Crucial role of vHNF1 in vertebrate hepatic specification

During embryogenesis, the primitive gut forms during gastrulation and is gradually patterned along its anterior-posterior axis into the fore-, mid- and hindgut domains. The foregut contains the common endodermal precursors of the liver, pancreas, thyroid and lung. Reciprocal tissue interactions between the endoderm and adjacent mesoderm further subdivide these broad primitive gut domains into smaller subdomains, and allow for the emergence of different endoderm-derived organs at precise locations. The mammalian liver is derived from both the endoderm of the ventral foregut and the adjacent mesenchyme of the septum transversum. In the mouse, its development begins at 8 days post-conception (E8), when the ventral foregut lies immediately adjacent to the cardiac mesoderm. Liver development is thought to occur in sequential steps, beginning with the acquisition of competence to respond to inductive signals, followed by hepatic specification and bud formation. FGFs from the cardiac mesoderm and BMPs from the septum transversum mesenchyme (STM)coordinately induce the underlying endoderm to adopt a hepatic fate(Jung et al., 1999; Rossi et al., 2001). The hepatic endoderm responds to this induction by expressing albumin(Alb), transthyretin (Ttr) and alpha fetoprotein(Afp), and generates a nascent hepatoblast population (bipotential precursors) from which the hepatocytes and biliary cells are derived(Lemaigre and Zaret, 2004). By E9, hepatoblasts proliferate to form the primary liver bud. The basement membrane surrounding the liver bud is then degraded, and hepatoblasts delaminate from the bud and invade the surrounding STM in a cord-like fashion(McLin and Zorn, 2006). Aside from the inductive signals involved in early liver development, little is known about the transcriptional regulatory circuit within the ventral endoderm itself that controls hepatic specification.

Gene-targeting studies have implicated several transcription factor genes in controlling distinct aspects of early hepatogenesis in the endoderm,including Hex (Hhex - Mouse Genome Informatics), Gata6 and Prox1. In Hex-deficient embryos, the hepatic diverticulum is specified but proliferation is reduced and subsequent migration of hepatoblasts into the STM fails to occur(Bort et al., 2004; Martinez Barbera et al.,2000). Similarly, in Gata6^-/- embryos, hepatic development is disrupted at the liver bud stage soon after hepatic specification (Zhao et al.,2005). The homeodomain transcription factor PROX1 acts at later stages. Hepatoblasts differentiate but are unable to delaminate into the adjacent STM; this is most likely owing to their inability to degrade the basement membrane surrounding the hepatic bud(Sosa-Pineda et al., 2000). Thus, although mutations in these transcription factors affect early liver development, in each case defects occur shortly after hepatic induction. Defective hepatic specification has only been observed in a double-mutant mouse model in which both Foxa1 and Foxa2 are deleted at the onset of liver induction (Lee et al.,2005). Interestingly, ventral endoderm explants of these compound mutants are unable to induce hepatogenesis in vitro, thus providing genetic support for the requirement of Foxa factors for hepatic competence. By contrast, constitutive inactivation of Foxa1 or ablation of Foxa2 specifically in the ventral endoderm does not affect normal hepatic specification or differentiation.

Recent studies in mouse and zebrafish indicate that the homeodomain transcription factor vHNF1 (HNF1β, TCF2) plays a crucial role in the early development of several endodermally derived organs(Haumaitre et al., 2005; Sun and Hopkins, 2001). vHnf1-deficient mouse embryos die before gastrulation owing to defective formation of extraembryonic visceral endoderm. This early embryonic lethality can be rescued by providing vHnf1-null embryos with a wild-type (WT) extraembryonic endoderm through the use of tetraploid embryo complementation (Barbacci et al.,1999; Haumaitre et al.,2005). Using this technique, we have previously shown that vHNF1 is required for ventral pancreas induction, as well as for normal dorsal pancreas morphogenesis and regional specification of the gut(Haumaitre et al., 2005). Rescued vHnf1-deficient embryos also exhibit a severe liver hypoplasia, the molecular basis of which remains essentially unexplored. Interestingly, zebrafish vhnf1 hypomorphic mutants also exhibit abnormal gut, liver and pancreas development(Sun and Hopkins, 2001),suggesting that the function of vHNF1 is conserved through vertebrate evolution. In this study, we have examined the role of vHnf1 in early hepatic development in both mouse and zebrafish vHnf1 mutant embryos. Our results uncover the requirement of vHNF1 for liver specification in both organisms. More importantly, using an explant culture system, we have also established that vHnf1-deficient mouse ventral endoderm is unable to respond to FGF signals, which have previously been shown to be sufficient to induce hepatic specification in vitro(Calmont et al., 2006; Serls et al., 2005). Therefore, vHnf1 is required for the initiation of liver development in vertebrates.

