Repression of Wnt/β-catenin signaling in the anterior endoderm is essential for liver and pancreas development

The liver and pancreas are derived from the foregut endoderm. Although we increasingly understand the genetic pathways regulating proliferation and differentiation of these organs once they are specified(McLin and Zorn, 2006; Zaret, 2002; Zhao and Duncan, 2005), the earlier events linking endoderm formation at gastrulation to organ induction are less clear. A detailed understanding of these early developmental decisions is likely to be important for designing efficient strategies to generate therapeutically useful tissue from stem cells.

Hepatic specification in the ventral foregut endoderm requires signals from the surrounding mesenchyme (Fukuda-Taira,1981; Gualdi et al.,1996; Le Douarin,1975). In the mouse and chick, these signals include Fgf2 from the cardiogenic mesoderm and BMPs from the septum transversum mesenchyme(Jung et al., 1999; Rossi et al., 2001; Zhang et al., 2004), whereas in zebrafish Wnt2bb in the lateral plate mesoderm is required(Ober et al., 2006). There is also evidence that prior to these signals, there are earlier developmental decisions that are crucial for the hepatic lineage. For example,transplantation and explant studies in amphibian and chick embryos have shown that only the foregut endoderm, but not the posterior endoderm, is competent to become liver in response to signals from the cardiogenic mesoderm(Fukuda-Taira, 1981; Le Douarin, 1975; Okada, 1954b; Okada, 1960; Takata, 1960). In addition,tissue recombination experiments have shown that an unknown signal from the trunk mesoderm represses hepatic potential in the posterior endoderm(Gualdi et al., 1996; Le Douarin, 1975; Takata, 1960). Together, these findings suggest that establishment of the foregut domain prior to hepatic induction is an essential prerequisite for liver development.

The regional expression of marker genes along the anterior-posterior (A-P)axis of the gut tube demonstrates that distinct foregut, midgut and hindgut domains exist prior to organ specification(Costa et al., 2003; Gamer and Wright, 1995; Grapin-Botton, 2005; Newman et al., 1997). However,studies on the mechanisms controlling this have been somewhat contradictory. Some experiments in amphibians suggested that regional identity in the endoderm is determined by gastrulation(Hamburger, 1996; Holtfreter, 1938), whereas other studies indicated that continual interactions with the mesoderm are required (Okada, 1954a; Okada, 1954b; Okada, 1960; Takata, 1960). For example,anterior vegetal tissue isolated from Xenopus blastulae will autonomously express foregut genes owing to intrinsic Nodal and Wnt signaling in the explants (Gamer and Wright,1995; Henry et al.,1996; Zorn et al.,1999). However Horb and Slack subsequently detected cardiac and lateral mesoderm markers in such vegetal explants(Horb and Slack, 2001),suggesting that prolonged mesodermal interactions might be required to maintain regional identify, although the molecular details of this remain poorly understood.

In this study we postulated that differential zygotic Wnt signaling might regulate endoderm patterning similar to the `activation-transformation' model of A-P patterning in the neural tube. In this model, neural tissue is initially anterior in character and is then progressively posteriorized by Wnts expressed in the trunk mesoderm, whereas anterior identity is maintained rostrally where there is a lack of Wnt signaling(Christian and Moon, 1993; Erter et al., 2001; Kiecker and Niehrs, 2001; Kim et al., 2000; Lekven et al., 2001; McGrew et al., 1995; Nieuwkoop, 1999; Takada et al., 1994). In the canonical pathway, Wnt ligands interact with a Frizzled-LRP5/6 receptor complex causing the inactivation of a Gsk3β-containing intracellular complex that would otherwise promote β-catenin degradation. This results in the accumulation of nuclear β-catenin, which interacts with Tcf/Lef transcription factors to activate target gene transcription(Clevers, 2006). Secreted Wnt-antagonists block signaling by binding to Wnt ligands in the extracellular space (sFRPs), or by binding to the LRP co-receptor (Dkk1)(Kawano and Kypta, 2003; Mao et al., 2001).

