Functions of the novel RhoGAP proteins RGA-3 and RGA-4 in the germ line and in the early embryo of C. elegans

Asymmetric cell division is a general mechanism by which to generate two non-identical daughter cells. One of the most-studied systems is the first asymmetric division of the Caenorhabditis elegans embryo(Cowan and Hyman, 2004a; Gönczy et al., 2001; Schneider and Bowerman, 2003). After fertilization, the egg completes the meiotic divisions and extrudes two polar bodies. During the meiosis events, right after sperm entry, the entire cortex of the embryo is highly contractile. The sperm initiates cortical and cytoplasmic movements, which are actin-dependent. These streamings are essential for the establishment of polarity. How they are controlled, however,has remained largely elusive. Upon movement of the maternal pronucleus towards the paternal pronucleus, the posterior part of the cortex smoothens, which is supposedly controlled by the centrosomes brought along by the sperm(Cheeks et al., 2004; Cowan and Hyman, 2004b; Munro et al., 2004). Posterior smoothening and anterior ruffling of the cortex lead to a membrane invagination, called pseudo-cleavage. The two pronuclei meet in the posterior part of the embryo and migrate to towards the middle of the embryo, where mitosis is initiated. Upon centration, cortical ruffling activity ceases, and resumes during anaphase of the first mitosis. After cytokinesis, the larger anterior AB cell and the smaller P1 cell prepare the next division cycle,which is again asymmetric.

The PAR proteins play a crucial role in the setup and maintenance of polarity right after fertilization (Cuenca et al., 2003). Whereas the PAR-3-PAR-6-aPKC complex determines anterior polarity, PAR-1 and PAR-2 occupy the posterior part. The two domains do not intermix and are mutually exclusive. PAR-5 is a 14-3-3 protein that seems to be essential to maintain the two different domains(Kemphues, 2000).

Contractility in most systems involves the acto-myosin system, which is controlled by GTP-binding proteins of the Rho superfamily(Etienne-Manneville and Hall,2002; Glotzer,2005). Rho GTPases are molecular switches that exist in an activated, GTP-bound form and an inactive GDP-bound conformation. The switch between the two conformations is achieved by guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs)(Bernards, 2003; Schmidt and Hall, 2002). In their activated state, the GTPases can interact with and signal to downstream effectors, which in turn control a variety of cell processes. For example,effectors of RhoA (RHO-1 in C. elegans), such as ROCK1 (LET-502)regulate the acto-myosin system. RhoGAPs are particularly interesting proteins, because they often contain multiple functional motifs in addition to the GAP domain that might regulate localization or GAP activity, or might recruit other signalling proteins (Moon and Zheng, 2003). They outnumber RhoGTPases by 4:1 in humans (and 5:1 in C. elegans), so, in principle, the action of a given RhoGTPase could be regulated by at least four different RhoGAPs in different spatial and temporal patterns. Their versatility is exemplified by the ARAP proteins,which contain both a RhoGAP domain and an ArfGAP domain, and can thus serve as converging signalling platforms (Krugmann et al., 2002; Miura et al.,2002).

In oocytes, the acto-myosin network is distributed over the entire cortex. Upon fertilization, due to cortical and cytoplasmic rearrangements, the acto-myosin network becomes restricted to the anterior cortex, giving rise to contractions in the anterior region of the embryo and a smooth posterior part. Local contractility in the posterior region appears to be inhibited by a signal from the centrosome (Cowan and Hyman, 2004b; Munro et al.,2004). Interestingly, a more recent study provides evidence for the RhoGAP CYK-4 being the signal for establishing the posterior PAR-2 domain and causing the retraction of the acto-myosin system to the anterior region of the embryo (Jenkins et al.,2006). How contraction is controlled in the anterior embryo,however, remains elusive.

