Inhibition of mesodermal fate by Xenopus HNF3β/FoxA2

Abstract

The winged helix transcription factor HNF3β/FoxA2 is expressed in embryonic organizing centers of the gastrulating mouse, frog, fish, and chick. In the mouse, HNF3β is required for the formation of the mammalian node and notochord, and can induce ectopic floor plate formation when misexpressed in the developing neural tube; HNF3β expression in the extraembryonic endoderm is also necessary for the proper morphogenesis of the mammalian primitive streak. In the frog Xenopus laevis, several lines of evidence suggest that the related winged helix factor Pintallavis functions as the ortholog of mammalian HNF3β in both axial mesoderm and neurectoderm; the role of Xenopus HNF3β itself, however, has not been clearly defined, and is the subject of this study. HNF3β is widely expressed in the vegetal pole but, as previously suggested, is excluded from the gastrula-stage mesoderm. We find that expression of an HNF3β-Engrailed repressor fusion protein induces ectopic axes and inhibits head formation in Xenopus embryos, while ectopic HNF3β inhibits mesoderm and anterior endoderm formation in explant assays and in vivo. Our studies suggest that HNF3β target genes function to limit the extent of mesoderm formation in the Xenopus gastrula, and point to related roles for Xenopus HNF3β and the extraembryonic component of mammalian HNF3β during vertebrate gastrulation.

INTRODUCTION

All tissues in the animal derive from the three primary germ layers. The mesodermal germ layer plays a critical role in organizing the embryonic body axes, and gives rise to the skeletal, muscular, and circulatory systems. In amphibia, perhaps the most widely utilized animals for the study of vertebrate germ layer formation, mesoderm is an inducible cell fate: in co-culture assays, factors secreted from the vegetal pole endoderm induce mesoderm in overlying ectodermal “animal cap” explants. These observations are consistent with the amphibian fate map: mesoderm forms in the “marginal zone,” at the border between the animal and vegetal poles (Dale and Slack, 1987). The primary mesoderm-inducing signal secreted by the vegetal pole is thought to be a TGFβ ligand, the zygotic expression of which is initiated by the transcription factor VegT in cells of the vegetal pole and marginal zone (Clements et al., 1999; Harland and Gerhart, 1997; Kofron et al., 1999; Slack, 1994).

Recent studies in the mouse suggest that inhibition of mesoderm-inducing signals is also critical in establishing the early vertebrate body plan. The TGFβ ligand Nodal plays a central role in mesoderm induction in the mouse, pointing to a conservation between formation of this germ layer in lower and higher vertebrates (Beddington and Robertson, 1999; Hill, 2001; Isaacs, 1997; Whitman, 2001). In an apparent departure from mechanisms employed in Xenopus, however, the mesoderm-inducing signal in mouse is initially very widespread: while expression of the known nodal-related factors in Xenopus are excluded from the animal pole ectoderm, mouse nodal is transiently expressed throughout the epiblast (Whitman, 2001). Recent studies have pointed to multiple mechanisms for limiting this widespread, early domain of Nodal activity to the primitive streak, a posterior structure that is roughly analogous to the Xenopus marginal zone, by the onset of gastrulation. A primary source of Nodal antagonism in the mouse appears to be the anterior visceral endoderm (AVE), an extraembryonic organizing center that overlies the anterior epiblast (Beddington and Robertson, 1999).

The winged-helix transcription factor Hepatocyte Nuclear Factor 3β (HNF3β/FoxA2) is expressed in embryonic organizing centers of the gastrulating mouse, frog, fish, and chick (Carlsson and Mahlapuu, 2002). At the initiation of mouse gastrulation, HNF3β is expressed in cells at the leading edge of the primitive streak; expression is subsequently seen in the definitive node, the floor plate of the neural tube, and in the cells of the definitive endoderm. HNF3β expression is also found throughout the extraembryonic visceral endoderm. Targeted deletion studies have demonstrated distinct requirements for the embryonic and extraembryonic components of this factor in early mammalian development. HNF3β expression in the node and its axial, mesendodermal derivatives are required for the formation of the notochord, a transient, dorsal mesodermal structure; embryonic HNF3β expression is also required for the dorsoventral patterning of the neural tube, and for the development of both the foregut and midgut (Ang and Rossant, 1994; Dufort et al., 1998; Weinstein et al., 1994). Loss of HNF3β in the visceral endoderm, on the other hand, disrupts normal morphogenesis of the primitive streak (Dufort et al., 1998). These studies suggest a cell-autonomous requirement for HNF3β in the axial mesoderm, and a non-cell autonomous role for this factor in primitive streak formation.

