The homeobox gene Siamois is a target of the Wnt dorsalisation pathway and triggers organiser activity in the absence of mesoderm

In amphibians, patterning of mesoderm is thought to be set up as a result of sequential inductive interactions during early embryogenesis. Several lines of evidence indicate that the formation of dorsal mesoderm in Xenopus laevis is controlled by signalling molecules produced by a small number of cells localised in the so-called ‘dorsalising centres’. Two dorsalis-ing centres have been identified, acting in a temporal hierarchy: the Nieuwkoop centre and Spemann’s organiser (Boterembrood and Nieuwkoop, 1973; Gimlich and Gerhart, 1984; Gimlich, 1986). Both centres were defined embryolog-ically by heterotopic transplantation of dorsal cells into ventral territories (see Hamburger, 1988 and references therein). According to these data, the Nieuwkoop or dorsal-vegetal organising centre, fated to become pharyngeal endoderm, acts during the blastula stages by inducing, in the overlaying equatorial region of the embryo, the Spemann organiser, fated to become axial and head mesoderm. In turn, this centre recruits more lateral mesoderm into the axial struc-tures during gastrulation. Understanding the ontogeny of the Spemann organiser therefore requires a better understanding of the events responsible for the creation of the Nieuwkoop centre and of its mode of action.

Several pieces of evidence suggest that creation of the Nieuwkoop centre results from the activation of a Wnt-sig-nalling pathway before the midblastula transition (referred in what follows as the Pre-MBT Wnt-signalling pathway). First, ventral injection of mRNA for several Wnt family members (see Moon, 1993 and references therein) or for molecules acting along the Wnt pathway (as defined from work in Drosophila) such as a Xenopus homologue of the Drosophila Dishevelled protein (Sokol et al., 1995), a mutant version of the GSK-3 kinase or β-catenin (see Gumbiner, 1995 and references therein), leads to the creation of an ectopic Nieuwkoop centre. Second, depleting the embryos of β-catenin, thereby blocking the Pre-MBT Wnt pathway, prevents the formation of the endogenous Nieuwkoop centre (Heasman et al., 1994). These experiments provide convincing evidence that activation of the pre-MBT Wnt pathway is an important step in the creation of the Nieuwkoop centre. However, this does not necessarily imply the involve-ment of a dorsalising Wnt molecule in axis determination, as several other molecules can directly activate more downstream components of the pathway (Heasman et al., 1994).

The mode of action of the Nieuwkoop centre has also been analysed in some detail (reviewed by Christian and Moon, 1993; Holowacz and Elinson, 1995; see also Dale and Slack, 1987). From these experiments, dorsal vegetal cells appear to emit a combination of factors, some of which have general mesoderm-inducing activity, while others, referred to as modifiers or dorsalisers, lack this activity but are able to cooperate with mesoderm inducers to generate the dorsal-most cell types. This cooperation could take place at different levels. For example, the binding of dorsalisers and inducers to the cell surface could: (1) alter the properties of receptors for mesoderm inducers; (2) activate a transduction pathway that converges with that activated by mesoderm inducers before reaching the nucleus; (3) directly activate or repress specific target genes that will in turn cooperate with target genes down-stream of mesoderm inducers. The distinction between these models has so far been hampered by our ignorance of the identity of the endogenous Nieuwkoop factors.

Recently, using an expression cloning strategy, Lemaire et al. (1995) have cloned a zygotic homeobox gene named Siamois, which induced a secondary axis when over-expressed in ventral-vegetal blastomeres of Xenopus embryos. The progeny of the injected blastomeres do not contribute to the ectopic axis, indicating that overexpression of Siamois in ventral-vegetal cells is sufficient to confer a Nieuwkoop centre identity. Consistent with this idea, the expression of Siamois is localised during normal development in the dorsal vegetal cells of Xenopus blastulae and young gastrulae. As Siamois codes for a transcription factor, it was proposed that it acts by regu-lating the expression of components of the Nieuwkoop centre signal. In the work reported in this paper, we have analysed the regulation and mode of action of Siamois. We show that Siamois is activated by the Pre-MBT Wnt pathway but not by mesoderm inducers. Furthermore, we demonstrate that over-expression of Siamois in animal caps confers two organiser properties to these cells in the absence of detectable mesoderm differentiation: they can dorsalise ventral mesoderm, and secrete a neural inducer. Our data also suggest that Siamois, in addition to antagonising the BMP-4 pathway, is activating a novel dorsalising pathway.