We investigated the role of two RhoGAP proteins, RGA-3 and RGA-4, in the early C. elegans embryo and in the germ line. Knockdown of the two RhoGAPs by RNA interference (RNAi) resulted in hyper-contractility of the anterior cortex in the zygote. Although no major defects were detected in polarity establishment, we observed fluctuations in the size of the PAR-6 domain at the anterior cortex. RGA-3 and RGA-4 (RGA-3/4) have overlapping functions in the early embryo, because concomitant knockdown of both proteins yielded the strongest phenotype. RGA-3/4 are GAPs for RHO-1, because RHO-1,but not CDC-42, was epistatic to RGA-3/4. The only other characterized RhoGAP in the early embryo, CYK-4, has non-redundant functions with RGA-3/4, because the phenotypes of the triple knockdown were additive. Furthermore, rga-3/4 RNAi treatment [rga-3/4(RNAi)] enhanced the gonadal defects in let-502(sb106) mutants, whereas cyk-4(RNAi) had no effect. RGA-3/4 are required to control the activation of the C. elegans ROCK LET-502 in the early embryo. We provide evidence that the specificity of RHO-1 is determined by the differential use of RhoGAPs during early C. elegans development.

C. elegans was cultured and maintained as described previously(Brenner, 1974) on NGM medium at 20°C unless noted otherwise. The C. elegans strains N2 and the alleles let-502(ok1283) and let-502(sb106)(Piekny et al., 2000) were used for time-lapse studies and for the counting of embryonic lethality rates. Transgenic GFP-expressing lines used in this study are listed below:

The following strains were obtained from the CaenorhabditisGenetics Center (CGC): strain AZ212, which expresses a GFP::H2B histone fusion in the germ line (Praitis et al.,2001); KK866, a strain with GFP::PAR-2(Wallenfang and Seydoux,2000); JJ1473, a strain expressing NMY-2::GFP(Munro et al., 2004); and WH204, a strain expressing GFP::β-tubulin(Strome et al., 2001). JH1512,a strain bearing GFP::PAR-6 fusion (Cuenca et al., 2003) was a gift from Geraldine Seydoux (John Hopkins University Medical School, Baltimore, USA), and SA115, a strain expressing GFP::RHO-1 (Motegi and Sugimoto,2006), was a gift from Fumio Motegi (RIKEN, Kobe, Japan).

For RNAi-by-feeding experiments, plasmid L4440, containing the desired sequence (see Table 1) was transformed into the Escherichia coli strain HT115(Timmons et al., 2001). NGM plates containing 0.5-2.0 mM IPTG and 25 μg/ml carbenicillin were inoculated with transformed HT115 bacteria. The expression of double-stranded RNA (dsRNA) was induced for 6-9 hours at room temperature on plate. Subsequently, L1, dauer larvae or L3 larvae were transferred to these plates. Animals were cultured at 15, 20 or 23°C and their progeny were analyzed.

For RNAi-by-injection experiments, C. elegans genomic or cDNA was used as a template to PCR-amplify the desired sequence(Table 1). PCR products were used as templates for in vitro transcription using T7 polymerase (Promega). dsRNA was produced according to the manufacturer's protocol (Promega),purified by phenol/chloroform extraction and resuspended in 20 μl DEPC-H₂O. dsRNA was injected into the gonad of young adult worms,which were subsequently incubated at 20°C. The progeny of injected animals was analyzed.

For RNAi-by-soaking experiments, L3 larvae were incubated for 24 hours in dsRNA solution in a wet chamber. The animals were transferred onto an agar plate seeded with OP50 bacteria for recovery. The progeny was analyzed.

Embryos were mounted in a drop of M9 buffer (`hanging drop' method) and covered with a cover slip. Embryos were imaged with a Zeiss Axioplan 2 microscope equipped with a Zeiss Axio Cam MRm camera (Carl Zeiss, Aalen Oberkochen, Germany) and a Plan Apochromat 63×/NA1.40 objective. Zeiss Axiovision 3.1 software was used to control hardware and to acquire and process images. Confocal images were captured with a Leica confocal microscope TCS SP2 (Leica, Bensheim, Germany) and an HCX PL APO 63×/1.32-0.6 oil objective. Laser intensities were adjusted to avoid any defects in the development of the embryo. Images were collected at 20-second or 45-second intervals over a period of 15-35 minutes and processed using Adobe Photoshop 7.0.

