Drosophila RhoA regulates the cytoskeleton and cell-cell adhesion in the developing epidermis

During embryonic development and organogenesis, coordinated changes in individual cell shape contribute to the large-scale cell sheet movements that are characteristic of morphogenesis. Many such movements are generated through local contraction of the actin cytoskeleton, driven by the molecular motor protein nonmuscle myosin II. This is clearly demonstrated in Drosophila where mutations in zipper (zip) and spaghetti squash (sqh), genes that encode nonmuscle myosin II subunits, cause lethality and are associated with defects in cell division (Karess et al., 1991; Ohshiro et al., 2000; Peng et al., 2000), cell polarity (Winter et al., 2001), muscle development (Bloor and Kiehart, 2001), and the morphogenetic processes of leg elongation, head involution and dorsal closure (Cote et al., 1987; Edwards and Kiehart, 1996; Halsell et al., 2000; Nusslein-Volhard et al., 1984; Young et al., 1993; Zhao et al., 1988).

Dorsal closure describes the process by which the developing embryo becomes enclosed by the epidermis and serves as an excellent model for studying morphogenesis. It begins when, on both sides of the embryo, the dorsal-most lateral epidermal cells become organized into a well-defined row that possesses at its dorsal edge a supracellular purse-string of actin and nonmuscle myosin II (Young et al., 1993). These ‘leading edge’ cells overlie and adhere to the peripheral cells of the amnioserosa, an epithelium that occupies the dorsal surface of the embryo. As dorsal closure proceeds, the leading edge actomyosin purse-string contracts (Kiehart et al., 2000) and, beginning with the leading edge, lateral epidermal cells elongate dorsally. In addition, the amnioserosa contracts and concomitantly a small number of cells drop out of this epithelial layer. As a result, the amnioserosa diminishes in size as dorsal closure proceeds (Kiehart et al., 2000). The overall effect is that, while the amnioserosa ultimately sinks internally where it undergoes apoptosis, the two halves of the lateral epidermis progress towards the dorsal midline. As they approach each other the opposing epithelial sheets suture together, a bidirectional process that appears to be aided by leading edge filopodia (Jacinto et al., 2000).

Genetic analysis of dorsal closure has been a particularly effective tool with which to identify and dissect pathways that regulate morphogenetic movements. For example, mutations in components of the JNK (Jun N-terminal kinase) transcriptional activation pathway (Glise et al., 1995; Hou et al., 1997; Kockel et al., 1997; Martin-Blanco et al., 1998; Riesgo-Escovar et al., 1996; Riesgo-Escovar and Hafen, 1997; Sluss et al., 1996) and the decapentaplegic (dpp) signaling pathway (Affolter et al., 1994; Arora et al., 1995; Brummel et al., 1994; Grieder et al., 1995; Letsou et al., 1995; Nellen et al., 1994; Penton et al., 1994; Ruberte et al., 1995; Staehling-Hampton et al., 1995) fail to complete dorsal closure, leaving the dorsal surface of the embryo open. Ectopic expression of RhoGTPase family constructs also disrupt dorsal closure (Harden et al., 1995; Harden et al., 1999; Ricos et al., 1999). Thus, expression of dominant-negative Rac^N17 results in the formation of a dorsal hole. Rac^N17 expression, like mutations in the JNK pathway, disrupts formation of the leading edge actomyosin purse-string. Moreover, constitutively active Rac induces ectopic epidermal expression of dpp and puc, known targets of JNK signaling in the leading edge (Glise and Noselli, 1997). As this induction is dependent on the JNK pathway and as Rac is an activator of JNK signaling in vertebrates (Coso et al., 1995; Minden et al., 1995), this suggests that Rac activates the JNK cascade in the leading edge. Epidermal expression of constitutively active Cdc42 also induces ectopic expression of dpp and puc (Glise and Noselli, 1997). However, loss of function Cdc42 mutations have no effect on leading edge dpp expression (Genova et al., 2000). In addition, although expression of dominant-negative Cdc42 disrupts dorsal closure, it does not cause complete disruption of the leading edge actomyosin purse-string (Harden et al., 1999; Ricos et al., 1999; Riesgo-Escovar et al., 1996). Rather, the leading edge defects observed phenocopy those exhibited by mutants for downstream components of the DPP pathway. DPP signaling acts subsequent to JNK activation and is thought to mediate elongation of lateral epidermal cells (Riesgo-Escovar and Hafen, 1997). Thus, while evidence on Cdc42 function in dorsal closure is somewhat ambiguous, the simplest interpretation puts Cdc42 signaling downstream of DPP. This is supported by the ability of constitutively active Cdc42 to partially rescue of mutations in the DPP receptor, thickveins (Ricos et al., 1999).

