Difference in XTcf-3 dependency accounts for change in response to β-catenin-mediated Wnt signalling in Xenopus blastula

Embryonic organisation requires cell-to-cell signalling. The Wnt family of secreted glycoproteins provide molecular signals that enable intercellular signalling. In the last few years, Wnt signalling functions have been described in many animal species (reviewed by Cadigan and Nusse, 1997), but most importantly, Wnt signalling has also been found to function in different tissues and at different stages of development within the same species (e.g. Cadigan and Nusse, 1997; Gavin et al., 1990; Gradl et al., 1999; Klingensmith and Nusse, 1994).

As Wnt signalling functions in different tissues, the question arises of how the tissue-specific response to Wnt signalling is carried out in the responding tissue? Maybe different tissues express different Wnt receptors at their cell membranes, which link extracellular Wnt signalling to different tissue-specific cytoplasmic signal transduction cascades. Perhaps, Wnt receptors trigger upstream events in the same cytoplasmic signal transduction cascade, but the tissue-specific cytoplasmic factors will channel the intracellular pathway to act upon different downstream components in order to elicit a tissue-specific response. Alternatively, however, Wnt signals could generally use the same cytoplasmic signal transduction pathway but interactions of nuclear Wnt pathway components with tissue-specific nuclear factors could specify the appropriate response in any given tissue.

In fact, different Wnt signal transduction cascades have recently been suggested for different Wnt ligands and for different tissues (Fig. 1). The canonical Wnt pathway (Fig. 1A, see recent reviews by Miller et al., 1999; Polakis, 2000; Zhurinsky et al., 2000) is the best known after being extensively described in Drosophila (Bhanot et al., 1996; Brunner et al., 1997; Siegfried et al., 1994) and in vertebrate tissues (Behrens et al., 1996; Molenaar et al., 1996; Yang-Snyder et al., 1996; Yost et al., 1996). The extracellular Wnt ligand is suggested to interact with transmembrane receptors that are related to the Drosophila frizzled gene. Intracellular effects of Wnt signals and Frizzled-related receptors are executed by the dishevelled protein. In response to Wnt signalling, dishevelled protein functions to inhibit GSK3β-mediated phosphorylation of β-catenin. This inhibition results in accumulation of cytoplasmic β-catenin, owing to increased protein stability. Binding of β-catenin to members of the Tcf family of transcription factors is thought to be required for the transcriptional regulation of Wnt-responsive target genes in the nucleus.

Alternative pathways downstream of Wnt ligands and Frizzled receptors have also been described. Wnt7a signalling controls axonal remodelling in the developing mouse brain by modulation of GSK3β function, as in the canonical Wnt pathway, but implicates a different GSK3β substrate from β-catenin (Fig. 1B; Lucas et al., 1998). The original frizzled gene controls tissue polarity during Drosophila eye development, via dishevelled as in the canonical pathway, but recruits different cytoplasmic kinases from GSK3β as downstream components to its signal transduction cascade (Fig. 1C; Boutros et al., 1998). Similarly, Wnt5a overexpression in early Xenopus embryos regulates the activity of both protein kinase C (PKC) and Ca²⁺/calmodulin-dependent protein kinase II, rather than glycogen synthase kinase β (GSK3β), via a Ca²⁺-dependent mechanism (Fig. 1D; Kuhl et al., 2000a; Kuhl et al., 2000b; Sheldahl et al., 1999). It is very interesting that Xwnt-5A is also perfectly able to function via the canonical Wnt pathway. The choice of pathway through which Wnt5a signals is dependent on the expression of different classes of Frizzled receptor (He et al., 1997). There is even an example whereby two different Wnt signals elicit distinct responses in the very same tissue by using different intracellular signal transduction pathways (Kengaku et al., 1998). Given these examples, it is easy to imagine that the response to Wnt signalling could be specified in different tissues or at different stages of development by selectively employing one of the signal transduction cascades mentioned above (Fig. 1) or other yet undiscovered pathways.