Tetraploid (4n) embryos were generated as previously described(Haumaitre et al., 2005). The genotype of 4n embryos was confirmed by PCR analysis of genomic DNA and no WT allele was detected in embryos derived from vHnf1^-/- ES cells. Control embryos were obtained by parallel matings (OF1 females ×B6D2F1 males). We used embryos with identical numbers of somites to ensure that they were at equivalent developmental stages.

Antibody staining was performed as previously described(Haumaitre et al., 2005). We used the following primary antibodies: mouse anti-E-cadherin at 1:100 (BD Transduction Laboratories), rabbit anti-laminin at 1:50 (Sigma), goat anti-HNF4α at 1:150 (C-19, Santa Cruz), rabbit anti-vHNF1 at 1:50 (H-85,Santa Cruz), rabbit anti-NKX2.1 at 1:100 (H-190, Santa Cruz) and rabbit anti-phosphohistone H3 at 1:250 (Upstate Biotechnology). Secondary antibodies were goat anti-mouse IgG Alexa Fluor 546 at 1:500, donkey anti-rabbit IgG Alexa Fluor 488 at 1:500 (Molecular Probes) and biotinylated goat anti-rabbit IgG, horse anti-mouse IgG and anti-goat IgG (Vector) followed by streptavidin-Alexa Fluor 488 at 1:500 (Molecular Probes). TUNEL staining was performed using the Cell Death Detection Kit (Roche). Whole-mount in situ hybridization or on sections was performed as previously described(Haumaitre et al., 2005), with the exception that embryos were fixed with 60% ethanol/11% formaldehyde and 10% acetic acid. The following cRNA probes were used: vHnf1(Haumaitre et al., 2005), Afp, Prox1 (G. Gradwohl, Inserm U682, Strasbourg, France), Foxa2,Foxa1 and Foxa3 (K. H. Kaestner, Department of Genetics,University of Pennsylvania School of Medicine, Philadelphia, PA), Shh(A. P. McMahon, Harvard University, Cambridge, MA), Hex (Bedford et al., 1993), Gata6 and Gata4 (D. B. Wilson, Washington University Medical Center, St. Louis, MO); Irx2, Hnf4a, Gata1 and Hlx probes were generated by PCR.

RNA from dissected ventral endoderm or after its in vitro culture was extracted using the RNeasy Micro Kit (Qiagen) and subjected to semi-quantitative RT-PCR as described(Barbacci et al., 1999). The volume of each cDNA preparation was adjusted to give similar exponential phase PCR signal for Gapdh. Primer sequences are available upon request. For explant culture experiments, ventral endoderm was dissected from 4- to 6-somite embryos and cultured for 48 hours at 37°C in 4 microwells (Nunc)coated with type I collagen (Collaborative Biomedical Products) in 5%CO₂ in DMEM medium supplemented with 10% calf serum, human recombinant bFGF (Invitrogen; 5 ng/ml) and heparan sulfate proteoglycan(Sigma; 50 ng/ml).

Zebrafish (Danio rerio) were raised and genotyped as described(Lecaudey et al., 2007). Whole-mount in situ hybridization was performed using the following probes: hhex and transferrin (M. Pack, The Children's Hospital of Philadelphia, Pennsylvania, PA), gata 4 and gata6 (T. Evans, Albert Einstein College of Medicine, New York, NY), foxa3, ceruloplasmin and prox1 (Y. R. Stainier, University of California, San Fransisco, CA) axial (foxa2) and shh (F. M. Rosa, Inserm U784,Paris, France).

As a first step towards addressing the role of vHNF1 in liver development,we carried out a detailed analysis of its expression in mouse embryos from the first stages of its development. In situ hybridization on 7- to 8-somite pair stage (7-8s) embryos revealed low levels of vHnf1 transcripts in the ventral endoderm adjacent to the developing heart, in a region that delineates the future hepatic and ventral pancreas endoderm(Fig. 1A). At the hepatic bud stage (E9.5; 20s), vHnf1 was strongly expressed in the hepatic primordium (Fig. 1B). At the stage during which the hepatoblasts begin to invade the adjacent STM, vHnf1 transcript levels decreased in these cells, whereas in the gall bladder primordium the levels remained very high(Fig. 1C). At this stage, vHnf1 transcripts were also detected in the gut, the lung buds, the ventral and dorsal pancreas and the posterior stomach(Fig. 1D). By E17.5-18.5, vHnf1 expression increased transiently in parenchymal hepatocytes and sharply decreased after birth [around post-natal day 10 (P10)]; in the adult, vHnf1 expression was restricted to cholangiocytes (data not shown)(Kyrmizi et al., 2006). Thus, vHnf1 expression demarcates the early hepatic and ventral pancreas endoderm, which is consistent with a function of this factor during the earliest stages of development.

