The Rac GTPase-activating Protein RotundRacGAP Interferes with Drac1 and Dcdc42 Signalling in Drosophila melanogaster *

RhoGTPases are negatively regulated by GTPase-activating proteins (GAPs). Here we demonstrate thatDrosophila RotundRacGAP is active in vitro on Drac1 and Dcdc42 but not Drho1. Similarly, in yeast, RotundRacGAP interacts specifically with Drac1 and Dcdc42, as well as with their activated V12 forms, showing a particularly strong interaction with Dcdc42V12. In the fly, lowering RotundRacGAP dosage specifically modifies eye defects induced by expressing Drac1 or Dcdc42 but not Drho1, confirming that Drac1 and Dcdc42 are indeed in vivotargets of RotundRacGAP. Furthermore, embryonic-directed expression of either RotundRacGAP, or dominant negative Drac1N17, transgenes induces similar defects in dorsal closure and inhibits Drac1-dependent cytoskeleton assembly at the leading edge. Expression of truncated forms of RotundRacGAP shows that the GAP domain of RotundRacGAP is essential for its function. Unexpectedly, transgenes encoding Drac1N17, Dcdc42N17, or RotundRacGAP do not affect the c-Jun N-terminal kinase-dependent gene expression ofdecapentaplegic and puckered, indicating that another Drac1-independent signal redundantly activates this pathway. Finally, in a situation where Drac1 is constitutively activated, RotundRacGAP greatly reduces the ectopic expression ofdecapentaplegic, possibly by negatively regulating Dcdc42.

class of regulatory proteins that bind activated forms of RhoG-TPases and stimulate GTP hydrolysis, thus down-regulating RhoGTPase-mediated signals (5)(6)(7). GAPs may also serve as effector molecules and play a role in signaling downstream of Ras-like GTPases (8 -11).
In Drosophila, the RhoGTPases are specifically required in many developmental events involving actin-mediated cell shape changes and the control of gene transcription. In particular, misexpression of RhoGTPases during eye development causes defects in ommatidial patterning (12)(13)(14)(15)(16)(17)(18). During embryogenesis, Drac1, Dcdc42, and Drho1 specifically contribute to the elongation of the epidermis during dorsal closure (for review, see Refs. 19 -21). Dorsal closure is a morphogenetic event of mid-embryogenesis in which the lateral epidermis on either side of the embryo stretches dorsally to cover the degenerative epidermis called amnioserosa and finally encloses the embryo in a continuous epidermis (22,23). The two major distinct signaling cascades operating during dorsal closure are a c-Jun N-terminal kinase (JNK) cascade (24 -34) and a transforming-growth factor-␤ (TGF␤) pathway (35)(36)(37)(38). The JNK pathway consists of a set of sequentially activated kinases that transmits a cytoplasmic signal to the nucleus, which generates, in cells situated at the leading edge, both the accumulation of an acto-myosin cytoskeletal network at the dorsal membrane (23,39) and the activation of target genes. Two major JNKtarget genes required for dorsal closure are puckered (puc), encoding a phosphatase that participates in a feedback loop negatively regulating JNK activity (22,32), and decapentaplegic (dpp), encoding a TGF␤ homologue. In mammalian cell systems (40) and during dorsal closure in Drosophila (24,41,42), Rac1 and Cdc42, but not Rho (41), function as upstream activators of the JNK cascade.
The rotund (rn) locus from Drosophila encodes a protein, which has been identified as a RacGAP based on 30 -50% similarity in its C-terminal region with the GAP domain of several RhoGAP proteins, including the human n-Chimaerin (6,43,44). RotundRacGAP (RnRacGAP) presents 60% similarity with its closest human homologue Male Germ Cell RacGAP (MgcRacGAP) protein, which specifically enhances the GTPase activity of Rac1 and Cdc42 but is inactive on Rho (45).