In Xenopus, a maternal Wnt pathway results in high levels of nuclear β-catenin in the dorsal-anterior endoderm of the blastula, which specifies the dorsal-anterior axis of all three germ layers(Heasman, 2006). However,during gastrula and early somite stages, nuclear β-catenin levels are rapidly reduced in the anterior endoderm relative to the posterior endoderm(Schohl and Fagotto, 2002). We postulated that this region of low β-catenin activity is important to maintain foregut identity and that Wnt ligands (Wnt8, Wnt8b, Wnt3, Wnt3a and Wnt1) expressed in the trunk mesoderm(Christian et al., 1991; Kemp et al., 2005; Moon, 1993) signal to the posterior endoderm to inhibit foregut/hepatic fates and promote intestinal development. By contrast, Wnt-antagonists (Dkk1, Frzb1, Crescent and Sfrp5)secreted by the anterior endoderm (Kemp et al., 2005; Leyns et al.,1997; Pilcher and Krieg,2002; Wang et al.,1997) would protect it from the posteriorizing Wnt ligands, keep nuclear β-catenin levels low, and maintain foregut identity. Although such spatial interactions between zygotic Wnt ligands and their antagonists were known to regulate mesoderm and neural tube patterning, their roles in the early endoderm had not been established.

Consistent with this hypothesis, we find that forced Wnt/β-catenin signaling in the anterior endoderm, between gastrula and early somite stages,inhibits foregut development. By contrast, blocking β-catenin activity in the posterior endoderm is sufficient to initiate ectopic liver and pancreas development. This suggests that the endoderm is indeed patterned by differential Wnt signaling, similar to what has been described in the nervous system. The homeobox gene hhex, one of the earliest foregut markers,is a target of this β-catenin-mediated patterning and Hhex function is required for both normal and ectopic liver and pancreas development. Analysis of the hhex promoter indicates that β-catenin/Tcf activity represses hhex transcription in the posterior endoderm indirectly via homeodomain transcriptional repressor Vent2. This is a novel function for Vent factors, which are best known as mediators of BMP signaling in the mesoderm. These results provide a molecular pathway linking endoderm patterning to the initiation of liver and pancreas development, and illustrate that the spatial and temporal activity of Wnt signaling must be tightly controlled during this process.

Xenopus laevis embryos were cultured as previously described(Zorn et al., 1999) and staged according to Nieuwkoop and Faber(Nieuwkoop and Faber, 1994). Explants were microdissected with tungsten needles and cultured in 0.5×MBS (Zorn et al.,1999) with antibiotics. For tissue separations, explants were cultured in 10 μg/ml dispase for 15-20 minutes. Gastrula-stage endoderm tissue was transplanted into host embryos through an incision in the blastocoel roof, as previously described(Slack and Isaacs, 1994).

Embryos with clear pigmentation differences between dorsal-anterior and ventral-posterior cells were selected for 32-cell stage injections. The antisense Hex morpholino oligo (HexMO, 80 ng) has been described previously(Smithers and Jones, 2002). Synthetic RNA for microinjection was transcribed using the Message Machine Kit(Ambion) and purified on Microspin-6 columns (BioRad). The following plasmids were used for mRNA synthesis or directly for microinjection (total amounts of DNA or RNA injected, and enzymes used to linearize DNA templates and synthesize RNA are indicated): pCSKA-Xwnt8 (250 pg)(Christian and Moon, 1993);pCS2+pt-β-catenin (250 pg) (Yost et al., 1996); pCS2+MT Xgsk3β (500 pg)(Yost et al., 1996);pcDNA3-ΔNXTcf3 (800 pg) (Molenaar et al., 1996); pRN3 GFP (100 pg; SfiI, T3); pCS2+β-gal(100 pg; NotI, Sp6); pCS107 tDkk1 (500 pg; AscI, Sp6);pCS2+HA-GR-ΔNLEF-βCTA (800 pg)(Domingos et al., 2001);pCS107-Xvent1 (500 pg; AscI, Sp6); pCS107-Xvent2 (500 pg; AscI, Sp6); pT7TS-HA-Xhex (500 pg; BamHI, T7); pCS2+XfoxA2(500 pg; NotI, Sp6); pT7TSHA-GR-ΔNTcf3 (800 pg; XbaI,T7); pT7TSHA-GR-Vent2 (800 pg, XbaI, T7).

A Xenopus laevis genomic λ-library was screened using standard procedures to isolate hhex genomic clones and a ∼6kb fragment upstream of the ATG start codon (Accession No. EF059707) was subcloned into the pGFP3 vector to create the phhex-6kb:GFP plasmid used in transgenesis. Transgenic Xenopus embryos were generated using sperm nuclear transplantation as previously described(Kroll and Amaya, 1996).

Three hhex promoter fragments -6.0, -1.56 and -0.44 kb were PCR amplified and cloned into the pGL2-Basic luciferase vector (Promega). The hhex:luciferase promoter constructs (300 pg) were microinjected along with a pRL-TK renilla (Promega) control vector (25 pg). At gastrula stage, five injected embryos were pooled together and homogenized in 100 μl of 100 mM Tris (pH 7.5) on ice, assayed for luciferase activity and normalized to renilla activity using standard kits (Promega). Each construct was assayed in triplicate.