ImageJ was used to measure the PAR-2 and PAR-6 domains in wild-type and RNAi-treated embryos. Pictures of embryos - from the beginning of polarization until reaching the two-cell stage - were acquired in the Nomarski and the GFP channel. For quantification, the length of the embryo was measured by drawing a line from the anterior to the posterior of the embryo with ImageJ; this line was set to 100%. The length of the respective PAR domain was measured along this line by using the cortical GFP localizations. The ratio between the length of the PAR domain and the length of the entire embryo was determined.

To quantify the GFP signals in wild-type and rga-3/4(RNAi)embryos, either confocal or epifluorescence pictures with the same exposure times were subjected to analysis with ImageJ. The length of NMY-2::GFP patches in the embryo before pronuclear meeting was determined by drawing cortical lines corresponding to the contractile patches with the highest fluorescence signal. The length of these lines was measured.

For quantification of the RHO-1::GFP data, the cortical fluorescence signal was measured. The ratio of the signal between the contractile cortex at the ingressing furrow and the non-contractile cortex was determined.

Young adult worms were cut open to release the gonads in a depression slide and gonads were fixed with 3.7% formaldehyde for 10 minutes at room temperature. After washing with egg buffer, incubation in 0.4% BSA, 0.1%Tween-20 for 10 minutes, followed by another washing step, the gonads were stained with rhodamin-phalloidin (Molecular Probes, 0.01 U/μl) and DAPI(Boehringer Mannheim, 0.5 μg/ml) in a wet chamber for 30 minutes at room temperature. Gonads were mounted onto an agarose pad and examined with a Zeiss Axioplan 2 microscope using a 40× Axioplan objective as described above.

Upon fertilization, the C. elegans zygote segregates specific proteins and RNAs to opposite poles to provide asymmetry and polarity for the subsequent cell division (Schneider and Bowerman, 2003). This polarity establishment involves cytoplasmic streaming and cortical contractions. The nature of the membrane dynamics and the cortical contractions upon fertilization are poorly understood. Therefore,we undertook a candidate approach and screened databases for genes that influenced membrane dynamics, especially cortical movement and membrane ruffling in the fertilized C. elegans oocyte. RNAi of K09H11.3 resulted in a pronounced increase in membrane ruffling and ectopic furrowing after fertilization, and increased pseudo-cleavage(Fig. 1A,B; compare Movies 1 and 2 in the supplementary material), similar to phenotypes that had been reported previously (Kamath et al.,2003; Simmer et al.,2003; Sonnichsen et al.,2005). The membrane dynamics in the RNAi embryos were strongly increased compared with wild type and the pseudo-cleavage was sometimes shifted towards the anterior pole. Upon RNAi, the number of ruffles increased at least two- to threefold over wild type(Fig. 1B). The depth of the ruffles was also increased [rga-3/4 (RNAi): 6.32±2.0 μm;wild type: 3.24±1.57 μm). Most importantly, ruffles formed until pronuclear meeting, which was not the case in wild type(Fig. 1B). Sequence analysis revealed, however, that another gene, Y75B7AL.4, which is 81% identical to K09H11.3 on the amino acid level, was probably also reduced upon RNAi treatment, because our RNAi construct - like those used in other screens -targeted the highly conserved 5′ region of the K09H11.3 transcript (see Fig. S1 in the supplementary material). Both genes encode putative RhoGAP proteins that accelerate GTP hydrolysis on members of the Rho family, namely on RHO, CDC42 and RAC (Fig. 1C). K09H11.3 and Y75B7AL.4 were named RGA-3 and RGA-4 for RhoGAP-3 and RhoGAP-4, respectively. RGA-1 and RGA-2 are RhoGAPs expressed in epithelial cells and are the worm orthologues of mammalian RhoGAP1 (ARHGAP1)and ARHGAP20, respectively (Jenna et al.,2005; Schwarz et al.,2006). RGA-3 and RGA-4 belong to a subfamily of RhoGAPs with their GAP domain located at the N-terminus. Other members of this family comprise the hitherto uncharacterized mammalian ARHGAP11A and the amphibian MGC83907(Fig. 1C). Both RGA-3 and RGA-4 are expressed in the germ line and in the early embryo (NEXTDB Version 4.0). Expression seems to cease for both genes around the 100- to 200-cell stage. These are not the only RhoGAPs expressed at this early stage: CYK-4 is essential for cytokinesis and polarity establishment in the P0 cell(Jantsch-Plunger et al., 2000; Jenkins et al., 2006). RNAi treatment of CYK-4 generated multinucleated cells because of cytokinesis failure (Jantsch-Plunger et al.,2000) and cortical anterior markers were distributed over the entire cortex (Jenkins et al.,2006).