In this study, we provide evidence that RnRacGAP displays GAP activity in vitro toward Drac1 and Dcdc42, but not Drho1, and interacts with these same two RhoGTPases in a two-hybrid assay in yeast. The existence of a specifically strong interaction with the activated form Dcdc42V12 suggests an additional function for RnRacGAP in Dcdc42 signaling. We further demonstrate by genetic analysis that changes in RnRacGAP dosage interfere with eye development and embryonic dorsal closure in the fly, two processes controlled by the RhoGTPases Drac1 and Dcdc42. * This work was supported by the "Association pour la Recherche contre le Cancer" (grant ARC 9392) and by the Ministère de la Recherche et de l'Enseignement (grant ACCSV-4 and fellowship to K. R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Production and Isolation of Recombinant Baculovirus and Time Course Analyses of rnRacGAP Transcript and Protein Production-A
1296-base pair fragment, including the entire rnRacGAP coding sequence, was generated by XmnI-NspV enzymatic digestion of the rnRacGAP(2) cDNA (46) and inserted into the pVL1393 baculovirus transfer vector under the control of the polyhedrin promoter. Recombinant baculovirus were produced according to the method proposed by Invitrogen Corp. (Kit Maxbac). For the time course analyses of rnRac-GAP RNA and protein production, Sf9 cells were infected with recombinant baculovirus purified by at least two rounds of plaque purification, and as control, wild type virus. At various intervals of infection (24,36,48,60,72, and 96 h), cells were harvested and pelleted by centrifugation (10 min at 2000 ϫ g). Total RNA was isolated by the RNAϩ method (Quantum) and analyzed by Northern blot analysis. For protein analysis, similarly prepared cells were directly resuspended in SDS sample buffer (125 mM Tris, 20% (v/v) glycerol, 1% (w/v) SDS, 2% (v/v) ␤-mercaptoethanol, and 0.001% (w/v) bromphenol blue, pH 6.8). Protein extracts were then analyzed by electrophoresis on SDS-polyacrylamide gel electrophoresis gels containing 12% polyacrylamide, followed by Coomassie blue staining. Cytosols were prepared from cell extracts by centrifugation for 30 min at 100,000 ϫ g at 4°C.
GAP Activity Assay-Cloned cDNAs in pGEX-2T vector encoding the recombinant GST-Dcdc42, GST-Drac1, and GST-Drho1 proteins (41,47) were transformed into the Escherichia coli BL21 strain. Proteins were expressed and purified by affinity on glutathione-Sepharose beads (Amersham Pharmacia Biotech) using classical protocols. The GTPases were then cleaved from GST by 4-h incubation at 4°C in the presence of 1 unit of thrombin (Sigma Chemical Co.) per 100 ml of culture. Thrombin action was blocked by 1 mM diisopropyl fluoro phosphate. The GTPases were loaded with 10 M GTP (containing 2 Ci of [␥-32 P]GTP in the presence of 10 mM EDTA as described previously (48)). Aliquots of the assay medium were loaded onto a nitrocellulose filter under vacuum and washed with 5 ml of Tris-HCl, 20 mM, pH 7.5/MgCl 2 , 5 mM (washing buffer). Binding of GTP was followed for 5 min at 30°C and found to be maximal at 2.5 min, ranging from 0.5 to 1 mol of GTP/mol of GTPase. After loading with GTP for 2.5 min, spontaneous and stimulated GTPase activities were initiated by adding MgCl 2 to 20 mM. Aliquots of the assay medium were incubated either with the RnRac-GAP-expressing Sf9 cytosol, a negative control cytosol, GST-human BCR-GAP domain, or GST-p50 Rho-GAP domain as positive controls, or with the washing buffer alone. All samples were run in parallel for a given GTPase. The amount of remaining GTP-bound GTPase was assessed by filtration under vacuum and liquid scintillation counting of the filters.
Two-hybrid Analysis-The rnRacGAP cDNA was amplified by polymerase chain reaction using the following primers: 5Ј-GCCTTAA-GAATACTAATCACCGAGACCCTC-3Ј from rnRacGAP cDNA and T7 primer, and subcloned into the pGAD3S2X vector (digestion EcoRI/ NotI). Yeast two-hybrid assays were performed as described previously (49). Detection of LexA-fused protein by Western blot analysis was performed with anti-LexA antibodies (CLONTECH) using standard procedures.
