and Stit^{Y751F, Y762F}) showed strong binding to the SH2-domains of Drk, Src42A and Src64B (Fig. 2A–C). This indicates that Y⁷⁶² is crucial for Stit binding to Drk and to SFKs. We next tested the significance of Y⁷⁶² for Stit function in vivo. Mutants that are null for stit die at late pupal stages. This phenotype can be partially rescued by transgenic expression of wild-type Stit, but not by Stit^KD, in ectodermal tissues of stit mutants (Wang et al., 2009). We generated stit mutants expressing similar levels of either wild-type Stit, or Stit^KD or Stit^Y762F transgenes (Fig. 2E), and scored adult survival. Whereas the wild-type Stit enabled survival in 39% of stit mutants, the Stit^Y762F version only rescued lethality in 6% of the mutants (Fig. 2D). Thus, Y⁷⁶² provides the major binding site for Stit effectors and has a crucial function in Stit signaling in vivo.

stit mutants show reduced diphosphorylated (dp)ERK accumulation at the wound site, and Stit overexpression induces ectopic ERK activation (Wang et al., 2009). To test the requirement for Y⁷⁶² in ERK activation, we compared the abilities of Stit and Stit^Y762F to induce dpERK accumulation. Upon overexpression in epidermal stripes, Stit^Y762F only induced a weak accumulation of dpERK in the expressing cells compared with the robust increase in the dpERK signal induced by Stit (Fig. 3A–B″). This indicates that Y762 is crucial for ERK activation. ERK phosphorylation following wounding leads to transcriptional activation of wound reporter genes. To investigate the impact of Y762 on the activation of the 1.4-kb Ddc1.4-GFP transcriptional reporter (Mace et al., 2005), we expressed Stit^Y762F and a Stit^Y751F version, which retains the ability to bind to Drk and Src. Consistent with its failure to activate ERK, overexpression of Stit^Y762F did not activate Ddc1.4-GFP, whereas overexpression of Stit^Y751F readily induced it (Fig. 3C,C′). Thus, Y⁷⁶² has a key role in ERK phosphorylation and the activation of the reporter.

The Drk adaptor binds to phosphorylated RTKs and activates the Ras–Raf–Mek pathway, which culminates in the dual phosphorylation of ERK (Gabay et al., 1997). We hypothesized that Drk binding to Y⁷⁶² on Stit leads to ERK phosphorylation and the transcriptional induction of injury repair genes. We first tested whether drk^eOA mutants (Simon et al., 1991) could activate Ddc1.4-GFP following wounding. This reporter faithfully reflects the rapid accumulation of Ddc transcripts at the wound sites of wild-type embryos (Mace et al., 2005). However, both Ddc mRNA and the Ddc1.4-GFP reporter were also expressed in the epidermis of unwounded stage 17 embryos, reflecting the developmental maturation of the epidermal barrier (supplementary material Fig. S2A–C; Movie 1). To elucidate the wound-response-specific roles of Drk and the other Stit effectors, we performed our injury assays during a defined interval prior to the developmental activation of Ddc1.4-GFP (supplementary material Fig. S2C,D; Movie 2). Following injury, 82% of the control embryos showed a broad strong Ddc1.4-GFP induction around the wound site. Only 17% of drk^eOA mutants showed comparable wound-induced Ddc1.4-GFP activation. Moreover, 39% of drk^eOA mutants showed a very weak induction of the reporter as opposed to 3% of control embryos (Fig. 4A–C). This weak activation of Ddc1.4-GFP after injury in drk^eOA mutants resembles the stit mutant phenotypes (Wang et al., 2009), and suggests that Drk is required downstream of Stit for the full activation of gene expression upon epidermal wounding. We further investigated the potential requirement of Drk in the transcriptional activation of Ddc by Stit signaling. We overexpressed Stit by using the enGal4 driver, and detected Ddc transcripts by in situ hybridization. Stit overexpression induced ectopic Ddc mRNA accumulation in epidermal stripes and in regions of the hindgut compared with wild-type embryos at stage 16. This ectopic Ddc expression was not detected in drk^eOA mutants overexpressing stit by the same driver (Fig. 4D), indicating that Drk is required downstream of Stit for the activation of wound-response genes.