To decipher the contribution of RGA-3 and RGA-4 to the membrane-ruffling phenotype, we constructed different RNAi probes specifically targeting either RGA-3 or RGA-4 (Fig. 2A). Knockdown of the individual genes caused, in both cases, cortical hyper contractility, albeit with a lower penetrance(Table 2, Fig. 2B). Thus, RGA-3 and RGA-4 have overlapping functions and, given the phenotype of the RNAi-treated worms,both RhoGAPs are involved in inactivation of a Rho family GTPase, most probably at the plasma membrane. The contraction phenotypes were additive and the loss of both GAPs resulted in embryonic lethality, indicating that these two proteins are at least partially redundant. By contrast, another early-expressed RhoGAP, CYK-4, either plays less of a role or is not involved in the same process, because no anterior ectopic furrowing or invaginations were observed when this GAP was knocked down, although cytokinesis failures were detected (Tables 2 and 4). Given the redundant function of RGA-3 and RGA-4, the N-terminal construct, which targets both proteins and resulted in the highest penetrance, was therefore used for all subsequent experiments.

Which Rho GTPase is regulated by RGA-3/4? Lack of RGA-3/4 GAP activity leads to sustained activation of its target GTP-binding protein. In the early embryo, two GTPases of the Rho family show severe phenotypes when knocked down by RNAi: RHO-1 (the orthologue of mammalian RhoA) and CDC-42. Both GTPases have been reported to control actin dynamics in numerous different cell types and organisms (Narumiya and Yasuda,2006; Raftopoulou and Hall,2004). In the early C. elegans embryo, however, they seem to also be essential for the establishment and maintenance of polarity(Motegi and Sugimoto, 2006; Schonegg and Hyman, 2006; Gotta et al., 2001; Kay and Hunter, 2001). Upon knockdown of the specific GTPase regulated by the RhoGAPs, the hyper-contractility phenotype should be rescued, because the GTPase would no longer be present to hyper-activate the acto-myosin network, which performs the contractions in the early embryo. By contrast, if the GTPase were not involved in the regulation of the acto-myosin system at the anterior cortex,loss of this specific GTPase activity would have no impact on the ectopic furrowing and membrane ruffling. Whereas RNAi of RHO-1 rescued the extensive-ruffling phenotype caused by knockdown of RGA-3/4, reducing the levels of CDC-42 had no effects on the extent of invaginations and contractions (Fig. 3A,B, Movies 3-6, Table 3). CDC-42 levels were reduced in the triple-RNAi experiment, because polarity defects were observed as described for cdc42(RNAi)(Gotta et al., 2001; Kay and Hunter, 2001). More importantly, no cortical contractions were observed at all upon rho-1(RNAi), indicating that RHO-1 is responsible for the regulation of the acto-myosin network at the anterior cortex. RHO-1 has been implicated in the control of all contractions at the cortex(Motegi and Sugimoto, 2006; Schonegg and Hyman, 2006). Our results strongly indicate that RGA-3/4 act as RhoGAPs on RHO-1 and that they control contractility at the anterior cortex.