Embryos were in vitro fertilised, dejellied, cultivated in 10% MBS, and injected with mRNA as previously described (Lemaire and Gurdon, 1994). 4.6 nl rhodamine lysinated dextran (RLDx; 5 mg/ml in water; M_r 10×10³, Molecular Probe) was injected into both blas-tomeres of two-cell embryos.

Synthetic capped mRNA was prepared as previously described (Lemaire et al., 1995). noggin and goosecoid mRNAs were synthe-sised as described by Lemaire et al. (1995). Synthetic Siamois mRNA was prepared from a plasmid, pBSRN3 XSia- orF, containing the Siamois open reading frame but lacking the 5′ and 3′ UTR. Synthetic chordin and Xbra mRNAs were synthesised as in Sasai et al. (1994) and Cunliffe and Smith (1992) respectively. Bmp-4 mRNA was prepared as described by Dale et al. (1992). Activin βB mRNA was prepared from pSP64TActivin βB (a gift from Doug Melton). bFGF mRNA was made from plasmid containing bFGF cDNA in pSP64TbFGF (a gift from Betsy Pownall). Xnr-2 mRNA was prepared as described by Jones et al. (1995). Xnr-3 mRNA was syn-thesized from pCS2/Xnr-3 (a gift from Chris Wright). bVg1 mRNA was prepared as described by Thomsen and Melton (1993). Xwnt-8 mRNA was prepared as described by Lemaire et al. (1995).

All explants were cultured in 1× MBS. To determine the competence of ventral mesoderm to be dorsalised by noggin-, Siamois- or chordin-expressing caps, ventral marginal zone explants, composed mostly of ventral mesoderm cells, were cultured until stage 12 or 13 as sand-wiches (two combined explants) in 1× MBS containing 0.1% BSA. At the appropriate stage, the two ventral marginal zone pieces were separated and combined with stage 9 animal caps (see also Fig. 6A). All sandwiches were cultured in 1× MBS with 0.1% BSA until sibling control embryos reached the indicated stage.

Tissue explants were fixed in MEMFA (Hemmati-Brivanlou and Harland, 1989) for 3 hours and kept (overnight or longer) in methanol at −20°C. 10 μM sections were cut from tissue explants or whole embryos embedded in Histoplast:beeswax (98:2). Immunostainings were performed using four monoclonal antibodies: 12/101 (muscle specific; Kintner and Brockes, 1984), MZ15 (notochord specific; Smith and Watt, 1985), D7F2 (recognising the muscle-specific protein MyoD; Hopwood et al., 1992) and 4d (recognising the pan-neural molecule N-CAM; Watanabe et al., 1986). Incubations, washes and colour reactions (using NBT-BCIP (Boehringer MA) or the alkaline phosphatase substrate I (Vector labs) were performed according to the method of Hopwood et al. (1992). When double staining was performed, the first colour reaction was stopped by incubating the sections in 1× MBS with 5 mM EDTA overnight at 4°C before incu-bation with the second mAb. Where indicated, the nuclei were stained with Hoechst 33258 (Boehringer, 2 μg/ml during 40 minutes).

RNase protection assays were performed as previously described (Lemaire and Gurdon, 1994) using 10-15 animal cap sandwiches per tube. Quantitations were performed using a Molecular Graphics phos-phorimager running the ImageQuant software. Antisense RNA probes for the FGF receptor, Xbra, gsc and Xlim-1 were prepared as described by Lemaire et al. (1995). The Bmp-4 probe was prepared as described by Dale et al. (1992). The Siamois probe was generated using T7 poly-merase from a plasmid named pXSia BglII 350 that contained the 5′ 312 bp BglII fragment of the Siamois cDNA (Lemaire et al., 1995) cloned in pBluescript SK(−). The chordin antisense probe was derived from the Xenopus chordin full-length clone (a gift from Dr E. De Robertis; Sasai et al., 1994) as described by Ryan, K., Garret, N., Mitchell, A. and Gurdon, J. B. (unpublished data). The noggin probe was prepared using T7 RNA polymerase from pnogginΔAS, a deriv-ative of the plasmid pnoggin Δ5′(a gift from Dr R. Harland; Smith and Harland, 1992) lacking an AvrII-SphI fragment in the 5′ end of the noggin cDNA. The Xnr-1 and Xnr-2 probes were prepared as described by Jones et al. (1995). The Xnr-3 probe was prepared from the plasmid pdor3 (a gift from Dr W. C. Smith; Smith et al., 1995). The plasmid pdor3 was linearised with PvuII and the antisense probe was synthesised using T7 polymerase.