To investigate whether Drk is also required for Stit signaling leading to re-epithelialization, we recorded wound closure in wild-type and drk^eOA mutant embryos expressing the spaghetti-squash-Moesin–GFP (sGMCA) reporter (Kiehart et al., 2000). We inflicted laser wounds to stage-15 control and drk^eOA embryos and monitored their closure. The epidermal cells in both the control and drk^eOA mutant embryos readily extended over the gap and closed it. Wound closure was completed in average within 161.5 min in wild-type embryos (n = 18) and within 176 min in drk^eOA mutants (n = 13) (supplementary material Fig. S3A). This minor delay in re-epithelialization contrasts with the notable reduction of Ddc1.4-GFP activation in drk^eOA mutants and suggests that Drk binding to activated Stit predominantly induces transcriptional activation of wound-response genes without influencing the cytoskeleton.

To elucidate how Stit signaling controls multiple responses in epidermal cells following wounding, we investigated the function of its remaining binding partners. We visualized wound closure in laser-wounded wild-type and src42A^E1 mutant embryos (Laberge et al., 2005), by using the sGMCA reporter at stage 15. Whereas all control embryos closed their wounds within ∼141 min, 19% of src42A^E1 mutants failed to heal within the 8-h period of imaging and the remaining src42A^E1 mutants showed delayed re-epithelialization, with a mean wound-closure time of 202 min (Fig. 5A). This suggests that Src42A is required for efficient wound re-epithelialization. By contrast, wounded btk29A²⁰⁶ and src64B single mutants expressing the sGMCA reporter repaired their epidermal wounds as efficiently as control embryos treated in parallel (supplementary material Fig. S3B,C). To assess whether the loss of Btk29A or Src64B could exacerbate the wound-closure defects of src42A^E1 mutants, we generated src42A^E1 btk29A²⁰⁶ and src42A^E1 src64B double mutants expressing the sGMCA marker. To avoid indirect effects caused by the developmental role of Src-like kinases in dorsal closure, we inflicted wounds in the ventral epidermis, the integrity of which is not affected in the double mutants (Tateno et al., 2000). Whereas 93% of wild-type embryos completely closed their wound within 166 min, only 40% of src42A^E1 btk29A²⁰⁶ double mutants managed to do so (Fig. 5A′). The average closure time for the remaining src42A^E1 btk29A²⁰⁶ mutants was 395 min, twice as long as control embryos. Similarly, in src42A^E1 src64B double mutants expressing the sGMCA marker, 50% of the embryos failed to close their wounds and the remaining mutant embryos showed a significant delay in wound closure compared with the control (supplementary material Fig. S3D). This suggested that Src42A, Src64B and Btk29A collectively mediate epidermal wound closure. To investigate which aspect of re-epithelialization requires the function of non-receptor tyrosine kinases, we imaged wound closure in control, src42A^E1, src42A^E1 src64B and src42A^E1 btk29A²⁰⁶ embryos by confocal microscopy. We did not detect gross differences in the cellular protrusions of the leading cells at the wound edges between the mutant and control embryos. However, the actin ring surrounding and closing the wound in wild-type embryos was discontinuous in src42A^E1 mutants and failed to form in the src42A^E1 btk29A²⁰⁶ and src42A^E1 src64B double mutants (Fig. 5B–D; supplementary material Fig. S3E,E′; Movies 3,4). We compared the relative fluorescence intensity (RFI) of sGMCA at the wound perimeter at different timepoints during wound healing in control, src42A^E1, src42A^E1 btk29A²⁰⁶ and src42A^E1 src64B mutants. The relative signal intensity of the actin-filament marker around the wound perimeter rose sharply during the entire process, both in controls and src42A^E1, but failed to reach its peak in src42A^E1 mutants. By contrast, the actin-cable RFI in double mutants remained flat during the entire wound-healing process (Fig. 5E). This suggested that Src42A, Src64B and Btk29A act collectively to establish and maintain the actin ring at the wound perimeter. We conclude that the SFKs bind to phosphorylated Stit and are required, like Stit (Wang et al., 2009), for intercellular actin-ring formation and re-epithelialization following wounding.

