Drosophila embryos close epithelial wounds using a combination of cellular protrusions and an actomyosin purse string

Wound repair is an essential physiological process that is compulsory for organism survival, tissue homeostasis, and to avoid infections. From invertebrates to mammals, organisms have developed robust and rapid responses to restore tissue continuity during embryonic and adult life. An efficient wound repair response is particularly important for tissues such as the epithelia, which serves as protective barriers from the surrounding environment and are constantly exposed to chemical and mechanical trauma. Epithelial wound repair has been studied in vivo in different model organisms and extensively in vitro in cell/tissue culture (Martin and Lewis, 1992; Bement et al., 1993; McCluskey et al., 1993; Brock et al., 1996; Danjo and Gipson, 1998; Fenteany et al., 2000; Davidson et al., 2002; Wood et al., 2002; Galko and Krasnow, 2004; Russo et al., 2005; Tamada et al., 2007). From these studies, it has been proposed that epithelial wounds heal through a variety of mechanisms, including lamellipodial crawling/cell migration, contraction of the cells underlying the wound, and actomyosin purse string-mediated contraction of the cells at the wound periphery (Fig. 1A) (Martin and Lewis, 1992; Bement et al., 1993; Brock et al., 1996; Wood et al., 2002).

Several factors have been proposed to affect the choice of repair mechanism to be employed, including the size and shape of the wound, the intrinsic tension in the tissue, the nature of junctions/connections between cells, and the developmental stage. Re-epithelialization by lamellipodial crawling, where the leading edge cells push themselves forward over the wound is commonly observed in adult tissues (Fig. 1A) (Odland and Ross, 1968; Pang et al., 1978; Buck, 1979). Epithelial wound closure in embryos is often mediated by the formation of a contractile actomyosin purse string linked intercellularly by adherens junctions (Martin and Lewis, 1992; Bement et al., 1993; Brock et al., 1996) (Fig. 1A). However, the developmental stage of an organism does not necessarily define the repair mechanism used, as purse string-mediated repair has been observed in certain adult tissues, including the adult cornea (Danjo and Gipson, 1998). Some types of epithelial cells in culture repair their wounds using a combination of purse string contraction and lamellipodial crawling (Bement et al., 1993; Russo et al., 2005). Likewise, cells that are tightly adhered to one another may prefer a purse string based mechanism, while less adherent cells may drag themselves forward over the wound matrix. In addition, large wounds in tissue culture wound assays tend to heal by lamellipodia extension and cell migration (Bement et al., 1993), whereas smaller wounds heal by purse string contraction of an actomyosin ring assembled at the wound edge (Bement et al., 1993). While these different repair mechanisms have been associated with specific contexts, a major challenge in the field is to identify the molecular components underlying each repair mechanism and to elucidate their roles in determining a particular tissue's means of repair.

Drosophila has recently emerged as a genetic model for studying epithelium wound repair and the subsequent inflammatory response (Wood et al., 2002; Galko and Krasnow, 2004; Mace et al., 2005; Stramer et al., 2005). Embryonic (epithelial) wound repair has been proposed to take place using an actin cable that operates as a purse string to draw the hole mostly closed followed by the use of filopodia for the final knitting together of the epithelium cells (Wood et al., 2002), whereas post-embryonic (tissue) repair has been proposed to occur through cell shape changes and cell migration (Galko and Krasnow, 2004; Lesch et al., 2010). Furthermore, genetic screens for components involved in embryonic (epithelial) and post-embryonic (tissue) wound repair have, with the notable exceptions of the JNK signaling cascade and the Rho family of small GTPases, yielded non-overlapping sets of mutants (Campos et al., 2010; Lesch et al., 2010), reinforcing the idea that the regulatory mechanisms employed in wound repair are context-dependent.

To determine the effect of different parameters on wound repair mechanism, we employed 4D in vivo microscopy along with mutant analysis to define the series of changes that occur during four distinctive phases in response to epithelial wounding in Drosophila embryos. We find that specific molecular components including myosin, E-cadherin, Echinoid, the plasma membrane, microtubules, and the Cdc42 small GTPase respond dynamically during wound repair, and demonstrate that perturbations of each of these components result in abnormal wound healing. Our results show that embryo epithelial wound repair requires a combination of partially redundant mechanisms involving intertwined actin elements (dynamic cellular protrusions and a contractile actomyosin cable), and provide new insight into the different cellular machineries and mechanisms required for these processes.