To provide further evidence in support of this hypothesis, we simultaneously knocked down the GEF for RHO-1, ECT-2, with RGA-3/4(Fig. 3C, Table 3). Again, the membrane-ruffling phenotype was rescued and contractility was greatly reduced even when compared to wild type. Without its GEF, RHO-1 cannot be activated and thus the concomitant loss of the GAPs has no impact on the phenotype. Finally, we knocked down the RHO-1-associated kinase (ROCK), which, in C. elegans, is encoded by let-502 and relays the signal of RHO-1 to the acto-myosin network (Piekny and Mains,2002). As predicted, co-knockdown of ROCK and RGA-3/4 rescued the cortical hyper-contractility caused by the loss of RGA-3/4 function. Furthermore, knockdown of RGA-3/4 in let-502^-/- animals gave a similar result (Fig. 3D,and see Movies 7 and 8 in the supplementary material).

To determine the effect of silencing GAP expression on the subcellular distribution of the Rho GTPase, we localized GFP::RHO-1 in wild-type and rga-3/4(RNAi) embryos. In wild type, RHO-1 was mostly cytoplasmic but was also visible at the cortex. Upon loss of RGA-3/4, GFP::RHO-1 was enriched at the plasma membrane, more precisely in the ruffles and furrows(Fig. 4). We determined the ratio of the concentration of RHO-1 at a given length at the contractile and at the non-contractile cortex in wild-type and rga-3/4(RNAi) embryos. The ratio in the RNAi-treated embryos was higher (1.55±0.21; n=11) compared with wild type (1.34±0.12; n=9; P<0.05). Given the increased number and persistence of ruffles in the knockdown, this ratio even underscores the enrichment of GFP::RHO-1 at the plasma membrane in rga-3/4(RNAi) zygotes. This result is exactly what one would expect for a GTPase that can no longer be deactivated and cannot stop signalling downstream. Therefore, the enrichment of GFP::RHO-1 in the ectopic cleavage furrows at the plasma membrane substantiates our finding that RGA-3/4 act on RHO-1. Taken together, these experiments strongly indicate that RGA-3/4 are novel RhoGAPs for RHO-1 during early embryonic development and that they regulate contractility at the anterior cortex.

RHO-1 and CDC-42 are both involved in cytokinesis. Therefore, by co-staining early embryos with the lipophilic dye FM4-64 and GFP::H2B, we investigated whether rga-3/4(RNAi) causes cytokinesis defects. FM4-64 served as a marker for the plasma membrane in these experiments. rga-3/4(RNAi) resulted in a moderate cytokinesis defect(Table 4 and see Fig. S2 in the supplementary material). Upon feeding, approximately 20% of the rga-3/4(RNAi) embryos showed multinucleated and anucleated cells. We observed cytokinesis defects, which were most probably a result of extra furrowing, that led to small anucleate cells (see Fig. S2 in the supplementary material). However, injection of RGA-3/4 dsRNA yielded in a higher level of over-pronounced cleavages and less cellularization defects(Table 4). Therefore, we concluded that failure in cytokinesis upon rga-3/4(RNAi) is most likely a secondary effect and that RGA-3/4 play no major role in cytokinesis.

The extensive contractility and ruffling caused by loss of RGA-3/4 function implied a role for actin and the non-muscle myosin NMY-2(Guo and Kemphues, 1996) in this process. To test whether NMY-2 acts downstream of RGA-3/4-dependent RHO-1, we first knocked down both NMY-2 and RGA-3/4(Fig. 5A). The ruffling phenotype of rga-3/4(RNAi) was rescued by the simultaneous loss of NMY-2 and RGA-3/4 function, confirming that indeed the contractile machinery in the early embryo is hyper-activated in the absence of the GAPs RGA-3/4. Hence, NMY-2::GFP should be enriched at the anterior cortex in rga-3/4(RNAi) embryos, which was what we observed: NMY-2::GFP was concentrated in invaginations and ectopic cleavages in the anterior part of rga-3/4(RNAi) embryos as well as in the pseudo-cleavage(Fig. 5B,C). Moreover,NMY-2::GFP covered a large extent of the anterior cortex, unlike the NMY-2::GFP speckles detected in wild-type embryos before pronuclear meeting(Fig. 5B-D). The NMY-2 patches at the cortex were at least twice as long in rga-3/4(RNAi) than in wild-type embryos (Fig. 5D). The concentration of NMY-2::GFP was also strongly increased in the cleavage furrow of the first cell division after RGA-3/4 knockdown (data not shown). Thus, the inability of the embryo to deactivate RHO-1 at the anterior cortex results in hyper-activation and aberrant recruitment of the acto-myosin system to the cortex.