We further tested whether Src kinases are also required downstream of Stit for transcriptional activation of wound-response genes. We compared the induction of the Ddc1.4-GFP and ple-DsRed transcriptional wound reporters in wild-type and src42A^E1 mutant embryos at 3 h post-injury. Wounding induced a broad strong Ddc1.4-GFP response in 75% of wild-type embryos. The remainder of these embryos showed a moderate increase in GFP in the cells surrounding the injury site (Fig. 6A,C). However, src42A^E1 mutants showed a reduced GFP induction (Fig. 6A′). Only 23% of the mutants showed strong and broad GFP accumulation at the wound site, whereas 37% showed a moderate induction and 40% displayed a weak and limited GFP accumulation at the wound (Fig. 6C). In contrast to src42A mutant embryos, src64B mutants did not show any significant reduction in the activation of the Ddc1.4-GFP reporter after wounding compared with control embryos (supplementary material Fig. S3F). This suggested that Src64B is dispensable for the activation of Ddc upon wounding. src42A^E1 mutants also showed a defect in the activation of ple-DsRed following injury (Fig. 6B,B′,D), suggesting that Src42A is required for the transcriptional induction of several barrier-repair genes after injury. To extend the analysis of the role of Src42A in gene activation downstream of Stit, beyond the level of wound reporters, we overexpressed stit in epidermal stripes in src42A mutants and wild-type control embryos, and analyzed the expression of Ddc mRNA. Control embryos expressing Stit under the control of enGAL4 showed the anticipated ectopic accumulation of Ddc mRNA in epidermal stripes at late stage 16. However, src42A^E1 mutants expressing stit under the control of the same driver did not show any ectopic Ddc expression (Fig. 6E–E″). This indicates that Src42A is required downstream of Stit for Ddc activation. Interestingly, we observed that src42A^E1 embryos at late stage 16 showed conspicuous patches of Ddc-expressing cells in the dorsal epidermis (Fig. 6E′″). This ectopic activation of Ddc in src42A^E1 mutants is Stit-independent, because it was also detected in src42A^E1 mutant embryos lacking the UAS-stit transgene. In 80% of unwounded src42A^E1 mutants (n = 34), there was widespread Ddc mRNA signal in the dorsal epidermis around the site of the failed dorsal closure at late stage 16 (supplementary material Fig. S4A,B). Consistently, the Ddc1.4-GFP marker became upregulated in the dorsal epidermis of src42A^E1 embryos at late stage 16 (supplementary material Fig. S4C,D). This suggests an additional developmental role of Src42A in restricting the epidermal expression of Ddc.

We further analyzed the activation of Ddc1.4-GFP in src42A^E1 btk29A²⁰⁶ double mutants to investigate whether Btk29A also contributes to the activation of wound-response genes following injury. Consistent with the developmental roles of Src42A and Btk29A in the embryonic epidermis (Tateno et al., 2000), the expression of the Ddc1.4-GFP marker was increased in the dorsal epidermis of unwounded src42A^E1 btk29A²⁰⁶ double mutants (supplementary material Fig. S4E). However, we did not detect any further reduction in GFP activation following wounding in the ventral epidermis of the double src42A^E1 btk29A²⁰⁶ mutants compared with the decrease observed in src42A^E1 embryos (supplementary material Fig. S4F–G). This suggests that among the non-receptor tyrosine kinases, Src42A is the prominent activator of Ddc expression after wounding. Collectively, our analysis suggests that Src42A and Btk29A have distinct effects on Ddc expression during epidermal morphogenesis and wound healing. During late embryogenesis, they are both required to restrict the developmental expression of Ddc independently of Stit signaling and injury. Following injury or Stit activation, Src42A is required to activate wound-response genes at the wound site.