Stage 15 Drosophila embryos have a single layer ectoderm covering the surface of the embryo (Fig. 1B–F). Previous studies have shown that while most of the embryo surface is subjected to tensile forces due to dorsal closure movements, the ventral epidermis is under lower net tension (Kiehart et al., 2000). Thus, we generated wounds in the ventral epithelium by ablating a circular patch of cells then followed the wound repair process by 4D confocal microscopy. The ablations were highly reproducible (Fig. 1H), and XZ views showed the damage was restricted to the epithelial cell layer, as the tissue underlying the epithelium remains intact (Fig. 1G′). In vivo analysis of epithelial wound repair in embryos expressing markers for actin (sGMCA; spaghetti squash driven, GFP, moesin-α-helical-coiled and actin binding site) and the nucleus (His2Av-mRFP), showed a range of dynamic morphological changes in the epithelium cells (Fig. 1G,H; Table 1). We found that the epithelial wound repair process can be divided into four phases, corresponding with the morphological changes observed: (i) expansion, (ii) coalescence, (iii) contraction and (iv) closure.

Upon wounding (4–5 cells, ∼800 µm² area), a clear gap is visible in the epithelium (Fig. 1G,G′). During this initial expansion phase, the tissue margins retract from the wound and cells at the edge of the injured area exhibit different levels of damage. Many cells at the edge of the wound showed high actin levels, particularly at cell junctions, and became the leading edge cells for the repair process (Fig. 1G; supplementary material Movie 1A). Other cells, however, showed either lower actin levels and fell below the plane of the epithelium, or became rounded (swollen) and were removed from the wound edge (Fig. 1G). These morphological changes suggest that some of the epithelium cells suffer irreparable damage upon wounding, and as a consequence need to be removed from the epithelium. We examined if caspase activity, one of the hallmarks of cell death by apoptosis, was activated in the cells at the wound border using the caspase fluorescent sensor, Apoliner (Bardet et al., 2008). Apoliner was expressed in the embryo epithelium, and its localization was followed upon wounding (Fig. 1I). None of the cells at the wound edge showed detectable translocation of GFP into the nucleus, suggesting that caspases are not activated, neither in the damaged cells nor the cells surrounding the wound area. Thus, cell removal likely occurs by necrosis and/or cell engulfment.

During the coalescence phase, the wound reaches its largest size then begins to assemble the cellular machineries necessary for closing the lesion. To be able to compare this phase in different genetic backgrounds, we statistically defined coalescence as the time in which the wound size remained relatively constant at greater than 90% of the maximum area. Actin foci were visible along the leading edge, particularly at cells junctions, and, by 5 minutes post-wounding, dynamic protrusions were detected on cells at the leading edge of the wound (Fig. 1G,G′,L; supplementary material Movie 1A). Actin accumulated initially in discontinuous patches at the apical edge of these leading edge cells (Fig. 1J–K′; supplementary material Movie 2A). During the next 10–15 minutes these discontinuous actin patches joined up to form a thick and continuous actin cable encircling the leading edge of the wound (Fig. 1G). As a result of these events, the wound was observed to decrease slightly in size and changed from a jagged irregularly shaped opening to a taut rounded hole.

The contraction phase is characterized by the rapid reduction in area of the rounded wound opening (Fig. 1G,G′; supplementary material Movie 1A). For medium sized wounds (∼800 µm²) this process takes ∼42 minutes on average (Fig. 1H; Table 1). Actin enrichment in the cable increased throughout this phase, indicative of the contractile force associated with a functional actomyosin purse-string. Apically-localized actin-rich protrusions, both filopodia and broad lamellipodia, were observed throughout the contraction phase in the leading edge cells. These protrusions actively probed the open space of the wound area, making frequent contact with apical protrusions from other leading edge cells, as well as occasional contact with the overlying vitelline membrane (Fig. 1K,K′). We did not observe basal protrusions or contact of apical protrusions with the underlying somatic muscle cells and/or ventral cord cells.

In the closure phase, the final ∼5% of the peak wound area was drawn closed to restore tissue integrity (Fig. 1G; supplementary material Movie 1A). Similar to what has been shown previously, we found that contacts between filopodia/lamellipodia from opposing edges mediate this final wound closure by knitting together the wound edges (Wood et al., 2002).

We observed both dynamic cellular protrusions and formation of an actomyosin cable during the coalescence phase, suggesting that Drosophila embryos may be able to repair lesions using more than one mechanism. One of the factors proposed to affect the mechanism by which wounds will be repaired is wound size. To determine if this was a deciding factor in the Drosophila embryo, we examined repair in different sized wounds (Fig. 2A–I, Table 1). Wounds fell in three different size ranges (area determined at fully expanded wound): (i) small: wound area <500 µm²; (ii) medium: wound area ranging from 500-1200 µm²; and (iii) large: wound area >1200 µm². Much larger sized wounds (>2500 µm²) can undergo successful repair, but were not used here since they often trigger an accompanying stress response complicating subsequent analyses. We found that in wild-type embryos, wound repair kinetics followed the same trends (see percentage of time in each phase) and different sized wounds displayed similar phenotypic repair dynamics (i.e. using an actin cable and actin-rich protrusions) (Fig. 2A–I; Table 1; and data not shown).