Fly Strains-All flies were maintained on standard medium at 25°C except where indicated. Full-length rnRacGAP cDNA (46,50) and the truncated forms UAS-Zn (encoding from amino acids 23 to 166) and UAS-GAP (encoding from amino acids 148 to 383) were cloned into the pUAST vector (51), and transgenic strains were generated following standard methods (52). The rn mutant used in this study is the deficiency rn 20 (43), which deletes the entire rn locus. The rn locus itself includes at least two transcription units: rnRacGAP and a second transcript, originally reported to be 5.3 kb in size, for which the size and direction of transcription have been recently revised. 2 RNA in Situ Hybridization-Embryonic distribution of dpp was determined by in situ hybridization to whole-mount embryos adapted from a previous study (53). Antisense probes labeled with digoxigenin-UTP were generated using the full-length dpp cDNA (gift from M. Affolter) cloned into the pBluescript vector.
Immunohistochemistry-Anti-␤-galactosidase staining was performed as described in (54) to detect expression of ␤-galactosidase in the puc-lacZ enhancer trap line puc E69 (22). Anti-Fas staining was performed as described previously (55). 7G10 anti-Fasciclin III monoclonal antibodies were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD (National Institutes of Health) and maintained by The University of Iowa, Department of Biological Sciences (Iowa City, IA). Anti-phosphotyrosine (PY) staining was performed with anti-PY antibodies diluted 1:500 (Transduction Laboratories) and Texas-red conjugated secondary antibodies (Vector laboratories). For F-actin staining, TRITC-labeled phalloidin (Sigma) was used at a final concentration of 1 g/ml. Fluorescent staining was viewed on a Leica TCS/SP2 confocal laser scanning microscope.

RESULTS
In Vitro Activity of RnRacGAP toward RhoGTPases-To determine the extent and specificity of the potential GAP activity of RnRacGAP, its expression was optimized in a baculovirus insect-cell system. Sf9 insect cells were infected with recombinant baculovirus carrying rnRacGAP cDNA, and cell extracts were generated at the post-infection times indicated (Fig. 1, A and B) to follow rnRacGAP expression and protein production. rnRacGAP RNA (Fig. 1A) and protein (arrowhead; Fig. 1B  RnRacGAP cytosol stimulated the GTPase activity of Drac1 and Dcdc42 as efficiently as the GAP domain from human BCR (Bcr-GAP) (Fig. 1C). In contrast, it had virtually no effect on Drho1 whereas the GAP domain from human p50-RhoGAP was active (Fig. 1C). Control cytosol displayed no GAP activity on any of these three RhoGTPases (data not shown). Thus, Rn-RacGAP displays the same specific GAP activity toward Drac1 and Dcdc42 as its closest human homologue MgcRacGAP giving further support to the hypothesis that they are true functional homologues (45).
Yeast Two-hybrid Tests for Direct Interaction between Rn-RacGAP and Rho GTPases-To test for direct interaction between RnRacGAP and Drac1, Dcdc42, and Drho1, we carried out a yeast two-hybrid interaction assay (49). Fusions of the Gal4 activation domain with RnRacGAP were expressed in yeast together with fusions of the LexA-DNA binding domain with either Drac1, Dcdc42, or Drho1, or with the corresponding mutated forms Drac1V12, Dcdc42V12, and Drho1V14, each of which is constitutively blocked in an active conformation (41,47). Direct interaction of RnRacGAP with these target proteins was measured as the ability of transformed yeast to activate the transcription of the histidine selector gene and the lacZ reporter gene, both under the control of the LexA-binding sequences. None of the constructs alone was able to activate the two genes allowing histidine prototrophy or lacZ expression (not shown).