The function of Src42A in the initiation of gene expression after wounding prompted us to investigate the transcription factors downstream of Src. The activity of Src kinases is controlled by phosphorylation. Csk phosphorylates Src on Y⁵²⁷, leading to an intermolecular interaction that keeps Src in an auto-inhibitory conformation. The substitution of a phenylalanine residue for Y⁵²⁷ in Src42A generates a constitutively active form (Src42A^CA), because it relieves this negative regulation (Nada et al., 1991). We expressed Src42A^CA either in lateral epidermal stripes using en-GAL4, or in the dorsal epidermis using pnr-GAL4, and asked whether Src42A^CA is sufficient to induce the transcriptional activation of wound reporters. Both the barrier-repair marker Ddc1.4-GFP and the stitγ wound reporter (Wang et al., 2009) were readily activated in Src42A^CA-expressing cells in the epidermis of unwounded embryos (Fig. 7A,B–C). This indicates that Src42A activation is sufficient to induce wound-response genes. We then asked whether Grh mediates the transcriptional activation of wound reporters by Src42A^CA. In contrast to control embryos, grh³⁷ mutants expressing Src42A^CA in epidermal stripes failed to induce an ectopic GFP signal (Fig. 7A,A′), suggesting that Grh is required downstream of Src42A for Ddc activation. Similarly, the stitδ reporter, which is identical to stitγ but lacks the Grh-binding sites, was not activated by Src42A^CA expression (Fig. 7C′). Collectively, these results indicate that Src42A^CA induces transcriptional activation of wound reporters by activating Grh. Grh is phosphorylated by ERK, and this modification is necessary for its function following wounding (Kim and McGinnis, 2011). Therefore, we tested the ability of Src42A^CA to activate ERK in the embryonic epidermis. Overexpression of either Src42A^CA or stit in epidermal stripes readily activated the Ddc1.4-GFP reporter but, in contrast to Stit, Src42A^CA failed to induce any detectable dpERK accumulation (Fig. 7D–D″). This suggests that Src activates Grh and the transcription of wound-response genes through an ERK-independent mechanism, presumably involving JNK or other members of the p38 family of kinases (Davis et al., 2008; Sekine et al., 2011; Tateno et al., 2000).

Our analysis highlights the crucial role of a single pY in the control of several aspects of wound healing by the Stit RTK. Four effectors bind to pY⁷⁶² to activate multiple parallel intracellular pathways, each controlling distinct aspects of wound healing. Similarly, the SH2 domains of both Grb2 and Src42A bind to the same pY of ErbB (Schulze et al., 2005), suggesting that this method of activating parallel pathways by docking several effectors to the same site is common to several RTKs.

Drk is an adaptor protein with one SH2 and two SH3 domains. In flies, it was first identified and characterized in the context of signaling from the Sevenless receptor tyrosine kinase, where it links the phosphorylated receptor at the plasma membrane (by its SH2 domain) to the Ras guanine-exchange-factor Sos (by its SH3 domain). Ras activation then leads to the activation of a phosphorylation cascade of MAPK kinase (Raf) and Rolled, the Drosophila homolog of ERK and MAPK (Biggs et al., 1994; Olivier et al., 1993; Raabe et al., 1995; Simon et al., 1991). Based on our analysis and on previously published work, we propose that the predominant role of Drk in wound healing is to mediate ERK activation and the transcriptional induction of wound-response genes (Fig. 8). In the context of wounding, Stit mutants show reduced dpERK activation, and overexpression of Stit induces ectopic ERK accumulation (Wang et al., 2009). Overexpression of Stit induces dpERK accumulation but embryos expressing Stit^Y762F, which does not bind to Drk (or to the other effectors), showed strongly diminished dpERK accumulation. Because the overexpression of the constitutively active version of Src42A can induce the transcriptional activation of wound-response genes without inducing detectable accumulation of dpERK, we suggest that Drk is the major linker between Stit and ERK. ERK phosphorylates Grh in vitro, and this phosphorylation is important for the Grh-dependent activation of wound reporters (Kim and McGinnis, 2011). This suggests that the binding of Drk to phosphorylated Stit induces ERK-mediated Grh phosphorylation and Ddc activation. Drk might activate an additional pathway downstream of Stit. The SH3 domains of Drk could recruit other downstream effectors, such as Disabled (Le and Simon, 1998) or Cbl (Wang and Pai, 2011), to fulfill the functions of Drk upon Stit activation.