We have rescued the early embryonic lethality of vHnf1 homozygous mutants by providing WT extraembryonic endoderm through tetraploid embryo complementation. In these experiments, the embryo proper is derived exclusively from vHnf1^-/- diploid ES cells and the visceral endoderm derives from WT tetraploid embryos(Barbacci et al., 1999). vHnf1^-/- ES cell-derived embryos (henceforth denoted vHnf1^-/-) exhibited apparently normal external characteristics up to E13.5, except for severe liver hypoplasia, evident by external inspection (Fig. 2A). Dissection of mutant livers indicated a 70% reduction in size; yet they presented the same number of lobes as WT livers(Fig. 2B). vHnf1^-/- embryos died around E14.5-15.5, probably because of the inability of the hypoplastic mutant liver to conduct its embryonic hematopoietic function.

Histological analysis at E13.5 confirmed that the vHnf1^-/- liver occupied a very reduced volume in the abdominal cavity. The liver architecture was severely disorganized with ventrally located hemorrhagic regions (Fig. 2D, arrow). Moreover, cells were rounded and lacked hepatoblast characteristics (Fig. 2D,insets). The absence of the cell-adhesion protein E-cadherin (cadherin 1) in vHnf1^-/- livers further suggested a defect in the epithelial characteristics of the hepatoblasts(Fig. 2F), or their absence. The gall bladder primordium, which normally develops in conjunction with the liver, was not discernible at any stage (data not shown).

Analysis of the percentage of cells positive for the mitotic marker phosphohistone H3 (P-H3) showed a 60% decrease in P-H3-positive cells in vHnf1^-/- relative to WT livers between E10.5 and E12.5,and a decrease of 60-70% by E13.5 (not shown). Whereas no increase in apoptosis was detected in mutant livers from E10.5 to E12.5, massive apoptosis was detected at E13.5 (Fig. 2H).

Considering that vHnf1 is selectively expressed in the gut endoderm and hepatoblasts, but not in the STM, we examined the expression of early endoderm differentiation markers, including Alb, Afp, Ttr, Prox1,Foxa1, Foxa2, Foxa3, Hnf1a, Hnf4a and Hnf6 (Onecut1 -Mouse Genome Informatics), at earlier stages, before liver damage was evident. The combined expression of these genes would reflect correctly differentiated hepatoblasts. Surprisingly, none of these markers was expressed at E10.5, nor at any stage (Fig. 3A-L, and data not shown), in the vHnf1 mutant liver.

We then examined the expression of markers of other cells types that compose the embryonic liver, namely the STM, endothelial and hematopoietic cells. In the embryonic liver, the homeobox transcription factor Hlxis expressed exclusively in cells derived from the STM(Hentsch et al., 1996). Once hepatoblasts invade the STM, Hlx expression essentially becomes restricted to the periphery of hepatic lobes (E10.5-11; Fig. 3M). Surprisingly, Hlx expression at E10.5 was observed homogenously throughout the entire lobes of vHnf1^-/- livers(Fig. 3N), along with reduced expression of Gata1 (Fig. 3P), a transcription factor restricted to hematopoietic cells(Palis et al., 1999). Together, these results show that mutant hepatic lobes are devoid of hepatoblasts and are mainly composed of mesenchymal and hematopoietic cells,which enter into apoptosis by E13.5. These observations further suggest a defect in hepatoblast formation and/or migration in vHnf1^-/- embryos.

Whole-mount in situ hybridization at E9.5 (20-22s) showed that Afp, one of the first markers expressed at high levels in the hepatic bud (Fig. 4A), was absent from the vHnf1^-/- mutant presumptive hepatic domain(Fig. 4B). Foxa2, a gene normally expressed in the foregut-midgut endoderm and necessary for its development (Dufort et al.,1998), was specifically absent from the presumptive hepatic endodermal region, while in the gut endoderm its expression was caudally expanded (Fig. 4C,D). In the vHnf1^-/- mutants, the expression of Shh, a foregut marker, was expanded and higher in the foregut region compared with WT embryos (Fig. 4E,F), which was more obvious on sagittal sections (not shown).