Coexpression of RnRacGAP with Drac1, Dcdc42, Drac1V12, or Dcdc42V12 allowed yeast to grow on the selective medium without histidine ( Fig. 2A, ϪUWLH). However, only the coexpression of RnRacGAP with Dcdc42V12 allowed the activation of the reporter gene lacZ ( Fig. 2A, ϪUWLϩX-Gal), although Drac1V12 and Dcdc42V12 proteins were expressed in yeast at a similar level (Fig. 2B). There was no detectable interaction between RnRacGAP and either Drho1 or Drho1V14 ( Fig. 2A), although these two forms of Drho1 were expressed in cotransformed yeast (Fig. 2B). The A35 mutation affects the RhoGT-Pase effector loop known to be essential for interaction with specific effectors, including RacGAP proteins (56). As expected, the two mutant forms Drac1V12A35 and Dcdc42V12A35 were not capable of interacting with RnRacGAP and did not grow on selective medium without histidine ( Fig. 2A, ϪUWLH). Thus, this result validates the specificity of the interaction observed with Drac1V12 and Dcdc42V12, although RnRacGAP interacts more strongly with Dcdc42V12 than with Drac1V12.
In Vivo Function of RnRacGAP with Respect to Rho GTPase Activity-Flies homozygous for the rn 20 deficiency, which deletes the rnRacGAP transcript sequence, together with other genetic units (see "Experimental Procedures"), present small rough eyes suggesting a possible role for RnRacGAP in eye development (57). Indeed, defects in retinal organization displayed by rn 20 /rn 20 flies were rescued by the expression of a transgene carrying genomic rnRacGAP-encoding sequence (58). To investigate whether RnRacGAP dosage would affect RhoGTPase activities in the eye, we analyzed the effect of the deficiency rn 20 on the phenotypes induced by expression of wild type Drac1, Dcdc42, or Drho1 under the control of the eyespecific promoter GMR in the transgenic flies GMRRac1, GM-RCdc42, and GMRRho1ϩGMRRho3, respectively (12). Although lowering RnRacGAP dosage in vivo in rn 20 /ϩ hemizygous flies does not affect eye morphology (not shown) compared with wild type eyes (Fig. 3, A and B), it strongly enhances the effects of expressing Drac1, consistent with Rn-RacGAP being a negative regulator of Drac1. Compared with GMRRac1/ϩ flies, which are moderately affected (Fig. 3, C and D), GMRRac1/rn 20 flies (Fig. 3, E and F) present eyes that are dramatically reduced, with flat ommatidia and almost no hairs. In the case of Dcdc42, the data are more complex and may reveal a dual function of RnRacGAP on Dcdc42 activity. On the one hand, observation with the binocular microscope (not shown) and lower magnification scanning electron microscopy showed the eyes of GMRCdc42/rn 20 flies to be systematically less rough and larger (Fig. 3J) when compared with those of GMRCdc42/TM3,Sb flies (Fig. 3G). On the other hand, ommatidial packing in GMRCdc42/rn 20 flies was more affected compared with that in GMRCdc42/TM3,Sb flies; in particular, ommatidia were much less closely associated with one another, resulting in contiguous gaps or channels between them ( Fig. 3; compare H and I with K and L). In contrast to Drac1 and Dcdc42, no change in the phenotype induced by Drho1 was observed in an rn 20 /ϩ context (data not shown), consistent with the absence of any direct interaction in the 2-hybrid system and the absence in vitro of GAP activity of RnRacGAP toward Drho1.