Expression of dominant-negative Rho^N19 also disrupts dorsal closure. However, it is unclear how Rho fits into the pathway described above. Thus, Rho^N19 expression results in loss of actin and nonmuscle myosin II from the leading edge, but only in leading edge cells that flank the segment borders (Harden et al., 1999). Alternatively, targeted expression of this Rho construct in the leading edge disrupts elongation of all leading edge cells (Lu and Settleman, 1999). This defect is similar to that seen in JNK pathway mutants, however dominant negative Rho does not disrupt dpp expression, suggesting that JNK pathway activation is unaffected. A role for Rho in dorsal closure is also supported by genetic evidence; zygotic mutations in the RhoA gene (Rho1 – FlyBase) result in the puckering of the dorsal midline (Magie et al., 1999). Depletion of maternally contributed Rho leads to severe defects early in development, suggesting that under normal conditions there is sufficient maternal contribution of Rho to affect an aberrant dorsal closure.

Rho is known to activate cellular contractility by modulating phosphorylation of the nonmuscle myosin II regulatory light chain (reviewed by Bresnick, 1999) and interacts genetically with nonmuscle myosin II in Drosophila (Halsell and Kiehart, 1998; Winter et al., 2001). As endogenous RhoA is ubiquitously expressed during embryogenesis (Hariharan et al., 1995), this suggests that Rho could function in dorsal closure by activating contraction within either the amnioserosa and/or leading edge actomyosin purse-string. Alternatively, in cultured vertebrate epithelial cells, Rho is required for formation and maintenance of E-cadherin mediated cell junctions (Braga et al., 1999; Braga et al., 1997; Zhong et al., 1997), suggesting that Rho may play a more fundamental role in the development of the Drosophila epidermis. Thus, a reduction in Rho function could disrupt the integrity of both the amnioserosa and epidermis through effects on the cytoskeleton and/or cell-cell junctions, thereby causing a defect in dorsal closure.

We have focused on the role of Drosophila RhoA in epidermal development by expressing the dominant-negative DRhoA^N19 construct (Strutt et al., 1997) in stripes of epidermal cells. This form of RhoA is thought to remain in an inactive GDP-bound state that binds to and sequesters endogenous guanine nucleotide exchange factors, thereby preventing activation of endogenous Rho. We confirm that striped epidermal expression of DRhoA^N19 causes dorsal closure to fail and we show that actin and nonmuscle myosin II localization is disrupted in all the leading edge cells that express RhoA^N19. Surprisingly, we find that rather than blocking the JNK transcriptional activation pathway, the inhibition of RhoA function leads to its ectopic activation in the lateral epidermis. RhoA^N19 expression also severely affects the integrity of the ventral epidermis. Live imaging shows that ventral epidermal cells lose adhesiveness, frequently fall out of the epithelial cell layer and appear to undergo apoptosis. By contrast, dorsal cells alter their adhesive properties, such that they form ectopic adhesions with cells within adjacent RhoA^N19-expressing stripes via extensive filopodia. We show that both phenotypes are associated with loss of cell polarity and Drosophila E-cadherin from the surface of RhoA^N19-expressing cells. These observations suggest that RhoA has distinct repertoires of function in different epidermal domains. Within the ventral and lateral epidermis, RhoA is essential for proper actomyosin cytoskeletal organization, is involved in regulating gene transcription and promotes cell-cell adhesion; within the leading edge, RhoA is required for the formation of the leading edge purse-string and acts to restrict inappropriate cell-cell adhesion.

The following mutations and transgenes were used in this study: p[w⁺,enGAL4] and p[w⁺ prdGAL4] (Brand and Perrimon, 1993); p[w⁺,UAS-GMA] (Bloor and Kiehart, 2001); p[w⁺,RhoA^N19] and RhoA^72O (Strutt et al., 1997); p[w⁺,UAS-p35] (Zhou et al., 1997); and the β-gal enhancer trap puc^E69 (Ring and Martinez Arias, 1993).