Organiser genes such as goosecoid, noggin or chordin have been shown to be activated in ectodermal (animal cap) cells by high but not by low doses of the mesoderm inducer activin (reviewed by Dawid, 1994; Kessler and Melton, 1994). To test if Siamois could also be activated by activin, we injected several concentrations of activin mRNA into animal blas-tomeres of two-cell stage embryos, excised the animal tissue at stage 9 and analysed for the presence of transcripts for Siamois and several mesoderm or organiser genes at stage 10. As described previously, Xbrachyury (Xbra) induction was strongest at low activin concentration while induction of dorsal genes such as goosecoid (gsc), noggin or chordin was highest at high concentrations (Fig. 1A and data not shown). In contrast, Siamois was not induced in animal caps injected with low or high amounts of activin mRNA (Fig. 1A). In addition, injection into animal blastomeres of high or low doses of mRNA for the mesoderm inducers, Xnr-2, Bvg1, encoding a processed form of Vg1, Bmp-4 or bFGF (Dawid, 1994; Kessler and Melton, 1994; Jones et al., 1995) led to the activation of Xbra but also failed to activate Siamois (Fig. 1B). Therefore Siamois, in contrast to gsc, noggin and chordin, is not a target of mesoderm inducers in animal cap cells.

To study the effect of the activation of the Pre-MBT Wnt pathway on the expression of Siamois in animal cap cells, we injected 1-100 pg of Xwnt-8 mRNA into the animal pole of 2-cell embryos and analysed the expression of Siamois in stage 10 explanted animal caps. Injection of as little as 1 pg of Xwnt-8 mRNA into animal blastomeres was sufficient to induce Siamois but none of the Xwnt-8 mRNA concentrations tested resulted in the induction of gsc or Xbra gene expressions (Fig. 1C). Therefore, Siamois, like Xnr-3 (Smith et al., 1995), is activated by the pre-MBT Wnt-signalling pathway but not by mesoderm inducers.

While injection of Xbra mRNA into ectodermal (animal cap) cells only gives rise to ventral mesodermal cell types (Cunliffe and Smith, 1992), Xbra mRNA can cooperate with Xwnt-8 mRNA to give rise to dorsal mesodermal cell types (Cunliffe and Smith, 1993). As the results described in the previous section indicated that Siamois acts downstream of the Pre-MBT Wnt pathway, we tested if Siamois can also cooperate with Xbra. Xbra mRNA, Siamois mRNA or a combination of both were injected into the animal pole of two-cell embryos, and animal caps obtained from these injected embryos were excised at stage 9.5 (Fig. 2A). At stage 17, animal caps injected with Xbra mRNA or Siamois mRNA alone showed no elongation, suggesting the absence of dorsal mesoderm (Fig. 2B-D). In contrast, substantial elongation was observed in all animal caps co-injected with both messages (n=6/6; Fig. 2E). To characterise the type of mesoderm present in the elongated caps, they were cultured until stage 32 and immunostained with the monoclonal antibodies 12/101 (Kintner and Brockes, 1984) and MZ15 (Smith and Watt, 1985) which recognise muscle and notochord cells, respectively. Muscle or notochord differentiation was not observed in Siamois-injected animal caps (n>20) (data not shown) and few or no muscle cells were present in animal caps injected with Xbra alone (Fig. 2F). However, injection of a combination of Xbra and Siamois mRNA led to the formation of large blocks of muscle (n=6/6) and notochord (n=2/6) (Fig. 2G). Therefore, Siamois, like Xwnt-8, can cooperate with Xbra to give rise to dorsal mesoderm.

To test if Siamois could also dorsalise normal ventral mesoderm, different amounts of Siamois mRNA were injected into the two ventral-vegetal blastomeres of 4-cell embryos. Siamois-expressing ventral mesoderm tissue were explanted at the early gastrula stage, cultured until stage 26, and then analysed for the presence of muscle or notochord (Fig. 3A). Injection of as little as 1 pg of Siamois mRNA was sufficient to lead to massive muscle differentiation. Injection of 10 pg of Siamois mRNA induced mostly notochord differentiation (Fig. 3B). Injection of 100 pg of Siamois mRNA suppressed all muscle differentiation and led to the presence of limited amounts of notochord (Fig. 3B), which may reflect the con-version of Siamois-expressing mesoderm cells into the most dorsoanterior mesoderm: the prechordal plate.