Embryonic Expression of RnRacGAP or Dominant Negative Drac1N17 or Dcdc42N17 Induces Similar Defects in Dorsal
Closure-Drac1 and Dcdc42 signaling are required to complete embryonic dorsal closure (59,60). Therefore, we assayed the in vivo action of RnRacGAP on its putative molecular targets Drac1 and Dcdc42 by overexpressing RnRacGAP in embryos using the Gal4-UAS system (51) and dorsal closure as a phenotypic test to look at the specificity of RnRacGAP with respect to these two RhoGTPases. We constructed transgenic lines containing complete or truncated forms of the rnRacGAP cDNA (46,50) under the control of UAS regulatory sequences and  4A) and completed embryogenesis, giving rise to normal larvae and adults, embryonic-directed expression of UASRnRacGAP caused 100% embryonic lethality in all the lines. A correlation between RnRacGAP expression levels and the cuticular phenotypes was observed. Indeed, cuticle examination of dead embryos from the lines in which rnRacGAP reached high levels (not shown) showed that only a residue of the two unfused lateral epidermis sheets remained and that the ventral epidermis was totally absent (Fig. 4B). In contrast, in the line expressing lower amounts of rnRacGAP (not shown), progeny exhibited less severe defects (Fig. 4C). In this case, the cuticular phenotype is clearly interpretable as an anterior open phenotype resulting from defective dorsal closure and absence of head involution. In addition, small ventral holes are frequently observed (not shown). Ventral holes were also found in Dcdc42 mutant embryos (16) suggesting that RnRacGAP also interferes with Dcdc42 function in the maintenance of the integrity of the ventral epidermis. Defects in dorsal closure were also observed with the embryonic epidermal driver 69BGal4 (51) (not shown) and with the leading edge cell-specific drivers pnrGal4 (61) (not shown) and LEGal4 (25) (Fig. 4D). In the latter case, RnRacGAP leading edge-restricted expression induces only about 6% embryonic lethality. Dorsal closure defects induced by high levels of RnRacGAP are reminiscent of those induced by expressing the dominant negative forms of either Drac1 or Dcdc42 under the control of PtcGal4 (Fig. 4, E and F, respectively, and Refs. 59, 60). As for the RnRacGAP transgene, leading edge-restricted expression of Drac1N17 or Dcdc42N17 induces only partial lethality of about 5%, associated with dorsal closure defects (data not shown). Similarities in the observed phenotypes are consistent with RnRacGAP acting as a negative regulator of these two proteins.
The GAP Domain of RnRacGAP Is Necessary and Sufficient to Prevent Dorsal Closure-The RnRacGAP protein contains two major domains: a PKC-C1 domain in the N terminus, including a zinc finger, and a domain homologous with the GAP domain of RhoGAP proteins in the C terminus (44). The N, and C, termini were separately cloned in the pUAS vector and used to generate transgenic lines named, respectively, UAS-Zn and UAS-GAP (see "Experimental Procedures"). The isolated GAP domain is able to produce defects in dorsal closure when expressed under the control of each of the drivers tested previously, as shown for PtcGal4 and LEGal4 (Fig. 4, G and H,  respectively). However, the GAP domain alone is less deleterious than the full protein, because only about half of the Ptc-Gal4/UAS-GAP progeny die as embryos compared with the death of all the PtcGal4/UASRnRacGAP embryos. In addition, Ptc-driven expression of the GAP domain of RnRacGAP induces less severe defects, closely resembling those induced by expressing Drac1N17 or Dcdc42N17 (Fig. 4; compare G with E and F). Ptc-driven expression of the N-terminal part of RnRac-GAP induces 30% lethality with defects restricted to the terminal structures (Fig. 4I), indicating that it affects develop-  mental processes other than dorsal closure. These results further suggest that RnRacGAP overexpression affects dorsal closure predominantly via the enhancement of the GTPase activity of one or several RhoGTPases.
Expression of RnRacGAP Disrupts Drac1-dependent Cytoskeleton Assembly at the Leading Edge-During dorsal closure, cell shape changes initiate at stage 13 in the dorsal-most rows of cells known as the leading edge cells, and their elongation acts as a motor for the subsequent stretching of lateral epidermis (39). Previous studies have shown that the closely related proteins Drac1 and Dcdc42 have separate functions in the organization of the leading edge cell cytoskeleton (42). We used various leading edge-specific markers to determine the kind of defects that are induced by embryonic overexpression of RnRacGAP. We first examined the localization of the glycoprotein Fasciclin III (FasIII) (55) to observe cellular morphology. In wild type stage 13 embryos, anti-FasIII immunostaining reveals leading edge cells that are well-ordered and columnar in shape (Fig. 5A). FasIII is absent from the dorsal-most edge of the leading edge cells (Fig. 5A, arrow) until they touch the leading edge cells of the opposing epithelia and fuse. In embryos overexpressing RnRacGAP, most of the cells along the dorsal midline remain polygonal while some cells either widen or pinch together, individually or in groups of two or three (Fig.  5B, white and black arrowheads). In addition, some cells flanking the amnioserosa show abnormal staining of FasIII on their dorsal sides (Fig. 5B, arrow). Most of the more lateral epithelial cells remain polygonal even late in embryogenesis (not shown).