We have investigated the mechanisms of stage-specific Wnt signalling using early Xenopus embryos as a model. There is a dramatic change in response to Wnt signalling from a dorsalising response in early blastula stages (Christian et al., 1991) to a ventrolateral-promoting response in late blastula embryos (Christian and Moon, 1993), which are only a few hours older. Very early in development, maternally encoded components of the canonical Wnt signalling pathway (see above and Fig. 1A) induce the dorsal side of the embryo (e.g. Heasman et al., 1994; reviewed by Moon and Kimelman, 1998). This effect can be reproduced experimentally by RNA-mediated misexpression of agonists of the canonical Wnt pathway, such as Wnt1 (McMahon and Moon, 1989), Xwnt-8 (Christian et al., 1991; Smith and Harland, 1991; Sokol et al., 1991), Dishevelled (Rothbacher et al., 1995; Sokol et al., 1995), β-catenin (Guger and Gumbiner, 1995) and the Tcf-family transcription factor LEF-1 (Behrens et al., 1996). Ectopic expression of these molecules on the prospective ventral side induces an ectopic dorsal axis. Although Xwnt-8 was found to be the most potent of these axis-duplicators, endogenous Xwnt-8 is expressed neither at the right time nor in the right place for inducing the endogenous dorsal axis. Instead, Xwnt-8 is expressed in the ventral and lateral mesoderm in late blastula and during gastrulation (Christian et al., 1991; Smith and Harland, 1991). Xwnt-8 is both sufficient and required to specify ventrolateral mesoderm and for restricting dorsal mesoderm to the prospective dorsal side (Christian and Moon, 1993; Hoppler et al., 1996; Hoppler and Moon, 1998). This effect can be experimentally reproduced by DNA-mediated misexpression of Xwnt-8 (Christian and Moon, 1993). Ectopic expression of Xwnt-8 on the prospective dorsal side during late blastula and gastrulation stages represses dorsal midline structures (e.g. notochord) and replaces them with more lateral structures (e.g. somitic muscles). How is the stage specificity controlled to switch from the dorsal-promoting in the early blastula to the ventrolateral-promoting Wnt response in the late blastula?

We show that ventrolateral mesoderm-promoting Wnt signalling in Xenopus embryos at the late blastula stage is facilitated by the same cytoplasmic components of the canonical Wnt pathway as dorsalising Wnt signalling in early blastula embryos. We find, however, that nuclear mechanisms vary between dorsal-promoting and the ventrolateral-promoting Wnt signalling, which could account for the shift in response from early to late blastula stage Xenopus embryos.

Xenopus laevis (Daudin) embryos were generated (Sive et al., 2000) and staged according to Nieuwkoop and Faber (Nieuwkoop and Faber, 1967). Embryos were treated with 0.3 M LiCl for 20 minutes at stage 9.5 (see Fig. 2) or at the stages indicated in Fig. 3. Control embryos were treated at the equivalent stages with 0.3 M NaCl for 20 minutes. All embryos were subsequently washed six times in 0.1×MMR (Marc’s Modified Ringers; Sive et al., 2000) and left to develop to gastrulation stage 11. Embryos were fixed in 1×MEMFA (MOPS, EGTA, MgSO₄, formaldehyde standard media, described in Sive et al., 2000) for 2 hours at room temperature then dehydrated in ethanol and stored at 4°C (Sive et al., 2000).