In situ hybridization of transversal sections at E9.5 (20-22s) showed that the expression of Hex, one of the earliest factors expressed in prehepatic endoderm and required for hepatic and ventral pancreatic bud development (Bort et al., 2004; Martinez Barbera et al.,2000), was barely detectable in the ventral endoderm of vHnf1^-/- embryos (Fig. 4G,H). Yet, Hex expression was not affected in the thyroid domain (data not shown). At this stage, migration of hepatoblasts is manifested by progressive disruption of the laminin-positive basal membrane that surrounds the liver bud of WT embryos(Fig. 4I). As in Hexmutants, in vHnf1^-/- embryos the formation of a pseudostratified epithelial hepatic bud was not observed. The ventral endoderm remained columnar as the epithelium lining the gut tube and was surrounded by an intact laminin basal membrane (Fig. 4J). In agreement with the lack of endoderm outgrowth and expansion, we observed a 40% decrease in the number of P-H3⁺ cells in the presumptive vHnf1^-/- hepatic endoderm(4.4±1.29%; 17 sections, n=2 embryos) as opposed to WT controls (7.49±1.35%; 16 sections, n=2 embryos) (data not shown). At E9.5, Shh is normally expressed in the lateral gut endoderm in a ventral-dorsal gradient but is excluded from the hepatic bud(Fig. 4K). In Hex^-/- embryos, ectopic expression of Shh in the hepatic bud has been proposed to impair the transition of the endoderm to a pseudostratified epithelium (Bort et al.,2006). Shh was expressed ectopically and at higher levels in vHnf1^-/- than WT embryos in the lateral gut epithelium(Fig. 4, compare E,K with F,L);it was however excluded from the presumptive hepatic bud domain.

Expression of Prox1 was barely detectable in the ventral endoderm(Fig. 4N). By contrast, Gata6 was expressed normally in the gut epithelium and adjacent STM of vHnf1^-/-embryos(Fig. 4Q,R). Gata4 is normally expressed transiently in undifferentiated gut endoderm, which becomes competent to activate liver genes (Bossard and Zaret, 1998), but from E9 (14s) is rapidly downregulated in the prehepatic endoderm while persisting in the gut duodenal region(Bort et al., 2006) (see also Fig. S1 in the supplementary material). Surprisingly, Gata4, which was, as in the WT, induced at 8s and downregulated by 14s in mutant ventral endoderm (see Fig. S1 in the supplementary material), was subsequently ectopically expressed in the entire ventral gut of vHnf1^-/- embryos (Fig. 4P and see S1 in the supplementary material). Interestingly, a few cells coexpressing GATA4 and the pancreatic-duodenal factor PDX1 were observed in a restricted part of the mutant ventral gut (see Fig. S1 in the supplementary material).

Remarkably, the expression of Foxa1 and Foxa2 was strongly decreased and/or absent in the ventral part of the gut and in the presumptive liver bud of vHnf1^-/- embryos(Fig. 4T,V), whereas their expression in the lateral gut epithelium was unaffected. Furthermore, the expression of Foxa3 was completely lost(Fig. 4X), which is in agreement with previous data indicating that vHNF1 is a direct transcriptional activator of this gene in visceral endoderm formation(Barbacci et al., 1999; Hiemisch et al., 1997).

These results indicate that vHNF1 is required in two steps of early liver development: the thickening of the hepatic bud and the expression of essentially all known hepatic genes.

The observation that the expression of Foxa factors is strongly reduced or absent in the presumptive hepatic endoderm of vHnf1^-/-embryos suggested a defective hepatic specification. To directly address whether the specification process was impaired, we determined the expression levels of several hepatic genes in ventral endoderm at the 6-8s stage by semi-quantitative RT-PCR (Fig. 5A). At this stage, the ventral foregut endoderm is instructed by FGF and BMP signals to be specified to the hepatic lineage. As expected, vHnf1 transcripts were detected in ventral endoderm of control embryos, but not in vHnf1^-/- ventral endoderm. Comparable Hex and Foxa2 transcript levels were detected in mutant and WT embryos. By contrast, no transcripts of Alb were detected in mutant ventral endoderm (Fig. 5A). Since the Alb gene encodes the earliest marker expressed in the prehepatic endoderm, our results demonstrate that vHNF1 is required for hepatic specification. The same results were obtained when these analyses were performed at the 8-10s (Fig. 5B) or 10-12s stage (not shown), thereby excluding the possibility of a delay in hepatic specification. Furthermore, transcripts of Ptf1a, a key transcription factor required for the acquisition of a pancreatic fate (Kawaguchi et al.,2002), were not detected in mutant ventral endoderm at the 8-10s stage (Fig. 5B). These results show that ventral pancreas specification does not occur in vHnf1^-/- mutants. This is in agreement with the reported absence in vHnf1^-/- mutants of a ventral pancreas bud at all stages examined (Haumaitre et al.,2005), and further indicates a global defect in ventral endoderm specification.