In addition to the anticipated roles of dpERK in gene activation following wounding, we reveal a key function of SFKs in the repair of epidermal injuries. The activation of SFKs has been investigated in several contexts. A standard model involves the de-phosphorylation of Y⁵²⁷ of chicken c-src, which unlocks the autoinhibitory interaction and allows trans-auto-phosphorylation (Yeatman, 2004). More recently, cysteine oxidation by reactive oxygen species has emerged as a mechanism for Src activation in epithelial cells and leukocytes during wound healing in zebrafish (Yoo et al., 2012). Our results reveal an additional RTK-dependent mechanism for the immediate SFK activation at the wound site. Phosphorylation of Y⁷⁶² of Stit might provide a high-affinity binding site for the SH2 domain of Src kinases, which would therefore compete with the autoinhibitory interaction of the SH2 domain with Y⁵¹¹ in Src42A. This would lead to full activation either by autophosphorylation or by trans-phosphorylation by Stit (Fig. 8). This is the mechanism by which Src is recruited to and activated on the docking site of the PDGF receptor in porcine aortic endothelial cells (Mori et al., 1993).

Following wounding or Stit activation, SFKs control both the local induction of wound-response genes by the conserved transcription factor Grh, and the assembly of a distinctive actin cable around the wound edge. Interestingly, Src42A is the predominant player in transcriptional activation. During late embryonic development, Src42A restricts Ddc expression in epidermal cells. However, following injury, it is required for the local induction of Ddc. Additionally, Src42A^CA can ectopically activate wound-response genes without causing any detectable accumulation of dpERK. This suggests a novel dpERK-independent mechanism of Grh activation during wound healing. Our results contrast with the observations of Juarez and colleagues (Juarez et al., 2011), who reported that injury of src42A mutant embryos results in widespread activation of wound-response genes along the entire epidermis. There are several explanations for this discrepancy. First, Juarez et al. (Juarez et al., 2011), recorded the GFP intensity of Ddc.47, a minimal wound enhancer reporter presumably lacking the cis-elements that drive the developmental expression of Ddc in the epidermis and other tissues (Mace et al., 2005). Second, the widespread activation of Ddc.47 was recorded several hours after the interval of our assays. Thus, our analysis reveals the immediate role of Src42A in the activation of wound-response genes at the injury site, whereas the results of Juarez et al. (Juarez et al., 2011) might suggest a later, direct or indirect, role of Src42A in restricting the spread of the response. The different mechanisms of Src activation by phosphorylation or cysteine oxidation might provide a clue as to how Src42A can act both as a repressor of epidermal Ddc expression during development and as an activator of Ddc and other genes upon injury. Basal low levels of Src42A phosphorylation in the epidermis might favor Ddc repression, whereas increased levels of p-Src by Stit or other RTKs following injury might lead to the Grh-dependent activation of wound-response genes.

We found that re-epithelialization after injury is controlled by all three non-RTKs, because double mutants show more pronounced defects in wound closure than single src42A^E1 mutants. The assembly of the actin ring requires the coordination of the cytoskeleton across the membranes of the epithelial cells surrounding the wound edge. Because Src42A has been implicated in the control of E-cadherin trafficking, it is tempting to speculate that its role in re-epithelialization is to control the adhesion of the leading cells, thereby controlling wound constriction (Abreu-Blanco et al., 2012; Förster and Luschnig, 2012; Nelson et al., 2012). By contrast, Btk29A and Src64B control the growth of the actin-rich ovarian ring canals (Lu et al., 2004) and microfilament contraction during cellularization (Thomas and Wieschaus, 2004), suggesting that they might preferentially control actin-filament assembly and contraction at the wound edge. Besides Stit and its effectors, the Rho GTPase and profilin and Karst (a Drosophila β-heavy spectrin homolog) have also been shown to participate in the formation of a continuous actin cable (Brock et al., 2012; Campos et al., 2010; Wood et al., 2002), suggesting that they might represent downstream targets of Src and Btk29A during re-epithelialization. Our analysis reveals distinct roles for RTK effectors in wound-healing responses, and provides a molecular framework towards understanding and manipulating RTK signaling during wound healing.

