Control of embryonic Xenopus morphogenesis by a Ral-GDS/Xral branch of the Ras signalling pathway

Mesoderm induction depends on multiple diffusible extracellular factors that cooperate to induce specific programs of gene expression and morphogenesis. Prior to overt gastrulation, a number of mesoderm-specific transcription factors are transcribed in induced mesoderm (reviewed by Stennard et al., 1997; Chan and Etkin, 2001). Shortly thereafter, mesodermal cells participate in stereotyped gastrulation movements (reviewed by Gerhart and Keller, 1986) that are achieved via coordinated changes in the cytoskeleton of individual cells (Selchow and Winklbauer, 1997; Wacker et al., 1998). Despite intensive study, the intracellular signalling pathways connecting mesoderm induction to the direct modification of the cytoskeleton by effector molecules remain unclear.

FGFs can induce the expression of many mesodermal markers (Kimelman and Kirschner, 1987; Green et al., 1992; Whitman and Melton, 1992; LaBonne and Whitman, 1994; Amaya et al., 1991) and expression of a dominant negative FGF receptor dramatically perturbs gastrulation (Amaya et al., 1991). Mesodermal gene induction by FGFs has been shown to be achieved via the Ras protein, acting through both the MAPK and PI 3-kinase pathways (Whitman and Melton, 1992; LaBonne and Whitman, 1994; Umbhauer et al., 1995; Carballada et al., 2001). Ras is also known to modulate the formation of the actin cytoskeleton in a variety of cell types (Bar-Sagi and Feramisco, 1986), but the specific Ras pathway involved is unknown. A role for FGF signalling in Drosophila embryonic cell motility has been shown in the requirement for the FGF receptor heartless in mesodermal cell migration (Gisselbrecht et al., 1996; Beiman et al., 1996). More recently, FGF4 and FGF8 have been shown to control directional movement of streak cells in the chick (Yang et al., 2002). These data demonstrate that FGFs are able to control both gene expression and cellular behaviour (for reviews, see Manske and Bade, 1994; Montell, 1999; Boilly et al., 2000) and they raise the possibility that during mesoderm development, Ras not only mediates the induction of gene expression by FGF signalling, but also controls the cellular behaviour.

As indicated above, the Ras protein initiates several cellular processes including proliferation, differentiation and modulation of the cytoskeleton (Reuter and Der, 2000). Although the Raf serine/threonine kinase is the best-studied Ras effector protein, at least two other effectors, PI 3-kinase (Rodriguez-Viciana et al., 1997) and Ral-GDS (Kikuchi et al., 1994), are known to mediate Ras signalling. Moreover, a Ras mutant with an inactive effector domain can be complemented by another distinct Ras effector-domain mutant (White et al., 1995), demonstrating that different Ras effectors can cooperate to transform cells (Urano et al., 1996). During early Xenopus development, PI 3-kinase was shown to synergise with the extracellular signal regulated kinase (ERK) pathway in the induction of mesoderm (Carballada et al., 2001). Recently, oncogenic activity was demonstrated for the Ras mutant 12V37G, which binds and activates Ral-GDS, but not Raf (Hamad et al., 2002). Much remains to be learned about the roles of Ras and its various effectors during early development. Of particular interest is whether or not Ras directly influences cellular morphology via modulation of the cytoskeleton. In previous studies, we have characterised XralB, a small G protein of the Ras family, isolated by differential screening of a subtractive cDNA library (Moreau et al., 1999). This protein cycles between an inactive GDP-bound and an active GTP-bound conformation. Ral proteins are activated by the Ras effector, Ral-GDS (Kikuchi et al., 1994; Wolthuis et al., 1998b) and have been directly implicated in various cellular mechanisms, such as endocytosis (Nakashima et al., 1999) and regulation of the actin cytoskeleton (Ohta et al., 1999). By microinjecting different mutant forms of XralB into Xenopus embryos, we have shown that expression of an inactivate form of XralB, XralB S28N, leads to abnormal gastrulation movements, whereas a constitutively active form, XralB G23V, causes disruption of the actin cytoskeleton (Moreau et al., 1999).

In this paper, we delineate a cascade linking FGF signalling to embryonic morphogenesis. We show that exposure of animal caps to bFGF causes activation of XralB, and that overexpression of an activated form of XralB causes cortical F-actin disassembly. We also show that early developmental arrest is caused by ectopic expression of a mutant form of Ras, which is able to bind Ral-GDS, but not Raf, or by ectopic expression of Ral-GDS itself. In contrast, we find that FGF-induced expression of a number of signature mesodermal genes requires signal transduction by Raf, but not XralB. Finally, we find that XralB is preferentially activated in the marginal zone of Xenopus embryos, and that disruption of the XralB pathway is most effective when targeted to this region.