Thus, like Xwnt-8 (Sokol et al., 1991; Cunliffe and Smith, 1993), Siamois can respecify, to dorsal fates, both normal ventral mesoderm tissue and that obtained by overexpression of Xbra in ectoderm. However, Siamois can dorsalise embryonic ventral mesoderm tissue more strongly than it can dorsalise Xbra-induced ventral mesoderm. This suggests that these two types of ventral mesoderm are qualitatively different and that, in addition to Xbra, Siamois cooperates with other mesodermalising factors present in the marginal zone.

The results above establish that Siamois can dorsalise ventral mesoderm of normal or ‘artificial’ nature, but does not dis-criminate between a cell-autonomous or non cell-autonomous mode of action. As during embryogenesis Siamois is mainly expressed in dorsal-vegetal cells underlying the mesoderm (Lemaire et al., 1995), Siamois probably respecifies mesoderm from ventral to dorsal fates in the embryo by activating a secreted dorsaliser(s). To confirm this hypothesis, we injected Siamois mRNA (25, 50 or 100 pg) into animal poles of 2-cell embryos and conjugated stage 9 animal caps derived from these embryos to pieces of early gastrula ventral mesoderm labelled with the fluorescent lineage tracer rhodamine-lysinated dextran (RLDx). The conjugates were cultured until the equivalent of stage 18 and analysed for the presence of the skeletal muscle determining protein MyoD using the D7F2 monoclonal antibody (Hopwood et al., 1992), or cultured until the equivalent of stage 26 and analysed for the presence of dif-ferentiated dorsal tissues such as mature skeletal muscle and notochord by immunostaining (Fig. 4A). Animal caps from embryos injected with 25, 50 or 100 pg of noggin mRNA were used as a positive control for dorsalisation in conjugate exper-iments (Smith et al., 1993).

When the ventral mesoderm tissue was cultured alone (data not shown) or conjugated with an uninjected animal cap, no 12/101 or MyoD staining was detected (Fig. 4B; Table 1). In contrast, both Siamois- and noggin-expressing animal caps can dorsalise early gastrula ventral mesoderm as demonstrated by the presence of muscle cells (MyoD+ and 12/101+) only in the mesodermal part of the conjugate (Fig. 4B; Table 1). MyoD or 12/101 staining was never detected in animal cap cells. As the injection of Siamois or noggin mRNA into the ventral marginal zone of early embryos results in the conversion of ventral mesoderm into axial mesoderm including notochord (Fig. 3 and data not shown), we were surprised to find no sign of notochord differentiation in conjugated ventral mesoderm explants (Table 1). A possible explanation for this difference is that, in our conjugate experiments, the ventral mesoderm is exposed to factors secreted by the animal caps from a later stage than when mRNAs are directly injected into the ventral marginal zone of early embryos. Alternatively, Siamois-induced dorsalising molecule could cooperate with additional factors present in ventral marginal zone but not in animal cap cells.

Activation of a secreted dorsaliser in animal caps is not a general property of dorsalising homeobox genes: animal caps injected with 100 pg or 500 pg of mRNA for the homeobox gene gsc (Cho et al., 1991) failed to dorsalise ventral mesoderm in our assay (Fig. 4B, Table 1) although gsc mRNA used here induced partial secondary axes when injected into ventral blas-³ tomeres of 4-cell stage Xenopus embryos (not shown).

Finally, we tested the effect of activating the pre-MBT Wnt ₂ pathway on the dorsalising properties of animal caps. Surpris-ingly, injection of 10 pg of Xwnt-8 mRNA in animal caps also conferred dorsalising activity to the injected caps (Fig. 4C). ¹

From these results, we conclude that, in the absence of any vegetal or marginal factor, Siamois, like Xwnt-8, activates the secretion of a dorsalising factor by animal cap cells.

In an attempt to identify target genes of Siamois involved in the dorsalising activity of Siamois-expressing caps, we over-expressed Siamois mRNA in animal caps and looked for effects on genes thought to play a role in the dorsoventral pat-terning of the embryo during the blastula and early gastrula stages (for a review see Dawid, 1994).