We were surprised, however, to find that the method of mounting the embryos for wounding and imaging greatly affected the outcome of the wound repair dynamics (Fig. 2J–L). To perform laser wound assays, embryos were initially mounted either: (i) in series 700 halocarbon oil between a hydrophilic gas permeable membrane and a coverslip (a popularly used mounting method; Wood and Jacinto, 2005), or (ii) on a glue-painted coverslip then covered with halocarbon oil (Foe et al., 2000). The epithelial wound response was highly reproducible using glue mounting (Fig. 2L). However, embryos mounted using the membrane exhibited a highly variable repair response and in some cases even failed to heal (Fig. 2L). This interference with wound repair is likely caused by the slight compression introduced when embryos are sandwiched between the membrane and coverslip (Fig. 2J,K). Thus, our wound assays were typically conducted using medium size wounds with embryos mounted on glue painted coverslips. Under these conditions, wound repair was completed on average by 50 minutes (Table 1).

Actomyosin-powered purse string contraction is a mechanism common to multiple morphogenetic movements in a variety of organisms (Young et al., 1993; Williams-Masson et al., 1997; Vavylonis et al., 2008), and has been suggested to power wound closure in Drosophila embryonic epithelium wounds (Wood et al., 2002). To further understand the role of the actomyosin cable, we examined the function of myosin II in vivo and its contribution to the wound repair process. We first monitored the dynamics of myosin II recruitment relative to actin by examining embryos co-expressing a myosin regulatory light chain (sqh) fusion to GFP (sqh-GFP) and actin (sChMCA). Myosin II can be detected at the wound edge 5 minutes after wounding, overlapping with actin accumulation along the leading edge (Fig. 3A–B′; supplementary material Movie 1B). During contraction, actin and myosin co-localized to form a robust actomyosin purse-string in both surface and orthogonal planes. A striking feature of this accumulation was the complete absence of myosin II from the actin-rich protrusions (Fig. 3A–B′), suggesting that myosin II is only associating with actin as part of the supracellular cable.

Given myosin dynamics in response to wounding, we anticipated that mutations affecting myosin II would impair actomyosin purse string assembly and contractility without affecting the actin-rich protrusions. In zipper¹ (zip¹) homozygous mutants, which disrupt the myosin II heavy chain gene, an actomyosin purse string is not assembled at the leading edge of the wound (Fig. 3C,C′; Table 2; supplementary material Movie 2B). The wound edges were irregular, with actin accumulated in patches that extended over a few cells and particularly at cell-cell junctions, but never forming the thick continuous cable that encircles wounds in wild-type embryos (Fig. 3C,C′). Interestingly, despite the lack of a contiguous actin cable, wounds in zip¹ mutant embryos healed, albeit with a severe delay especially in the contraction phase (P = 0.0051; Fig. 3D,E; Table 2). In vivo analysis showed that wound closure was achieved by the actin-rich cellular protrusions (Fig. 3C,C′,F; supplementary material Movie 2B). Quantification of the wound area covered by protrusions showed that zip¹ embryos displayed three times more protrusions than wild-type embryos at similar stages of repair (Fig. 3G). This increase could be detected from the coalescence phase (100% of max wound area) (Fig. 3F,G). These protrusions pulled the wound closed by capturing protrusions from adjacent or opposing cells and tugging these areas closer together. As a consequence, reduction of the wound area occurred in jumps (Fig. 3D; supplementary material Fig. S1C). Our data suggests that the contractile force of the actomyosin purse string was necessary for the rapid closure of the wound during the contraction phase, but in the absence of this component, both local and long-distance zippering by dynamic filopodia and lamellipodia can mediate wound closure.

In order for the actomyosin purse string to contract to close a wound it needs to be linked from cell to cell to generate a supracellular cable. Cadherin-based adherens junctions are a major site of cell-cell adhesion and actomyosin network anchoring to the plasma membrane within a cell (reviewed in Tepass et al., 2001; Hartsock and Nelson, 2008). Analysis of epithelium wound repair in embryos expressing markers for E-Cadherin (DE-Cadherin-GFP) and actin (sGMCA) showed that E-Cadherin accumulated at the apical most side of the leading edge cells and in particular at the sites of lateral cell-cell junctions, consistent with its reported role (Fig. 4A,A″′; supplementary material Movie 1C) (Wood et al., 2002). By midway through the contraction phase, cadherin expression was largely lost along the apical most side of leading edge cells, remaining at the sites of lateral cell-cell junctions (Fig. 4B,B′). Interestingly, we found that immediately following injury, some of the cells at the wound site retained DE-Cadherin expression but lacked actin (Fig. 4A). These cells slowly disappeared from the surface of the embryo as the leading edge cells moved forward to close the wound (Fig. 4A,A′).