Impaired hepatic specification in vHnf1 mutants could be due to a defective positioning of the definitive endoderm beyond the hepatogenic influence of the cardiac mesoderm, as described in Hex^-/-mutants (Bort et al., 2004). We therefore compared the expression of vHNF1 and HNF4α proteins, which at 8s are both strongly expressed in the extraembryonic visceral endoderm. We found, however, a similar positioning of this tissue relative to the ventral foregut endoderm and cardiac mesoderm in vHnf1^-/- and WT embryos (Fig. 5, compare C with D and E with F).

Another possible explanation for defective hepatic specification is a switch in the identity of the ventral endoderm to another, more-anterior fate of the foregut, as might be expected from the posterior expansion of Foxa2 (Fig. 4D) and Shh (Haumaitre et al.,2005) expression in the gut of vHnf1^-/-embryos. We analyzed at 8s the expression of several factors typically expressed in different ventral foregut regions, including Nkx2.1,which is expressed in the thyroid and lung domain(Deutsch et al., 2001), Hex, which is expressed in the thyroid and hepatic endoderm domain(Martinez Barbera et al.,2000), and Irx2, which is expressed in the lungs and tracheapharyngeal domain (Becker et al.,2001). No differences in the expression pattern of these markers were observed between mutant and WT embryos, indicating that at this stage, vHnf1^-/- ventral endoderm is apparently correctly regionalized (Fig. 5G-L). We also observed correct expression of Hex in the presumptive hepatic domain of mutant ventral endoderm. However, and in contrast to the similar levels of Foxa1/2 transcripts observed by RT-PCR analysis of control and mutant ventral endoderm, the expression of both Foxa1 and Foxa2 appeared significantly reduced specifically in the mutant ventral endoderm region (Fig. 5N,P). One possible explanation for this discrepancy is that during ventral endoderm microdissection, some foregut-adjacent tissue,expressing normal levels of Foxa1/2 genes, was inadvertently included and therefore biased our quantifications. More importantly, a further decrease in Foxa1/2 expression was observed in 14s stage mutant ventral endoderm (see Fig. S2 in the supplementary material). Moreover, the induction of Foxa3 expression was severely perturbed as indicated by rare Foxa3-positive cells distributed throughout the vHnf1^-/- ventral endoderm(Fig. 5R). These data together clearly indicate a requirement of vHNF1 for maintaining the expression of Foxa1 and Foxa2 specifically in the ventral endoderm, as well as for Foxa3 induction.

We then addressed the competence of vHnf1^-/- endoderm to be specified into a hepatic fate. Ventral foregut endoderm was dissected from embryos at the 2-6s stage and cultured for 48 hours in the presence of bFGF (FGF2) and heparan sulfate proteoglycan, under conditions reported to be sufficient to induce hepatic specification in the absence of cardiac mesoderm(Calmont et al., 2006; Deutsch et al., 2001; Lee et al., 2005). Our experiments show that, in contrast to WT embryos, vHnf1^-/-ventral endoderm (n=3 embryos) was not competent to induce the expression of Alb or Ttr(Fig. 5S). These data suggest that the failure of hepatic specification in the vHnf1^-/-embryo is, at least in part, owing to a loss of competence to respond to FGF inductive signaling.