Crosses were performed at 25°C on standard medium unless indicated otherwise. The w¹¹¹⁸ strain was used as the wild-type control. We expressed the actin-binding domain of Moesin fused to GFP under the control of the sqh (myosin II regulatory light chain) promoter (sGMCA) (Kiehart et al., 2000) to examine both epithelial repair and actin-cable assembly after wounding. The UAS-stit^Y751F and UAS-stit^Y762F strains were generated by cloning the mutated stit versions into the pUAST vector, followed by standard P-element-mediated transformation. The mutant alleles used were: src42A^E1, src64B, drk^eOA, btk^k206 and grh^B37. en-GAL4 (Bloomington Stock Center) and pnr-GAL4 (Calleja et al., 2000) were used to drive UAS-Src42A^CA (Kyoto Stock Center). The 2-kb stitγ-GFP, 2-kb stitδ-GFP, UAS-stit, UAS-stit^K504A, UAS-stit^Y751F and UAS-stit^Y762F were expressed in epidermal stripes at 25°C. The 1.4-kb Ddc1.4-GFP and the 3.0-kb ple-DsRed wound reporters (Mace et al., 2005) were used to assess Ddc and Ple induction following wounding. In the stit-lethality rescue experiment, the UAS-stit, UAS-stit^K504A and UAS-stit^Y762F transgenes were expressed in stit mutant embryos under the control of the epithelial driver 69B-GAL4. Dfd-GFP-marked CyO and TM3 (or TM6) balancers were used to identify homozygous mutant embryos.

The screen was carried out by Hybrigenics using a prey library constructed from RNA from embryos that were 0–24-h old. A fragment encoding the intracellular domain of Stit (amino acids 337–773) was inserted into the pB27 vector (N-Lex–bait-C fusion) and was used to screen 72.7 million clones. 185 positive clones were sequenced and classified according to the interaction confidence by Hybrigenics.

Immunohistochemistry was performed as described previously (Hemphälä et al., 2003). We used the following primary antibodies: rabbit polyclonal anti-Src (phospho-Y418, 1∶500, Invitrogen), guinea-pig anti-Stit (1∶500, Wang et al., 2009), mouse anti-diphosphorylated ERK (dpERK) (1∶100, Sigma), mouse anti-phosphotyrosine (pY) (1∶100, Upstate Cell Signaling), mouse anti-GFP (1∶400, Sigma). Secondary antibodies conjugated to Cy2, Cy3 (Jackson Immunochemicals) or Alexa Fluor 488 (Molecular Probes) were used at 1∶400. Whole-mount RNA in situ hybridization using digoxigenin-labeled probes was conducted as described previously (Lehmann and Tautz, 1994). Heterozygous mutant embryos carrying the LacZ balancer were distinguished from homozygous mutant embryos by staining with rabbit anti-β-gal primary antibody (1∶1000, Cappel).

PCR fragments comprising the full-length of Stit, Stit^K504A, Stit^Y553F, Stit^Y572F, Stit^Y660F, Stit^Y717F, Stit^Y751F, Stit^Y762F and Stit^Y751F/Y762F were sub-cloned in frame with a C-terminal V5 tag into pMT expression vector (Invitrogen), and transfected into Drosophila S2 cells using the Effectene kit (Qiagen). At 72 h after transfection, cells were harvested in 125 mM pyrophosphate, 1 M sodium fluoride, 1 M β-glycerophosphate and 500 mM sodium orthovanadate. PCR fragments comprising the sequences coding for the SH2 domain of Drk, Src42A and Src64B were sub-cloned into the pGEX-4T-1 vector.

Expression of the GST–Drk fusion protein and GST–Src64 fusion protein in Escherichia coli BL21 cells were induced with isopropyl-β-d-thiogalactopyranoside (IPTG) for 2 h at 25°C. Expression of the GST–Src-42 fusion protein in E. coli BL21 cells was induced with IPTG for 2.4 h at 25°C, followed by purification on glutathione–Sepharose beads (GE Healthcare). The beads were washed three times in GST lysis buffer (20 mM Tris-HCl pH 8.0, 200 mM NaCl, 1 mM EDTA, 0.1% NP-40). AEBSF [4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride] (Sigma) and complete protease inhibitor cocktail tablets (Roche) were added to the GST lysis buffer before use. Proteins bound to the beads were eluted by addition of GSH buffer (20 mM glutathione, 50 mM Tris-HCl pH 8.0) and were analyzed by SDS-PAGE (10% SDS) followed by exposure to an FLA 1000 phosphorimager (Fuji). The following antibodies were used at the indicated dilutions: anti-GST (1∶2500, Santa Cruz), anti-V5 (1∶5000, Invitrogen).