In Xenopus, the related winged-helix gene pintallavis, but not HNF3β, is expressed in the axial mesoderm; misexpression studies suggest that Pintallavis functions as the early Xenopus equivalent of the embryonic component of murine HNF3β (O’Reilly et al., 1995; Ruiz i Altaba and Jessell, 1992; Ruiz i Altaba et al., 1993; Sasaki and Hogan, 1994). The functional equivalent of the extraembryonic component of murine HNF3β in Xenopus has not been defined, nor has the precise region of the Xenopus gastrula embryo that corresponds to the visceral endoderm been identified. Conservation of expression domains and, in some cases, function between frog and mouse suggest, however, that the Xenopus deep anterior endoderm functions, like the AVE, as a mesoderm-antagonizing and anterior-determining center (Bouwmeester et al., 1996; Brickman et al., 2000; Jones et al., 1999).

Previous reports have described low levels of Xenopus HNF3β expression during gastrula stages (Ruiz i Altaba and Jessell, 1992). In an effort to better characterize the role of HNF3β during Xenopus gastrulation, we re-examined the expression pattern of this gene using a sensitive reverse transcription polymerase chain reaction (RT-PCR) assay. We have found that Xenopus HNF3β is expressed in both the anterior and posterior endoderm at early gastrula stages, suggesting that Xenopus HNF3β is distributed in a pattern similar to the extraembryonic component of murine HNF3β. We then examined the effects of HNF3β target gene activation and repression in Xenopus embryos, using a variety of HNF3β wild-type and chimeric fusion constructs. We find that activation of HNF3β target genes inhibits mesoderm formation; conversely, repression of HNF3β targets inhibits head formation and induces mesoderm formation and axial duplication. Our findings suggest that Xenopus HNF3β functions to inhibit ectopic mesoderm formation, a role that shares some similarity with the function of HNF3β in the murine visceral endoderm.

MATERIALS AND METHODS

RNA preparation, explant dissection, and cell culture

RNA was synthesized in vitro in the presence of cap analog using the mMessage mMachine kit (Ambion). Microinjection, explant dissection, cell dissociation, and culture were performed as described (Hemmati-Brivanlou and Melton, 1994; Wilson and Hemmati-Brivanlou, 1995).

Preparation of HNF3β constructs

Xenopus HNF3β was isolated by PCR from a gastrula stage cDNA library (Ruiz i Altaba et al., 1993; Weinstein et al., 1998); HNF3β fusion constructs were generated by PCR. For EnR-HNF3β, residues 1–298 of the Drosophila Engrailed repressor were fused upstream of Xenopus HNF3β residues 52–343 (Kessler, 1997; Ruiz i Altaba et al., 1993). For VP-HDNAB, residues 410–490 of the VP16 activator were fused upstream of Xenopus HNF3β residues 132–257 (Kessler, 1997; Ruiz i Altaba et al., 1993). HNF3β residues in VP-HDNAB encode only the winged-helix DNA-binding domain (conserved region I); EnR-HNF3β retains the HNF3β DNA-binding domain and transactivation domain V, and lacks the transactivation domains II, III, and IV (Pani et al., 1992; Qian and Costa, 1995). As domain V requires the presence of domain IV for transactivation in vitro, EnR-HNF3β and VP-HDNAB are both predicted to lack endogenous transactivation function (Qian and Costa, 1995).