In zebrafish, the signaling molecules and the tissue interactions involved in early liver induction are less well documented than in chick or mouse. Hepatoblast specification in zebrafish is thought to occur at approximately 22 hours post-fertilization (hpf) as marked by the localized expression of hhex and prox1, which are among the earliest markers of liver development (Ober et al.,2003). Between 24 and 28 hpf, two thickened regions emerge from the intestinal rod, the anterior of which corresponds to the liver bud and the posterior to the pancreas bud (Field et al., 2003). Recent studies point to a conserved molecular program orchestrating liver formation in this organism. As in amniotes, FGF and BMP signaling pathways have been shown to be required for liver induction(Shin et al., 2007). Moreover,a role for Wnt signaling from the lateral plate mesoderm in zebrafish liver induction has been recently described(Ober et al., 2006). Additionally, in contrast to in amniotes, Hedgehog signaling appears to regulate the proliferation of already specified liver progenitors(Wallace and Pack, 2003). It has also been reported that hhex is required for liver development and normal gut looping in zebrafish(Wallace et al., 2001),whereas gata4 and gata6 play an important, yet redundant,role in liver expansion and differentiation(Holtzinger and Evans,2005).

Several vhnf1 zebrafish mutants presenting abnormal liver and pancreas development, in addition to defective hindbrain segmentation, have been isolated (Song et al.,2007; Sun and Hopkins,2001). These mutants present a wide range of hepatic phenotypes from reduced liver and abnormal biliary development(Matthews et al., 2004; Sun and Hopkins, 2001) to the lack of a discernible liver at 3 dpf (Song et al., 2007). These observations prompted to us to examine in this organism the role of vhnf1 at the stage of hepatic specification. For this analysis we chose the vhnf1^hi2169 mutants because they are highly hypomorphic leading to null alleles (Bagnat et al., 2007; Sun and Hopkins,2001).

Although the general expression pattern of vhnf1 in zebrafish embryos has been reported previously (Song et al., 2007; Sun and Hopkins,2001), its early expression in the endoderm before and during hepatic budding is less well documented. As shown in Fig. 6A,B, at both 18s and 22-24s stages, vhnf1 was strongly expressed in the pronephric duct,while weakly expressed in the gut endoderm. By 48 hpf, vhnf1 was expressed in the gut and in the liver and pancreas anlagen(Fig. 6S), displaying an endoderm expression pattern similar to that of foxa3(Fig. 6O)(Field et al., 2003).

In situ hybridization analysis at the time of liver induction showed that the expression of both hhex and prox1, which at 24 hpf are normally expressed in the nascent liver and pancreas buds(Fig. 6C,E), was completely abrogated in these structures in the vhnf1^hi2169 mutant(Fig. 6D,F). Unlike our observations in vHnf1^-/- mouse embryos(Fig. 4), in zebrafish vhnf1^hi2169 mutants expression of gata6 was lost from the entire endoderm (Fig. 6H). By contrast, gata4 expression at 27 hpf, or foxa3 at 30 hpf, was maintained in the gut endoderm, but was not detected in the hepatic or pancreatic domain(Fig. 6J,L). Similar results were obtained at 30 and 48 hpf indicating that vhnf1 mutants exhibit a profound defect in liver induction rather than a delay in its formation(Fig. 6M-P). As expected,expression of the specific liver marker ceruloplasmin (cp)was absent in mutant embryos (Fig. 6R). As vhnf1 transcripts are produced from the vhnf1^hi2169 mutant allele, we performed an in situ hybridization on 48 hpf mutant embryos using a full-length vhnf1probe and detected the presence of mutant transcripts only in the gut endoderm(Fig. 6T). This observation further confirms the complete absence of both the liver and pancreas buds. It also excludes the possibility that hepatoblasts or pancreatic cells, lacking the expression of key regulatory factors, were still formed in the absence of vhnf1 function. Interestingly, the mutant gut endoderm did not exhibit the typical leftward bending observed in WT embryos(Fig. 6, compare S with T), and was abnormally regionalized as manifested by the caudal expansion of foxa2 and shh expression (see Fig. S3 in the supplementary material) (Sun and Hopkins,2001), indicating that vhnf1 is also required for gut morphogenesis. Together, these data strongly suggest that vhnf1^hi2169 mutant embryos exhibit defective hepatic specification, whereas the gut endoderm integrity appears to be maintained.

Our analysis of the function of the transcription factor vHNF1 reveals its essential role in the initiation of liver development in vertebrates. We show that vHNF1 is required for hepatic specification of the mouse ventral endoderm and for subsequent bud formation. A consequence of this defect is the absence of hepatoblast migration into the adjacent STM. Despite this, mutant liver lobes are correctly formed, although highly hypoplastic. In Prox1mutant mouse embryos, liver lobes are also correctly shaped despite a highly reduced number of hepatoblasts(Sosa-Pineda et al., 2000). Unlike Prox1 mutants, in E13.5 vHnf1^-/- embryos the liver architecture is severely disrupted and followed by massive apoptosis. Since vHnf1 is not expressed in the STM, these observations imply that signals from hepatoblasts play a crucial role in the maintenance and generation of the hepatic architecture required for normal liver function.