In situ hybridizations were preformed as previously described(Costa et al., 2003) using the following probes: for1 (Seo et al., 2002), pdx1/xlhbox8(Wright et al., 1989), nkx2.1 (Small et al.,2000), endocut/darmin(Costa et al., 2003), hhex and ampb (Zorn and Mason, 2001), vent2(Onichtchouk et al., 1998), ptfa1 (Jarikji et al.,2007), amylase and elastase(Horb and Slack, 2002).

Real time RT-PCR analysis was performed on an Opticon PCR machine (MJ Research) as previously described (Sinner et al., 2004). The following primers were used (F, forward; R,reverse): for1 (F, 5′-CTTAATGTGACTGAAGCAGAG-3′ and R,5′-TTTCCATCTGTAGAGCCACAA-3′); c-troponin (F,5′-AAGTCTCCATGGATCTAC-3′ and R,5′-CAGCTCTAACCATTTCAG-3′); endodermin (F,5′-AGCAGAAAATGGCAAACACAC-3′ and R,5′-GGTCTTTTAATGGCAACAGGT-3′); foxa2 (F,5′-CCTATCATGAACTCCTCATAG-3′ and R,5′-GGCCAGAATACATACAGCAGTC-3′); hhex(Zorn et al., 1999); and odc (Sinner et al.,2004).

Maternal Wnt/β-catenin is active in the dorsal-anterior endoderm of the Xenopus blastula, but is dramatically downregulated at gastrulation (Schohl and Fagotto,2002). It is this clearance of zygotic β-catenin in the anterior endoderm that we postulate is crucial to maintain foregut fate. To test this, we microinjected a Wnt8-plasmid into the D1 cells of 32-cell stage Xenopus embryos (Fig. 1A), which are fated to contribute to the foregut(Dale and Slack, 1987; Moody, 1987). The injected Wnt8 plasmid is not transcribed until the gastrula stage(Christian and Moon, 1993) and therefore only activates zygotic β-catenin signaling. At stages 35-37,when organ buds are first visible (Chalmers and Slack, 1998; Nieuwkoop and Faber, 1994), the embryos were assayed by in situ hybridization with the liver marker for1 (Seo et al., 2002), as well as with markers for pancreas/duodenum(pdx1/xlhbox8) (Wright et al.,1989), lung/thyroid (nkx2.1)(Small et al., 2000) and intestine (endocut/darmin) (Costa et al., 2003). In almost all of the treated embryos, foregut gene expression was inhibited and an examination of the gut tube at stage 42 revealed an absence of foregut organ buds in the Wnt8-injected embryos(Fig. 1B).

To determine if increased repression of Wnt signaling was sufficient to expand the foregut, we injected RNA encoding the canonical Wnt-antagonist Dkk1(Glinka et al., 1998) into cells that contribute to the posterior-lateral endoderm(Dale and Slack, 1987; Moody, 1987). Dkk1 overexpression is known to inhibit zygotic but not maternal Wnt signaling(Glinka et al., 1998). The resulting embryos exhibited an expansion of for1 and pdx1expression at the expense of the intestine marker, and at stage 42 the liver and pancreas buds were conspicuously enlarged(Fig. 1B).

As Wnt8 and Dkk1 are secreted, we next addressed whether their effects on the endoderm were direct, or secondary to changes in mesoderm patterning. To cell-autonomously activate Wnt signaling in the anterior endoderm we microinjected RNA encoding a stabilized form of β-catenin(Yost et al., 1996) into the D1 blastomeres at the 32-cell stage. The stabilized β-catenin should persist in the gastrula anterior endoderm long after the time when endogenousβ-catenin is downregulated, and this resulted in the repression of foregut markers as well as ectopic intestinal gene expression in the foregut domain (Fig. 1B and see Table S1 in the supplementary material). Unlike Wnt8 plasmid injections, there was little or no effect on head development in β-catenin-injected embryos,consistent with this acting in the endoderm and not affecting mesoderm or neural tissue. Co-injection of β-galactosidase (β-gal) RNA as a lineage-tracer confirmed that the injections targeted the anterior endoderm(data not shown).