To monitor the expression of injected mRNA, proteins were extracted from embryos and analysed as previously described (Gusse et al., 1989). Protein extracts were separated by SDS-PAGE on a 15% acrylamide gel and transferred to a Hybond P membrane (Amersham). The membrane was probed overnight at 4°C with anti-rat RalB goat antibodies (Sc1531, Santa Cruz) diluted 1:125 in 1× PBS – 1% BSA (wt/vol). The secondary antibody, peroxidase-linked rabbit anti-goat Ig was diluted 1:1250 (Chemicon). Signals were detected by chemiluminescence (ECL+, Amersham).

To analyse Ral-GTP levels, protein extracts from embryos were incubated with RalBP1 agarose (Upstate Biotechnology, ref. 14-415) for 30 minutes at 4°C. Immunoprecipitates were separated on a 15% acrylamide gel and transferred to a P Hybond membrane (Amersham). Precipitated endogenous Ral was probed with an anti-RalA antibody (Upstate Biotechnology, ref. 20-189) diluted 1:750 in 1× PBS – 0.05% Tween 20 and 5% milk powder (wt/vol). Signals were detected by chemiluminescence (ECL, Amersham) and quantified by ImageQuant (Molecular Dynamics).

Xenopus frogs were imported from South Africa (South Africa Farms, Fish Hoek) or from the CNRS frog colony (Rennes). Animals were housed and fed as described previously (Gurdon et al., 1984). Eggs were fertilised in vitro and the embryos were chemically dejellied with 2% (wt/vol) cysteine-HCl, pH 7.8/7.9, then maintained in 1× MBS (Modified Barth's solution pH 7.6 (Gurdon and Wickens, 1983) until microinjection. Microinjection was performed in a solution of 1× MBS and 3% Ficoll (wt/vol), followed by overnight incubation at 16°C in 0.1× MBS and 3% Ficoll, then at 15-22°C until they reached appropriate stages. Embryonic stages were determined according to Nieuwkoop and Faber (Nieuwkoop and Faber, 1956). In vitro transcribed RNA (4.6 nl) was microinjected into two-cell stage embryos in the animal hemisphere or into four-cell stage embryos in the marginal zone, animal pole or vegetal pole.

pSP64-Ras 12V35S and 12V37G were obtained by cloning the SalI-BamHI insert from pDCR-12V35S and 12V37G (White et al., 1996) into the SalI-BamHI sites of the pSP64 vector (Promega). pSP64-Ral-GDS was obtained by subcloning of the BamHI insert from pCEP4-Ral-GDS (White et al., 1996). Ras 17N and Rap1A 17N (kindly provided by J. De Gunzburg) were amplified by PCR and cloned into the EcoRI-NotI sites of the pRN3 vector. pRN3-RalBD-GST was obtained by subcloning of the HindIII-SalI insert from pGEXT4T3-GST-RalBD (Wolthuis et al., 1998b). The Raf KD (a gift from E. Amaya), a kinase-defective Raf protein, containing a serine to alanine substitution at position 621, was cloned into the BglII-SmaI sites of the pSP64 vector (Fabian et al., 1993). The 3′ untranslated regions of the pRN3 RalB clones S28N, G23V were deleted by removal of BamHI/NotI, followed by end-filling and religation. mRNAs were transcribed from pSP64T (Promega) or pRN3 (Lemaire et al., 1995) vectors using the SP6 or T3 mMessage mMachine (Ambion).

Animal caps were observed using a Leica TCS-4D confocal imaging system (Leica Instruments, Heidelberg, Germany) fitted with a 63× objective (NA 1.4). For FITC and TRITC, an argon-krypton ion laser was used set to 488 and 568 nm, respectively. For each optical section, double fluorescence images were acquired sequentially to avoid potential signal contamination by linkage-specific fluorescence emission `cross-talk'. Signal noise was reduced by line averaging of four frames. A focal series was collected for each specimen. The focus step between sections was 1 μm.

Embryos were prepared for confocal microscopy analysis as described previously (Moreau et al., 1999).

Animal caps were manually dissected from midblastula transition (MBT)-stage embryos. These explants were incubated with or without FGF (100 ng/ml) in 1× MMR, 1% BSA for 4 hours at 16°C.