Animal caps injected with increasing doses of Siamois mRNA were dissected at stage 9, cultured until stage 10.25 or 10.5/11, and analysed by RNase protection for the expression of early genes with dorsalising, ventralising, or mesoderm-inducing activity. First, Siamois activates neither Xbra, an early trunk mesodermal marker (Smith et al., 1991), nor Xlim.1, a marker for both trunk and head mesoderm at this stage (Taira et al., 1992 and D. Caillol and P. L., unpub-lished results), thus strengthening our proposition that Siamois does not induce mesoderm in animal caps (Fig. 5A). Second, overexpression of Siamois induced the dorsal genes gsc, chordin, Xnr-3 and to a lesser extent noggin, and repressed the ventralising gene Bmp-4 (Fig. 5), demonstrat-ing the acquisition by the injected caps of a dorsal character. Interestingly, as gsc, chordin, noggin and Xnr-3 mark both mesodermal and endordermal dorsal cells (Sasai et al., 1994; Vodicka and Gerhart, 1995; D. Caillol and P. L., unpublished results), activation of gsc, chordin, Xnr-3 and noggin is not incompatible with our proposition that Siamois-expressing animal cells do not form mesoderm and may suggest that they form anterior endoderm. In contrast to Siamois, noggin failed to regulate the expression of any of the above genes in animal caps (Fig. 5).

It has been previously reported that overexpression of gsc in the equatorial region of Xenopus embryos inhibits Bmp-4 gene expression (Fainsod et al., 1994). Furthermore, gsc is able to induce chordin, which can antagonise Bmp-4 activity (Sasai et al., 1995). We tested the possibility that repression of Bmp-4 by Siamois in our system may be mediated by gsc and/or chordin. In animal caps, increasing concentrations of gsc or chordin mRNA did not affect the level of Bmp-4 mRNA (Fig. 5B) suggesting that Siamois represses Bmp-4 independently of the activation of gsc and chordin.

We conclude therefore that Siamois triggers an early dorsal genetic programme in animal caps in the absence of mesoderm. Furthermore, the activation of noggin and chordin, whose proteins can antagonise BMP-4 signalling (Sasai et al., 1995; Re’em Kalma et al., 1995) and the down-regulation of Bmp-4, suggest that the repression of the BMP-4 pathway may play a crucial role in the ability of Siamois-injected caps to dorsalise ventral mesoderm. This is further supported by the observation that the strength of this repression of Bmp-4 increased with the amount of injected Siamois mRNA, thereby paralleling the efficiency of Siamois-expressing animal caps to dorsalise ventral mesoderm in our conjugate assay (Fig. 5B; Table 1).

To test if secretion of chordin or noggin could account for the dorsalising activity of Siamois-injected caps, we have used the fact that ventral mesoderm progressively looses its competence to become dorsalised during gastrulation (Lettice and Slack, 1993). If Siamois acts in animal caps mainly through the secretion of noggin or chordin, one would expect that ventral mesoderm would simultaneously lose its competence to be dor-salised by Siamois-, noggin- or chordin-expressing caps. The assay used to address this issue is schematised in Fig. 6A. Animal caps expressing Siamois, noggin or chordin were isolated at stage 9 and conjugated with RLDx-labelled ventral mesoderm of increasing age. The competence of early (stage 10.25), mid (stage 12) or late (stage 13) gastrula ventral mesoderm to become dorsalised by the injected caps was then determined by looking for the presence of muscle cells in the conjugates. Stage 9 animal caps were used in these experi-ments because the ability of Siamois-injected caps to dorsalise early mesoderm was maximal at that stage (data not shown). Nearly all (n=58/60) ventral mesoderm explants combined at stage 10.25 or stage 12 with Siamois- or noggin-expressing animal caps were dorsalised as demonstrated by the presence of muscle cells in the mesodermal part of the explant (Fig. 6B). The proportion of muscle cells in the labelled mesoderm ranged from about 30 to 60% depending on the amount of RNA injected (25 pg or 100 pg) (Fig. 6B). Animal caps injected with 600 pg of chordin mRNA could also dorsalise stage 10.25 or 12 ventral mesoderm. This activity, however, appeared to be weaker than that elicited by noggin or Siamois as conjugates presented about 30% of differentiated muscle cells in the labelled mesoderm (Fig. 6B). Animal caps overexpressing Siamois markedly failed to dorsalise stage 13 ventral mesoderm (Fig. 6B): in 9 out of 10 explants analysed, the con-jugates contained only ∼1-5% of muscle cells. In contrast, animal caps dissected from embryos injected with noggin mRNA or chordin mRNA dorsalised stage 13 ventral mesoderm (Fig. 6B).