For the HNF3β DNA-binding domain mutants HNF3β-MT and EnR-HNF3β-MT, residue 195 of Xenopus HNF3β was mutated from Asparagine to Aspartic Acid (AAC→GAC), residue 199 was mutated from Histidine to Cysteine (CAT→TGT), and residue 241 was mutated from Arginine to Glutamic acid (CGA→GAA). Mutagenesis was performed using the QuikChange Multi Site-Directed Mutagenesis Kit (Stratagene).

Gel Mobility Shift Assays

HNF3β, HNF3β-MT, EnR-HNF3β and EnR-HNF3β-MT proteins were generated by in vitro translation using a rabbit reticulocyte lysate system (Promega). Mobility shift assays were performed using the Gel Shift Assay System (Promega) with a modified binding buffer supplemented with 0.01 mg/ml Poly-L-Lysine (Sigma) and 0.2 mg/ml Herring Testes DNA (Clontech). For the radioactive DNA probe, −109 to −70 bp of the mouse Transthyretin (TTR) promoter (5′-TGA CTA AGT CAA TAA TCA GAA TCA GCA GGT TTG GAG TCA G-3′, 3′-CTG ACT CCA AAC CTG CTG ATT CTG ATT ATT GAC TTA GTC A-5′) (Costa et al., 1989) was end-labeled with ³²P according to the manufacture’s recommendations. 50-fold excess of unlabeled TTR DNA was used as a specific competitor. A DNA fragment derived from the AP2 gene and supplied by the kit manufacturer, was used as a non-specific competitor: 5′-GAT CGT ACT GAC CGC CCG CGG CCC T-3′, 3′-CTA GCA TGA CTG GCG GGC GCC GGG A-5′. DNA-protein complexes were separated by electrophoresis on nondenaturing polyacrylamide gels (Novex 6% retardation gel, Invitrogen).

RT-PCR

RT-PCR was performed as described (Wilson and Hemmati-Brivanlou, 1995). Primers used in this study are as follows: cerberus (forward, 5′-GCTGAACTATTTGATTCCACC; reverse, 5′-ATGGCTTGTATTCTGTGGGGC); chordin (forward, 5′-CAGTCAGATGGAGCAGGATC; reverse, 5′-AGTCCCATTGCCCGAGTTGC); collagen type II (forward, 5′-GGATTCAAGGACTCTAGTGC; reverse, 5′-GATCTCAGCATTGGGGCAAT; dkk (forward, 5′-ACAAGTACCAACCTCTGGATGC; reverse, 5′-ACAGGGACACAAATTCCGTTGC); EF1-α (forward, 5′-CAGATTGGTGCTGGATATGC; reverse, 5′-ACTGCCTTGATGACTCCTAG); globin (forward, 5′-TTGCTCTGTGGGGCAAAATC; reverse, 5′-GGTTGTAGGCATGCAGGTCA); goosecoid (forward, 5′-AGAGTTCATCTCAGAGAG; reverse, 5′-TCTTATTCCAGAGGAACC); HNF3β (forward, 5′-CTTGAAAGCCTACGAACAGG; reverse, 5′-ATAAACAGAGCCCAGGTGAC); muscle actin (forward, 5′-GCTGACAGAATGCAGAAG; reverse, 5′-TTGCTTGGAGGAGTGTGT); ODC (forward, 5′-AATGGATTTCAGAGACCA; reverse, 5′-CCAAGGCTAAAGTTGCAG); otx2 (forward, 5′-CGGGATGGATTTGTTGCA; reverse, 5′-TTGAACCAGACCTGGACT); sox17β (forward, 5′-GTCATGGTAGGAGAGAAC; reverse, 5′-TCTGTTTAGCCATCACTGG); vent1 (forward, 5′-AGGCAACAGATGGAAAGGAC; reverse, 5′-CGCTGGAGATCAGATTTGG); Xbrachyury (forward, 5′-GGATCGTTATCACCTCTG; reverse, 5′-GTGTAGTCTGTAGCAGCA); Xhex (forward, 5′-TCAAGCAGGAGAACCCAC; reverse, 5′-TCGCCTTCAATGTCCACC); Xwnt8 (forward, 5′-GTTCAAGCATTACCCCGGAT; reverse, 5′-CTCCTCAATTCCATTCTGCG).

