Hex acts with β-catenin to regulate anteroposterior patterning via a Groucho-related co-repressor and Nodal

In Xenopus, axis specification is initiated maternally. Dorsoventral polarity is established after fertilisation, by a canonical Wnt signal that leads to the accumulation of maternal β-catenin on the prospective dorsal side of the embryo(Heasman et al., 2000; Marikawa et al., 1997). With the onset of zygotic transcription, the vegetally localised maternal transcription factor VegT induces the expression of the Nodal-related genes Xnr1, Xnr2, Xnr4, Xnr5 and Xnr6(Clements et al., 1999; Takahashi et al., 2000; Xanthos et al., 2001; Yasuo and Lemaire, 1999),which synergise with β-catenin activity to establish the Nieuwkoop centre. Canonical Wnt/β-catenin targets such as Siamois(Carnac et al., 1996; Lemaire and Kodjabachian,1996) and Xnr3(McKendry et al., 1997) are expressed in the Nieuwkoop centre region and are important for induction of the Spemann organiser, also referred to as mesendoderm (reviewed by De Robertis et al., 2000; Harland and Gerhart, 1997). Mesendoderm induction is also mediated by Nodal-related molecules, which act in a dose-dependent manner. The highest levels of Nodal signalling induce the most anterior mesendoderm derivative (anterior endoderm) and consequently the dose of Nodal-related activity correlates with anteroposterior positional information (Chen and Schier,2001; Dyson and Gurdon,1998; Green and Smith,1990; Vincent et al.,2003).

Following mesendoderm induction, the anterior endoderm expresses antagonists of both the canonical Wnt and the Nodal-related pathways, such as Lefty (Meno et al., 1999),Cerberus (Piccolo et al.,1999), Dickkopf (Glinka et al., 1998) and Frzb (Leyns et al., 1997; Wang et al.,1997), which insulate the anterior part of the embryo from the posteriorising influence of continued mesoderm induction(Niehrs, 1999; Piccolo et al., 1999). Accordingly, the anterior endoderm itself may prevent the expansion of trunk into the head field.

A large component of the spatiotemporal patterns of gene expression generated during embryogenesis are mediated by homeodomain transcription factors. However, how these factors integrate their intrinsic activities with specific signalling pathways is largely unknown. Hex, a homeodomain protein required for the specification of anterior mesendoderm(Brickman et al., 2000; Martinez Barbera et al.,2000), has been shown to suppress posterior mesodermal gene expression in the anterior endoderm via the repression of promoters associated with mesendodermal genes such as Goosecoid(Brickman et al., 2000). In addition, overexpression of Hex leads to the expansion of anterior endoderm expressing Wnt and Nodal antagonists such as Cerberus(Brickman et al., 2000; Jones et al., 1999; Zorn et al., 1999). Thus, Hex plays a pivotal role in anterior specification.

There are two possible explanations for the ability of Hex to induce anterior endoderm. It either suppresses more posterior identities in the mesendoderm or it promotes the inductive events upstream of anterior endoderm formation. To examine these possibilities, we asked whether Hex influenced signalling by the pathways responsible for mesendoderm induction. Here we show that Hex promotes anterior endoderm induction by amplifying early Wnt signalling. To elucidate the mechanism involved in this process, we looked for Hex target genes and identified the Groucho-related co-repressor/Wnt antagonist Tle4 as a Hex target in both mouse and Xenopus. We also identified four other Hex targets including Nodal. Gain and loss-of-function experiments in Xenopus suggest that anterior specification by Hex is dependent on Tle4 and a subset of Xenopus Nodals. Together, our work supports a conserved model in which Hex promotes anterior endoderm through both amplification of transduced Wnt signals followed by suppression of Nodal-mediated mesoderm induction.

Xenopus laevis embryos were obtained by in vitro fertilisation,cultured according to Slack et al. (Slack et al., 1984) and staged according to Nieuwkoop and Faber(Nieuwkoop and Faber, 1994). mRNAs for injection were synthesised from linearised plasmids according to Smith (Smith, 1993).β-catenin, mouse Hex, Hex-λVP2 and nuclear GFP were linearised with NotI; Xnr1 was linearised with SmaI; Siamois was linearised with SacII; Xtle4 (IC 7011138) was linearised with HpaI. All were transcribed with SP6 polymerase. Activin A (kindly provided by J. C. Smith)was added to explant culture based on units defined by Cooke et al.(Cooke et al., 1987). A morpholino oligonucleotide (GeneTools) was used against the Hex 5′UTR:CCTTAGCTGTACGTCATGGTCGTGG. The control morpholino was the standard provided by GeneTools.