The different capacities of Siamois-injected caps on the one hand and of chordin- or noggin-injected caps on the other hand to dorsalise stage 13 ventral mesoderm suggest that secretion of these two latter molecules is not sufficient to account for the dorsalising properties of Siamois-injected caps. Three TGF-β related molecules with dorsalising potential, Xnr-1, Xnr-2 and Xnr-3, have been isolated recently (Jones et al., 1995; Smith et al., 1995). Of these, only Xnr-3 is activated by Siamois in animal caps (Fig. 5A and data not shown).

As Xnr-3 is not activated by noggin in animal caps (this study; Smith et al., 1995), we tested whether the state of acti-vation of Xnr-3 in animal caps may account for the qualitative difference in dorsalising activity of caps injected with Siamois or noggin mRNA. Injection of up to 1 ng of Xnr-3 mRNA promoted elongation of the caps but failed to confer dorsalis-ing activity (data not shown), suggesting that Xnr-3 dorsalis-ing activity does not account for the difference in dorsalising activity of Siamois- or noggin-expressing animal cap cells.

In addition to being able to dorsalise ventral mesoderm cells, organiser cells have the property of inducing neural tissue in overlaying ectoderm (reviewed in Kessler and Melton, 1994). We tested if Siamois-expressing caps also had this property. Injection of as little as 10 pg of Siamois mRNA in animal blas-tomeres and explantation of the injected animal caps at stage 9 resulted in the appearance in the injected caps of large cement glands, an anterior proneural structure (Sive et al., 1989) (16/17 caps analysed, Fig. 7A). In addition, a large proportion of injected animal caps contained cells expressing the pan-neural marker N-CAM (7/17 caps injected with 10 pg of Siamois mRNA and 3/3 caps injected with up to 100 pg of Siamois mRNA, Fig. 7B).

To test if neural or proneural differentiation is a cell-autonomous property of Siamois-expressing cells or results from the secretion by these cells of a neural inducer, we con-jugated stage 9 animal caps injected with various amounts of Siamois mRNA with stage 9 animal caps injected with the lineage tracer RLDx alone and looked for the presence of RLDx-positive cells in the induced cement glands or N-CAM-positive cells (Fig. 7C-E). Analysis of 10 conjugates injected with up to 100 pg of Siamois mRNA and positive for N-CAM staining showed that N-CAM-positive cells were never found in RLDx-positive cells. In contrast, in all animal conjugates injected with 10 pg, 33 pg or 100 pg of Siamois mRNA (n=30), the induced cement gland contained RLDx-positive cells (Fig. 7D,E), indicating that overexpression of Siamois in animal cap cells conferred on them cement gland-inducing ability.

Formation of a functional Organiser has been proposed to result from the synergistic action of the pre-MBT Wnt-sig-nalling pathway and mesoderm induction. Activation of Siamois (this study) and Xnr-3 (Smith et al., 1995) in animal caps in response to Wnt signalling but not in response to mesoderm induction demonstrates that activation of the pre-MBT Wnt signalling on its own is sufficient to trigger a dorsal genetic programme during the blastula stages. Furthermore, as animal caps injected with Xwnt-8 mRNA have dorsalising properties, our results suggest that, while a synergy with mesoderm induction is required for the differentiation of organiser cells into axial structures, activation of the pre-MBT Wnt pathway alone is sufficient for early embryonic cells to acquire organiser properties. In addition, our finding that Siamois, a target of the pre-MBT Wnt-signalling pathway, can cooperate with Xbra, an early target of the mesoderm induction pathway, to induce dorsal mesoderm strongly suggests that the synergy between the pre-MBT Wnt pathway and mesoderm induction takes place at the level of the early target genes of these two pathways.

In keeping with a possible role of Siamois as a mediator of the pre-MBT Wnt-signalling pathway, Xwnt-8 mRNA and Siamois mRNA have several properties in common: unlike Xnr-3, another target of the pre-MBT-signalling pathway, both can induce complete secondary axes when injected into the ventral marginal zone and both confer similar dorsalising properties on injected animal caps (Sokol et al., 1991; Lemaire et al., 1995; this study). However, several differences exist. For example, animal caps co-injected with Siamois and Xbra mRNA elongate (Fig. 2), but animal caps injected with Xwnt-8 and Xbra mRNA do not (Cunliffe and Smith, 1993). Similarly, injection of Siamois, but not of Xwnt-8, mRNA into animal caps results in neural differentiation of ectodermal cells (this study, Cunliffe and Smith, 1993). However, activation of the pre-MBT Wnt-signalling pathway by overexpression of β-catenin in animal caps results in the activation of a cement gland marker (McGrew et al., 1995). An explanation for these differences may be that, although we used Xwnt-8 mRNA as a convenient tool to mimick the activation of the pre-MBT Wnt-signalling pathway, it may also have other effects. Overex-pression of Xwnt-8 after the midblastula transition leads to a partial ventralisation of dorsal structures (Christian and Moon, 1993). Thus, injection of Xwnt-8 mRNA may lead to two antagonistic sequential effects in animal tissue: until the mid-blastula transition the pre-MBT Wnt-signalling pathway is activated, thereby promoting dorsal development, while after the MBT, residual Xwnt-8 mRNA or protein antagonises dorsal development. The differences observed between injection of Siamois and Xwnt-8 mRNA could reflect the fact that overex-pression of Siamois has the same consequences as the activa-tion of the pre-MBT Wnt-signalling pathway without having the subsequent ventralising effects of Xwnt-8.