Total RNA extraction was carried out as previously described (Moreau et al., 1999), except that extracts were treated with 1 U of DNase I RQ1 (Promega) for 30 minutes at 37°C after the LiCl precipitation step. First strand cDNA was prepared from total RNA using the reverse transcriptase M-MLV (Gibco-BRL, ref. 28025-013) and Oligo(dT) primer (Pharmacia). PCR and analysis of cDNA was performed in 50 μl of solution containing 5 μl of cDNA, 1× Taq buffer, 3 mM MgCl₂, 0.1 mM (each) dNTPs, 200 ng of each specific primer, and 1 U of Taq polymerase (Invitrogene), using the following PCR conditions: 93°C for 3 minutes, followed by a variable number of cycles at 93°C for 45 seconds, 52°C for 50 seconds and 72°C for 1 minute. Ornithine decarboxylase (ODC), Xbra, Xwnt11, Xnot and Xty2 PCR amplification was carried out for a total of 27 cycles, and Xwnt8, Xsnail and Xcad3 amplification was for a total of 35 cycles. The primer sequences were as follows: ODC forward 5′-GTCAATGATGGAGTGTATGGATC-3′, reverse 5′-TCCATTCCGCTCTCCTGAGCAC-3′; Xbra forward 5′-GTGTAGTCTGTAGCAGCA-3′, reverse 5′-GGATCGTTATCACCTCTG-3′; Xwnt 11 forward 5′-GTGAGAGAGGTCTGAGCTGG-3′, reverse 5′-ACATGACATAGCAGCACC-3′; Xnot forward 5′-CAGAGCAGCTGGAGAAGCTG-3′, reverse 5′-CAGTGTGATCTGAGCTGTTC-3′; Xty2 forward 5′-ATCTGCTCCACGACGGAC-3′, reverse 5′-GAAGAACTGCTACTTGTCC-3′; Xcad3 forward 5′-CAGCGCAGAATACGTCTCC-3′, reverse 5′-CCCAGTCCCAGATGGATGTG-3′; Xwnt8 forward 5′-AGATGACGGCATTCCAGA-3′, reverse 5′-TCTCCCGATATCTCAGGA-3′.

Activation of Ral proteins by growth factors, including those signalling through receptor tyrosine kinases (EGF, FGF, etc.), has been documented in a variety of contexts (Wolthuis et al., 1998a; Suzuki et al., 2000; Nakashima et al., 1999). Considering the role of FGFs in mesoderm induction, and our previous observations that Xral function is required for gastrulation, we wished to determine whether the FGF pathway regulates Xral. Animal cap explants were dissected at the blastula stage and cultured for 4 hours in the presence or absence of bFGF (Fig. 1A). After incubation, mRNA was extracted to verify the activation of the pan-mesodermal gene marker Xbra by bFGF (Fig. 1B). To determine the relative abundance of endogenous GTP-bound Xral, the GTP-bound Xral form was pulled down by affinity purification on glutathion beads, which bind to the GST fused to the Ral binding domain of RalBP1 (GST-RalBP1) (Wolthuis et al., 1998a) (Fig. 1C). Animal caps incubated in the presence of bFGF became elongated (Fig. 1A), and Xbra transcript levels (Fig. 1B) and GTP-bound Xral increased (Fig. 1C). Thus, the Ral pathway can be activated by FGF signalling during early Xenopus development.

Ras stimulates multiple, distinct signalling cascades. Indeed, White et al. (White et al., 1995) have shown that a mutant form of Ras, Ras 12V37G, does not transform cells, but can be complemented for transformation by another Ras mutant, Ras 12V35S. Ras 12V35S binds specifically to Raf kinase but not to Ral-GDS, whereas Ras 12V37G only binds to Ral-GDS (White et al., 1996). We utilised these Ras mutants to assess the relative roles of Ras signalling via Raf and Ral-GDS in the activation of Xral in Xenopus embryos.

Whole embryos were injected with Ras 12V35S (0.5 ng/blastomere), or Ras 12V37G (0.5 ng/blastomere) RNA at the two-cell stage, grown until the 128/256-cell stage and protein was extracted and analysed by GST-RalBD pull-down. We found the level of activated XralB to be specifically increased in embryos injected with Ras 12V37G (Fig. 2A). Xral-GTP levels in embryos injected with Ras 12V35S were similar to the basal levels observed in uninjected embryos. To confirm the specificity of XralB activation by Ras 12V37G, we determined whether a dominant-negative form of RalB, XralB S28N, can interfere with Xral activation when co-injected with Ras 12V37G. In embryos treated this way, levels of Xral-GTP were similar to those of control embryos (Fig. 2A,B). As Ras is able to activate Ral, we examined whether Ras is required to activate Ral by FGF. The dominant negative forms of Ras (Ras 17N) and, Rap1 (Rap1A 17N) used as control, were injected into two-cell stage embryos. Embryos were cultured until stage 9, then the animal cap was dissected and cultured for 4-5 hours in the presence of FGF (100 ng/ml). Proteins were then extracted and analysed by GST-RalBP1 pull-down. Whereas the Rap1A 17N did not seem to interfere with the activation of Ral protein (Fig. 2C), Ras 17N severely impaired RalBP1 binding activity. However, the RalBP1 interaction with Ral, which was dependent of FGF, was not completely abolished by Ras 17N. The analysis presented above indicates that Ras can activate Xral independently of Raf kinase and that Ras is an element of the FGF pathway to activate Ral.