To inhibit protein synthesis, whole embryos were incubated with 10 μg/ml cycloheximide (CHX) from stage 8 to 10.5 and analysed by in situ hybridisation. To monitor protein synthesis inhibition, embryos were cultured in media containing [³⁵S]methionine with and without CHX, and incorporation of radiolabelled methionine into Xenopus tissue was measured.

Total RNA was isolated using Absolutely RNA RT-PCR (Stratagene). RNA (0.5μg) was used for cDNA synthesis according to Brickman et al.(Brickman et al., 2000). Real-time RT-PCR was performed using a LightCycler (Roche). For Xenopus, cDNA from uninjected DMZ tissue from stage 10.5 embryos was serially diluted and used as template for a standard curve assayed in each run. For ES cells, standard curves were generated using varying amounts of plasmid containing the appropriate gene. Samples were normalised to Ornithine decarboxylase (Odc) for Xenopus embryos and to β-actin for ES cells. PCR primers and conditions are listed in Table 1. Melting curve analysis was performed on all PCR runs to ensure that quantification was based on the accumulation of the specific PCR product.

In situ hybridisation was performed according to Harland(Harland, 1991), except that BMpurple (Roche) was used as a substrate. Xnr3, Goosecoid, Chordin,Cerberus and Derrière were linearised with EcoRI; Siamois was linearised with StyI; nuclear GFP was linearised with BamHI; Xnr1, Xnr2 and Xtle4 (IC 7011138) were linearised with SmaI; Xnr5and Xnr6 were linearised with NotI; Xlim1 and BF1 were linearised with XhoI; and Xbra was linearised with EcoRV. All were transcribed with T7 polymerase,except for BF1, which was transcribed with SP6. Double in situ hybridisation was performed based on a protocol developed in N. Papalopulu's laboratory (Hardcastle and Papalopulu,2000).

HEK 293T cells (2×10⁵) were plated on a 24-well plate and transfected a day later with 10 ng of reporter plasmids (TOPflash or FOPflash from Upstate Biotechnology) and 90-100 ng of test plasmids according to Brickman et al. (Brickman et al.,2001).

ES cells were cultured as described by Li et al.(Li et al., 1995). All cell lines were derived from CGR8s (Nichols et al., 1990). Details for targeting vector construction are available on request. Plasmids are described elsewhere(Brickman et al., 2000; Soriano, 1999; Srinivas et al., 2001). Vectors were electroporated into ES cells(Schmidt et al., 1997),G418-resistant colonies expanded and genotyped by Southern blot(Soriano, 1999). Two independent clones of Rosa26 HexERT2-λVP2 ES cells and one clone of Rosa26 Hex ES cells were electroporated with pCAGCreIP (gift of A. Smith) and plated at clonal density. Individual clones were picked, expanded and genotyped by PCR (5′-CAAACTCTTCGCGGTCTTTC, with either 3′-GCGTAATCAGGAACGTCATA to detect non-recombined locus or 3′-CCAGAGGCCACTTGTGTAGC for recombined), and analysed by Southern and western blot.

ES cells (2×10⁶) were plated on 10 cm²gelatinised dishes and cultured for 8-12 hours. Cells were then cultured with or without 600 nM 4-hydroxytamoxifen (4-OHT) for 7.5 hours. RNA was prepared using RNeasy and QIAshredder (Qiagen). RNA quality was tested with an Agilent Bioanalyzer and a minimum of three samples applied to Affymetrix MGU74Av2 GeneChip Oligonucleotide Arrays as described in Frye et al.(Frye et al., 2003). Data were normalised using the RMA algorithm in Genespring 7.2. Expression measurements for each gene were represented as average fold-change between each treatment condition group. The groups used for comparison were Rosa-Hex compared with Cre-recombined Rosa-Hex (RRosa-Hex); non-recombined cell lines Rosa-HexERT2-λVP2 (clone1) and Rosa-HexERT2-λVP2 (clone2) with and without 4-OHT; and the recombined versions of these lines(RRosa-HexERT2-λVP2 clones 1 and 2) with and without 4-OHT(Fig. 4A,B). Fold filtering was performed to select for profiles that differed in both RRosa-HexERT2-λVP2 clones upon 4-OHT treatment and did not change in the non-recombined controls. The list of genes obtained was further filtered to remove candidates with low absolute expression levels. During the analysis,some individual samples expressed lower levels of the pluripotency marker Oct4. The candidate list was therefore re-examined excluding these samples for an added level of refinement. All microarray data are available from GEO(series accession number GSE5141).