Single or double combinations of transgenes were expressed in epidermal stripes by crossing either w;p[w⁺,enGAL4] or w;p[w⁺,prdGAL4] virgins to males of the following genotypes: w;p[w⁺,UAS-GMA], w;p[w⁺,RhoA^N19], w;p[w⁺,UAS-GMA];p[w⁺,RhoA^N19] or w;p[w⁺,UAS-p35];p[w⁺,RhoA^N19]. The puc^E69 enhancer trap was co-expressed with RhoA^N19 by crossing p[w⁺,enGAL4]/+; puc^E69/+ males to w;p[w⁺,RhoA^N19] virgins. The presence of puc^E69 was detected by assaying expression of β-gal; RhoA^N19 expression was inferred by detection of epidermal stripes showing actin organization defects. Crosses were cultured in small population cages and eggs were collected on grape juice plates and aged at 25°C.

Further information can be obtained from FlyBase (http://flybase.bio.indiana.edu/).

Two-hour egg collections were aged for 46 hours at 25°C. Unhatched, fertilized embryos were hand dechorionated and transferred to an agar slab where they were appropriately oriented. They were transferred to an embryo glue coated cover slip, covered with Hoyer’s medium and placed on a glass slide. A small weight was placed over the cover slip and the slides were incubated at 65°C for 24 hours to allow embryonic tissues to clear and cuticles to flatten.

Antibody staining was performed on whole embryos using standard methods with the exception of staining with the anti-nonmuscle myosin II heavy chain antibody, 656, which was performed as described (Bloor and Kiehart, 2001; Young et al., 1993).

Staining with rhodamine-phalloidin also followed standard methods with the exception that dechorionation was performed with 80% ethanol. Embryos were incubated for 2 hours at room temperature in a solution containing a rhodamine-phalloidin concentration of 4×10^–4 units/μl.

The following primary antibodies were used in this study: a 1:1000 dilution of rabbit anti-nonmuscle myosin II heavy chain, 656 (Kiehart and Feghali, 1986; Kiehart et al., 1990), a 1:1000 dilution of mouse anti-β-gal (Roche Diagnostics Corporation, Indianapolis, IN), a 1:125 dilution of rat anti-DE-cadherin, DCAD2 (Oda et al., 1994) and a 1: 25 dilution of rabbit anti-β_Heavy-spectrin, 243 (Thomas and Kiehart, 1994). Affinity-purified, directly conjugated fluorescent secondary antibodies from commercial sources were used at the following concentrations: FITC-conjugated anti-mouse, 1:1000; rhodamine-conjugated anti-rabbit, 1:1000; CY3-conjugated anti-rat, 1:2000.

Digital images of cuticle preparations were collected on a Spot camera (Diagnostic Instruments, Sterling Heights, MI) mounted on a Zeiss Axioplan using a 20×, 0.5 NA objective. Confocal microscopy for fixed specimens was performed either with a Zeiss LSM410 and Axiovert S100 TV or a Zeiss LSM510 and Axioplan 2 using 25×, 0.8 NA, 40×, 0.9 NA and 63×, 1.4 NA objectives. Live embryos were imaged using a modified Teflon window chamber (Kiehart et al., 1994) and a Perkin Elmer Ultraview Confocal Scanner fitted to a Zeiss Axioplan with a 40×, 1.2 NA objective. Images were collected every 30 seconds on a Sony Ultrapix CCD camera. Images were processed in Adobe Photoshop (Adobe Systems, San Jose, CA); videos were processed using NIH Image (http://rsb.info.nih.gov/nih-image/).

We have investigated RhoA function in the developing embryonic epidermis of Drosophila by using the GAL4/UAS system (Brand and Perrimon, 1993) to express the dominant-negative RhoA^N19 construct in stripes of epidermal cells. This approach has two advantageous features. First, RhoA^N19 expression should be limited to the developing epidermis; secondly, in fixed samples the phenotype of RhoA^N19-expressing epidermal cells can be directly compared with wild-type epidermal cells within the same embryo.