Given the accumulation of DE-Cadherin at the sites of cell-cell junctions and its overlap with actin accumulation, perturbations in the levels of cadherin would be expected to disrupt cable function and possibly its assembly. Cadherin loss of function mutants (shotgun; shg^k03401) showed defects in wound repair during the contraction phase: actin cable assembly was impaired leading to initially jagged wound boundaries and irregularly shaped wounds (Fig. 4C,C′,E,F; supplementary material Fig. S1A; Table 2). Remarkably, wound closure was eventually achieved by non-uniform contraction of leading edge areas (a few cells wide) that accumulated actin. These areas of localized pinching often corresponded with cells exhibiting active protrusions (Fig. 4D; supplementary material Movie 2C). Our data shows that DE-Cadherin is required for both the assembly and stability of the actin purse-string at the wound leading edge.

Another adherens junction-associated cell adhesion molecule, the Drosophila nectin ortholog Echinoid (Ed), has been shown to interact with Canoe (afadin) and Jaguar (unconventional myosin VI) to modulate the accumulation of actin regulators at the leading edge during dorsal closure (Wei et al., 2005; Laplante and Nilson, 2006; Lin et al., 2007; Chang et al., 2011; Laplante and Nilson, 2011). In this context, Ed has been proposed to mediate cell sorting by promoting actomyosin cable formation and filopodial protrusions in the dorsal most leading edge cells. In the context of epithelial wound repair, we anticipated that loss of Ed might prevent epithelial wound repair by destabilizing adherens junction complexes, disrupting actomyosin purse string assembly and stabilization, and by reducing cellular protrusions from leading edge cells. Consistent with this, ed^M/Z mutant embryos failed to form a continuous actomyosin purse-string at the leading edge of wounds (Fig. 4G,G′,K–L′), and exhibited significant delays in both the coalescence and contraction phases (P<0.0001 and P = 0.0005, respectively) (Fig. 4I,J; Table 2; supplementary material Fig. S1A and Movie 2D). However, similar to cadherin mutants, wound closure took place by local contraction of actin patches and protrusions. Indeed, we observed protrusions that joined with those from nearby and/or opposing cells to pinch the wound shut (Fig. 4H). Thus, Ed in the context of wound repair may be required for the stabilization of adherens junctions needed to link the actomyosin cable intercellularly at the wound leading edge cells.

Previous studies suggested that an actomyosin cable was needed to close the edges of the wound until the fronts were close enough for cellular protrusions to reach across and knit closed the small remaining hole (Wood et al., 2002). In contrast to previous reports, we observed filopodia and lamellipodia throughout the wound repair process in wild-type embryos making contact with other protrusions (Fig. 1I–J′; Fig. 4B′). To determine the role of cellular protrusions during the contraction phase when the actomyosin cable is thought to be the major mechanism for closing the hole, we followed the dynamics of cellular protrusions during the wound repair process. In particular, we examined two membrane markers: GFP-PH-PLC, the pleckstrin-homology domain of PLC fused to GFP, which specifically binds to the phosphoinositidine PI(4,5)P2 and is enriched in the apical side of epithelial cells (Pinal et al., 2006), and Venus-GAP43, the palmitoylation signal of the growth-associated protein 43 fused to Venus fluorescent protein, which tightly anchors to the plasma membrane (Mavrakis et al., 2009). In addition, we examined a marker for microtubules as they have previously been reported to be reorganized in response to injury (Etienne-Manneville, 2004).

Cells at the leading edge of the wounds show highly dynamic filopodia and lamellipodia enriched in both membrane markers, particularly when the PH-PLC reporter was used (Fig. 5A,B). These cellular extensions reach across the wound to make both nearby and long-distance connections to the protrusions of other cells (Fig. 5A,B; supplementary material Movie 3). In addition, GAP43 was observed to accumulate in the cell-cell contacts (Fig. 5C–C″).

Time-lapse movie analysis of Drosophila embryos expressing GFP-α-tubulin in the epithelium shows that, unlike actin, microtubules do not accumulate at the leading edge of the wound (Fig. 5D–F). However, consistent with previous observations during the dorsal closure process (Jankovics and Brunner, 2006; Liu et al., 2008), microtubules are recruited to the filopodial protrusions assembled by the leading edge cells (Fig. 5D–F). We wounded embryos mutant for the microtubule capping protein EB1 (eb1²), which disrupts microtubule dynamics (Elliott et al., 2005). Interestingly, we observed defects only in the coalescence phase, where eb1² mutant embryos showed a significant delay before beginning contraction (P = 0.045; Fig. 5G–I; Table 2). We are not able to rule out, however, that disruption of microtubule dynamics in this context is indirect and may not affect cellular protrusions per se, but rather affect repair during the coalescence phase by impairing the recruitment of the repair machineries/components thereby delaying initiation of the contraction phase.