We have identified four potential target genes for Siamois: gsc, chordin, Xnr3 and Bmp-4. Although our experimental system does not provide us with proof of a direct regulation, this pos-sibility is not excluded: chordin and gsc accumulate shortly after Siamois in the dorsal-vegetal region of the late blastula (Sasai et al, 1994 ; Lemaire et al., 1995 ; Smith et al., 1995; D. Caillol and P. L., unpublished results) and the domain of expression of Xnr3 also overlaps with that of Siamois in the dorsal vegetal epithelium (Lemaire et al., 1995; Smith et al., 1995). The amounts of injected Siamois mRNA necessary to regulate the expression of these genes (25-100 pg) are much higher than the level of endogenous Siamois mRNA present in dorsal vegetal cells (see Fig. 1 for a comparison between the levels of gsc and Siamois mRNA in whole embryos). While it is difficult to relate the amount of injected mRNA to the amount of protein produced in the embryo, this may suggest that, in vivo, Siamois cooperates with other vegetal factors to regulate these genes.

While the regulatory sequences for chordin, Xnr-3 and Bmp-4, have not been characterised, the regulatory sequences of gsc have been analysed recently (Watabe et al., 1995). In this study, the authors identified a proximal and a distal regulatory element responsive to Wnt and activin signalling respectively. As Siamois is not activated by mesoderm inducers, regulation of gsc by Siamois is likely to be mediated by the Wnt-respon-sive element. This element, both in Xenopus and in the mouse, contains two potential ATTA homeodomain consensus binding sites (Watabe et al., 1995). It will therefore be interesting to test if Siamois can bind to these sequences and to analyse the effect of their mutation on gsc regulation.

Activation of gsc by Siamois is unlikely to play a significant role in the dorsalising activity of Siamois-injected caps: while the dorsalising ability of Siamois-injected animal caps increases with the amount of injected Siamois mRNA, the acti-vation of gsc by Siamois is maximal at low concentrations of injected Siamois mRNA and decreases with higher doses. However, gsc dorsalising potential has been shown to be restricted to the vegetal and equatorial cells (Niehrs et al, 1994). Activation of gsc by Siamois in the marginal zone could therefore contribute to the normal function of Siamois in the embryo.

Antagonising the BMP-4 pathway is sufficient to create an organiser (reviewed in Lemaire and Kodjabachian, 1996). For instance, noggin and chordin, two Spemann organiser genes, act as inhibitors of BMP-4 protein activity (Re’em Kalma et al., 1995; Sasai et al., 1995). The concerted repression of Bmp-4 and activation of chordin and noggin gene expression by Siamois may account for the organising properties of this gene. However, two pieces of evidence suggest that Siamois also activates a parallel dorsalising pathway. First, the time of loss of competence of ventral mesoderm for dorsalisation by animal caps expressing either noggin, chordin, or Siamois differs (this study). This suggests that, although one of the consequences of the overexpression of Siamois in animal caps is to repress the BMP-4 pathway, the main dorsalising activity detected in our assay may be due to the activation of another dorsalising pathway. Second, secondary axes generated by overexpression of noggin, chordin, or a truncated BMP receptor in whole embryos generally lack the anterior-most structures, while Siamois efficiently induces secondary axes with a complete head (Graff et al., 1994; Sasai et al., 1994; Lemaire et al, 1995). This suggests that antagonising the BMP-4 pathway results in the creation of a trunk organiser while overexpression of Siamois induces both head and trunk organisers. Surprisingly, Xlim.1, the only gene shown to be required for head formation (Shawlot and Behringer, 1995) and a gene displaying similar effects as Siamois when overexpressed in animal caps (Taira et al., 1994; Sasai et al, 1995), is not activated by Siamois in these cells. Xnr-3, because of its ability to dorsalise gastrula ventral mesoderm, its expression in dorsal vegetal cells and its activation by Siamois in animal caps may account for the dor-salising activity of Siamois. However, Xnr-3-expressing animal caps fail to dorsalise ventral mesoderm (data not shown), suggesting that if Xnr-3 plays a role in the organising properties conferred by Siamois it probably requires a co-factor.