Whole-mount immunohistochemistry and β-gal detection

Whole-mount antibody staining was performed as described (Hemmati-Brivanlou and Melton, 1994). The 12/101 antibody (ascites, Developmental Studies Hybridoma Bank) was used at a 1:1 dilution. Secondary antibody was a donkey anti-mouse IgG coupled to horseradish peroxidase (Jackson Laboratories), and was used at 1:1000 dilution. Color reactions were performed using the Vector SG kit (Vector Laboratories). Whole-mount β-gal detection was performed as described (Smith and Harland, 1991).

RESULTS

Xenopus HNF3β is expressed in deep endodermal cells during gastrulation

Previous whole-mount in situ hybridization studies demonstrated that Xenopus HNF3β is weakly expressed during early gastrulation in cells that appear to correspond to the suprablastoporal endoderm, the epithelial layer of Spemann’s organizer fated to give rise to the pharyngeal endoderm (Ruiz i Altaba et al., 1993). Given the very low levels detected during gastrulation, we decided to re-examine expression of this gene using a more sensitive RT-PCR assay. As described previously, HNF3β transcripts are first detected after the midblastula transition (stage 9), indicating that this gene is expressed zygotically, but not maternally (Fig. 1A) (Ruiz i Altaba et al., 1993). The initial expression of HNF3β is quite low: 30 cycles of PCR were required to detect a signal comparable to that achieved with 21 cycles for the dorsal marker chordin (Sasai et al., 1994). We next explanted dorsal regions of stage 10+ gastrulae and, immediately after dissection, processed these tissues for RT-PCR analysis; these dissections are shown schematically in Fig. 1B. “meso” explants express both chordin and the pan-mesodermal marker Xbrachyury (Xbra) (Fig. 1B; duplicate experiments are shown) (Sasai et al., 1994; Smith et al., 1991). This explant, which contains both deep and superficial (SBE) layers of the organizer, was positive for HNF3β in one out of six experimental trials (data not shown), suggesting that the expression of HNF3β in the SBE, described earlier, is extremely transient (Ruiz i Altaba et al., 1993). Similar transient expression of the Xhex gene in the SBE has been reported previously (Jones et al., 1999) (Fig. 1B). In contrast, we observed persistent expression of HNF3β in “endo” explants, which contain both endomesodermal cells expressing chordin and deep, anterior endoderm cells expressing Xhex (Newman et al., 1997; Sasai et al., 1994; Zorn et al., 1999); this region may function as the equivalent of the mouse anterior visceral endoderm (AVE) (Bouwmeester et al., 1996; Brickman et al., 2000; Jones et al., 1999). Thus, HNF3β is excluded from the mesodermal component of the early gastrula organizer.

FIG. 1.

RT-PCR analysis of HNF3β expression during early Xenopus development. (A) Temporal expression of Xenopus HNF3β. ornithine decarboxylase (ODC), expressed uniformly throughout early embryogenesis, is used as a loading control (Bassez et al., 1990). chordin expression, like that of HNF3β, is initiated zygotically. (B, C, D) Spatial localization of HNF3β transcripts during early gastrula stages. Schematic diagrams of stage 10+ embryos, showing regions isolated for expression analysis, are shown above gels. (B) HNF3β transcripts are excluded from gastrula-stage dorsal mesoderm. meso: epithelial and deep dorsal mesoderm; endo: dorsal endomesoderm. (C) HNF3β is expressed throughout the endoderm during early gastrula stages. AE: anterior endoderm; PE: posterior endoderm. (D) HNF3β transcripts are excluded from the ectoderm and ventral mesoderm during early gastrula stages. AnC: animal cap; PC: posterior cortex; PD: posterior deep. EF1-α is used as a loading control (Krieg et al., 1989). The –RT lane contains all reagents except reverse transcriptase, and is used as a negative control.