To further investigate the role of dynamic cellular protrusions, we examined the function of the Cdc42 small GTPase in vivo and its contribution to the wound repair process. Cdc42 constitutive active or dominant negative mutants have been shown to regulate filopodial assembly in tissue culture cells and in both dorsal closure and wound repair in Drosophila (Nobes and Hall, 1995; Jacinto et al., 2000; Wood et al., 2002). We first generated a ChFP-Cdc42 reporter to follow Cdc42 localization in vivo simultaneously with actin (ChFP-Cdc42; SGMCA) (Fig. 6A–A″). Throughout wound repair, Cdc42 did not accumulate at the wound leading edge or in the actin-rich protrusions, suggesting that upon wounding Cdc42 subcellular localization does not change significantly.

To modulate the levels of Cdc42, we used the heteroallelic combination Cdc42⁴/Cdc42⁶. These Cdc42 loss of function mutant embryos exhibit severe developmental defects as previously reported (Genova et al., 2000), where epithelial cells fail to elongate into columnar shaped cells, resulting in failure of germ band retraction and, in most cases, random holes appeared in the epithelium. Upon wounding, Cdc42⁴/Cdc42⁶ embryos were not significantly delayed in the expansion or coalescence phases (Fig. 6B–D; Table 2). By contraction, these wounds became quite rounded consistent with the presence of circumferential tension provided by a functional actomyosin cable (Fig. 6B,B′; supplementary material Fig. S1B and Movie 2E). Strikingly, we observed a severe reduction of filopodia and lamellipodia protrusions in the leading edge cells throughout the wound repair process: Cdc42⁴/Cdc42⁶ mutant embryos displayed three times less filopodia/lamellipodia than wild-type controls (Fig. 6E,F). As a consequence, these embryos spend 95 minutes in the contraction phase, over double the time spent by wild-type embryos (Fig. 6D; Table 2). As previously described, Cdc42⁴/Cdc42⁶ mutant embryos are unable to seal the wound during the final closure phase and a small hole persists (n = 4/5 embryos) (Fig. 6B) (Wood et al., 2002). Our data shows that the contractile force of the actomyosin cable was sufficient to draw the wound closed, however the kinetics of this phase were not the same as those observed for repair in wild-type embryos, indicating that cellular protrusions are playing an active role in the contraction phase.

Our in vivo mutant analysis allowed us to dissect the individual contributions of cytoskeleton components to the wound repair process. In particular, our results show that disruption of the actomyosin cable leads to increased protrusions (Fig. 3C–G), whereas disruption of cellular protrusions leads to defects in both active wound closure during contraction and the final wound zippering (Fig. 6B–E). To determine if both components – the actomyosin cable and cellular protrusions – are required simultaneously for epithelial wound repair, we generated double mutants that disrupts both structures by targeting myosin II and Cdc42 small GTPase. Since generating double genetic mutants was not possible, we used the combination of a genetic mutant, zip¹ and dsRNA interference (RNAi) to specifically reduce Cdc42 (Cdc42^RNAi). We injected dsRNA for Cdc42 into zip¹ mutant and wild-type embryos, or buffer alone into zip¹ mutants (see Material and Methods). Cdc42^RNAi embryos display similar wound repair dynamics when compared to Cdc42⁴/Cdc42⁶ mutants (supplementary material Fig. S1D), and buffer-injected zip¹ homozygous mutant embryos display similar wound repair dynamics to uninjected zip¹ homozygous mutants (supplementary material Fig. S1E). zip¹ Cdc42^RNAi double mutants fail to close epithelium wounds: 100% of the wounds are still open at 220 minutes post-wounding, the time by which zip¹ and Cdc42⁴/Cdc42⁶ single mutants are closed (Fig. 6G,H). zip¹ Cdc42^RNAi double mutant embryos show defects in all the wound repair phases, particularly in the coalescence and contraction phases (Fig. 6G,H). During the coalescence phase, wound edges were jagged and wound shape was irregular, with actin failing to accumulate in the leading edge cells. Strikingly, patches of continuous actin are not observed until 60 minutes post-wounding (four times longer than wild-type embryos) (Fig. 6G). The contraction phase is severely delayed, due to the lack of both a proper actin cable and cellular protrusions (Fig. 6G). Significantly, zip¹ Cdc42^RNAi double mutant embryos stall during the contraction phase, and a hole persists with an average area of 159.1±31.5 µm² (n = 4). Thus, our data shows that a combination of functional actomyosin cable and cellular protrusions are required for proper embryo epithelial wound repair.