We also examined the effect of the Ras 12V35S and 12V37G mutants on development and morphogenesis. All embryos injected with Ras 12V35S (Fig. 3A and Table 1) appeared normal through to the end of gastrulation and 37% survived to stage 32, although many displayed developmental defects such as a shorter axis or open neural folds. Embryos injected with Ras 12V37G displayed earlier defects. Development was normal through the MBT, but then blastomeres near the injection area became necrotic, characterised by large bleached cells (Fig. 3A), and 30.6% of embryos (n=186) underwent developmental arrest prior to gastrulation (stage 10.5) (Fig. 3A). More than 96.6% of Ras 12V37G-injected embryos failed to survive beyond neurulation (Table 1) and only 1.1% reached the tadpole stage. To determine whether Ras 12V37G had the same target as XralB, we examined whether the expression of this Ras mutant disrupted the integrity of the actin cytoskeleton, as does the constitutively active form of XralB, XralB G23V (Moreau et al., 1999), and we compared the effects of Ras 12V37G on the actin cytoskeleton with those of Ras 12V35S. Apical blastomeres from the animal hemisphere, before and after stage 8, were observed by confocal microscopy following F-actin staining with Rhodamin-Phalloidin. Actin cytoskeleton analysis in embryos injected with Ras 12V37G was carried out on embryos that did not display cell necrosis.

As expected, embryos injected with Ras 12V37G had severely disrupted cortical and nuclear actin cytoskeletons (Fig. 3B), whereas blastomeres of embryos injected with Ras 12V35S were similar to uninjected controls. To better characterise the timing of F-actin disruption, we performed a kinetic study (Fig. 4). We found that F-actin is intact until the 2000-cell stage (Fig. 4D) and becomes damaged near MBT (Fig. 4F). This timing of Ras 12V37G-induced disruption differs from that of constitutively active XralB G23V, which induces disruptions as early as the 256-cell stage (corresponding to the technical limit of analysis). Nevertheless, the ability of either Ras 12V37G or XralB G23V to disrupt the actin cytoskeleton, together with the ability of Ras 12V37G to induce activation of Xral, suggests that Ras 12V37G-induced disruption of the cytoskeleton is mediated by Xral. To confirm this apparent specificity, we coinjected the dominant-negative XralB S28N mutant with Ras 12V37S at a ratio of six mRNA molecules to one. Co-injection of XralB S28N rescued early morphological defects, as embryos had no necrotic cells (Fig. 3A) and survived longer. The development of these embryos was blocked during gastrulation, but this was expected, as gastrulation arrest results from injection of XralB S28N alone. We also observed partial rescue of the cortical actin array when XralB S28N was co-injected with Ras 12V37G (Fig. 3B). The dependent activation of Ral by Ras was confirmed by the effects of Ras 17N on the activation of Ral by FGF. However Ras 17N did not abolish completely the activation of Ral by FGF, so we do exclude the possibility that Ral can also be regulated by a Ras-independent pathway as already observed in fibroblasts (Hofer et al., 1998) and platelets (Wolthuis et al., 1998a). We conclude that Ras signalling via Xral, independently of Raf, can cause alterations in the actin cytoskeleton. As Raf is known to be an essential mediator in the activation of several Xenopus mesodermal genes by Ras (McNicol et al., 1993), our results indicate that Ras/Raf/MAPK and Ras/Xral signalling act on distinct targets during early Xenopus development.

Having established that Ras can affect early Xenopus morphogenesis via Xral activation, we wished to determine which molecule mediates this event. A good candidate for this function is Ral-GDS, a Ral-GEF-related molecule, which specifically binds to Ras GTP and activates Ral in a Ras-dependent manner (reviewed by Wolthuis and Bos, 1999). To evaluate the role of Ral-GDS during Xenopus embryonic development, we injected murine Ral-GDS mRNA (1.5 ng) into each blastomere of two-cell stage embryos. From the 64-cell stage, depigmentation of the animal hemisphere was observed in these embryos and during blastula stages ectodermal lesions arose, leading to arrested development at the end of blastula stage (Fig. 5A). A lower dose of Ral-GDS mRNA (750 pg) induced hyperpigmented cells. The same spectrum of phenotypes, including opposing effects on cell pigmentation, are seen in response to high and low doses of the constitutively active XralB G23V (Moreau et al., 1999). As described for XralB G23V, in embryos injected with low doses of Ral-GDS, patches of hyperpigmentation are maintained throughout development until the tadpole stage. At hatching, some hyperpigmented cells are often lost with the vitelline membrane. Thus, the phenotypic consequences of either Ral-GDS or XralB G23V misexpression are indistinguishable.