Capped ΔN-XTcf-3 RNA was synthesised from pT7TS ΔN-XTcf-3 (Molenaar et al., 1996, linearised with XbaI) with T7 RNA polymerase. The N-XTcf-3 DNA template was generated by ligating a BglII fragment from pT7TS N-XTcf-3 (Molenaar et al., 1996), into BamHI-digested pCS2⁺ (pCS2⁺ N-XTcf-3, the sense orientation was confirmed by using an internal SacI site). Capped N-XTcf-3 RNA was synthesised from this pCS2⁺ N-XTcf-3 template (linearised with Asp 718) with SP6 RNA polymerase. Capped β-galactosidase RNA was synthesised from pSP6nβ-gal (Smith and Harland, 1991, linearised with XhoI) with SP6 RNA polymerase. Capped β-catenin69 RNA (a mutated more stable form of β-catenin; Yost et al., 1996) was synthesised from pXBC69 (Yost et al., 1996, linearised with Not I) with SP6 RNA polymerase. RNA was synthesised using the mMessage mMachine kit (Ambion), purified by passing over a ProbeQuant G-50 Micro Columns and injected into embryos (Sive et al., 2000). Approx. 0.9 ng ΔN-XTcf-3 RNA, 1.2 ng N-XTcf-3 RNA, 1 ng β-galactosidase RNA, 0.05 ng β-catenin69 RNA or 0.3 ng CSKA Xwnt-8 plasmid DNA (Christian and Moon, 1993) were injected (alone or in combination) per embryo into the marginal zone of both dorsal blastomeres or both ventral blastomeres at the two-cell or four-cell stage, as indicated in the text and the figure legends. The dorsal or ventral side of the embryo was identified by pigment and cell size differences between dorsal and ventral blastomeres at this stage. Embryos were then left to develop to stage 11 (for analysis of gene expression with whole-mount RNA in situ hybridisation) or to tailbud stages (approx. stage 30, for morphological analysis) and fixed in 1×MEMFA (Sive et al., 2000).

β-catenin overexpression experiments were carried out in embryos transgenic with the plasmid pHS1β-catHSG. β-catenin69 from pXBC69 (Yost et al., 1996) was inserted as a BamHI/NotI fragment into the heat-shock promoter vector pHS1 (Wheeler et al., 2000) to produce pHS1β-cat69. pHS1β-catHSG was created by combining sequences from pHS1β-cat69 (ScaI/NotI (blunt)) with sequences from pHS1GFP3 (Wheeler et al., 2000, HindIII (blunt)/ScaI). Control experiments were carried out with the plasmid pHS1GFP3 (Wheeler et al., 2000). Transgenic embryos were produced according to procedures previously described (Amaya and Kroll, 1999; Kroll and Amaya, 1996). The plasmid was linearised with SalI restriction enzyme. Induction of gene expression was essentially as described previously (Wheeler et al., 2000). Embryos were allowed to develop to stage 8 at 16°C. For induction of gene expression, embryos were transferred between 0.1× MMR at 16°C and 0.1× MMR at 34°C. They were treated four times for 15 minutes at 34°C with intervals of 45 minutes at 16°C. Heat-shock treatment were strictly limited to between stages 8 and 10 to avoid heat-induced gastrulation artefacts (see also, Wheeler et al., 2000). At stage 11 embryos were sorted according to GFP expression (Wheeler et al., 2000) and either fixed in 1× MEMFA for analysis by whole-mount RNA in situ hybridisation or allowed to proceed for morphological analysis.

Whole-mount RNA in situ hybridisation was performed (Harland, 1991) with modifications as described in McGrew et al. (McGrew et al., 1999). Digoxigenin-labelled antisense RNA probes used were Xpo (Sato and Sargent, 1991), Xnot (von Dassow et al., 1993), XmyoDa (Frank and Harland, 1991) and Xenopus chordin (Sasai et al., 1994).