The epidermal cell shape defects caused by the overexpression of RnRacGAP are accompanied by a partial disruption of the actin cytoskeleton. Normally, F-actin is accumulated at the dorsal side of leading edge cells (Fig. 5C and Ref. 23) in a Drac1-dependent manner (42,59,60). In RnRacGAP-overexpressing embryos, actin staining is very weak, and only small stretches of polymerized actin are sometimes detectable (Fig.  5D, arrow). Finally, we observed a concomitant loss of the triangular phosphotyrosine (PY) nodes (Fig. 5F, arrow) normally present in putative focal complexes at the leading edge dorsal side (Fig. 5E, arrow (60)). The cytoskeletal defects observed are similar to those induced by the expression of Drac1N17, therefore suggesting a role for RnRacGAP in inhibiting Drac1-mediated cytoskeleton assembly during dorsal closure.
Embryos Expressing RnRacGAP, Drac1N17, or Dcdc42N17 Transgenes Show Normal Levels of dpp and puc Expression-JNK cascade signaling is required for the complete assembly of the leading edge cytoskeleton and for the production of both the TGF␤ morphogen Dpp and the phosphatase Puc (25-28, 30 -32). Because JNK activation is controlled by Drac1 (25,26), we speculated that expression of RnRacGAP would disrupt the JNK-dependent gene expression of dpp and puc. Surprisingly, embryonic overexpression of RnRacGAP does not reduce leading edge-specific expression of puc and dpp ( Fig. 6; compare A and D with, respectively, B and E). We tested whether this expression was dependent upon hemipterous (hep), which encodes the JNKK essential for JNK activation (25). Normally expressed in early germ band-retracted embryos, puc and dpp are no longer detectable in either hep 1 mutant embryos (24,25) or in hep 1 mutant embryos overexpressing RnRacGAP (Fig. 6, C and F), indicating that activation of both puc and dpp still requires Hep function. Levels of dpp and puc are also normal in embryos expressing the Drac1N17 or the Dcdc42N17 transgenes and only slightly reduced in embryos expressing both Drac1N17 and Dcdc42N17 (shown for dpp in Fig. 6, G-I). We therefore conclude that the dorsal open phenotype induced by the expression of Drac1N17, Dcdc42N17, or RnRacGAP is attributable to defects in cytoskeletal assembly rather than to the loss of puc and dpp expression.
RnRacGAP Partially Rescues the Expression of Drac1V12 but Not That of Dcdc42V12-To determine to what extent Drac1 and Dcdc42 downstream signaling depends on RnRacGAP function, we examined the cuticle of embryos expressing jointly RnRacGAP and one of the constitutively activated forms Drac1V12 or Dcdc42V12. Embryos expressing Drac1V12 under the control of PtcGal4 are completely disorganized and produce only random pieces of cuticle (Fig. 6J (60)). When RnRacGAP is coexpressed with Drac1V12, one-third of the embryos show very significant cuticular rescue, presenting only limited defects in the anterior region (Fig. 6M). Similarly, the ectopic activation of dpp and puc observed in Drac1V12-expressing embryos (Fig. 6, K and L, and Ref. 25) is significantly rescued by coexpression of RnRacGAP (Fig. 6, N and O). In contrast, neither the strong cuticular defects nor the ectopic expression of dpp and puc induced by Dcdc42V12 expression were rescued by coexpression of RnRacGAP (data not shown). Finally, the ectopic activation of puc and dpp induced by the constitutively active form of the JNKK encoded by hep (UASHep ca (62)) is also not modified by coexpressing RnRacGAP (not shown). Thus, rescue of the ectopic activation of JNK-dependent gene transcription by RnRacGAP operates upstream of Hep.