We went on to address whether, like Ras 12V37G or XralB G23V, Ral-GDS misexpression causes disruptions to the actin cytoskeleton. Animal blastomeres of embryos injected with 1.5 ng of Ral-GDS indeed lose their cortical F-actin cytoskeleton (Fig. 5B). In contrast to Ras 12V37G, Ral-GDS-induced damage is already apparent at cleavage stages, as is also seen in XralB G23V-injected embryos (Moreau et al., 1999). To confirm that this Ral-GDS-induced phenotype was dependent on the XralB protein, we examined whether it could be rescued by co-injection of XralB S28N RNA. Embryos that were co-injected with XralB S28N and Ral-GDS, at a ratio of seven mRNA molecules to one, displayed a fainter depigmented phenotype (Fig. 5A). As expected, the high quantity of XralB S28N also blocked embryonic development beyond the gastrula stage (Fig. 5A), however, no early ectodermal lesions appeared in rescued embryos and survival up to the time of the XralB S28N-induced block was significantly enhanced. Co-injection of XralB S28N with Ral-GDS also protected embryos against F-actin cortical disruption (Fig. 5B). In addition, coinjection of XralB S28N rescues the hyperpigmentation phenotype arising from lower doses of injected Ral-GDS (data not shown). Finally, we examined the activation of Xral by Ral-GDS, using the pull-down assay to precipitate GTP-bound Xral from embryonic lysates. Embryos injected with 750 pg and 1.5 ng of Ral-GDS RNA, displayed 1.47-fold and 2.8-fold increases in Ral-GTP levels, respectively, relative to uninjected embryos (Fig. 5C). These data demonstrate that disruption of the actin cytoskeleton and cellular depigmentation are induced by Ral-GDS, that these events require functional Xral signalling and that the XralB protein is activated by Ral-GDS. The Ras pathway is therefore able to achieve morphogenetic changes during early development through the activation of XralB via Ral-GDS.

In Xenopus, all mesoderm and some endoderm arises from equatorially situated cells of cleavage-stage embryos, a region known as the marginal zone. Furthermore, large molecules residing in- or delivered to the marginal zone cytoplasm of early-stage blastomeres are inherited by this cell lineage. Consistent with this, mesoderm-specific gene products and activities are often concentrated in the marginal zone. Because the mesoderm inducer bFGF activates Xral, and because the Ras/Raf effector, MAPK is specifically activated in the marginal zone (Christen and Smith, 1999), we wished to assess the relative activity of Xral in this region. We divided stage 10.5 embryos into three parts, corresponding to the animal hemisphere, marginal zone and vegetal hemisphere (Fig. 6A). Protein was extracted from pools of 30 such explants, and Xral-GTP levels were assessed in pull-down assays. Levels of endogenous Xral-GTP were clearly highest in medial zone explants (Fig. 6B). To confirm this, we increased the level of XralB in targeted zones by direct microinjection of XralB mRNA (Fig. 6C). RNA encoding wild-type XralB (500 pg/embryo) was injected into the animal hemisphere, marginal zone or vegetal hemisphere of four-cell stage embryos. Protein from whole embryos (n=10) at stage 10.5 was extracted and analysed in pull-down assays. The pull-down results clearly show that the highest levels of Xral-GTP were obtained from embryos injected in the marginal zone. This indicates that, like MAPK activation, XralB activation is most concentrated in the mesodermal region of Xenopus embryos.

We previously reported that microinjection of XralB S28N mRNA into Xenopus embryos causes abnormal embryonic development (Moreau et al., 1999). Developmental alterations caused by XralB S28N were seen to vary with the XralB S28N RNA concentration injected, and a diversity of phenotypes was even observed to arise from a fixed quantity of microinjected RNA. In more severely affected embryos, the blastopore failed to close during gastrulation. The least affected embryos either had incompletely closed neural tubes or were bent dorsally at the tailbud stage. In light of our finding that endogenous Xral activation is concentrated in the embryonic marginal zone, we decided to assess the relative effects of disrupting Xral signalling with ectopic XralB S28N that was targeted to different embryonic regions. Four-cell embryos were injected with 750 pg of XralB S28N in each blastomere at the apical animal hemisphere, the marginal zone or the bottom of the vegetal hemisphere (Table 2). We confirmed the location of XralB S28N using β-galactosidase mRNA as a co-injected tracer. When RNA encoding XralB S28N was microinjected into animal (Fig. 7A) or vegetal hemispheres (Fig. 7C), the majority of embryos survived gastrulation and beyond. Only 38.1% (n=79) of embryos injected animally and 25% (n=45) injected vegetally failed to develop normally through the neurula stage. However, when XralB S28N was microinjected into the marginal zone, most embryos (81%; n=153) exhibited incomplete blastopore closure (Fig. 7B,F) and developmental arrest between stages 10.5 and 11.5. Histological sections of marginal-zone-injected embryos reveal that β-galactosidase is expressed in mesoderm and that arrest occurs when these labelled cells are undergoing invagination (data not shown). These observations support the idea that the most severe gastrulation defects arise when XralB S28N is targeted to prospective mesoderm and further suggest that XralB function is specifically required by mesodermal cells to achieve the morphological changes associated with invagination. To verify that the effect of XralB S28N during gastrulation was not due to a specific toxicity of XralB protein in the marginal zone, wild-type XralB (4× 750 pg) was injected. No developmental disturbances were observed in these control embryos (Fig. 7E) and over 90% (n=35) developed normally to the tadpole stage.