Epithelial wound repair in the Drosophila embryo proceeds through a distinct set of phases and requires the coordinated efforts of several cytoskeletal components/machineries. We show, for the first time, the specific contributions of myosin, E-cadherin, Echinoid, and plasma membrane components in response to wounding (Fig. 7A–C). Actin-rich protrusions are the first cytoskeleton structure observed early on in the wound repair process, followed by the formation of an actomyosin purse-string at the wound leading edge (Fig. 7A–C). While each of these components by themselves makes distinct functional contributions, they function together in the wild-type wound repair process to ensure rapid and complete healing.

A major observation while studying embryo epithelial wound healing in vivo is the existence of defined phases occurring at characteristic intervals along the wound repair process, and requiring specific molecular components. The expansion phase encompasses the initial clearance of the wound area and the establishment of the wound leading edge. Interestingly, none of the mutants we examined significantly affected this phase, suggesting that the factors affecting expansion may be inherent to the tissue. Notably, the expansion phase was not affected in zip mutant embryos, despite the epithelial cells in this mutant being smaller on their apical surfaces. In addition, in severe E-cadherin (shg²) mutant embryos in which only non-continuous patches of epithelium exist by later stages of development, wounds expand with normal dynamics, precluding morphogenetic tension as the explanation for the initial wound expansion (data not shown). We anticipate that mutants for components of the mechanosensory system and/or signals that initiate the wound repair response will impair the expansion phase.

We initially hypothesized that the coalescence phase corresponded to a passive transition where the forces exerted by the expanding wound were gradually overcome by the contraction machinery. However, the delay we observed in eb1 mutants is consistent with this phase being active. Furthermore, the coalescence phase is not extended in all mutants affecting the actomyosin cable, as would be expected if the contractile forces needed to overcome expansion are impaired.

The majority of wound closure is achieved during the contraction phase, as the wound is actively closed by a combination of cable contraction and cellular protrusions. The wound area exhibits exponential rather than linear reduction, consistent with a steady rate of wound margin ingression towards the center. Interestingly, in mutants lacking the actomyosin cable, the wound area is reduced in a more stepwise fashion corresponding to local zippering events. This suggests that the circumferential tension of the actomyosin cable not only provides direct force for repair, but also transmits the force generated by local zippering around the wound edge.

The final stage of repair is closure, in which filopodia and lamellipodia on opposing epithelial faces contact and engage to fuse the wound fronts and restore the continuity of the epithelium. Consistent with previous reports, our genetic analysis shows that the Cdc42 small GTPase is crucial in this phase of the process, by modulating the formation of polarized actin-rich protrusions (Wood et al., 2002). It will be interesting to see if trafficking of adhesion molecules, or perhaps integrins, along with filopodia are necessary for the protrusions to mediate epithelial fusion and the formation of new cell-cell junctions.

In the absence of a functioning actomyosin cable, epithelium wounds heal by extension and zippering of actin-rich cellular protrusions. Previous studies have suggested that interactions between filopodia and lamellipodia of neighboring cells were only responsible for wound closure during the final steps of repair, when the edges of the wound were close to one another (Wood et al., 2002). In their study, wounding of Rho1 small GTPase embryos, in which the actomyosin cable is disrupted, showed a two-hour delay in repair during which no changes were observed in the leading edge cells, followed by the wound closing with normal wound kinetics mediated by local zippering (Wood et al., 2002). In contrast, our in vivo studies show that actin-rich protrusions are present throughout the wound repair process, from the onset of the coalescence through the closure phase. These apically-localized protrusions are highly dynamic, as observed with actin and plasma membrane markers, and reach across the wound fronts throughout the repair process to contact neighboring cells, as well as the overlying vitelline membrane. When we disrupted the actomyosin cable using myosin (zip) and E-cadherin (shg) mutants, we find that wound repair is significantly delayed, and showed that actin-rich protrusions are used for contraction in these mutants. Indeed, we observed long-range events where filopodia reaching far across the middle of the wound made contact and pulled the wound closed at the middle. Consistent with this, we observe a delay in contraction when the ability to form cellular protrusions, but not a functional actomyosin cable, was impaired using Cdc42 mutants (Table 2). The results observed with the Cdc42 mutants highlight two important aspects of the mechanisms involved in the repair process. First, protrusions are necessary for normal wound repair kinetics during contraction. Second, intact protrusions are crucial for the closure phase. In the absence of protrusions, the wound fronts fail to fuse indicating that the contractile force of the actomyosin cable cannot compensate for the lack of protrusions in the final steps of wound closure.