After fertilization and meiosis II of the female pronucleus, the entire cortex of the embryo is contractile. Via the cortical and cytoplasmic streaming, ruffling later becomes restricted to the anterior part of the embryo. These contractions and ruffles are thought to be important for the establishment of polarity in the early embryo. In a subset of rga-3/4(RNAi) embryos, extensive ruffling and furrowing was not restricted to the anterior part and seemed to extend also towards the posterior region of the embryo. Therefore, we wanted to determine whether polarity establishment occurred normally upon depletion of RGA-3/4. First, we tested whether posterior polarity was established in rga-3/4(RNAi)embryos in the same manner as in wild type. Despite the extensive ruffling,the polarity establishment seemed to be mostly correct and we did not detect a significant difference in the posterior cortex area upon rga-3/4(RNAi) compared to wild type, as judged by the localization of the posterior marker GFP::PAR-2 (Fig. 6A,C). Because the ruffling in wild type occurs at the anterior cortex, we next tested the localization of GFP::PAR-6 in rga-3/4(RNAi) embryos. PAR-6 normally localizes together with PAR-3 and the atypical PKC, PKC-3, to the anterior cortex(Hung and Kemphues, 1999). Again, the onset of polarity was unperturbed. However, the GFP::PAR-6 signal seemed to be restricted to the hyper-contractile region and within the cleavage furrow (Fig. 6B and see Movie 9 in the supplementary material).

Taken together, our data indicate that there is no primary defect in polarity onset and maintenance per se. However, we observed greater fluctuations in the anterior-domain size as signified by GFP::PAR-6,indicating that the stable posterior domain prevented overshooting of the anterior domain into the posterior part of the embryo(Fig. 6C). Conversely,knockdown of the other GAP, CYK-4, caused a polarity defect, because GFP::PAR-6 and NMY-2::GFP could be detected over the entire cortex and GFP::PAR-2 was mostly cytoplasmic (Jenkins et al., 2006) (Fig. 7, Table 4 and see Movie 10 in the supplementary material). The fact that knockdown of RGA-3/4 did not cause any major polarity defects suggests that the GAPs fulfil different functions in the early embryo. If this were the case, the phenotypes caused by the loss of the three GAPs should lead to extended membrane ruffling and to a polarity defect. When we performed simultaneous knockdown of the three GAPs, we observed a hyper-contractile cortex [caused by the rga-3/4(RNAi)], and NMY-2::GFP was spread over the entire cortex and GFP::PAR-2 was cytoplasmic, indicating a loss of polarity [caused by the cyk-4(RNAi)] (Fig. 7). Furthermore, we detected a strong cytokinesis defect unlike the effect we observed upon rga-3/4(RNAi) (Table 4). Hence, CYK-4 and RGA-3/4 appear to fulfil different functions in the P0 cell.

To gain insight into the subcellular localization of RGA-3, we expressed a GFP::RGA-3 fusion in the germ line under the control of the pie-1promoter. The expression of GFP::RGA-3 seemed to be toxic, because it was very difficult to obtain and maintain transgenic lines. Furthermore, the transmission of the extrachromosomal array was very poor, the transgenic worms had a reduced brood size and only a few offsprings contained the transgene. Given these difficulties, it was impossible to analyze the early embryonic phenotype in a quantitative manner. However, in adult transgenic animals, we observed robust defects in gonad proliferation and migration, whereas the gonad architecture seemed to be unaffected (see Fig. S3 in the supplementary material, data not shown). Gonad arms were too long and crossed from one side of the body to the other. GFP::RGA-3 was barely detectable in the gonad, which might be due to low expression because of germline silencing of the transgene.