Hex is normally expressed in embryonic regions where signalling pathways responsible for mesendoderm induction are active. One of the functions of a transcriptional repressor such as Hex may be to ensure that specific targets of these pathways are activated only in the appropriate cells. We tested this hypothesis with respect to Wnt signalling by co-expressing Hex alongsideβ-catenin in regions where neither Hex nor this pathway is active, i.e. ventral-vegetally. Although ectopic expression of β-catenin induced a complete secondary axis with both head and trunk structures, the combination of Hex and β-catenin induced only an ectopic head and in some cases dorsalised the primary axis. Embryos were scored both morphologically and by expression of the neural marker BF1 (forebrain/otic vesicle area)(Fig. 1A, see Table S1 in the supplementary material). Injection of Hex on its own produced a small ventral outgrowth of the gut (Fig. 1A,inset).

The ability of Hex to anteriorise β-catenin activity can be interpreted in two ways; Hex suppresses the induction of trunk byβ-catenin or Hex anteriorises β-catenin signal. To distinguish between these possibilities, we examined whether Hex could anterioriseβ-catenin activity in ventral marginal zone (VMZ) explants. Although modest doses of β-catenin (500 pg) induced dorsal morphology, it never induced key anterior features such as cement glands or eyes(Fig. 1B, see Table S2 in the supplementary material). However, when this dose of β-catenin was combined with Hex, VMZ explants were anteriorised and appeared indistinguishable from uninjected dorsal marginal zone (DMZ) explants. Interestingly, a similar phenotype was observed in response to higher doses ofβ-catenin (1000 pg), suggesting that Hex increased the ability of low levels of β-catenin to transduce the canonical Wnt signal. Together,these results support a role for Hex in anteriorising a β-catenin signal.

To better understand the level at which Hex regulated Wnt signalling, we looked at Wnt-responsive gene expression. We quantified the expression of theβ-catenin direct targets Siamois and Xnr3 in VMZ explants by real-time RT-PCR (Fig. 1C,D). We included in the analysis Hex-λVP2, a VP16 fusion protein that turns Hex into a transcriptional activator(Brickman et al., 2000). Hex and Hex-λVP2 showed very little or no effect in the absence ofβ-catenin. However, the levels of Siamois and Xnr3induced by β-catenin were significantly increased by co-injection of Hex(Fig. 1C) and suppressed by co-injection of Hex-λVP2 (Fig. 1D). Similar results were obtained by in situ hybridisation where the upregulation of Xnr3 expression was particularly pronounced in the animal hemisphere (Fig. 1E, insets and see Table S3B in the supplementary material).

To reveal the consequence of activating β-catenin and Hex locally, we injected Hex, Hex-λVP2 and/or β-catenin mRNA into specific blastomeres at the four-cell stage, alongside nuclear GFP (nucGFP) used as a lineage tracer and examined by double in situ hybridisation (light blue, Fig. 1E). When embryos were injected on the dorsal side, where β-catenin is normally active, Hex strongly increased endogenous Xnr3 expression, whereas Hex-λVP2 inhibited it within the injected cells(Fig. 1E, see Table S3A in the supplementary material). Moreover, co-injection of Hex and β-catenin into a single-ventral blastomere led to a pronounced enhancement of the ability ofβ-catenin to induce ectopic Xnr3(Fig. 1E, see Table S4A in the supplementary material).

To determine whether endogenous Hex was required forβ-catenin/Nieuwkoop centre activity, Hex was knocked-down by injection of an antisense morpholino oligonucleotide (MO). We used a previously described Hex MO known to induce an anterior phenotype in Xenopus laevis(Foley and Mercola, 2005; Smithers and Jones, 2002). Siamois and Xnr3 expression was strongly reduced by Hex depletion, and this depletion also led to a reduction in the ability ofβ-catenin to induce ectopic expression of both Siamois and Xnr3 (Fig. 1F, see Table S7 in the supplementary material). The phenotypes obtained were rescued by mouse Hex mRNA (Fig. 1F),which differs with Xenopus Hex in DNA sequence around the ATG initiation codon, but is indistinguishable in terms of activity(Brickman et al., 2000; Smithers and Jones, 2002).