In one case only, among five UAS-GAP lines tested, expression of UAS-GAP partially rescued the Drac1V12-cuticular phenotype (not shown) but to a lesser extent than the full protein and with less than 5% penetrance indicating that the full protein is required for efficient cuticular rescue. In contrast, expression of the isolated GAP domains reduced dpp ectopic expression to the same extent as the full protein (not shown). The expression of the N terminus of RnRacGAP had no effect on the Drac1V12-induced phenotype. This indicates that the ability of RnRacGAP to lower ectopic dpp activation downstream of Drac1V12 depends upon its GAP domain and, therefore, upon its capacity to down-regulate a RhoGTPase.

Drac1 and Dcdc42
Are Direct Targets of RnRacGAP-We present here the results of independent experimental approaches that are consistent with Drac1 and Dcdc42 being direct targets of RnRacGAP. In vitro, RnRacGAP, like its human homologue MgcRacGAP, activates the spontaneous GTP hydrolysis of both the Drac1 and Dcdc42 RhoGTPases but has no effect on Drho1. In accordance with this function, in the yeast two-hybrid assay RnRacGAP interacts directly with either wild type Drac1 or Dcdc42 or with their mutated V12 forms, which are blocked in the GTP-bound conformation, an interaction that depends on their effector loops. However, Dcdc42V12 was the only activated form that established a stable interaction with RnRacGAP, thereby allowing expression of the non-vital reporter gene lacZ. Enhancement of GTPase activity may result from relatively brief interaction between the two partners, whereas certain effector functions may require more durable interactions such as that observed with Dcdc42V12, thus suggesting that RnRacGAP may act as a Dcdc42 effector in some cases. In the fly eye, the deficiency rn 20 , which deletes the rnRacGAP transcription unit, strongly enhanced the phenotypes induced by overexpressing Drac1 whereas it displayed both suppressor and enhancer functions with respect to specific phenotypes induced by expressing Dcdc42 and had no effect on Drho1-mediated phenotype. All results support the hypothesis that both Drac1 and Dcdc42 are direct targets for RnRacGAP in vivo. Furthermore, in eye development, RnRacGAP behaves as a negative regulator of Drac1 whereas it may assume a dual regulatory/effector function toward Dcdc42, acting either as a GAP or as an effector, depending on the process involved.
Toxic Activity of Embryonic-directed Expression of RnRac-GAP-There is unlikely to be an essential role in dorsal closure for RnRacGAP, because the homozygous deficiencies survive to adulthood (57). Alternatively, the presence of two additional RacGAPs recently described in Drosophila (63, 64) may compensate for the absence of RnRacGAP. Indeed, both are expressed throughout development and, like RnRacGAP, are specific for Drac1 and Dcdc42 (64,65). In contrast to the absence of embryonic defects in loss-of-function mutants, overexpression of RnRacGAP during embryogenesis proved highly toxic. It has already been described that the actin cytoskeleton is particularly sensitive to changes in the dose of rnRacGAP during development (46) and we report here that embryonic overexpression of RnRacGAP leads to embryonic lethality associated with defects in dorsal closure. This toxic activity is likely to result from ectopic negative regulation of the two RhoGTPase proteins Drac1 and Dcdc42 for several reasons. First, the overexpression of RnRacGAP induces dorsal closure defects similar to those induced by expressing the dominant negative transgenes Drac1N17 or Dcdc42N17, consistent with RnRacGAP similarly blocking the corresponding pathways. Second, its isolated GAP domain is capable of producing the same kind of developmental defects as the complete protein indicating that RnRacGAP may act through its ability to enhance the GTPase activity of its target GTPases. Third, embryonic-directed expression of RnRacGAP induces defects in leading edge cytoskeleton assembly similar to those induced by a Drac1N17 transgene, strongly suggesting that RnRacGAP disrupts Drac1 signaling driving cytoskeleton assembly in these cells. In turn, the resulting deregulation of this Drac1-mediated cytoskeletal assembly would preclude the epidermal bunching characteristic of Dcdc42N17-expressing embryos (42); therefore, we cannot exclude an effect of RnRacGAP on Dcdc42-mediated shape changes in segment border cells, which would be masked by this primary disruption of Drac1 signaling. Taken together, these data provide additional evidence supporting the notion that Drac1 and Dcdc42 are direct targets of RnRacGAP; as such, their activities are sensitive to RnRacGAP dosage during fly development.