We inferred from our studies in the zygote that RGA-3/4 are involved in the control of the C. elegans ROCK, LET-502, and given the phenotype of the expression of GFP::RGA-3, we tested the effect of RGA-3/4 RNAi treatment on the temperature-sensitive (ts) let-502(sb106) mutant in the germ line. Although let-502^-/- animals are sterile and have severe gonadal defects, such as unusually large spaces devoid of nuclei and actin (`holes'), or nuclei fallen into the rachis, the ts mutant let-502(sb106) showed barely any gonadal defects at the restrictive temperature of 23°C (Fig. 8A,B). Knockdown of RGA-3/4 aggravated the let-502(sb106)phenotype such that the gonadal defects were very similar to those observed in let-502<sup>-/-</sup> animals; `holes' and strong defects in distal and proximal rachis formation were observed. Conversely, rga-3/4(RNAi) by itself had no effect on gonad migration or brood size (Fig. 8A,C). The only noticeable phenotype was a problem in the formation of the distal rachis. A similar, mild gonad defect was observed upon cyk-4(RNAi). However, no enhancement of the gonad phenotype was observed in let-502(sb106)mutants. Taken together, our data suggest a role for RGA-3/4 as GAPs for RHO-1 in the germ line and in the early embryo, and that these functions might be,at least in part, temporally and spatially distinct from the functions of CYK-4.

We have identified two novel RhoGAPs, RGA-3/4, that are essential to turn off RHO-1 signalling to the acto-myosin network at the anterior cortex in the C. elegans zygote. Furthermore, RGA-3/4 are required for proper germline formation. The two novel RhoGAPs belong to an uncharacterized subfamily. They carry the GAP domain at the N-terminus and the remainder of the proteins do not contain any known interaction or signalling motifs, which is unusual for RhoGAP proteins (Jaffe and Hall, 2005; Moon and Zheng,2003). The two vertebrate orthologues, the mammalian ARHGAP11A and the amphibian MGC83907, have the same domain structure and also lack known motifs, and might therefore fulfil similar functions. How do these proteins interact with effector molecules or provide specificity for Rho GTPases? It seems rather likely that these GAPs contain yet unidentified interaction domains. Further analysis of this novel class of regulators is likely to reveal interesting and new roles for Rho during development.

In C. elegans, RGA-3/4 are expressed in the germ line and in the early embryos, probably up to the 100- to 200-cell stage [Nematode Expression Pattern DataBase (NEXTDB), Ver. 4.0]. Another RhoGAP, CYK-4, is expressed in a similar pattern. Although these GAPs are expressed in the same cells, and target the same GTPase, their biological functions are largely non-overlapping. RGA-3/4 seem to control the contractility of the acto-myosin network, whereas CYK-4 is involved in cytokinesis, and in polarity establishment and maintenance (Jenkins et al., 2006) (Fig. 7and Table 4). By contrast,RGA-3/4 do not seem to play any primary role in meiosis, mitosis, polarity establishment or cytokinesis. The occasional failure in these processes upon rga-3/4(RNAi) might be due to the destruction or compromised assembly of cellular structures by the extensive furrowing and the hyper-contractility. Polarity setup and maintenance were also mostly normal. We observed, however,a significant fluctuation in the size of the GFP::PAR-6 domain, indicating that the boundaries separating the anterior and posterior domains might not be stable. RGA-3/4 clearly control contractility-related functions of RHO-1, and seem to be specific for this process. However, the bulk of CYK-4 is located at the mitotic spindle and only some paternal CYK-4 is found at cortex at the sperm entry site (Jantsch-Plunger et al.,2000; Jenkins et al.,2006).