Embryos were homogenised in modified RIPA buffer with protease inhibitors (complete™, Roche; single embryos were lysed in 50 μl for Fig. 4A; 10 embryos were lysed in 200 μl for Figs 5Q, 6Q) then incubated on ice for 30 minutes. Lysates were centrifuged at 13,000 g for 30 minutes in a tabletop centrifuge at 4°C and the supernatant transferred to a new tube. Protein concentration was determined using a BCA Protein Assay Kit (Pierce, 23225). The lysates were mixed with the same volume of 2×sample buffer, 0.05% Bromophenol Blue and β-mercaptoethanol to a final concentration of 10%. The samples were then boiled for 10 minutes and placed immediately on ice. The equivalent of three quarters of one embryo was loaded per lane on an SDS-polyacrylamide gel (SDS-PAGE; 12% gel for Figs 4A, 5Q; 15% gel for Fig. 6Q). After separation by electrophoresis, the protein samples were transferred to nitrocellulose. Duplicates of gels were run to confirm equal protein loading per lane by staining one gel with Coomassie Blue. The gel was incubated with the Coomassie Blue solution for 30 minutes and destained and fixed in destain overnight. For documentation the gels were blotted with a heated vacuum pump onto 3MM filter paper (Whatman). Membranes were blocked at room temperature for 2 hours in 5% milk powder in PBST (PBS with 0.1% Tween 20, Sigma) before being incubated in a humidified chamber with the primary antibody (anti-GFP (a gift from Dr F. Barr, Glasgow University), anti-Myc (clone 9E10, Sigma, to detect β-catenin69 protein), anti-HA-tag (Sigma, to detect ΔN-XTcf-3 and N-XTcf-3 protein), all 1:1000 dilution, for 4 hours at room temperature to overnight at 4°C. The blot was washed five to six times with PBST and incubated with the secondary antibody (peroxidase conjugated anti-sheep (anti-GFP) or rabbit anti-mouse (anti-Myc and anti-HA-tag), all diluted 1:1000, for 4 hours in a humidified chamber at room temperature to overnight at 4°C. Four washes with PBST were followed by three washes with PBS to remove all detergent, which could inhibit the detection procedure. Enhanced chemiluminescence detection was carried out incubating the membranes for 5 minutes before the reaction was detected with X-ray film (Konica medical X-ray film AX). To detect anti-Myc, the blot was stripped for 2 hours at 65°C, after detection of GFP, before being washed five times in PBST and blocked again in 5% milk powder in PBST following the same protocol as described above.

In order to test whether late blastula stage Wnt signalling employs a different cytoplasmic signal transduction machinery from early blastula Wnt signalling, we tested the involvement of GSK3β function at both stages of development. GSK3β function is known to be involved in dorsalising Wnt signalling in early Xenopus embryos (Pierce and Kimelman, 1995). We tested the involvement of GSK3β function in late blastula Wnt signalling by using lithium as an inhibitor (for details see Materials and Methods). Lithium as an inhibitor of GSK3β (Lucas et al., 1998; Stambolic et al., 1996) has been widely used to study GSK3β function during Xenopus development (Fredieu et al., 1997; Hedgepeth et al., 1997; Klein and Melton, 1996; Seufert et al., 1999).

We treated late blastulae (stage 9.5) with lithium to test the involvement of GSK3β function in late blastula Wnt signalling. Similar manipulations have been previously performed to study the effects of lithium on larval body patterning (Yamaguchi and Shinagawa, 1989) and on neural development (Fredieu et al., 1997). In our experiments we analysed the phenotypes not only by morphology (data not shown) but also with region-specific mesodermal molecular markers (Fig. 2) in order to detect the immediate effect of lithium treatment. We find that inhibition of GSK3β by lithium in late blastula causes a ventrolateral-promoting effect. Ventral and lateral mesodermal markers (Xpo and XmyoD) are ectopically expressed and a notochord-specific dorsal marker (Xnot) is repressed, whereas, as a control, a dorsal marker that does not respond to late blastula Wnt signalling (chordin; Hoppler and Moon, 1998) remains unchanged. These results are identical to the ones in DNA-mediated Wnt misexpression experiments (Christian and Moon, 1993; Hoppler and Moon, 1998). Although lithium could conceivably affect other cellular processes in addition to inhibiting GSK3β function (Berridge et al., 1989; Rogers and Varmuza, 2000), the most straightforward interpretation of our results suggests a late blastula Wnt signalling cascade that involves GSK3β.