Although the immediate early response to β-catenin signalling is the induction of direct targets such as Siamois and Xnr3, these regulators have complex activities and induce a range of different cell types,including both the anterior endoderm expressing Cerberus and Hex itself, but also more posterior axial mesoderm expressing genes such as Goosecoid. Interestingly, anterior gene products such as Hex and Cerberus then block the induction of more posterior cell types, creating a complex network of antagonistic interactions(Brickman et al., 2000; Piccolo et al., 1999). Based on the anteriorised phenotypes observed in Fig. 1, we hypothesised that amplification of Wnt signalling by Hex should preferentially induce anterior endoderm, suggesting a role for Hex in specifying regional identities within the mesendoderm. To test whether Hex-amplified Wnt signalling preferentially induced anterior endoderm and determine how this early activity of Hex relates to its later known role in the suppression of posterior mesendodermal identity, we looked at the expression of a number of mesendodermal markers in response to a Hex-amplified β-catenin signal. Fig. 2 shows the response of the anterior endodermal marker Cerberus, alongside the axial mesendodermal markers Goosecoid and Chordin to a Hex-amplified Wnt signalling.

As Hex has been previously found to induce Cerberus within dorsal endoderm (Fig. 2B)(Jones et al., 1999; Brickman et al., 2000; Zorn et al., 1999), we were not surprised to find that β-catenin and Hex synergised to induce ectopic Cerberus expression (Fig. 2A,B, see Table S4 in the supplementary material). Consistently,morpholino-mediated depletion of Hex resulted in both a reduction in endogenous Cerberus and in the ability of β-catenin to induce ectopic Cerberus (Fig. 2F). However, in contrast to the results obtained with Cerberus, co-expression of Hex alongside β-catenin led to a context-dependent enhancement of Wnt-stimulated ectopic Goosecoid and Chordin. Thus, in an explant culture that gives an indication of the net effect in a population of cells, the ability of Hex to stimulate Wnt signalling appeared to compete with the ability of Hex to inhibit axial mesoderm (Fig. 2C,D). Modest levels of Hex (200 pg) inhibited β-catenin-mediated stimulation of these genes, whereas at higher doses of Hex (500 pg), β-catenin-stimulated gene expression recovered somewhat more pronouncedly in VMZ explants(Fig. 2D). However, when this effect is examined at a cellular level, by in situ hybridisation on an injected clone of cells, Hex-stimulated levels of Wnt signalling appeared to predominate (Fig. 2E, see Tables S5 and S6 in the supplementary material). This observation was particularly striking with respect to Goosecoid, a defined Hex target, where Hex and β-catenin synergise to induce Goosecoid,but the effect is predominately cell non-autonomous(Fig. 2E, rightmost panels). As with Goosecoid, Chordin is also synergistically induced byβ-catenin and Hex, but in this case the induction is primarily cell-autonomous (Fig. 2E,rightmost panels), implying that the inhibition of Chordin expression observed in explant culture (Fig. 2C,D) was probably an indirect effect arising from the ability of Hex to suppress axial mesoderm.

These observations are consistent with the view that Hex is required for the maximal level of early β-catenin signalling to induce anterior endoderm and then in the mesendoderm to specify/maintain anterior endoderm by suppressing axial mesoderm. Remarkably, we found that the ability ofβ-catenin to induce all three markers was impaired in Hex-depleted embryos (Fig. 2F, see Table S7 in the supplementary material). Moreover, in Hex-depleted embryos whereβ-catenin was not overexpressed, Cerberus and Chordinexpression was lost, whereas that of Goosecoid was expanded(Fig. 2F, see Table S7 in the supplementary material). These phenotypes were rescued with mouse Hex RNA and presumably reflect the ability of Hex to both regulate early Wnt signalling and suppress axial mesoderm. The similar behaviour of both Chordinand Cerberus may reflect an absolute dependence on Wnt signalling for the expression of both markers at blastula stages(Agius et al., 2000). However,blastula stage expression of Goosecoid and the majority of mesendoderm, is induced by a combination of both Wnt and Nodal-related signalling and thus may be less dependent on the ability of Hex to amplify early Wnt signalling and more sensitive to Hex-mediated suppression of axial mesoderm. Consistent with this interpretation, we found that other mesendodermal markers that are dependent on both Nodal-related and Wnt signalling for their initial induction such as Xlim1 and Brachyury (Xbra) were increased in Hex-depleted embryos (see Fig. S1A and Table S7 in the supplementary material). Moreover, the ability of Hex to synergise with β-catenin to preferentially induce anterior endoderm and suppress more posterior mesendoderm was also evident in their synergistic ability to suppress Derrière, a gene involved in posterior axial patterning (see Fig. S1B in the supplementary material).