Repression of the BMP-4 pathway in ectoderm by overex-pressing noggin, chordin or dominant negative forms of a BMP-4 receptor or of the ligand itself is sufficient to convert this tissue into cement gland and neural tissue (reviewed in Lemaire and Kodjabachian, 1996). Neuralisation by Siamois may therefore reflect the repression of the BMP-4 pathway in Siamois-expressing ectoderm. It is so far unclear whether Siamois-expressing cells are directly converted into neural tissue: while our data indicate that Siamois-expressing caps do not form mesoderm, they do not exclude the possibility that they contain some dorsal endoderm. Indeed, activation of the dorsal mesendodermal genes, gsc, chordin, noggin and Xnr-3, but not of the mesodermal genes, Xlim.1 and Xbra, may suggest the presence of early endodermal cells in Siamois-injected caps, a possibility that will need further investigation. Our finding that cement gland induction by Siamois can occur in a non cell-autonomous fashion is in keeping with a role for Siamois in neural induction during normal develop-ment: the progeny of cells expressing Siamois at the early gastrula stage will form during gastrulation the anterior mesendoderm, located underneath the ectoderm most highly specified for cement gland formation (Sive et al., 1989) and expressing chordin. However, the inducing ability of anterior mesendoderm is questionable as Sive and colleagues (1989) have shown that explanted anterior mesendoderm from mid-gastrula lacks neural and cement gland-inducing ability, while Sharpe and Gurdon (1990) showed that it harboured anterior neural inducing ability. Therefore further studies will be necessary to determine precisely the potential role of Siamois in neural induction during development.

On the basis that injection of Siamois mRNA into ventral-vegetal cells could induce a complete secondary axis to which the injected cells do not participate, we proposed that Siamois may confer a Nieuwkoop centre-like activity on embryonic cells (Lemaire et al., 1995). This is in keeping with our finding that Siamois is a target of Wnt signalling but not of mesoderm induction (this study). However, while the Nieuwkoop centre is thought to act during the blastula stages, the results presented in this article indicate that signals activated by Siamois are able to dorsalise gastrula mesoderm and induce neural structures. While it is not ruled out that Nieuwkoop centre signals have properties different from expected, an alternative interpretation of our results is that Siamois is involved in conferring organ-ising properties on late blastula or early gastrula vegetal cells. Indeed, these cells have been proposed by Gerhart and collab- orators (1991), to constitute the late blastula organiser, an inter-mediate organising centre acting between Nieuwkoop centre and Spemann organiser.

In summary, the findings reported in this article suggest that Siamois may be an important mediator of the Pre-MBT Wnt-signalling pathway and indicate that overexpression of Siamois in embryonic cells confers upon them two organiser properties in the absence of recognisable mesoderm: ability to dorsalise ventral mesoderm during gastrulation and to induce neural tissue. Therefore mesoderm induction, while being necessary for the differentiation of organiser cells into dorsal mesoderm, may play a more minor role than the Pre-MBT Wnt-signalling pathway in the acquisition by these cells of early organising potential.

We thank J. Brockes, D. Melton, J. Smith, L. Dale, E. De Robertis, R. Harland, F. Watt and C. V. E. Wright for reagents, and Nigel Garrett for cDNA constructs. Special thanks go to B. Rodbard, R. Coulson and M. Thoday for their help and E. Tweed and G. Tétart for keeping our frog colonies. We thank Agnes Chan and Daniel Mahony for encouragement during the course of this project. This manuscript was improved by critical reading of C. Henderson, K. Kato, C. Niehrs, A. Zorn, F. Stennard and K. Ryan. This work was supported by a Cancer Research Campaign programme grant SP 2184/0101 to J. B. G., CNRS ATIPE no. 7 and Ligue Nationale contre le Cancer (Comité départemental des Bouches du Rhône) grants to P. L, a Fondation pour la Recherche Médicale postdoctoral fellowship to L. K., and European Science Foundation and European Human Capital and Mobility (BMH1-CT94-7091) postdoctoral fellowships to G. C.