Wound repair by lamellipodial crawling was thought to be an exclusive repair mechanism of adult tissues, whereas embryonic epithelium wound closure was mediated solely by actomyosin purse string contraction (reviewed in Jacinto et al., 2000). However, more recent studies have shown an overlap between these two types of epithelial events and have suggested that wound size and intrinsic tissue dynamics may also influence the mode of wound closure (Bement et al., 1993; Wood et al., 2002; Russo et al., 2005). We find that the repair of epithelial wounds in the Drosophila embryo integrates both repair mechanisms. By 5 minutes post-wounding the major structures observed at the apical side of the leading edge cells are small protrusions, not the actomyosin purse string. During the contraction phase, dynamic filopodia extending all around the wound edge area can be seen along with the prominent actomyosin cable. Impairing the actomyosin cable leads to healing through cellular protrusions. Likewise, impairing cellular protrusions leads to healing through actomyosin cable contraction. Interestingly, shg mutant embryos, which have an impaired actomyosin cable, and Cdc42 mutants, which have reduced actin-rich cellular protrusions, spend almost identical amounts of time in the contraction phase suggesting that the two mechanisms are equally important to generate the contractile force required for wound closure. In both cases, the repair kinetics are severely delayed compared to wild type, again suggesting that both mechanisms are working synergistically in wild-type repair. Consistent with this simultaneous requirement for both an actin cable and cellular protrusions for proper repair, embryos in which both wound repair mechanisms are functionally disrupted (by combining mutants for Myosin II and Cdc42) fail to close epithelium wounds.

While our knowledge of the dynamics and components of epithelial wound repair has rapidly increased in recent years with the addition of new model organisms and imaging techniques, major questions still remain concerning the intrinsic factors modulating the repair response, as well as the signaling molecules involved in triggering and terminating wound repair. In particular, what are the signals that lead to the context-dependent choice of wound repair mechanism? The Drosophila embryo is an excellent genetic model in which these questions encompassing the complex process of epithelial wound repair can be systematically addressed.

Flies were cultured and crossed at 25°C on yeast-cornmeal-molasses-malt medium. The following stocks containing fluorescence fusion proteins were used: sGMCA (Kiehart et al., 2000), sChMCA (Abreu-Blanco et al., 2011), P{His2Av-mRFP1}III.1 and P{UASp-GFPS65C-alphaTub84B} (Bloomington Stock Center), P{UAS-Apoliner}8/TM3, Sb (Bardet et al., 2008), sqh^AX3; P{sqh-GFP}42 (Royou et al., 2004), ubi-DE-Cadherin-GFP (Oda and Tsukita, 2001, Kyoto Stock Center), P{UASt-GPF-PH-PLC} (Pinal et al., 2006), UASp-Venus-GAP43 (Mavrakis et al., 2009), and ChFP-Cdc42 (this study). Double fluorescently tagged lines were generated using standard genetic methods. Expression of UAS lines was driven in the embryo epithelium with P{en2.4-GAL4}e22c (Bloomington Stock Center).

The following mutants alleles were used: Cdc42⁴ and Cdc42⁶ (Fehon et al., 1997), ed^F72 (Laplante and Nilson, 2006), and from the Bloomington Stock Center: zip¹ (Young et al., 1993), shg^k03401 and shg² (Tepass et al., 1996), and eb1² (Elliott et al., 2005). Mutant alleles were crossed to the sGMCA; CyO-ChFP or sGMCA; TM3-ChFP balancer stocks, to screen for homozygous mutants by selecting against the ChFP balancer. ed^M/Z mutants are germ-line clones, generated using the FLP-DFS system with ed^F72 FRT40A (Laplante and Nilson, 2006). Mutant embryos for Cdc42 were generated using the heteroallelic combination Cdc42⁴/Cdc42⁶ (Fehon et al., 1997). All mutant embryos express sGMCA allowing the actin cytoskeleton to be followed.

ChFP-Cdc42 is a fusion of ChFP with the Cdc42 cDNA, expressed under the control of the spaghetti squash (sqh) promoter. The construct was generated as follows: the Cdc42 ORF was amplified by PCR from DGCr1 cDNA HL08128, then cloned 3′ of ChFP as a EcoRI and XbaI fragment. The ChFP-Cdc42 fusion was PCR amplified and cloned into the StuI and XbaI sites of the pSqh5′+3′UTR plasmid (Abreu-Blanco et al., 2011). To generate CyO-ChFP and TM3-ChFP balancers, ChFP was cloned into the StuI and XbaI sites of the pSqh5′+3′UTR plasmid (Abreu-Blanco et al., 2011). These constructs were used to make germline transformants as previously described (Spradling, 1986). The resulting transgenic lines were mapped to a single chromosome and shown to have non-lethal insertions.

To generate the Cdc42 double strand RNA (dsRNA), the template was amplified by PCR using primers that include the T7 promoter sequence. The sequence used for the Cdc42 dsRNA was selected based on previously reported RNAi lines by the TRiP stock center (DRSC25134). The dsRNA was synthesized and injected into embryos as previously described, at a final concentration of 2.5 µM (Magie et al., 2002; Magie and Parkhurst, 2005). Embryos were injected and aged at room temperature for 12–14 hours before conducting the wound healing assays. Control (sGMCA) and zip¹ (sGMCA, ChFP balancer) were injected with the Cdc42 dsRNA, generating Cdc42^RNAi and zip¹ Cdc42^RNAi embryos, respectively. Control (buffered injected) zip¹ was also included in the assay.