Analysis at the time of liver induction indicates that vHNF1 is required for the acquisition of competence of the ventral endoderm to respond to inductive signals and for hepatic specification. Remarkably, the liver induction defect exhibited by vHnf1^-/- embryos is basically identical to that reported in compound Foxa1/Foxa2 mutants. In both cases, the ventral endoderm fails to acquire a hepatic fate when cultured in vitro under conditions reported to be sufficient to induce its specification. The acquisition of hepatic competence is disrupted even though Foxa1 and Foxa2 transcripts are present, albeit at lower levels, in vHnf1^-/- ventral endoderm. Thus, vHNF1, like Foxa factors, mediates regional competence of the ventral endoderm to respond to FGF signaling. Because of their ability to open highly condensed chromatin structure upon binding, it has been proposed that FOXA and GATA factors behave as `pioneer factors' that mark their target genes as competent to be expressed when exposed to the appropriate inductive signals(Cirillo et al., 2002). Given the structural properties of the protein it is unlikely that vHNF1 mediates competence by relieving chromatin condensation. One explanation that we explored was whether vHNF1 is a direct target of the FGF signaling pathway. Prior studies in zebrafish supported this hypothesis, as injected vhnf1 mRNA can rescue the abnormal pancreatic and liver phenotype caused by the disruption of either FGF or BMP signaling pathways(Song et al., 2007),indicating that vhnf1 is downstream of these pathways. Moreover,recent studies in mice have shown that hepatic gene induction is elicited by an FGF/MAPK pathway, whereas the FGF/PI3K pathway is required for growth and morphogenesis of the hepatic bud (Calmont et al., 2006). In vHnf1^-/- mouse embryos, both processes are affected as hepatic specification is disrupted and the liver bud fails to form. However, when ventral endoderm was cultured under different conditions, we found that vHnf1 expression is influenced neither by FGF signaling nor by pharmacological inhibitors of FGF signaling, indicating that vHNF1 is not a direct target of FGF signals. Moreover, Fgfr2,sprouty 2, Dusp6 (Mkp3) and Socs3 genes were induced at similar levels in vHnf1^-/- and WT ventral endoderm (see Fig. S4 in the supplementary material), suggesting that the hepatic competence defect of our mutants is not due to a global disruption of FGF signaling. Based on these observations, we propose that vHNF1 could mediate competence of the endoderm to respond to FGF signaling by controlling the expression of a downstream target of this pathway. An alternative,non-mutually exclusive possibility is that the reduced expression of Foxa1 and Foxa2, along with the absence of Foxa3,lead to a level of Foxa proteins that is below the threshold required for relieving chromatin condensation and subsequent acquisition of the hepatic competence. Considering the broad expression of both vHnf1 and Foxa1/2 genes within the gut endoderm, it is difficult to explain how these factors, acting either in synergy or in a linear cascade, could mediate hepatic specification in a restricted domain of the ventral endoderm. It remains possible that local signals induce a hepatic specification factor(s)or co-factor(s) that either cooperates with vHNF1 or is an obligatory partner of this gene.

Following the competence step, vHNF1 appears to function through sequential and complementary mechanisms. One of them concerns the proper regional specification of the gut and the subsequent accurate acquisition of cell fates. Abnormal regional gut specification in vHnf1-deficient mutants at E9.0 is manifested by the caudal expansion of Foxa2 and Shh expression, along with the ectopic expression of Gata4in the presumptive hepatic domain, which otherwise essentially lacks all early hepatic markers examined. One interesting possibility is that ectopic expression of Gata4 reflects a change in the identity of the ventral endoderm into a duodenal-like fate. In agreement with the acquisition of a duodenal fate is the observation of sparse cells coexpressing PDX1 and GATA4 in the presumptive hepatic endoderm at E9. Since a ventral pancreas bud is not formed at any stage in vHnf1 mutant embryos(Haumaitre et al., 2005),these PDX1-positive cells might represent duodenal precursors. A similar conversion of ventral pancreas progenitor cells into duodenal cells has been observed in mouse Ptf1a mutants(Kawaguchi et al., 2002). Interestingly, Ptf1a is not induced in vHnf1^-/-ventral endoderm at any developmental stage(Fig. 5)(Haumaitre et al., 2005). Such cell-fate conversion would account, at least in part, for the abrogated morphogenesis and the lack of liver bud outgrowth.