Embryos were collected during early stage 15 (11 h after egg laying at 25°C), were dechorionated in bleach and placed on a gas-permeable membrane stretched over two silicon bars on a slide (Tsarouhas et al., 2007). Embryos were mounted in halocarbon 700 oil and covered with a cover glass (Menzel-Gläser). Embryos were wounded on the ventral side using a nitrogen pulsed laser (VSL-337ND-S; Laser Science) tuned to 440 nm. The laser was connected to an upright microscope (Axioplan2; Zeiss) equipped with a ×63/1.2NA C-Apochromat water-immersion objective (Zeiss). Ablations were monitored by using the MicroPoint Ablation System (Andor Technologies). Our laser-wounding procedure resulted in mean wound sizes of 1147.29 µm² (±98.37) and 1197.21 µm² (±83.95) (mean±s.e.m.) in wild-type and mutant embryos, respectively.

Embryos carrying the Ddc1.4-GFP or the Ple-DsRed reporters were puncture-wounded with a sterile injection capillary as described previously (Mace et al., 2005). Pricked embryos covered with halocarbon 700 oil were allowed to develop for 3 h before imaging. The extent of GFP or DsRed induction at the wound edge was classified as strong, moderate or weak by two independent observers.

Live imaging using wide-field fluorescence microscopy was performed to estimate the time taken for wound closure as described previously (Wang et al., 2009). Wounded wild-type and mutant embryos were placed on the same slide and imaged in parallel. Confocal live imaging was used to examine the actin cytoskeleton during re-epithelialization. Laser-wounded embryos were transferred to a laser-scanning confocal microscope (LSM 510 META or LSM780; Zeiss) and imaged with an Argon 2 488-nm laser using a ×63/1.3NA C-Apochromat water-immersion objective (Zeiss). Individual z-stacks with a step size of 1–2.5 µm were taken every 6 min over a 3–8-h period. Recording was initiated approximately 30 min after wounding. All recorded embryos developed to late stage 17. Time-lapse movies were created from confocal z-stack projections (maximum intensity) by using NIH ImageJ software from http://rsb.info.nih.gov/ij/index.html.

Semi-quantitative analysis of the fluorescence intensity of GFP–Moesin during re-epithelialization was performed essentially as described previously (Wang et al., 2009). Confocal z-stacks, 22–30 µm in total thickness, were collected to cover the entire puncture depth of the epidermis. Intensities were estimated from the sum of signals of the z-stack. RFI at the wound edge (Bement et al., 1993) was calculated from projected stack images at defined time-points in each embryo by using the formula RFI = FIw/FIep, where FIw represents the average intensity within the area of a 1.2-µm ring around the wound edge and FIep represents the average intensity in a concentric ring area located at a distance of 8 µm from the wound border. Mean RFI values were calculated for each time-point and were plotted in the graph shown in Fig. 5E. The time-points in wild-type embryos represent the start, quarter, half and three quarters of the mean closure time and the final closure time. src42A^E1, src42A^E1 btk29A and src42A^E1 src64B mutant embryos required more time to close their wounds, or failed wound closure. Therefore, two time-points were added in experiments with these embryos – one at the mean of src42A^E1 wound-closure time and one at the final recording time. The xy projections and the quantification of fluorescence intensity measurements were performed in ImageJ.

Error bars in all graphs indicate s.e.m. An unpaired two-tailed Student's t-test was used to estimate statistical significance, unless indicated.

We are indebted to Naumi Nautiyal, who initiated the analysis of Stit effectors and performed several of the experiments. We would like to thank members of our laboratory and Mattias Mannervik (Stockholm University, Stockholm, Sweden) for support, discussions and critical reading of the manuscript. We thank Monika Björk for help with the generation of transgenic strains. We thank William McGinnis (University of California at San Diego, San Diego, CA) for wound reporters, Bloomington and Kyoto Stock Centers for fly strains, and the imaging facility at Stockholm University (IFSU). The two-hybrid screen was carried out by Hybrigenics.