We next asked if reducing canonical Wnt signaling in the posterior endoderm was sufficient to induce foregut fate. To cell-autonomously repressβ-catenin signaling we injected RNA encoding Gsk3β into the D4 posterior endoderm cells at the 32-cell stage. Gsk3β overexpression promotes β-catenin degradation (Yost et al., 1996) and as maternal β-catenin is not active in the ventral-posterior cells, Gsk3β is expected to repress zygoticβ-catenin signaling. We observed ectopic for1 and pdx1expression in the posterior endoderm of 50-60% of the Gsk3β-injected embryos, along with inhibition of the intestinal marker endocut(Fig. 1B and see Table S1 in the supplementary material). Similar results were obtained with a dominant-negative ΔNTcf3 construct(Molenaar et al., 1996) that represses β-catenin/Tcf target genes (see Table S1 in the supplementary material). At stage 42-46, we observed ectopic organ buds in the intestines of Gsk3β- and ΔNTcf3-injected embryos. Some of the ectopic buds expressed early liver and pancreas markers (hhex, pdx1 and ptf1a) as well as hepatic (for1, ambp and transferrin) and exocrine pancreas (elastase and amylase) differentiation markers(Fig. 1C and see Table S1 in the supplementary material). However, we did not observe ectopic insulin expression suggesting that endocrine pancreas development requires additional signals.

To confirm that the injected Gsk3β acted autonomously in the endoderm,we co-injected Gsk3β with β-gal RNA as a lineage label. In none of the cases in which β-gal staining was restricted to the mesoderm(n=17) was ectopic for1 expression observed. By contrast,ectopic for1 expression was observed in 19 cases in which theβ-gal staining was in the endoderm and not the mesoderm(Fig. 1D), indicating that Gsk3β acts in the endoderm.

Since our injections did not disrupt head or axial mesoderm structures,this argues that we did not interfere with the known role of Wnt signaling in early axial patterning, which occurs during blastula and early gastrula stages. To rigorously test this, we repeated our experiments with hormone-inducible constructs that could be temporally controlled.

To confirm that β-catenin activity inhibits foregut fate after gastrulation, we used a GR-LEFΔN-βCTA construct, which has the hormone-binding domain of the glucocorticoid receptor fused to the Lef1 DNA-binding domain and the β-catenin transactivation domain. In the presence of dexamethasone (Dex), this fusion protein constitutively activatesβ-catenin target genes (Domingos et al., 2001). By treating GR-LEFΔN-βCTA-injected embryos with Dex at progressively later stages, we determined that β-catenin signaling must be excluded from the anterior endoderm between stage 11(midgastrula) and stage 20 (6-7 somites). Dex treatment during this period inhibited for1 and pdx1 expression and expanded the intestinal marker into the foregut territory(Fig. 2A). Addition of Dex at stages 25 and 30 had no obvious effect on foregut development when assayed at stage 35. However, when Dex was added from stage 30 to 42, about half of the embryos exhibited enlarged liver buds (Fig. 2B). This is consistent with recent reports in zebrafish and mice suggesting that later in development, β-catenin promotes hepatoblast specification and/or proliferation (Ober et al., 2006; Tan et al.,2006).

To determine the window in time when ectopic foregut development could be induced in the posterior endoderm, we injected GR-ΔNTcf3 RNA into D4 blastomeres at the 32-cell stage. GR-ΔNTcf3 consists of the glucocorticoid receptor ligand-binding domain fused to a dominant-negative Tcf3, which represses β-catenin/Tcf target genes in the presence of Dex(Darken and Wilson, 2001). Activation of GR-ΔNTcf3 at midgastrula (stage 11) or early neurula(stage 13) efficiently induced ectopic for1 and pdx1expression (Fig. 2C), whereas addition of Dex between stages 15 and 20 had no effect. When GR-ΔNTcf3 was injected into D1 anterior endoderm cells and Dex was added between stages 30 and 42, some embryos exhibited smaller liver buds(Fig. 2D) consistent with the late induction of GR-LEFΔN-βCTA. These experiments reveal aβ-catenin function that is distinct from its role in early axial patterning.

We conclude that between the gastrula and early somite stages, repression of β-catenin activity in the endoderm is necessary and sufficient to initiate liver and pancreas development, probably by regulating the formation of early foregut progenitors. However, during later embryogenesis,β-catenin has a second role in promoting hepatic development.