In this study, we used two GAL4 drivers, enGAL4 and prdGAL4, to drive expression in epidermal stripes. To test whether expression driven by these GAL4 lines is limited to the epidermis, we first used them to express GMA, an actin marker in which GFP is fused to the Drosophila Moesin actin-binding domain (Bloor and Kiehart, 2001; Edwards et al., 1997; Kiehart et al., 2000). By following GFP fluorescence in live wild-type embryos, we find that both GAL4 lines drive expression in epidermal stripes from before germband extension through dorsal closure (Fig. 1A-C). The enGAL4 driver directs GMA expression in 14, narrow epidermal stripes, while prdGAL4 driven GMA expression remains in eight wide stripes throughout dorsal closure. In neither case does GMA expression exactly recapitulate the expression pattern of the endogenous genes, most probably because of the high stability of GMA (Edwards et al., 1997). In addition, both GAL4 lines drive variable GMA expression in some amnioserosa cells, although this is more apparent with prdGAL4 (Fig. 1B,C; see Movie 1). Prior to germband elongation, both en and prd are expressed in the dorsal ectoderm (DiNardo et al., 1985; Kilchherr et al., 1986), the tissue fated to become the amnioserosa. This amnioserosal expression is again likely to be an effect of GMA stability. Interestingly, the amnioserosal cells that express GMA always lie above and continue dorsally the epidermal stripe of GAL4-driven expression (Fig. 1B,C). Furthermore, during germband retraction they move in register with their corresponding epidermal stripe (see Movie 1). Importantly however, as dorsal closure begins, amnioserosal expression fades, leaving a pattern of stripes restricted to the epidermis.

At the beginning of dorsal closure GMA expression, driven by either GAL4 line, clearly demonstrates a parallel band of actin at the leading edge, as well as dorsally extending filopodia (Movie 2). Leading edge cells, followed by lateral epidermal cells, then extend dorsally as closure proceeds. As opposing stripes come together at the dorsal midline, leading edge filopodia from opposite epidermal hemispheres appear to adhere and rapidly pull the epidermal fronts together as previously reported (Jacinto et al., 2000). However, we observe frequent initial mismatches between opposing stripes (Fig. 1D; see Movie 2). These are always resolved, but resolution can occur after dorsal closure has extended well beyond the mismatched stripes (Fig. 1E-G), suggesting that opposing epidermal cells can adjust their relative positions at the dorsal midline subsequent to closure.

To determine the effect of expressing RhoA^N19 in epidermal stripes, we crossed virgin females homozygous for either enGAL4 or prdGAL4 to males homozygous for UAS-RhoA^N19. In both cases no embryos were observed to hatch (n=123 and 219 embryos, respectively), demonstrating that striped expression of RhoA^N19 causes embryonic lethality. Cuticle preparations show that each GAL4 line generates a similar, but distinct terminal phenotype (Fig. 2A-F). All embryos have a large anterior hole, indicating failure of head involution and reflecting a GAL4-driven stripe of expression in the head ectoderm, as well as variable lateral, ventral and dorsal defects. Regardless of the GAL4 driver used, ∼40% of embryos have a hole in the dorsal cuticle (Fig. 2A,B,E), indicative of a disruption in dorsal closure. The rest exhibit puckering at the dorsal midline (Fig. 2C,F), which probably represents a defect in the suturing stage of dorsal closure. GAL4-specific differences in phenotype relate to the magnitude of the dorsal and ventral defects. RhoA^N19 expression driven by enGAL4 generates small ventral holes (Fig. 2D) and posteriorly positioned dorsal holes (Fig. 2A,B) that range in size from single segments to the length of the embryo. By contrast, embryos that express RhoA^N19 under the control of prdGAL4 exhibit wide ventral holes that extend laterally, frequently merging with the lateral defects (Fig. 2F). Furthermore, the dorsal holes exhibited by these embryos, while still posteriorly positioned, never extend the length of the embryo. These differences probably reflect the different expression patterns of the two GAL4 lines. Interestingly, with both GAL4 lines we observe a reverse correlation between the size of the dorsal cuticular hole and that of the ventral holes: the larger the dorsal hole, the smaller the ventral holes (compare Fig. 2A,B with 2C,D and compare 2E with 2F).

As RhoA regulates actin organization and activation of nonmuscle myosin II, disruption of its function during dorsal closure might be expected to affect localization of actin and nonmuscle myosin II both at the leading edge and in the lateral epidermis. To test this hypothesis, we first stained embryos expressing RhoA^N19 via prdGAL4 or enGAL4 with rhodamine-phalloidin to visualize F-actin distribution (Fig. 3). At germband retraction stages, cell shape is clearly abnormal within stripes of RhoA^N19 expression (Fig. 3A,B). RhoA^N19-expressing epidermal cells possess apical filopodia and appear generally larger and more rounded than cells in stripes of wild-type tissue. By comparing optical, confocal sections at the surface (Fig. 3A) with sections deeper within the epithelium (Fig. 3B), it is clear that some cells at the surface have flattened and spread over underlying RhoA^N19-expressing cells and neighboring wild-type epidermis. After completion of germband retraction, but prior to the formation of the leading edge actomyosin purse string, RhoA^N19-expressing leading edge cells put out extensive filopodia (Fig. 3C) and these become more prominent as dorsal closure progresses (Fig. 3D-G). During dorsal closure, wild-type leading edge cells possess a clear band of actin that runs parallel to the leading edge. In addition, small spikes of staining (under these fixation conditions, we observe approx. two for every five wild-type leading edge cells, averaging ∼4 μm in length, n=7) extend perpendicular to the leading edge, presumably representing leading edge filopodia. The leading edge cells within stripes of RhoA^N19 expression do not possess an organized band of actin. However, filopodal spikes are both more common (averaging one per cell) and larger (averaging 8.4 μm, n=12).