Three RhoGAPs expressed in the germ line have been characterized so far:CYK-4, and now RGA-3/4; all of which act on RHO-1. No GAP for CDC-42 in the early C. elegans embryo has been identified yet. However, CYK-4 and RGA-3/4 act on RHO-1, which is upstream of CDC-42 in the zygote. Therefore,despite the result of the epistatic analysis, we cannot rule out that one or more of the three GAPs also turns off CDC-42. Another possibility is the presence of at least one other GAP, which is likely to be the case (NEXTDB,Ver. 4.0). But why would we need so many different RhoGAPs? Perhaps the GAPs provide the specificity for Rho activity(Moon and Zheng, 2003) and hence provide the basis for temporal and spatial control of RHO-1 activity. This concerted action by the GAPs might be extremely important because there is only one RhoGEF expressed in the C. elegans zygote.

The RhoGEF ECT-2 might be the first target of the polarity cue provided by the sperm (Jenkins et al.,2006; Motegi and Sugimoto,2006). Then, ECT-2 activates RHO-1, which in turn leads to the rearrangement of the acto-myosin network. To control the contractions performed by the acto-myosin network, the RhoGAPs RGA-3/4 inactivate RHO-1. If RHO-1 cannot be inactivated at the anterior cortex, cytoplasmic and cortical streaming still occurs, and therefore polarity establishment and maintenance are normal. Motegi and Sugimoto suggested that CDC-42 maintains polarity by controlling the acto-myosin network in a second phase in the one-cell embryo(Motegi and Sugimoto, 2006). Our data are in agreement with this interpretation, because after centration,the entire cortex becomes smooth even upon rga-3/4(RNAi), before the anterior part starts to contract again during cytokinesis. Therefore, CDC-42 still functions normally in the absence of RGA-3/4. CDC-42 might be involved in turning off RHO-1 signalling at the cortex during centration, perhaps by interacting with a RhoGAP. This GAP would be different from RGA-3 or RGA-4,because cortical ruffling stopped at centration upon loss of RGA-3/4 function. Interestingly, the original cyk-4 mutant displayed less-pronounced furrowing, which is the opposite phenotype than that observed after rga-3/4(RNAi) (Jantsch-Plunger et al., 2000).

A possible scenario for the control of RHO-1 function is that first, upon fertilization, CYK-4 initiates cytoplasmic and cortical streaming, allowing the establishment of polarity. In a second step, RGA-3/4 control membrane contractility and the pseudo-cleavage at the anterior cortex by keeping RHO-1 activity at a certain threshold. Finally, CYK-4 takes over again; however,this time at the mitotic spindle, where CYK-4 controls cytokinesis. This rather simple model provides an explanation concerning the control of RHO-1 action by spatially and temporally separated RhoGAP activities.

In addition to the function of RGA-3/4 at the anterior cortex in the zygote, we found a requirement for RGA-3/4 in gonad development. Expression of GFP::RGA-3 led to over-proliferation of the gonad and to gonad migration defects. More importantly, whereas neither rga-3/4(RNAi) nor let-502(sb106) alone, or cyk-4(RNAi) let-502(sb106) showed severe gonad defects, rga-3/4(RNAi) let-502(sb106) worms contained non-functional and abnormal gonads, similar to those observed in let-502^-/- animals. RGA-3/4 could either act in the same pathway or in a parallel pathway as LET-502 in the gonad. The manner in which RGA-3/4 and CYK-4 control RHO-1 activity in the early embryo and in the gonad might be different. Further components of these signalling pathways must be identified to shed light on the precise function of the RhoGAPs RGA-3/4 and CYK-4.

Stephanie Schonegg et al. have obtained similar results to those reported here (Schonegg et al.,2007).

We thank G. Seydoux, F. Motegi, P. Mains and the CaenorhabditisGenetics Center for C. elegans strains. We are grateful to M. Labouesse, T. Sandmann, D. Poteryaev and I. G. Macara for critical reading of the manuscript. We are indebted to R. Ciosk for help with the identification of the gonad phenotypes. This work was supported by the Max Planck Society(Germany) and the Biozentrum of the University of Basel (Switzerland).