As the amplification of Wnt signalling by Hex occurs at the level of immediate-early targets of the Wnt pathway, we asked if it could be recapitulated in vitro. We tested the ability of Hex to regulate the activation of TOPflash, a minimal Wnt reporter in HEK 293T cells(Fig. 3A). Co-transfection of Hex with β-catenin enhanced the ability of β-catenin to activate TOPflash transcription (Fig. 3B). As Hex is known to act as a transcriptional repressor and the activator Hex-λVP2 suppressed β-catenin activity in vivo, we addressed whether in vitro activation of TOPflash was indirect. Consistent with an indirect mechanism, co-transfection of Hex-λVP2 dramatically reduced the ability of β-catenin to induce the TOPflash reporter(Fig. 3B). In no case did Hex,Hex-λVP2 or β-catenin significantly affect the levels of FOPflash,a mutant version of TOPflash used as a negative control.

Normally, induction of the Wnt signal suppresses the ability of GSK3 to phosphorylate β-catenin (Yost et al.,1996). The unphosphorylated form of β-catenin is then stabilised and enters the nucleus to interact with the Tcf/Lef DNA-binding proteins and induces Wnt targets. To determine at which level Hex regulates this pathway, we asked whether Hex or Hex-λVP2 could modulate the activity of: (1) stabilised β-catenin, a constitutively activated variant of β-catenin that had been rendered insensitive to GSK3 regulation through mutations of GSK3 phosphorylation sites(Munemitsu et al., 1996); (2)LefdnβCTA, a nuclear localised fusion of the activation domain ofβ-catenin with Lef1 that is still dependent on the ability of the activation domain of β-catenin to stimulate transcription(Vleminckx et al., 1999); and(3) Xtcf3-λVP2, a fusion of Xtcf3 to two copies of the VP16 activation domain that acts independently of any β-catenin specific activation function. Activation of TOPflash by all three molecules was potentiated by Hex and efficiently suppressed by Hex-λVP2(Fig. 3C), strongly suggesting that Hex modulates the expression of a Wnt nuclear antagonist downstream ofβ-catenin that can be recruited by the Xtcf/Lef complex.

To gain insight into the mechanism by which Hex positively acts on Wnt signalling, we sought to identify Hex transcriptional targets. To facilitate this process, we used a mouse ES cell-based model, as it enabled us to work with large numbers of cells that reflect an undifferentiated early embryonic phenotype that can be genetically manipulated in vitro.

ES cells expressing Hex and HexERT2-λVP2, an inducible version of Hex-λVP2 that responds to 4-hydroxytamoxifen (4-OHT)(Indra et al., 1999) were generated alongside clonally related control cell lines, by targeting Hex or HexERT2-λVP2 to the Rosa26 locus by homologous recombination(Soriano, 1999; Srinivas et al., 2001)(Fig. 4A). These cell lines contain a stop cassette flanked by LoxP sites that can be removed by transfection with Cre-recombinase, allowing recombination-mediated induction of gene expression. Following excision of the stop cassette, Hex or HexERT2-λVP2 were expressed from the Rosa26 promoter and detected by western blot (see Fig. S2A in the supplementary material). HexERT2-λVP2 activity was confirmed using a Goosecoid reporter gene (see Fig. S2B in the supplementary material).

We used affymetrix MGU74Av2 murine gene chips to profile gene expression in the different Rosa26 ES cell cultures as outlined in Fig. 4B. We reasoned that Hex targets should obey three criteria: (1) induction by 4-OHT in the two independent lines expressing HexERT2-λVP2; (2) regulation by Hex; and(3) invariant expression in parental lines that do not express a Hex derivative. Although the third criterion may exclude some Hex targets, it was an essential noise filter for analysis of heterogeneous ES cell cultures(Fig. 4C). When all three criteria were applied, a list of five candidates presented in Fig. 4D was obtained, and we conclude they represent the most likely Hex targets from this analysis.