As an independent test of the essential role of Xral activity in the marginal zone, we examined the effect of another protein capable of disrupting Ral signalling. The Ral binding domain (RalBD) region of the effector protein RLIP, corresponding to amino acids 397-518 of RLIP76 (Wolthuis et al., 1998a), has been shown to be sufficient for direct binding to GTP-bound Ral in vitro. This RLIP peptide fused to either the glutathion S-transferase (GST) protein or the myc peptide was used in injection experiments. As a control, we co-injected embryos at the four-cell stage with either 4× 500 pg of GST-RalBD or 4× 300 pg of myc-RalBD mRNAs, together with β-galactosidase, into the apical animal hemisphere, the vegetal hemisphere and the marginal zone. Whole embryos injected with β-galactosidase, GST, or myc-tag mRNA alone did not show any morphological changes (data not shown). However, as observed with XralB S28N, RalBD injected into the marginal zone caused an arrest of gastrulation (Fig. 7D) between stages 10.5 and 11.5. The specific effect of the dominant negative form of Ral was also confirmed by co-injection of XralB S28N with wild-type XralB. Whereas XralB S28N, expressed in the marginal zone, resulted in the incomplete closure of the blastopore and blocking of the embryos at stage 10.5-11, gastrulation was rescued when the dominant negative form of Ral was co-injected with wild-type Ral (Fig. 7F). The embryos co-injected with Xral S28N and wild-type Xral showed a delay in blastopore closure, similar to that occurring when embryos where injected with wild-type XralB alone. Nevertheless 88.6% (n=45) of embryos closed their blastopore and developed until neurulation.

In summary, these data demonstrate a region-specific effect of titrating XralB targets (with constitutively inactivated XralB S28N) or Xral effector sites (with RalBD). When expressed in the marginal zone, these proteins disrupt development at the time of mesodermal cell migration, whereas expression in other regions has much milder consequences.

In Xenopus, FGF is able to induce the conversion of prospective animal cap ectoderm into mesoderm (Slack et al., 1987; Kimelman and Kirschner, 1987) and FGF signalling through the Ras pathway is required for formation of most mesoderm (Amaya et al., 1991; LaBonne and Whitman, 1994; Kroll and Amaya, 1996). Induction of the expression of a number of mesodermal genes by FGFs or Ras has been shown to depend on the MAPK and PI 3-kinase pathways (Whitman and Melton, 1992; LaBonne and Whitman, 1994; Umbhauer et al., 1995; Carballada et al., 2001). We wished to assess whether or not XralB signalling is similarly required for the activation of these signature mesodermal genes. Four-cell stage embryos were injected in the animal hemisphere with mRNA encoding either XralB S28N or Raf KD dominant-negative constructs to disrupt Ras/XralB signalling or Ras/Raf signalling respectively (Fabian et al., 1993). When embryos reached stage 8, animal caps were excised and explants were cultured for 4 hours in the presence or absence of bFGF. Expression levels of several mesodermal genes including the pan-mesodermal marker Xbra (Smith et al., 1991), Xcad3 (Northrop and Kimelman, 1994), Xsnail (Sargent and Bennett, 1990), Xty2 (Nutt et al., 2001) Xnot (Dassow et al., 1993), Xwnt8 (Christian et al., 1991) and Xwnt11 (Saka et al., 2000) were assessed by RT-PCR. Whereas the dominant-negative form of Raf inhibited expression of all markers tested, except expression of Xcad3 (Fig. 8), their expression was not significantly diminished by overexpression of XralB S28N (Fig. 8). Thus, Raf signalling, but not XralB signalling, is essential for the induction of several key mesodermal genes by bFGF. Considering our other data implicating Xral signalling in the control of morphogenesis, these results suggest that mesoderm patterning and mesoderm morphogenesis are controlled to a degree by independent pathways.