As a control, we treated embryos with lithium at early blastula stage (stage 6) and analysed the phenotype by morphology and with molecular markers (data not shown, but see Fig. 3A,G,M,S). Lithium treatment of early blastulae results in dorsalisation of the embryo, which manifests itself in a repression of ventral and lateral mesodermal markers (Xpo and XmyoD) and an expansion of dorsal molecular markers (chordin, Xnot). These findings are consistent with the effects of RNA-mediated Wnt misexpression (e.g. McMahon and Moon, 1989; Sokol et al., 1991) and confirm the involvement of GSK3β in dorsalising blastula stage Wnt signalling (Pierce and Kimelman, 1995).

Yamaguchi and Shinagawa have suggested that the effect of lithium on the larval body plan changes dramatically at mid-blastula transition (MBT; Yamaguchi and Shinagawa, 1989). In order to determine the stage of development at which the response of the tissue switches from the early dorsalising to the later ventrolateral-promoting response, we performed a timecourse of lithium treatment that spanned the whole blastula stage and analysed the phenotypes with region-specific mesodermal markers (Fig. 3). To our surprise, we do not find a sudden shift in the embryonic tissue from a dorsal- to the ventrolateral-promoting response, instead we find that the stage for the shift from the dorsal- to the ventrolateral-promoting response differs, depending on the molecular marker with which we analyse the experiment. At mid-blastula stages of development (stage 8.5), lithium treatment (and therefore presumably Wnt signalling) causes some molecular markers to respond in a dorsalised manner (chordin, Fig. 3C; Xnot, Fig. 3I) and others in a ventralised manner (Xpo, Fig. 3U).

Yamaguchi and Shinagawa noticed in their experiments that the effects on the larval body plan were less pronounced when the embryos were treated with lithium during the mid-blastula stage of development (Yamaguchi and Shinagawa, 1989). They interpreted this, however, as a decrease in sensitivity to lithium and therefore proceeded to suggest an exact stage of development when the effect of lithium on the larval body plan changed. Our analysis of individual marker gene expression (Fig. 3) suggests that the decreased sensitivity that Yamaguchi and Shinagawa observed may simply reflect the opposing activities that lithium has on different downstream genes. Our findings however, do not contradict the conclusion drawn by Yamaguchi and Shinagawa: that the change of the effect of lithium on embryonic patterning coincides approximately with MBT and is likely to be a consequence of MBT (Yamaguchi and Shinagawa, 1989).

We do not know whether the marker genes we have analysed in our experiments are direct or indirect targets of Wnt signalling, but our results suggest that some targets can respond to Wnt signalling in a dorsalising and others in a ventrolateral-promoting way at the same stage of development. This is evidence to suggest that there is not a general shift to a different Wnt signal transduction pathway in the cytoplasm of this tissue, but rather that the difference between early and late blastula Wnt signalling lies in nuclear mechanisms that occur in close proximity to individual Wnt-responsive promoters (see below). Although there is no direct evidence, it could be imagined that individual Wnt-responsive promoters or enhancers have cis-regulatory elements that are bound with different affinities by an activated downstream component of the Wnt signal transduction cascade. These Wnt-responsive promoters or enhancers could therefore respond to different levels of Wnt signalling or to different levels of expression of such a downstream component of the Wnt signal transduction cascade.

β-catenin functions downstream of GSK3β in early blastula stage Wnt signalling (Yost et al., 1996) and is sufficient for the dorsalising response (Guger and Gumbiner, 1995). Cytoplasmic injection of a DNA construct containing Xwnt-8 cDNA under the control of a ubiquitous promoter has been shown to be sufficient for the ventrolateral-promoting Wnt response (Christian and Moon, 1993; Hoppler and Moon, 1998). We wanted to test whether β-catenin also functions in ventrolateral-promoting Wnt signalling by overexpressing β-catenin during late blastula stages of development. We found however that expression from DNA constructs with ubiquitous promoters was either mosaic if injected into the cytoplasm or too low if integrated into transgenic embryos (data not shown).