Staining these embryos with an antibody that specifically recognizes nonmuscle myosin II heavy chain demonstrates that RhoA^N19 expression also causes defects in nonmuscle myosin II localization (Fig. 4). This is first seen at stage 11, where RhoA^N19-expressing mesectodermal cells fail to invaginate at the ventral midline (Fig. 4A) and instead spread along the midline groove generated by the invagination of their wild-type neighbors. Within these cells, nonmuscle myosin II staining is partially lost from the cell cortex and is more cytoplasmic (Fig. 4B). This defect in localization is more distinct in later embryos undergoing dorsal closure (Fig. 4C-F). At these stages and particularly at the leading edge, RhoA^N19-expressing cells lose much of their cortically localized nonmuscle myosin II. Instead, although excluded from the nucleus, it is present diffusely within the cytoplasm (Fig. 4C,E). The leading edge cells also show clear differences in cell morphology when compared with wild-type neighbors. They appear to spread dorsally over the amnioserosa (Fig. 4C,E) and laterally, such that connections are made between adjacent RhoA^N19-expressing stripes (Fig. 4D,F).

The effect of RhoA^N19 on the organization of actin and nonmuscle myosin II in the leading edge could be due to disruption of the JNK pathway. However, it has been reported that expression of RhoA^N19 within the leading edge has no effect on JNK-dependent dpp expression (Lu and Settleman, 1999). The puc^E69 enhancer trap directs expression of β-gal exclusively in the leading edge (Fig. 5A) in a manner that is dependent on JNK signaling (Martin-Blanco et al., 1998). Thus, we assayed puc^E69 directed β-gal expression in embryos expressing RhoA^N19 under the control of enGAL4. These embryos express β-gal in all the cells of the leading edge. However, within stripes of RhoA^N19 expression, β-gal can also be detected in cells just lateral to the leading edge (Fig. 5B,C). We have shown that RhoA^N19 expression causes cell rearrangement, thus the presence of β-gal-positive cells in the lateral epidermis could reflect rearrangement of leading edge cells. Alternatively, these cells could be lateral epidermal cells that now ectopically express puc. Each stripe of enGAL4 driven GMA expression is five cells wide at the leading edge. By counting β-gal-positive nuclei in our experimental embryos, we find that up to 10 cells are labeled with β-gal in individual RhoA^N19 expression stripes (Fig. 5C). In other words, the number of β-gal-positive nuclei present within a domain of RhoA^N19 expression can be greater than the number of leading edge cells in that domain. Furthermore, we detected β-gal-positive cells within the ventral epidermis (Fig. 5D). We conclude that, although cell rearrangement is certainly a factor, RhoA^N19 can induce ectopic puc expression. This suggests that in the absence of RhoA function the JNK signaling pathway can become ectopically activated in the lateral and ventral epidermis.

To determine the effect of RhoA^N19 expression on epidermal cell shape and F-actin distribution in real time, we co-expressed GMA and RhoA^N19 using either enGAL4 or prdGAL4. Similar observations were made with both GAL4 lines. At the leading edge the most striking defect is the formation of ectopic cell ‘bridges’ that link stripes of RhoA^N19-expressing cells. In some embryos, these appear during germband retraction (Fig. 6A; see Movie 3). As posterior segments retract, dorsal epidermal cells constrict, allowing them to make a tight turn around the end of the embryo. This constriction does not occur in RhoA^N19-expressing stripes. Thus, when intervening wild-type cells constrict, leading edge cells from RhoA^N19-expressing stripes come into close proximity. Filopodia extending from these cells then initiate ectopic cell contacts. Whether or not these early ectopic contacts occur, extensive and dynamic dorsally oriented filopodia project from the leading edge of RhoA^N19-expressing stripes prior to the start of dorsal closure, consistent with our observations on fixed embryos (Fig. 6B, compare to Fig. 3C). As dorsal closure proceeds, RhoA^N19-expressing leading edge cells spread over the amnioserosa, while continuing to extend long filopodia (Fig. 6C). As in the wild type leading edge, these filopodia continuously probe their embryonic environment. However, they are not restricted in their adhesive behavior. In wild-type leading edge cells, filopodia only make adhesive contacts with their counterparts from the opposite epidermal hemisphere (see Movie 2) (Jacinto et al., 2000). When filopodia from RhoA^N19-expressing stripes touch, either during germband retraction or dorsal closure, they form ectopic cell contacts (Fig. 6D-F; see Movie 4). Thus, when development in these embryos arrests, many stripes of RhoA^N19-expressing cells are connected at the leading edge.