Stage 15 laser wounded embryos were recovered, fixed with 37% paraformaldehyde/heptane for 5 minutes, and hand devitellinized. Immunofluorescence was performed as described previously (Abreu-Blanco et al., 2011) with anti-nonmuscle myosin antibody (1∶500; Young et al., 1993) and anti-rabbit Alexa Fluor 568 (1∶1000; Invitrogen). Alexa-Fluor-488-labeled phalloidin (Invitrogen) was used at 5 Units/assay and added with the secondary antisera. Samples were mounted in SlowFade Gold (Invitrogen/Molecular Probes).

Confocal microscopy was performed using a Zeiss LSM-510M microscope (Carl Zeiss Inc., Jena, Germany) with excitation at 488 nm or 543 nm, and emission collection with BP-500-550 or LP560 filters, respectively. A Plan-Apochromat 20× 0.75 dry objective was used for imaging. Images were processed in ImageJ (http://rsb.info.nih.gov/ij/), and assembled with Canvas 8 software (Deneva Systems, Inc.).

All imaging was performed at room temperature (23°C). Stage 15 embryos were hand dechorionated, dried for 5 minutes then transferred individually with forceps onto strips of glue dried onto No. 1.5 coverslips, and covered with series 700 halocarbon oil (Halocarbon Products Corp.) (Foe et al., 2000). The following microscopes were used: (i) Nikon TE2000-E stand (Nikon Instruments, Melville, NY), with 40× 1.4 NA objective lens, controlled by Volocity software (v.5.3.0, PerkinElmer, Waltham, MA). Images were acquired with 491 nm and 561 nm lasers, with a Yokogawa CSU-10 confocal spinning disc head equipped with a 1.5× magnifying lens, and a Hamamatsu C9100-13 EMCCD camera (PerkinElmer, Waltham, MA). (ii) UltraVIEW VoX Confocal Imaging System (PerkinElmer, Waltham, MA), in a Nikon Eclipse Ti stand (Nikon Instruments, Melville, NY), with 60× 1.4 NA or 100× 1.4 NA objective lens and controlled by Volocity software (v.5.3.0, PerkinElmer, Waltham, MA). Images were acquired with 491 nm and 561 nm, with a Yokogawa CSU-X1 confocal spinning disc head equipped with a Hamamatsu C9100-13 EMCCD camera (PerkinElmer, Waltham, MA). (iii) Nikon LiveScan Swept Field Confocal (for Nikon by Prairie Technologies Inc., Middleton, WI) mounted on a Nikon Eclipse Ti (Nikon Instruments, Melville, NY); with 60× 1.4 NA objectives lens, using the NIS-Elements AR 3.0 as acquisition software (Nikon Instruments, Melville, NY). Images were acquired with a 491 nm laser, and a Photometrics QuantEM: 512SC EMCCD camera (Photometrics, Tucson, AZ). All images acquired with a 40×or 60×objective lens are 25 µm stacks/0.5 µm steps, for the 100×images the stacks correspond to 1.5 µm/0.25 µm steps.

Laser ablation experiments used the Photonic Instruments Micropoint® Computer Controlled system (Photonic Instruments, St Charles, IL), as previously described (Abreu-Blanco et al., 2011).

Image series were either analyzed with Volocity software (v.5.3.0, PerkinElmer, Waltham, MA), or were exported as TIFF files then imported into ImageJ for processing. XY projections of 1–5 µm were generated. Wound areas were measured manually with ImageJ or NIS-Elements AR software (version 3.0, Nikon Instruments, Melville, NY). Protrusion area was measured manually at 100, 75 and 50% of the maximum wound area with ImageJ. Circularity was measured using the following formula: 4π × area/perimeter², in which a perfect circle has a value of 1. A Student's t test was used to analyze the data; P<0.05 was considered to be statistically significant. All graphs present values ± s.e.m. All measurements were downloaded into Microsoft Excel and the data were graphed using Prism 5.0c (GraphPad Software).

We thank Parkhurst laboratory members for their interest, advice and comments on the manuscript. We are very grateful to D. Brunner, R. Fehon, R.E. Karess, D.P. Kiehart, J. Lippincott-Schwartz, L. Nilson, H. Oda, F. Pichaud, J.P. Vincent, and the Bloomington/Kyoto Stock Centers for flies and other reagents used in this study. We thank the M. J. Murdoch Charitable Trust for the spinning disk microscope used for imaging.