In order to achieve high-level and non-mosaic expression of β-catenin, we turned to the method of heat-induced gene expression that we had recently developed in transgenic embryos (Wheeler et al., 2000). We refined the protocol to obtain high level heat-induced expression during late blastula stages, while avoiding gastrulation artefacts (for details, see Materials and Methods). The results from heat-induced β-catenin overexpression in transgenic late blastulae were very clear (Fig. 4). Embryos with strongly induced GFP marker gene expression have also induced exogenous β-catenin expression. β-catenin-expressing embryos can therefore be recognised by their GFP expression and separated from non-expressing controls. Embryos with exogenous β-catenin develop with a shortened dorsal axis. When such embryos are assayed during gastrulation for marker gene expression, all show ectopic expression of ventral and lateral mesodermal markers (Xpo, XmyoD) and a repression of the notochord-specific dorsal marker (Xnot). These results are strictly dependent on β-catenin overexpression because neither heat-shock treatment nor GFP overexpression is able to produce this phenotype.

β-catenin overexpression in late blastulae therefore affects mesodermal patterning in the same way as inhibition of GSK3β by lithium (see above) and ectopic Xwnt-8 expression from a DNA construct (Christian and Moon, 1993; Hoppler and Moon, 1998). These results show that β-catenin is sufficient to mediate the ventrolateral Wnt response, as well as the earlier dorsal Wnt response (Guger and Gumbiner, 1995). This finding suggests that the canonical β-catenin-dependent Wnt pathway functions in early and late blastula Wnt signalling.

In order to test whether β-catenin function was not just sufficient but also required in late blastulae to promote ventrolateral mesoderm, we expressed a mutated XTcf-3, lacking the β-catenin-binding domain (ΔN-XTcf-3, Fig. 5A), which had previously been shown to inhibit β-catenin-mediated axis duplication (Molenaar et al., 1996). We expressed ΔN-XTcf-3 via RNA injection (see Materials and Methods) into either the prospective ventral side, or as a control into the prospective dorsal side and analysed the phenotype at morphological level and with molecular markers (Fig. 5). In the control experiment, ΔN-XTcf-3 expression on the prospective dorsal side resulted in a reduction of dorsal mesoderm (see also Molenaar et al., 1996), confirming a requirement for β-catenin function in dorsalising early Wnt signalling (Heasman et al., 1994). ΔN-XTcf-3 expression on the prospective ventral side, however, had no discernible effect at all, either on the morphology or on marker gene expression. This difference in activity is not due to differences in ΔN-XTcf-3 protein stability, either between the dorsal or the ventral side of the embryo or between the early and late blastula stages (Fig. 5Q). This experiment therefore suggests that β-catenin function is either not required for ventral-promoting Wnt signalling or not functioning via a XTcf-3 dependent pathway.

ΔN-XTcf-3 lacks the N-terminal putative β-catenin-binding domain and is thought to inhibit early blastula Wnt signalling by binding to XTcf-3 DNA binding sites without being able to activate transcription (Fig. 5A, Molenaar et al., 1996). In order to inhibit β-catenin function more directly, we designed a different mutated XTcf-3 construct. Our new construct, N-XTcf-3 (Fig. 6A), includes the N-terminal sequence domain for β-catenin binding but lacks the C-terminal DNA-binding domain. As above with the ΔN-XTcf-3 construct we expressed N-XTcf-3 via RNA injection (see Materials and Methods) in either the prospective dorsal side (as a control) or the prospective ventral side (to test β-catenin requirement in the ventrolateral mesoderm) and analysed the phenotype at the morphological level and with molecular markers (Fig. 6). In the control experiment, N-XTcf-3, expressed on the prospective dorsal side, reduces dorsal mesoderm development (Fig. 6G-K). This inhibition is similar to the one caused by ΔN-XTcf-3, but N-XTcf-3 is not as strong an inhibitor as ΔN-XTcf-3 (compare Fig. 6G with Fig. 5G). Unlike ΔN-XTcf-3, however, N-XTcf-3 expression on the prospective ventral side inhibits ventral mesodermal development (compare Fig. 6L with Fig. 5L). This difference between the phenotype caused by N-XTcf-3 and ΔN-XTcf-3 is not due to differing protein stability (compare Fig. 5Q with Fig. 6Q) but reflects a genuine difference in functional activity between these two mutant XTcf-3 constructs. N-XTcf-3 also inhibits the expression of ventral and lateral marker genes (Xpo and XmyoD) during gastrulation in those ventral cells to which N-XTcf-3 expression is targeted (Fig. 6O,P). These embryos are affected in a similar way as expression of the inhibitory dnXwnt-8 construct (Hoppler et al., 1996; Hoppler and Moon, 1998). However, while dnXwnt-8 functions as a non cell-autonomous factor (Hoppler et al., 1996), N-XTcf-3 functions only cell-autonomously (see below and Fig. 7E) in those ventral cells in which it is expressed in our experiment.