Ventrally, RhoA^N19-dependent defects are first seen at the ventral midline. Normally mesectodermal cells invaginate; however, within RhoA^N19 expression stripes, invagination does not occur (Fig. 4A,B). Instead, actin within these and neighboring cells compacts, reminiscent of cells entering apoptosis (see Movie 5). To test the hypothesis that apoptosis underlies RhoA^N19-dependent loss of ventral epidermal integrity, we co-expressed RhoA^N19 and the apoptosis inhibitor, p35 using the enGAL4 driver. Embryos expressing RhoA^N19 alone exhibit 5.4 ventral cuticular defects per embryo; co-expression of p35 reduced this number to 3.7 (n=24 and 21 respectively, P=0.011). This partial suppression suggests that, while programmed cell death contributes to RhoA^N19-dependent loss of ventral epidermal integrity, other mechanisms also play a role. Cells undergoing apoptosis are clustered close to the ventral midline, slightly more lateral cells are unaffected. However, at later stages these cells occasionally drop out of the epidermal epithelial layer (see Movie 6). Furthermore, RhoA^N19-expressing cells that remain within the epithelium no longer adhere tightly to each other. This is demonstrated by the ability of hemocytes, highly migratory phagocytotic cells, to transmigrate between these cells (see Movie 6) thereby generating or augmenting defects within the epidermal layer. Thus, even in the presence of p35, ventral cuticular defects may result from RhoA^N19-expressing epidermal cells losing contact with each other and falling out of the epidermis.

Clearly RhoA^N19-expressing epidermal cells exhibit altered adhesive properties. Cell-cell adhesion in the Drosophila embryonic epidermis is mediated by DE-cadherin. To test whether the altered adhesive properties of RhoA^N19-expressing cells are accompanied by changes in DE-cadherin mediated cell-cell adhesion, we stained embryos expressing RhoA^N19 via enGAL4 with an antibody that specifically recognizes this adhesion molecule. In embryos undergoing dorsal closure, DE-cadherin is completely lost from the surface of all RhoA^N19-expressing cells both dorsally and ventrally (Fig. 7A-C), although surface staining persists on dorsal epidermal cells longer than it does on ventral cells. Indeed, defects in DE-cadherin localization are observed from stage 11 onwards (Fig. 7D). As DE-cadherin localizes to apical cell junctions (Uemura et al., 1996), loss of cell surface staining could be due to loss of cell polarity. In this scenario, DE-cadherin should become evenly distributed around the cell surface. However, epidermal cells expressing RhoA^N19 no longer stain for DE-cadherin suggesting that failure to localize leads to its degradation. We also stained these embryos with an antibody that recognizes β_H-spectrin, a marker for the apicolateral domain of epidermal cells (Thomas and Kiehart, 1994). We observe clear apicolateral staining in wild-type epidermal cells; however, within stripes of RhoA^N19 expression, β_H-spectrin is lost from this domain (Fig. 7E,F) and uniform weak staining can be observed all around the cell surface. This defect in β_H-spectrin localization is also initially observed at stage 11 (Fig. 7G). Thus, our data suggest that inhibition of RhoA function results in loss of cell polarity, but do not distinguish between this being the result or the cause of DE-cadherin loss from the cell surface.

To determine whether loss of DE-cadherin was also a feature of zygotic RhoA mutants, we stained embryos homozygous for a severe RhoA allele with the DE-cadherin antibody. Maternally contributed RhoA supports early development of these embryos and most complete dorsal closure. Despite this, mutant embryos can be detected by their clear head involution defects and disorganization of the leading edge (Fig. 7H). No defects in DE-cadherin localization are seen at this stage. However, RhoA mutant embryos that have completed dorsal closure exhibit leading edge puckering and this can be associated with loss of DE-cadherin staining (Fig. 7J).