Unlike the strongly downregulated expression of early hepatic genes at E9.0, at the 8-10s stage, Foxa1, Foxa2 and Hex are expressed, albeit at different levels, in the presumptive hepatic domain of vHnf1 mutants. Thus, vHNF1 is not required for the initial induction of these genes but rather for maintenance of their expression. These observations further suggest that vHNF1 and FOXA1 and FOXA2 factors do not act in a simple linear cascade that leads to the activation of liver-specific genes, but rather in synergistic pathways(Fig. 7) to activate these genes. This phase of induction of Foxa genes is shortly followed by a maintenance phase in which vHNF1 is required both for their sustained expression and for subsequent induction of other hepatic transcription factors, thus establishing a complex cross-regulatory network involved in the determination of the hepatic fate(Cereghini, 1996). Consistent with this model (Fig. 7),functional HNF1 (HNF1A - Mouse Genome Informatics) binding sites in the regulatory sequences of both Foxa2(Kyrmizi et al., 2006) and Foxa3 (Hiemisch et al.,1997) have been identified. In addition, in silico analysis has indicated the presence of potential HNF1 binding sites within the upstream sequences of the Foxa1 gene.

More importantly, a comprehensive analysis of promoter occupancy during mouse hepatic development from E14 has recently shown that vHNF1, but not the structurally related protein HNF1, is recruited to the promoter regions of a`core' group of transcription factors including FOXA2, HNF1, HNF4 (HNF4A),HNF6 and GATA6, which compose autoregulatory and cross-regulatory circuits(Kyrmizi et al., 2006). Interestingly, the complexity and stability of this network increase gradually during organogenesis (Kyrmizi et al.,2006).

As in mice, in zebrafish vhnf1 mutant embryos neither the liver nor the pancreas bud is formed and the gut endoderm is abnormally regionalized(see Fig. S3 in the supplementary material)(Sun and Hopkins, 2001). Despite the fact that in these two organisms the positioning of the gut endoderm relative to the adjacent tissues and the inductive signals are not the same (Ober et al., 2006),our analyses in zebrafish confirm an evolutionary conserved role of vHnf1 in hepatic specification and development. Interestingly, the later roles of vHNF1 in intrahepatic biliary morphogenesis are also conserved in zebrafish (Matthews et al.,2004). Therefore, vHNF1 represents the earliest transcription factor involved in zebrafish hepatic specification.

Together, these studies support the notion that vHNF1 acts at two levels during vertebrate organogenesis. First, in the early acquisition of a hepatic and pancreatic fate from the multipotent ventral endoderm. Second, in the normal epithelial morphogenesis of tubular structures [i.e. biliary duct(Coffinier et al., 2002),stomach epithelium (Haumaitre et al.,2005) and gut lumen in zebrafish(Bagnat et al., 2007)]. Interestingly, these dual functions of vHNF1 correlate with its dynamic embryonic expression pattern, being initially high throughout the entire liver and pancreas buds and subsequently restricted to the branched pancreatic ductal network and the biliary system.

The strategies used to differentiate embryonic stem cells into endodermal cells are based on the knowledge of the conserved molecular network that controls endoderm formation in different vertebrate embryos(D'Amour et al., 2005). From a therapeutic point of view, it is important to now precisely define how the endoderm is patterned and how particular organs are induced. Our studies provide insights into the early molecular events of liver specification in vertebrates and may contribute to the development of in vitro strategies for the generation of hepatic cells for regenerative medicine, either from embryonic stem (Gouon-Evans et al.,2006) or somatic stem cells.

We thank K. H. Kaestner (University of Pennsylvania, Philadelphia, PA), A. P. McMahon (Harvard University, Cambridge, MA), M. Pack (The Children's Hospital of Philadelphia, Pennsylvania, PA), T. Evans (Albert Einstein College of Medicine, New York, NY), Y. R. Stainier (University of California, San Francisco, CA), F. M. Rosa (INSERM U784, Paris) for reagents and C. Lesaulnier and C. Vesque for comments on the manuscript. This work was supported by Ligue Contre le Cancer, Association pour la Recherche sur le Cancer (ARC, Contrats 3231 and 3911), Institut National de la Santé et de la Recherche Médicale, CNRS, Agence National de la Recherche (ANR) and Université Pierre et Marie Curie. L.L. is a recipient of a PhD studentship from Ministère de la Recherche et de la Technologie and ARC.