Although the lineage-tracing experiments suggested that β-catenin signaling acts directly on the endoderm to regulate lineage commitment, it was important to rigorously test this. We therefore isolated anterior endoderm from gastrula embryos injected with β-catenin and GFP RNA or posterior endoderm injected with Gsk3β and GFP RNA. This tissue was transplanted into the blastocoel of uninjected sibling host embryos and the fate of the cells was determined at stage 42 (Fig. 3A). By the neurula stage, the transplanted cells had incorporated into the host endoderm near the foregut/hindgut boundary(Fig. 3B,C). At stage 42,control transplants injected with GFP only respected their original anterior-posterior identity. Anterior endoderm was incorporated in the liver and anterior intestine ∼80% of the time, whereas control posterior transplants contributed to the intestine but rarely to the liver. By contrast,transplanted anterior endoderm overexpressing β-catenin almost never contributed to the liver, whereas posterior endoderm tissue injected with Gsk3β behaved like anterior endoderm and populated the liver bud ∼80%of the time (Fig. 3D-H). Importantly, GFP-labeled cells were not observed in the heart or other mesoderm tissue demonstrating that differential β-catenin activity in the endoderm directly restricts regional identity.

We occasionally observed ectopic cardiac gene expression in the mesoderm adjacent to Gsk3β-injected posterior endoderm (see Table S1 in the supplementary material), consistent with a recent report that Wnt-antagonists promote cardiogenesis indirectly by acting on the endoderm(Foley and Mercola, 2005). Since cardiac mesoderm induces liver fate in mouse and chick, we asked whether the mesoderm was also required for normal and Gsk3β-induced liver specification in Xenopus. We isolated foregut explants at stage 18(3-4 somites) and cultured these with or without their associated cardiac mesoderm (Fig. 4A). RT-PCR analysis of the explants at stage 35 revealed that, as in mouse and chick, the mesoderm was required for liver gene expression(Fig. 4B). By removing the mesoderm at progressively later times, we found that liver specification requires an interaction with the mesoderm until stage 28 (20-22 somites) (data not shown). Similarly, an association with the adjacent mesoderm was also required for Gsk3β-injected posterior endoderm to express liver markers(Fig. 4C). We conclude that repression of β-catenin activity in the endoderm initiates a cascade of reciprocal signaling with the mesoderm leading to hepatic induction.

To determine whether low β-catenin activity establishes an early foregut domain, we examined the expression of the homeobox gene hhex,one of the earliest foregut markers(Newman et al., 1997; Thomas et al., 1998) that is essential for early liver and ventral pancreas development in mice(Bort et al., 2004; Keng et al., 2000; Martinez Barbera et al.,2000).

Forcing β-catenin activity in the anterior endoderm by injection of either stabilized β-catenin or GR-LEFΔN-βCTA RNA resulted in a dramatic downregulation of hhex expression at stage 18(Fig. 5A). Conversely,repressing β-catenin activity in the posterior endoderm by microinjecting Gsk3β plus β-gal RNA cell-autonomously induced ectopic hhexexpression in the presumptive hindgut. A similar result was obtained by injecting either ΔNTcf3 RNA or an antisense β-catenin morpholino oligo (Fig. 5A and see Table S1 in the supplementary material), indicating that low β-catenin activity is necessary and sufficient to induce early foregut progenitors.

We also examined the expression of foxa2 and gata4, gata5and gata6 in β-catenin- and Gsk3β-injected embryos, as these transcription factors have been implicated in early mouse liver development (Lee et al., 2005; Zaret, 2002; Zhao et al., 2005). Although we did not observe obvious changes in gata4-6 expression, foxa2 foregut expression was downregulated by anterior β-catenin injection and modestly increased in the posterior endoderm by Gsk3βinjection (see Fig. S1 in the supplementary material). Since the effects on foxa2 expression were subtle, we focused our attention on hhex in our subsequent experiments.

We tested whether Hhex is required for liver and pancreas development in Xenopus. Embryos were injected with an antisense morpholino oligo complementary to the 5′ untranslated region of the hhex mRNA(HexMO), which blocks translation and has been shown to specifically knockdown Hhex function (Smithers and Jones,2002). HexMO-injected embryos exhibited very little if any for1 or pdx1 expression, demonstrating that Hhex is required for liver and pancreas development (Fig. 5B and see Table S1 in the supplementary material). Furthermore,injection of HexMO completely blocked the ectopic for1 and pdx1 expression induced by Gsk3β(Fig. 5B and see Table S1 in the supplementary material), indicating that Hhex function is required downstream of Gsk3β. Surprisingly, both the ventral and dorsal pancreas were absent from HexMO Xenopus embryos (as judged by pdx1expression), in contrast to the mouse Hhex^-/- mutant in which only the ventral pancreas is compromised(Bort et al., 2004). The reason for this difference is not known, but we note that the anterior neural phenotype in HexMO Xenopus embryos is also more severe than in the mouse mutant (Smithers and Jones,2002).

Finally, Hhex overexpressed in the posterior endoderm was not sufficient to induce ectopic for1 or pdx1 expression (data not shown),suggesting that additional factors are required to promote foregut development downstream of repression of Wnt.