These experiments suggest that there is a difference in activity between ΔN-XTcf-3 and N-XTcf-3 in their effects on endogenous Wnt signalling. While both inhibit dorsalising Wnt signalling, only N-XTcf-3 inhibits ventrolateral-promoting Wnt signalling. In order to test whether these constructs also have these different effects on ectopic Wnt signalling, we tested them together with RNA-mediated misexpression of β-catenin (Guger and Gumbiner, 1995), to mimic dorsalising Wnt signalling and with DNA-mediated misexpression of Xwnt-8 (Christian and Moon, 1993; Hoppler and Moon, 1998), to mimic ventrolateral-promoting Wnt signalling. Consistent with our findings about their effect on endogenous Wnt signalling, we see that both constructs inhibit axis duplication by β-catenin RNA (Fig. 7C,D), but that only N-XTcf-3 inhibits the effect on dorsal development caused by the Xwnt-8 DNA construct (compare Fig. 7I with 7H). In the same series of experiments, we also confirmed that N-XTcf-3 functioned cell-autonomously by showing that β-catenin-mediated axis duplication is inhibited if N-XTcf-3 RNA is injected into the same blastomere as β-catenin RNA (Fig. 7D), but not if injected into the adjoining blastomere (Fig. 7E).

Several alternative signal transduction pathways downstream of Wnt/Frizzled signalling have recently been described (see Introduction and Fig. 1). The differential use of one or the other of these distinct pathways could have accounted for the dramatic change of response to Wnt signalling within a few hours of development from dorsalising in morula and early blastula stages to ventrolateral promoting in the mesoderm of late blastula and gastrulation stage Xenopus embryos. However, our experiments indicate that both early and late blastula Wnt signalling involve GSK3β function. Our experiments also show that β-catenin executes not just dorsalising Wnt signalling in morulae and early blastulae, but also ventrolateral-promoting Wnt signalling in late blastula and gastrulating embryos. β-catenin overexpression in transgenic Xenopus embryos during late blastula stages is sufficient to promote ventrolateral mesodermal markers and to inhibit a notochord-specific dorsal marker. Inhibition of β-catenin by expression of a β-catenin-binding XTcf-3 construct that lacks the DNA-binding domain shows that β-catenin is also required for promoting ventrolateral mesodermal markers and for inhibiting a notochord-specific dorsal marker. Heasman and colleagues have recently developed a novel antisense approach (i.e. morpholinos) to deplete β-catenin protein in Xenopus embryos (Heasman et al., 2000). The focus of their investigation was the requirement for β-catenin in dorsal and anterior development. However, depletion of β-catenin protein from vegetal ventral blastomeres inhibited ventral development (compare Fig. 2b in Heasman et al., 2000 with Fig. 6L in this study). This finding supports a requirement for β-catenin in ventrolateral mesoderm. Therefore, both gain-of-function and loss-of-function experiments advocate that β-catenin functions in late blastulae to promote ventrolateral mesoderm. These findings clearly suggest that the canonical Wnt pathway functions downstream of both early and late blastula Wnt signalling. The instrument by which the tissue decides whether to respond in a dorsalising or a ventrolateral mesoderm-enhancing manner will therefore not be found in the cytoplasm but rather in differing nuclear mechanisms. In fact this is also borne out by the finding that the way individual marker genes respond to dorsal- or ventrolateral-promoting Wnt signalling changes at slightly different stages of development in mid-blastula embryos.