The Ral-GDS factor interacts with Ras (Hofer et al., 1994) and has been shown to be the guanine nucleotide dissociation stimulator protein for Ral in COS cells (Urano et al., 1996; Kishida et al., 1997) and fibroblasts (Matsubara et al., 1999). Another small G protein, Rap1, contains the same effector domain as Ras (Spaargaren and Bischoff, 1994). The study of the activation of Ral in human platelets suggested that Ral-GDS could be the effector protein of Rap1 rather than Ras (Wolthuis et al., 1998a). The potential interaction of Ral-GDS with either activated Ras or Rap1 suggests that Ral can be activated by at least two pathways. However, although Rap1/Ral-GDS interaction has been seen in the two-hybrid system and in vitro, Rap1 fails to co-immunoprecipitate with Ral-GDS in co-transfected cells, so compartmentalisation of the proteins was suggested (Nancy et al., 1999). Here we have demonstrated that Ras activates Ral through the Ral-GDS effector in Xenopus embryos. It has been reported that Ras recruits Ral-GDS to the plasma membrane, which can then induce the activation of Ral in COS cells (Matsubara et al., 1999). Constitutive binding of Ral-GDS to membranes has also been reported (Vojtek and Der, 1998). We have confirmed these latter data with the observation that mutation of the RalB membrane targeting sequence inhibits the depigmentation activity of RalB G23V (data not show). It is therefore conceivable that these phenotypes arise from overexpression of wild-type Ral-GDS via titration of factors necessary for the membrane localisation of endogenous Ral-GDS.

We demonstrate that bFGF, Ras and Ral-GDS can all activate Xral. Furthermore, misexpression of Ral-GDS or the constitutively active XralB G23V causes the same phenotype. This suggests that in early Xenopus embryos, Xral is the key target of Ral-GDS. The effect of Ras 12V37G misexpression is seen at a different time to those induced by Ral-GDS or XralB G23V. Thus, it is difficult to compare the effect of each individual injected mRNA on blastomeres appearance. However, for both Ras12V37G as Ral-GDS or Xral G23V, we observed the same alteration of the actin cytoskeleton. The complete mesoderm induction by bFGF is probably achieved through its activation of multiple parallel pathways. In addition to activating Xral, bFGF activates Raf/MAPK and PI 3-kinase via Ras. The Ras 12V37G misexpression phenotype is qualitatively similar to the Ral-GDS/RalB G23V phenotype, but its onset is later. This could reflect a regulatory mechanism that normally delays Ral-GDS activity, such as late regulation of maternal Ral-GDS mRNA translation or a derepression of its activity by a post-translational modification of the Ral-GDS protein. Indeed, a negative regulation of Ral-GDS has been demonstrated to occur through the phosphorylation of its catalytic domain (Rusanescu et al., 2001), and positive regulation of Ral-GDS occurs via formation of a complex with the N terminus of PI 3-kinase-dependent kinase 1 (Tian et al., 2002). We can also make the assumption that another exchange factor of Ral, such as Rlf (Wolthuis et al., 1996; Wolthuis et al., 1997) or Rgl (Kikuchi et al., 1994; Murai et al., 1997) interferes with Ral-GDS during cleavage stage and could be responsible for the delayed phenotype.

Ras 12V37G and Ras 12V35S have distinct effects on development (Fig. 3 and Table 1). Ras 12V37G induces large necrotic blastomeres in the injected area and embryos undergo developmental arrest by the end of the blastula stage, whereas embryos injected with Ras 12V35S have normal blastomeres and survive until the neurula stage. Hence, two distinct signalling pathways are controlled by Ras. One, corresponding to Ras 12V35S, activates mesoderm induction by the classic MAP kinase pathway and causes post-gastrulation phenotypes. The other, corresponding to Ras 12V37G, destabilises the actin cytoskeleton from the MBT stage onward, followed by cell necrosis and early arrest. Consistent with this, a Raf/MAPK-independent target of Ras has been described to induce membrane ruffling (Joneson et al., 1996) (for a review, see Ridley, 1994).

It is now clear that Ras controls several cellular functions by acting on various downstream pathways. However, only the Raf/MAP kinase cascade and more recently Ras/PI 3-kinase signalling (Carballada et al., 2001) have been investigated in early development. Previously, we demonstrated the involvement of the Ral protein in embryogenesis. Our current data strongly suggests a requirement for the Ral cascade during gastrulation. We show that expression in the marginal zone of either XralB S28N or RalBD, two distinct inhibitors of Ral signalling, causes developmental defects during gastrulation. The cellular mechanisms through which the Ral pathway participates in early development are not yet understood but our results clearly suggest that the Ral pathway affects cell behaviour by controlling actin cytoskeleton integrity. The presence of a distinct blastopore groove in embryos injected with XralB S28N indicates that gastrulation is successfully initiated in the absence of Ral signalling. Gastrulation defects arise later, and only when Ral signalling is disrupted in presumptive mesoderm. Cells expressing injected XralB S28N are found in the region of involuting mesoderm of such embryos. These blocked embryos also frequently acquire ectodermal folds in their animal hemispheres. This could be a secondary consequence of blocked invagination. If invagination were reduced, an excess of surface tissue that would normally invaginate or replace invaginated tissue might be expected to arise, and constraint of this excess tissue within a limited surface would inevitably lead to folding or lesion formation. The fact that gastrulation arrest is only observed when XralB S28N is injected around the marginal zone suggests that XralB S28N does not disrupt gastrulation movements by modulating early induction events but rather through a more direct effect on the cellular mechanism necessary for gastrulation, as confirmed by its action on the actin cytoskeleton. We therefore conclude that Ral signalling is essential for the involution of mesoderm at the blastopore lip, in an autonomous fashion. Considering that Ral signalling can induce actin cytoskeleton disassembly, we further propose that Ral is required to effect cytoskeletal changes that drive or enable the cellular shape changes required for involution.