To examine how zygotic β-catenin represses hhextranscription, we isolated a -6 kb fragment of Xenopus laevis genomic DNA upstream of the hhex transcription start site and used this to generate germline hhex:gfp transgenic animals. Confocal analysis of gastrula- and presomitic-stage hhex:gfp transgenic embryos showed GFP fluorescence in the anterior endoderm and ventral foregut(Fig. 6A,B), recapitulating endogenous hhex expression. To determine if the -6 kb transgene was regulated by β-catenin, we injected Gsk3β into the posterior D4 cells of transgenic embryos, isolated hindgut explants at the gastrula stage and scored these for GFP at stage 18. As expected, GFP was observed in∼50% of control foregut explants (one parent was heterozygous for hhex:gfp) but was undetected in uninjected hindgut explants(Fig. 6C,D). By contrast, we observed ectopic GFP in 4 of 12 Gsk3β-injected hindgut explants(Fig. 6E), indicating that the-6 kb promoter contains elements that confer correct A-P expression and Gsk3β responsiveness.

To better resolve the Gsk3β/β-catenin-responsive elements, we microinjected plasmids containing deletion fragments of the hhexpromoter driving luciferase expression (hhex:luciferase) into either D1 or D4 cells of 32-cell embryos (Fig. 6F) and the relative luciferase activity was determined at stage 11. The -6 kb and -1.56 kb reporters were highly expressed in the D1 anterior endoderm, but were much less active in posterior D4 cells. Gsk3βinjection robustly stimulated the -6 kb and -1.56 kb constructs in the posterior endoderm. By contrast, the -0.44 kb reporter was more active in the posterior endoderm and was not stimulated by Gsk3β(Fig. 6G). This indicates that promoter elements between -1.56 and -0.44 kb repress hhextranscription in the posterior endoderm and mediate Gsk3βresponsiveness.

Because β-catenin-Tcf complexes normally stimulate transcription, we postulated that hhex expression must be repressed indirectly, by activating a transcriptional repressor. Sequence analysis of the hhexpromoter revealed ten putative DNA-binding sites for the Vent class of homeodomain repressors (Friedle et al.,1998; Trindade et al.,1999), with eight of the Vent-sites located between -1.56 and-0.44 kb. Vent1 (PV.1) and Vent2 (Xbr-1, Vox and Xom) are closely related transcriptional repressors, best known as mediators of BMP signaling during dorsal-ventral mesoderm patterning (Ault et al., 1996; Onichtchouk et al.,1998; Papalopulu and Kintner,1996; Schmidt et al.,1996). Although a function for Vents in the endoderm has not been described, recent evidence suggests that Wnt signaling promotes venttranscription in the mesoderm (Friedle and Knochel, 2002; Ramel and Lekven, 2004).

To test whether Vents could repress the hhex promoter, we injected the hhex:luciferase reporter constructs into D1 anterior cells, with or without RNA encoding Vent1 or Vent2, and assayed luciferase activity. Co-injection of either Vent repressed the activity of the -6.0 kb and-1.56 kb reporters to levels observed in the posterior endoderm(Fig. 7A; data not shown),whereas the -0.44 kb promoter, lacking eight of the Vent-binding sites, was not efficiently repressed. Thus, the -1.56 to -0.44 kb hhex promoter region exhibits all three activities: (1) it mediates hhex repression in the posterior endoderm; (2) it can be activated by Gsk3β; and (3) it can be repressed by Vents.

Next, we re-examined the vent1 and vent2 expression patterns by in situ hybridization of bisected embryos, which exposes the deep endoderm. In addition to their previously documented mesodermal expression, we found that vent2 (and to a lesser extent vent1) was highly expressed in the posterior endoderm during gastrula and early somite stages in a pattern reciprocal to hhex. Moreover, the vent2-expressing endoderm was adjacent to wnt8-expressing lateral plate mesoderm throughout this period of development (Fig. 7B-D and see Fig. S2 in the supplementary material). To test whether vent2 expression in the posterior endoderm required Wnt/β-catenin activity, we injected either Gsk3β, ΔNTcf3 or antisense β-catenin morpholino oligos into the D4 posterior cells, all of which reduced vent2 mRNA levels(Fig. 7E and see Table S1 in the supplementary material). Injection of Vent2 RNA into the D1 anterior endoderm at the 32-cell stage inhibited hhex and foxa2expression as well as subsequent liver and pancreas organogenesis(Fig. 7F; see Fig. S2 and Table S1 in the supplementary material). In addition, co-injection of Vent2 plus Gsk3β RNA in the D4 posterior endoderm blocked the ectopic hhex,for1 and pdx1 expression induced by Gsk3β alone(Fig. 7F and see Table S1 in the supplementary material).