But what are the nuclear mechanisms that change the response to Wnt signalling in general and to β-catenin function in particular? β-catenin interaction with the Tcf family of transcription factors has been described in Xenopus (Molenaar et al., 1996) and other species (Behrens et al., 1996; Brunner et al., 1997). The binding of β-catenin to members of the Tcf family of transcription factors is thought to be required for the transcriptional regulation of Wnt-responsive target genes. This is certainly true for dorsalising Wnt signalling in early Xenopus embryos. This evidence is based on experiments with a mutated XTcf-3 fragment (ΔN-XTcf-3), which lacks the β-catenin-binding domain but retains the DNA-binding domain. Expression of ΔN-XTcf-3 inhibits β-catenin-mediated axis duplication and endogenous dorsal development (Molenaar et al., 1996). However, we found that expression of the same construct on the prospective ventral side had no effect on ventrolateral-promoting Wnt signalling, despite ΔN-XTcf-3 being a stronger inhibitor of Wnt signalling on the dorsal side than the N-XTcf-3 construct (compare Fig. 5G with 6G). Whereas the N-XTcf-3 construct inhibited β-catenin function directly by binding it, the ΔN-XTcf-3 construct more specifically inhibited XTcf-3-mediated β-catenin function, by suppressing the activation of transcription by the β-catenin-XTcf-3 protein complex (Molenaar et al., 1996).

Our results cannot completely rule out the possibility of two fundamentally different, but still XTcf-3-dependent, mechanisms in early and late blastula embryos. XTcf-3 is clearly expressed not just at early but also at late blastula and gastrula stages (Molenaar et al., 1998; Molenaar et al., 1996), but expression alone cannot be evidence for a function at these later stages. XCtBP, which has recently been identified as a co-repressor of XTcf-3, can in certain constructs mimic some aspects of ventrolateral-promoting Wnt signalling (Brannon et al., 1999), but CtBP molecules have been shown to interact with a large number of transcription factors apart from the Tcf family of proteins (e.g. Nibu et al., 1998; Poortinga et al., 1998; Turner and Crossley, 1998).

The most straightforward interpretation of our results is certainly that the role of XTcf-3 is restricted to early dorsalising Wnt signalling. Consequently, β-catenin would be expected to function via an XTcf-3-independent nuclear mechanism to promote ventrolateral mesoderm. Different transcription factors might therefore interact with β-catenin at late blastula and gastrula stages to accomplish ventrolateral-promoting Wnt signalling. Functional interactions of β-catenin with other Tcf family members (e.g. LEF-1; Behrens et al., 1996) and even with nuclear factors other than the Tcf family of transcription factors are being discovered (e.g. retinoid receptor (RAR), Easwaran et al., 1999; XSox17, Zorn et al., 1999).

Although Wnt signalling can function through several different signal transduction cascades, we have shown here that in the early Xenopus embryo, a dramatic shift in the response to Wnt signalling is not brought about by differential use of distinct signal transduction pathways, but rather by changing nuclear mechanisms. While specific responses in other Xenopus tissues or in other species might still be produced by a switch from one Wnt signal transduction cascade to another, our results highlight the potential versatility of the canonical Wnt pathway to interact with tissue-specific factors downstream of β-catenin in order to achieve tissue specific effects.

We thank all those colleagues who shared their molecular reagents with us; Andrew Bain and Megan Davey for technical assistance; Olivier Destree, Aaron Zorn, Andrea Münsterberg, Maike Schmidt and Inke Näthke for discussions; and Cheryll Tickle for comments on the manuscript. We thank the Cold Spring Harbor Xenopus Course for help and stimulation and the BBSRC (F. S. H.) and The Wellcome Trust (G. N. W. and S. H.) for their financial support.