Two of the identified genes directly impact on early embryonic signalling:the Wnt antagonist/Groucho-related co-repressor Tle4 and the mesoderm inducer Nodal. Induction of these genes by HexERT2-λVP2 in ES cells in the presence of 4-OHT was confirmed by real-time RT-PCR(Fig. 4E). Based on the sequence of previously identified Hex-binding sites(Brickman et al., 2000; Crompton et al., 1992), we searched the mouse genome for putative sites in regions around Nodaland Tle4. In Tle4, we found several clusters of putative Hex-binding sites located within highly conserved blocks in introns 6, 12, 14 and 16 (see Fig. S3C in the supplementary material; data not shown). The majority of these sequences are conserved in Xenopus in addition to other mammalian genomes. In Nodal, Hex-binding sites were found both upstream and within intron 1 (see Fig. S3A in the supplementary material). This binding site pattern appears to be conserved in the XenopusNodal homologues Xnr1 and Xnr2 (see Fig. S3B in the supplementary material).

The Xenopus homologue of Tle4, Xtle4, has been shown to be a potent antagonist of Wnt/β-catenin signalling at the level of Xtcf in early Xenopus embryos (Roose et al., 1998). Based on the phenotypes obtained in Fig. 1, we reasoned that Hex-mediated repression of Tle4 could provide the molecular basis for the regulation of Wnt signalling by Hex in Xenopus embryos. Fig. 5 shows expression of Xtle4 in response to expression of Hex-λVP2 and Hex depletion by MO. At stage 10.5, Xtle4 is normally expressed throughout the animal hemisphere and in the superficial regions of the marginal zone(Fig. 5A,B). Hex-λVP2 injection in deep dorsal tissue resulted in ectopic induction of Xtle4 (Fig. 5B, see Table S8 in the supplementary material). Moreover, Hex MO injection led to a significant upregulation of Xtle4 in the marginal zone, including the superficial dorsal lip mesendoderm, while injection of mouse Hex RNA rescued this phenotype (Fig. 5A, see Table S7 in the supplementary material).

While there is a single Xenopus homologue of Tle4, there are multiple Xenopus Nodals. Fig. 5 shows the expression of the Xenopus Nodal-related genes Xnr1, Xnr2, Xnr5 and Xnr6 at stage 10.5. As previously described, Xnr5 and Xnr6 are expressed earlier than Xnr1 and Xnr2; however,Xnr1 and Xnr2, but not Xnr5 and Xnr6 are autoregulatory(Adachi et al., 1999; Hyde and Old, 2000; Osada et al., 2000; Vincent et al., 2003). Interestingly, Hex depletion by Hex MO injection resulted in a very pronounced upregulation of Xnr1 and Xnr2, but not of Xnr5 and Xnr6. Moreover, this ectopic expression of Xnr1 and 2 was rescued by mouse Hex, which had no effect on Xnr5 and Xnr6 (Fig. 5A, see Table S7 in the supplementary material). The autoregulatory activity of Xnr1 and Xnr2 also explains why Hex depletion results in ectopic Xnr expression in cells that do not normally express Hex. Although we were unable to detect a role for Hex in the regulation of Xnr5 and Xnr6 by morpholino depletion, Hex-λVP2 mRNA injection induced all the Xnrs tested (Fig. 5B, see Table S8 in the supplementary material). Differential regulation between these subsets of Xnrs may be due to Xnr-specific co-factors required for Hex repression.

To address whether Hex repression of candidate genes was direct, we performed experiments in the presence of cycloheximide (CHX) to block zygotic translation. In Xenopus, the majority of zygotic transcription does not begin until mid-blastula transition (MBT); thus, we were able to transfer Hex-depleted embryos rescued with mouse Hex RNA to media containing CHX prior to MBT to block protein synthesis. As mouse Hex RNA was injected into embryos at the two-cell stage, it was translated before protein synthesis was blocked. Thus, the ability of mouse Hex RNA to inhibit transcription of its targets in the presence of CHX suggests that these targets are direct, and independent of any intermediate molecule. Fig. 5A shows a reduction in Xtle4, Xnr1 and Xnr2expression in the presence of CHX as a consequence of mouse Hex injection,indicating that these genes are direct targets of Hex. We also observed that CHX alone had some effect on embryonic transcription of these genes in addition to developmental timing. Similar observations have been reported previously (Clements and Woodland,2003; Sinner et al.,2004) and these factors can complicate interpretation. However,taken together with the in silico identification of conserved Hex binding sites and the fact that these genes are induced after a short period of treatment of 4-OHT in the array data set, these findings support the notion that Xtle4, Xnr1 and Xnr2 are direct Hex targets.

To determine whether Hex regulation of Xtle4 was relevant to Hex-mediated amplification of Wnt signalling, we overexpressed Xtle4 and asked how this impacted on the ability of β-catenin and Hex to stimulate the Wnt targets Xnr3 and Cerberus in Xenopus embryos. Fig. 6A shows that overexpression of Xtle4 blocked the ability of Hex to potentiateβ-catenin induction of both Xnr3 and Cerberus (see Table S3B and Table S4B in the supplementary material).