Although we did not observe any changes in axial mesoderm development in these experiments, it was important to determine whether we were observing a Vent2 function distinct from its known role in gastrula-stage mesoderm patterning. We therefore repeated the experiments with a hormone-inducible GR-Vent2 construct that we could activate after early mesoderm patterning. We found that for1 and pdx1 expression was repressed even when the GR-Vent2 construct was activated by Dex as late as stage 20 (see Fig. S2 in the supplementary material). This indicates that Vent2 has a novel function in the endoderm well beyond the gastrula stage. Together, these data suggest a model in which Wnt/β-catenin activity promotes vent2transcription in the posterior endoderm during gastrula and early somite stages, and Vent2 protein then represses hhex transcription and foregut fate in the presumptive intestine(Fig. 7G).

We have identified a molecular pathway linking endoderm patterning to the initiation of liver and pancreas development in Xenopus. Our data supports a model in which the anterior-posterior axis of the endoderm is patterned by differential β-catenin activity during gastrula and early somite stages (Fig. 8). Lowβ-catenin in the anterior endoderm maintains foregut fates, whereas highβ-catenin in the posterior promotes intestinal development. Immunostaining studies indicate that an anterior-to-posterior gradient of nuclear β-catenin is indeed present in Xenopus embryos(Schohl and Fagotto,2002).

Our model predicts that Wnt ligands in the posterior mesoderm (probably Wnt8, Wnt8b and/or Wnt3a) signal to the adjacent endoderm to activeβ-catenin-Tcf complexes that promote expression of the homeobox gene vent2. The presence of Tcf-binding sites in the vent2promoter suggests that it might be a direct β-catenin/Tcf target(Friedle and Knochel, 2002; Rastegar et al., 1999). The transcriptional repressor Vent2 would then inhibit foregut fate by repressing hhex transcription, probably directly via Vent-binding sites in the hhex promoter. By contrast, the anterior endoderm secretes a number of Wnt-antagonists (Frzb1, Sfrp5, Crescent and Dkk1) that protect it from the Wnt ligands, resulting in low β-catenin levels and the expression of early foregut genes such as hhex. This appears to be essential for the development of early foregut progenitors(Deutsch et al., 2001), as both the liver and pancreas fates were affected in our experiments.

Our data also suggest that the repression of β-catenin activity is sufficient to initiate a cascade of reciprocal signaling that controls some,but not all, foregut organogenesis. For example, we observed ectopic liver and exocrine pancreas development, but not endocrine pancreatic fates. In the case of the liver, it appears that low β-catenin activity imparts anterior character to the endoderm and this anteriorized endoderm then induces cardiac fate in the adjacent mesoderm (Foley and Mercola, 2005). The cardiogenic mesoderm then signals back to the foregut endoderm, instructing a subset of the cells to become liver(Zaret, 2002).

We found that Hhex was required, but not sufficient, for Xenopusforegut development. Although the direct targets of Hhex in this context are unknown, recent data in the Xenopus blastula suggest that Hhex acts in part by repressing the expression of the Groucho-family transcriptional co-repressor Tle4 (Zamparini et al.,2006). Repression of Tle4 by Hhex might also be important for maintaining the foregut precursors in a transcriptionally permissive state. However, as ectopic Hhex was not sufficient to initiate foregut development,other factors must also function downstream of the repression of Wnt. One candidate is the transcription factor Foxa2, which was also negatively regulated by β-catenin. Foxa2 is required for foregut development in mice(Dufort et al., 1998) and Zaret and colleagues have shown that Foxa binding to liver gene promoters correlates with hepatic competence (reviewed by Zaret, 2002). Consistent with this, liver development fails to initiate in mice in which Foxa1 and Foxa2 have been deleted from the endoderm(Lee et al., 2005). Another possibility is that even in the presence of Hhex or Foxa2, foregut development requires that intestine-promoting Wnt target genes such as Cdx genes must also be turned off. The Cdx homeobox genes are expressed in the posterior endoderm(and mesoderm), they are required for posterior development(Chawengsaksophak et al., 2004)and Cdx1 has been shown to be a direct β-catenin/Tcf target in mouse and zebrafish (Lickert et al.,2000; Shimizu et al.,2005).