Ral and its partners, Ral-GDS and the putative effector RLIP, are not present in the Saccharomyces cerevisiae genome or in other unicellular eukaryotes (Bauer et al., 1999). This pathway, therefore, may have a general role in the regulation of multicellular behaviour, as we have seen for Xenopus gastrulation. Indeed, in Drosophila development, RalB has also been implicated in the control of cell shape via regulation of the actin cytoskeleton (Sawamoto et al., 1999a; Sawamoto et al., 1999b) and is also required for the initiation of border cell migration during Drosophila oogenesis (Lee et al., 1996).

In conclusion, we propose that the Ral pathway is activated in the marginal zone during gastrulation and participates in the control of the dynamic equilibrium between F- and G-actin and we further propose that this function is essential for mesoderm involution during gastrulation. It appears, therefore, that at least three Ras-dependent pathways are required during development including the interactions between Ras/Ral-GDS/Ral, the Ras/Raf cascade and the Ras/PI 3-kinase cascade (Fig. 9). Whereas the Ras/Raf and Ras/PI 3-kinase pathways regulate mesodermal gene expression, the Ras/Ral pathways are likely to act directly on F-actin, independent of gene transcription. This last point is strongly supported by the facts that the constitutively active RalB G23V perturbs the organisation of F-actin at cleavage stages, well before the onset of zygotic gene transcription, and that the dominant negative form of RalB fails to inhibit expression of most mesodermal genes tested. It has been demonstrated in mammalian cells, that downstream the Ras/PI 3-kinase pathway can control the actin cytoskeleton and induce membrane ruffling (Rodriguez-Viciana et al., 1997). Although we cannot totally exclude a participation of PI 3-kinase in controlling the embryonic actin cytoskeleton, this must be minor, because the study of ectopic expression of the constitutively active PI 3-kinase catalytic subunit in embryos, by Carballada et al., showed that PI 3-kinase synergises with ERK on mesodermal induction but did not reveal pigment or mesodermal cells alterations (Carballada et al., 2001).

The idea that distinct pathways control cell behaviour and gene expression is supported by a variety of data. In mouse fibroblasts migration and cellular morphogenesis in response to FGF signalling occurs independently of Ras/Raf/MAP kinase (Liu et al., 1999). Also, cell migration during tracheal morphogenesis (Skaer, 1997; Ribeiro et al., 2002) or directional migration of mesodermal cells (Gisselbrecht et al., 1996) can occur in the absence of gene activation. We propose that the actin cytoskeleton dynamics controlling these morphogenetic changes depends in part on Ral signalling in response to FGFs, without gene activation. Understanding the molecular events connecting Ral activation to modification of the actin cytoskeleton is a key area for future research. A key player will probably be RLIP, a putative effector of Ral, which is a modular protein containing a Rac/CDC42-GAP domain and an interacting domain with the μ2 protein of AP2 heterotetrameric complex (Jullien-Flores et al., 2000). AP2 is an adaptator component of the endocytic machinery (Collins et al., 2002) that recruits clathrin to the plasma membrane. Activated Ral protein might recruit RLIP to the membrane where it interacts with the Rac/CDC42 protein. Rac/CDC42 is known to control rearrangements of F-actin to induce filopodia and lamellipodia (Nodes and Hall, 1995; Tapon and Hall, 1997). However, the mechanisms through which Rac/CDC42 exert their effects are not well defined. We can hypothesise that activated Ral may cause localised alterations to the actin cytoskeleton via membrane targeting and activation of an RLIP/Rac/CDC42 cascade. Thus, through its interaction with the AP2 complex, Ral could play a function in cell migration, in connection with endocytosis, as a co-ordinator of membrane trafficking and cytoskeletal reorganisation. Recently, endocytic markers have been seen polarised towards the leading edge of migrating cells, suggesting a role of clathrin-dependent endocytosis in cell migration (Rappoport and Simon, 2003). In this way, Ral signalling downstream of FGF, Ras and Ral-GDS may regulate morphogenesis at the cellular and ultimately multicellular level, in a transcriptionally independent manner.

We thank Michael White, Jacques Camonis, Jean De Gunzburg and Johannes Boss for the gift of plasmids used in this study. We thank Gerard Geraud for the confocal analyses and Anne-Lise Haenni and Benjamin Feldman for critical reading of this manuscript. This work was supported by the CNRS, the Association pour la Recherche sur le Cancer and the Fondation pour la Recherche Médicale.