A similar effect was observed when Xtle4 was transfected in the context of the TOPflash reporter in HEK 293T cells(Fig. 6B). Co-transfection of Xtle4 significantly reduced the stimulated expression levels obtained withβ-catenin and Hex co-transfection. Interestingly, while co-transfection of Xtle4 alongside β-catenin led to a complete suppression of the reporter, the presence of Hex allowed this activity to recover slightly,although in no case to the level of β-catenin on its own. Together, these data suggest that Xtle4 has a quantitative effect on Wnt-responsive transcription.

Nodal autoregulation is important for the sustained, high-level stimulation of the pathway itself (Norris and Robertson, 1999; Osada et al.,2000). Moreover, conserved Hex-binding sites are located in intron 1 alongside the Nodal autoregulatory element (see Fig. S3 in the supplementary material) (Norris et al.,2002; Osada et al.,2000). Thus, we reasoned that Hex should interfere with the relay,behaving as a Nodal antagonist. We next tested whether Hex misexpression reduced Xnr1 activity. Thus, while ectopic expression of Xnr1 induced a partial secondary axis marked by a small region of BF1 expression in Xenopus embryos, this axis was completely suppressed by co-injection of Hex (Fig. 7A, see Table S9 in the supplementary material). Similarly, the expression of increasing amounts of Xnr1 in VMZ explants led to a degree of dorsalisation and concomitant neural induction as evidenced by patchy BF1 staining,while co-expression of Hex alongside Xnr1 inhibited these phenotypes over a range of concentrations tested (Fig. 7B, see Table S10 in the supplementary material).

As the induction of Cerberus is not only dependent onβ-catenin but also on Nodal signalling, we asked whether Hex stimulates induction of Cerberus by Xnr1. As previously reported and also shown here, Xnr1 is a potent inducer of Cerberus expression(Fig. 7C, see Table S11 in the supplementary material) (Agius et al.,2000; Niederlander et al.,2001). However, when Hex was co-injected with Xnr1, this induction was dramatically reduced (Fig. 7C). This result is particularly surprising, as Hex did induce Cerberus expression on its own and combinatorially withβ-catenin. Thus, high levels of Nodal signalling presumably rely on the ability of the Nodal pathway to stimulate Xnr1 and Xnr2 transcription and Hex might interfere with this process. In support of this notion, we observed a general increase in the activity of ectopic Xnr1 in Hex-depleted embryos(Fig. 7D, see Table S7 in the supplementary material).

The ability of Hex to suppress Nodal-related activity was also tested with Activin, a related TGFβ. Hex inhibited the ability of Activin protein to induce convergent extension phenotypes in animal cap explants(Fig. 7E, see Table S12 in the supplementary material). A quantitative reduction in the ability of Activin to induce Goosecoid, Chordin, Xbra and the endodermal marker Mixer was also observed in these explants(Fig. 7F).

In this paper, we have shown that Hex promotes anterior identity at two levels. Hex quantitatively enhances Wnt signalling by regulating the expression of the Wnt antagonist/Groucho co-repressor Tle4. By elevating levels of Wnt signalling, Hex induces anterior character. We have also shown that Hex suppresses the activity of a subset of Nodal-related ligands. The suppression of Nodal-related transcription by Hex would interfere with the continued expression of Xnr1 and Xnr2 in response to activation of the Nodal pathway and may also contribute to the segregation of anterior endoderm from the continual process of mesendoderm induction. This activity of Hex would also be an important component of the ability of anterior endoderm to insulate the anterior neural plate from the posteriorising influence of continued Nodal signalling (Meno et al., 1999; Perea-Gomez et al., 2002; Piccolo et al., 1999). The model in Fig. 8 illustrates the juxtaposition of these two activities. As Tle4 and Nodal are targets of Hex in both mouse and frog, these regulatory interactions may be conserved.

Our findings that levels of a Groucho-related co-repressor such as Tle4 are specifically regulated by Hex represent a novel regulatory point for the canonical Wnt pathway. Both gain- and loss-of-function experiments demonstrated that Xtle4 is a potent antagonist of Wnt signalling in early Xenopus embryos and represses both Siamois and Xnr3 (Roose et al.,1998). Although Xtle4 is expressed in the animal hemisphere and superficial regions of the marginal zone at gastrula stage(Molenaar et al., 2000), Hex is expressed in the deep and early involuting endoderm. These complementary expression domains support the idea that Hex represses Xtle4 to establish an embryonic region that is responsive to Wnt signalling.