Drac1V12-mediated JNK Activation May Be Partially Mediated by Dcdc42-The constitutively activated forms Drac1V12 and Dcdc42V12 have both been described as inducing overactivation of the JNK pathway leading to ectopic expression of the two target genes dpp and puc outside the leading edge cells (25). We have shown that Drac1V12-dependent dpp and puc overexpression is rescued by coexpressing RnRacGAP, but not Dcdc42V12. The activated forms of Drac1V12 and Dcdc42V12 are stabilized in the active conformation and completely insensitive to a RacGAP regulator. Therefore, rescue of Drac1V12 is likely attributable to the down-regulation of a target downstream of Drac1. Two lines of evidence suggest that this target may be Dcdc42: first, the isolated GAP domain of RnRacGAP is necessary and sufficient to reduce dpp overexpression to the same extent as the full protein, indicating that GAP activity reduces JNK overactivation; second, RnRacGAP is not able to modify gene expression in embryos expressing Dcdc42V12 or Hep CA (62), indicating that there is no target sensitive to Rn-RacGAP dosage downstream of Dcdc42 or Hep. These data support the hypothesis that Drac1V12 overactivates the JNK pathway partly through overactivation of Dcdc42, and that Dcdc42 is sensitive to negative regulation by high levels of RnRacGAP in this pathway.
At the cuticular level, the isolated GAP domain provides very limited rescue of Drac1V12-induced defects compared with the full-length protein, indicating that the N terminus containing a PKC-C1-like domain may contribute to this function, possibly by driving RnRacGAP to the membrane. Differences in the capacity of the isolated GAP domain to induce rescue of cuticular development versus JNK-dependent gene expression suggest the existence of different effectors in the two Drac1signaling pathways in leading edge cells: one, controlling cytoskeletal rearrangement, the other, JNK activation.
Blocking RhoGTPase Activity Does Not Disrupt JNK Activation in Leading Edge Cells-We observed that embryonic expression of RnRacGAP induces defects in dorsal closure without loss of JNK-dependent gene expression. The same phenomenon was previously observed in Rho1 and Dcdc42 lossof-function mutant embryos and in embryos expressing a Rho1N19 dominant negative transgene (16,41,66). Furthermore, a mutation in Myoblast city (Mbc), encoding a Rac-specific activator, also prevents cytoskeletal rearrangement but not dpp and puc gene expression at the leading edge (15). Taken together, these results indicate that the cellular morphological changes and JNK-dependent gene activation required for dorsal closure are two separable events. Both events were described as being under the control of the RhoGTPases Drac1 and Dcdc42. However, our experiments suggest that blocking Drac1 and Dcdc42 signaling does not inhibit JNK activation, because neither the expression of RnRacGAP nor the expression of the two dominant negative forms Drac1N17 and Dcdc42N17 prevents puc and dpp expression at the leading edge. Another Drac1/Dcdc42-independent JNK activation pathway might exist that would be up-regulated in response to the inhibition of RhoGTPase activity. In fact, several proteins have been described to regulate the JNK pathway independently of Drac1 (67)(68)(69)(70). One of these, the junction protein Canoe (Cno), which colocalizes with Zo-1 at the adherens junctions (67), may activate the JNK in response to abnormal cytoskeletal organization at the leading edge.