The model of endoderm patterning that we present here has striking similarities to Nieuwkoop's classical `activation/transformation' model of neural patterning (Nieuwkoop,1999). In this model, an initial `activation' signal induces neural tissue of anterior character, followed by a subsequent `transformation'signal that imparts posterior character to the neural tube in a graded fashion, with the highest signal generating the most-posterior fates. There is now good evidence that the posteriorizing signals are a combination of Wnts,FGFs and retinoic acid (Domingos et al.,2001; Kudoh et al.,2002; Sasai and De Robertis,1997), and recent evidence, including our study, suggests that these pathways also regulate endoderm patterning. Wells and colleagues found that high levels of Fgf4 promote intestinal fate and repress foregut development in mice and avian embryos similar to the Wnt activity we have described here (Dessimoz et al.,2006; Wells and Melton,2000). In zebrafish, BMP signaling has been implicated in regulating anterior-posterior endoderm patterning(Tiso et al., 2002) and retinoic acid appears to specify the position of the pancreatic domain in Xenopus, zebrafish and mice (Chen et al., 2004; Martin et al.,2005; Stafford et al.,2004). In the future it will be interesting to determine how the Wnt, FGF, BMP and retinoic acid pathways functionally interact in the endoderm.

Our results are consistent with classical tissue recombination experiments in chick, which have shown that from the definitive-streak stage, the anterior endoderm but not the posterior endoderm is competent to become liver(Fukuda-Taira, 1981; Le Douarin, 1975). Our interpretation is that the posterior endoderm had already received a Wnt signal to repress foregut fate, which is also consistent with explant studies showing that the posterior axial mesoderm emits signals that inhibit liver development (Gualdi et al.,1996; Le Douarin,1975).

Although it remains to be determined whether β-catenin patterns the endoderm in other vertebrates, some genetic studies are consistent with this model. First, TOPGal transgenic mice, which contain a transcriptional reporter of β-catenin/Tcf activity, indicate that the anterior definitive endoderm experiences little if any Wnt signaling(Merrill et al., 2004). In addition, mice lacking both Tcf1 and Tcf4 have hindgut defects and an expansion of the anterior intestinal tract(Gregorieff et al., 2004), and ectopic activation of Wnt signaling in later-stage lung and pancreas development results in ectopic intestinal tissue(Heller et al., 2002; Okubo and Hogan, 2004). The role of individual Wnt ligands or Wnt-antagonists in the endoderm has been difficult to determine in mice because of genetic redundancy. Mutations in Wnt3 or Wnt3A result in posterior truncations(Liu et al., 1999; Yamaguchi et al., 1999),whereas deletion of Dkk1 results in anterior truncations(Mukhopadhyay et al., 2001);however, the endoderm in these mutant embryos has not been described in detail. Sfrp5 mutant mice do not exhibit foregut defects, but the overlapping expression patterns of Dkk1, Sfrp5, Sfrp1, Sfrp2 and Frzb (Sfrp3) suggest that they might act redundantly to repress β-catenin in the anterior endoderm(Finley et al., 2003; Heller et al., 2002; Kemp et al., 2005; Leaf et al., 2006).

The results we describe here, together with previously published reports,indicate that Wnts regulate multiple events in the hepatic lineage. Our previous studies have shown that the maternal Wnt pathway stimulates the initial expression of hhex and Wnt-antagonists in the anterior mesendoderm of the Xenopus blastula(Zorn et al., 1999). At this stage, β-catenin and hhex appear to cooperate to promote the formation of the anterior organizer (Zamparini et al.,2006). Now we show that just hours later, between the gastrula and early somite stages, zygotic β-catenin must be repressed to allow hhex expression and foregut development. Then, by stage 30, the function of β-catenin once again changes and it appears to enhance liver development. This later effect is consistent with recent reports that Wnt2bb is required for liver specification in zebrafish(Ober et al., 2006) and thatβ-catenin signaling promotes hepatoblast proliferation in the mouse and chick liver bud (Hussain et al.,2004; Tan et al.,2006). Finally, the inappropriate activation of β-catenin is linked to liver cancer (Park et al.,2001). Thus, one of the keys to understanding foregut organ development is to determine when a signaling pathway has to be turned on or off in a lineage and to characterize the resulting genetic program initiated in each case.

We thank Monica Lee, Cheng-Hui Hu and Rina Shah for expert technical assistance; Alan Kenny, Gail Deutsch and Jim Wells for discussions and comments on the manuscript; and Drs Horb, Itasaki, Kim, Krieg, Moon, Niehrs and Slack for generously providing reagents. This work was supported by grants from the NIH to V.A.M. (T32DK007727 and F32 HD47121) and A.M.Z. (HD42572).