Previous studies on RhoA function during dorsal closure have focused on its role in the formation of the leading edge actomyosin purse-string. Here we expand upon this initial work and show that RhoA is required for proper organization of actin and nonmuscle myosin II throughout the epidermis. In addition we find that inhibition of RhoA causes misregulation of the JNK transcription activation pathway, loss of DE-cadherin from the cell surface and disruption of the apicolateral distribution of β_Heavy-spectrin. We find that these changes in cell physiology have differential effects on cell behavior that depend upon the position of the cell within the dorsal-ventral axis. In particular, cell-cell adhesion in the ventral and lateral epidermis is severely compromised, but at the leading edge RhoA^N19-expressing cells form new, ectopic cell-cell adhesions.

Tension generated in the amnioserosa and the leading edge of the lateral epidermis independently contributes to the forces that drive dorsal closure (Kiehart et al., 2000). It has been proposed that nonmuscle myosin II activation generates tension in the leading edge and that this causes a leading edge intracellular actomyosin purse-string to shorten (Young et al., 1993). Signaling downstream of RhoGTPase activates nonmuscle myosin II by modulating the level of myosin regulatory light chain phosphorylation (reviewed by Bresnick, 1999). As such, expression of RhoA^N19 in epidermal stripes might disrupt contraction of the leading edge purse-string. In our experiments, defects in actin and nonmuscle myosin II organization caused by RhoA^N19 expression are first observed at germband extension, up to 2 hours before purse-string formation. Thus, while actin and nonmuscle myosin II are localized at the leading edge in wild-type tissue, a purse-string structure is never formed in leading edge cells that express RhoA^N19. RhoA^N19 expression therefore effectively cuts the leading edge purse-string at multiple sites. We find that this does not necessarily prevent progression of dorsal closure, confirming previous experiments that demonstrate that the integrity of the leading edge is not required for dorsal closure to continue to completion (Kiehart et al., 2000). We conclude that small independent regions of leading edge in wild-type epidermal stripes can, in conjunction with contraction of the amnioserosa, migrate dorsally with relative normalcy.

The question that arises is how do epidermal cells expressing RhoA^N19 move dorsally in the absence of a leading edge purse-string? These cells could hitchhike, i.e. they are pulled dorsally by the amnioserosa or dragged along with neighboring wild-type cells. We show that, although spread and disorganized, dorsal RhoA^N19-expressing cells do maintain adhesion with wild-type neighbors and this might then allow passive RhoA^N19-expressing cells to move dorsally with wild-type tissue. This is consistent both with the inverse correlation between integrity of the ventral epidermis and the extent to which dorsal closure proceeds, as well as with our observations on the distribution of tension at the embryo surface during dorsal closure (Kiehart et al., 2000). Thus, as the epidermis lateral to the leading edge opposes dorsal closure, ventral failure of epidermal integrity (and hole formation) would release the tensional restraints on the remaining lateral epidermis, allowing it to move dorsally with more success. Similarly, in the absence of this release (i.e. the ventral epidermis retains its integrity and opposes dorsal movement of the epidermis), the leading edge is presumably no longer capable of generating sufficient force to drive dorsal closure to completion.

We demonstrate that RhoA^N19 expression causes ectopic activation of the JNK pathway in the lateral epidermis, suggesting that RhoA normally functions to inhibit JNK signaling. JNK activation is antagonized by the protein phosphatase encoded by puc, and in puc mutants JNK signaling is increased at the leading edge and is activated in the lateral epidermis (Martin-Blanco et al., 1998). Thus, JNK signaling in wild-type embryos is not maximal and basal JNK activity in the lateral epidermis is revealed in the absence of either puc or RhoA mediated repression. Interestingly, RhoA^N19 expression does not increase JNK signaling in the leading edge. This difference between the effect of puc mutations and RhoA^N19 expression could be due to ectopic RhoA^N19 suppressing an upstream JNK activator that is itself maximally activated in the leading edge.