Evidence from a number of organisms suggest that high levels of Nodal signalling are required for the induction of endoderm, particularly anterior endoderm (Aoki et al., 2002; Vincent et al., 2003). However although Nodal signalling is seen as required early for anterior endoderm and later for the continued process of more posterior mesoderm induction,Wnt/β-catenin signalling is generally thought of as a posteriorising signal. Yet, the role of nuclear β-catenin in establishing the Nieuwkoop centre makes it one of the earliest determinants of dorsoanterior identity in both Xenopus and zebrafish (Kelly et al., 2000; Xanthos et al.,2002) and our data support an early, dose-dependent role for Wnt signalling in anterior endoderm specification.

A role for early Wnt signalling in anterior specification is also consistent with recent revisions to the Xenopus fate map that demonstrate an anterior fate for the progeny of cells experiencing high levels of Wnt signalling (Lane and Sheets,2000). Moreover, in Xenopus, both β-catenin(Sinner et al., 2004) and Siamois (Zorn et al.,1999) have been implicated in the induction of anterior endoderm and early Cerberus induction has been shown to be dependent onβ-catenin (Agius et al.,2000; Xanthos et al.,2002). In the mouse, the anterior endoderm is divided into two populations, the anterior definitive endoderm (ADE), formed during gastrulation and the anterior visceral endoderm (AVE), an extra-embryonic tissue formed at peri-implantation stages(Beddington and Robertson,1999). Hex is initially expressed throughout primitive endoderm at the hatched-blastocyst stage and then becomes restricted to the distal region of the visceral endoderm lineage. These Hex-positive visceral endoderm cells migrate anteriorly to become the AVE a day prior to gastrulation. Hex is later expressed in the ADE (Thomas et al.,1998). Genetic evidence links the induction of these endodermal tissues to Wnt signalling in both the ADE(Hsieh et al., 2003; Kelly et al., 2004) and AVE(Huelsken et al., 2000). In zebrafish, Hex is initially expressed asymmetrically as part of a Wnt-mediated signalling centre in the extra-embryonic yolk syncytial layer(Ho et al., 1999). This region is related to the Xenopus Nieuwkoop centre as it induces the earliest mesendoderm (Fekany et al.,1999; Kelly et al.,2000; Yamanaka et al.,1998) and Hex has also been implicated in this activity(Ho et al., 1999). Our data create a molecular framework for these observations and imply that early expression of Hex may represent a conserved component of the Nieuwkoop centre.

The anterior endoderm will continue to express Hex, whereas the axial mesoderm will not. The continued expression of Hex in the anterior endoderm represses both Goosecoid and Nodal-related gene expression. Repression of Nodal transcription by Hex would add to the inhibitory effects of Nodal antagonists such Cerberus and Lefty. A major component of Nodal-related activity occurs via an autoregulatory loop(Adachi et al., 1999; Osada et al., 2000; Vincent et al., 2003) and we have shown that Hex inhibits Nodal signalling by interfering with Nodal transcription. Moreover, the two Xenopus Nodal-related genes that are Hex targets are also the two genes that exhibit autoregulation(Hyde and Old, 2000). Thus,Hex may inhibit Nodal signalling by blocking Nodal production without interfering with Nodal-mediated signal transduction. The implications of this idea are that a Hex-expressing cell could receive the high levels of Nodal signalling required to induce anterior endoderm, but would then no longer transmit Nodal signals to its neighbours. By blocking the continued Nodal autoregulation while still leaving cells competent to respond to the signal,Hex may be in effect defining the end of the anterior axis.

We thank Osmany Larralde and Melanie Clements for technical assistance;Enrique Amaya, Nobue Itasaki, Daniel Metzger, Roger Patient, Shankar Srinivas,Mike Sargent, Andrew Smith, Jim Smith and Paul Wilson for plasmids; Mike Jones for plasmids and the gift of the HexMO; Nancy Papalopulu for plasmids and invaluable technical advice on double in situ hybridisation; Masazumi Tada for technical advice on CHX; Tom Burdon, Tilo Kunath, Richard Meehan, Gillian Morrison, Tristan Rodriguez, Austin Smith and Val Wilson for critical discussions of this manuscript. This work was supported by the Wellcome Trust(062965) and the MRC. J.M.B. was a recipient of a fellowship from the Wellcome Trust.