Co-expression of RhoA^N19 and GMA demonstrates that, in addition to effects on the actin cytoskeleton, inhibition of RhoA has profound effects on the adhesive properties of epidermal cells. In accordance with this, we show that DE-cadherin is lost from the surface of RhoA^N19-expressing cells. Rho is required for E-cadherin-mediated epithelial cell-cell adhesion in cultured vertebrate cells: in keratinocytes and MDCK cells, blocking Rho function prevents formation of E-cadherin-based junctions and causes preformed junctions to breakdown (Braga et al., 1999; Braga et al., 1997; Takaishi et al., 1997). This effect is dependent on cell-cell junction maturity; blocking Rho causes E-cadherin to be lost rapidly (within 1 hour) from immature junctions, but E-cadherin can persist for several hours at mature junctions. We also observe this differential affect, as removal of DE-cadherin from the cell surface is not uniform throughout the RhoA^N19-expressing stripe; ventral cells lose surface staining sooner than dorsal cells. This most probably reflects regional differences in the maturity of epidermal cell junctions. Cells of the dorsal epidermis form a compact epithelium early in stage 10, while neuroblast delamination in the ventral neurectoderm delays formation of the ventral epidermis proper until well into stage 11, by which time enGAL4- and prdGAL4-driven protein expression is apparent. Alternatively, the differential effect might be due to a dorsoventral gradient in enGAL4- or prdGAL4-driven expression of RhoA^N19. This seems unlikely, as no regional differences in fluorescence are observed when these GAL4 lines are used to drive GMA expression.

Drosophila E-cadherin is encoded by shotgun (shg) (Tepass et al., 1996; Uemura et al., 1996). If the primary defect associated with the epidermal expression of RhoA^N19 is loss of DE-cadherin, then the phenotypes induced by RhoA^N19 should phenocopy those of shg mutants. The defects exhibited by shg embryos are difficult to compare with those shown by embryos expressing RhoA^N19 in epidermal stripes. However, genetic analysis of shg demonstrates that the embryonic dorsal epidermis is less sensitive than the presumptive ventral epidermis to a reduction in DE-cadherin levels (Tepass et al., 1996; Uemura et al., 1996); embryos mutant for null shg alleles are missing head and ventral cuticle, while the dorsal cuticle appears unaffected. Thus, as with epidermal expression of RhoA^N19, shg mutants disrupt ventral epidermal integrity and cells undergo apoptosis, while dorsally epidermal cells remain adherent and secrete cuticle. However, there is at least one clear distinction between the genetic reduction of DE-cadherin and the defects induced by expression of RhoA^N19. That is, both null and dominant-negative mutations in shg do not affect epithelial cell polarity (Tepass et al., 1996), while inhibition of RhoA activity does. It seems likely that this difference reflects the additional function of RhoA in generating cell polarity, possibly through organization of the actin cytoskeleton. We conclude that the epidermal defects caused by RhoA^N19 expression cannot be explained simply on the basis of loss of DE-cadherin mediated adhesion.

Maintenance of dorsal epithelial integrity in the absence of detectable surface DE-cadherin suggests that a secondary adhesion system must function in the dorsal epidermis. As in vertebrate cells different classical cadherins exhibit cell type dependent sensitivity to Rho inhibition (Braga et al., 1999) this could involve another member of the cadherin family. The possibility that two cell-cell adhesion systems function in the dorsal epidermis may explain the behavior of RhoA^N19-expressing cells. Embryos that express RhoA^N19 in epidermal stripes differ from shg mutants in the uniformity of DE-cadherin loss from the cell surface. In shg mutants, zygotic DE-cadherin is lost from all cells, while striped expression of RhoA^N19 results in two populations of dorsal epidermal cells, those with cell surface DE-cadherin and those without. Differential adhesion properties of cell populations are the molecular basis for the classical phenomenon of cell sorting. Thus, the ectopic cell bridges formed by RhoA^N19-expressing cells in our experimental embryos could be due to activation of a cell sorting mechanism between populations of dorsal epidermal cells that associate via different adhesion molecules.

In summary, by expressing the dominant negative RhoA^N19 construct in epidermal stripes in the developing Drosophila embryo, we show that the small GTPase Rho has multiple functions throughout this tissue. It is therefore clear that experiments designed to test the function of Rho family GTPases in the epidermis cannot be interpreted simply in terms of their effect on the leading edge. The challenge ahead lies in dissecting the pathways downstream of Rho and determining how these Rho-dependent processes contribute individually to epidermal function and morphogenesis.

We thank Amy Bejsovec, Olga Nikiforova, Marek Mlodzic and Masatoshi Takeichi for the gift of fly stocks and antibodies. We thank Jackie Swain, Rick Fehon, Amy Bejsovec and Paul Martin for critical reading of the manuscript. This work was supported by NIH grants GM33830 and GM61240, as well as grants from the March of Dimes and Muscular Dystrophy Association awarded to D. P. K.