We next asked whether HNF3β was also expressed in ventral/posterior endoderm during gastrulation by performing the microdissections shown schematically in Figs. 1C and 1D. As expected, anterior endoderm (AE) and posterior endoderm (PE) explants express the pan-endodermal marker sox17β (Fig. 1C) (Hudson et al., 1997); chordin, Xhex, and the multifunctional antagonist cerberus are all expressed in the AE and excluded from the PE, as described (Bouwmeester et al., 1996; Newman et al., 1997; Sasai et al., 1994; Zorn et al., 1999). Strikingly, HNF3β is expressed in both AE and PE explants. Posterior expression is excluded from the embryonic cortex (posterior cortex, PC), and co-localizes with the expression of Sox17β (posterior deep, PD; Fig. 1D). Finally, HNF3β expression is not observed in gastrula stage ectoderm (animal cap, AnC; Fig. 1D). These data indicate that the deep endoderm is a primary site of HNF3β expression during early Xenopus gastrulation; this expression pattern is reminiscent of the extraembryonic domain of mouse HNF3β, expressed throughout the visceral endoderm.

Ectopic HNF3β suppresses mesoderm formation

The mouse anterior visceral endoderm is thought to be a source of signals that locally antagonize primitive streak formation, restricting mesoderm formation to the posterior side of the gastrula embryo (Beddington and Robertson, 1999; Brickman et al., 2000). The presence of HNF3β transcripts throughout the deep endoderm prompted us to examine the effects of HNF3β misexpression on mesoderm formation. Ectopic expression of HNF3β in the dorsal marginal zone dramatically effects subsequent embryonic development (Fig. 2A). 70% of injected embryos showed severe reductions in both anterior and posterior structures, and defects in gastrulation (Fig. 2B; n=27); affected embryos showed a marked reduction or complete loss of staining with a somite-specific antibody (Kintner and Brockes, 1984). Furthermore, embryos injected with both β-galactosidase and low doses of HNF3β showed a reduction or absence of somitic mesoderm in cells expressing β-gal (Fig. 2B). These results suggest that marginal zone expression of HNF3β inhibits dorsal mesoderm formation in Xenopus embryos.

FIG. 2.

Effects of ectopic expression of HNF3β. (A, B) HNF3β disrupts gastrulation and inhibits mesoderm formation in intact embryos. (A) 8-cell stage embryos were injected dorsally with 2 ng HNF3β RNA and cultured until stage 32; top embryo is an uninjected control. (B) 8-cell embryos were injected with both 100 pg HNF3β and 200 pgβ-Gal RNA; top embryo is an uninjected control. Embryos were probed with the somite-specific antibody 12/101 (Kintner and Brockes, 1984). (C) Gel shift analysis of HNF3β and HNF3β-MT; in vitro translation products are shown at left. TTR, specific competitor; AP2, non-specific competitior; control, no RNA included in in vitro translation. (D) HNF3β-MT does not affect dorsal development. 8-cell stage embryos were injected dorsally with 250 pg of either HNF3β RNA (top) or HNF3β-MT RNA (bottom). (E) HNF3β is a more potent inhibitor of dorsal development than Pintallavis. 8-cell stage embryos were injected dorsally with 250 pg of either HNF3β RNA (left) or pintallavis RNA (right). All embryos are lateral views, anterior is to left in (A), (B); anterior is to right in (D), (E).

To verify that HNF3β function is dependent on DNA binding, we constructed and tested the activity of HNF3β-MT, in which we mutated three conserved residues (N195, H199, R241) that directly contact DNA and are thought to be critical for the activity of the winged-helix proteins (Clark et al., 1993; Liu et al., 2002). Unlike the wild-type protein, HNF3β-MT does not interact with a HNF3β binding site in mobility shift assays (Fig. 2C). Consistently, dorsal injection of HNF3β-MT RNA, has little effect on early embryonic development (90% wild-type, n=30), suggesting that the effects of exogenous HNF3β are mediated primarily through the DNA-binding activity of the protein.

Several forkhead genes are expressed during early Xenopus development, including pintallavis, thought to be the partial functional equivalent of mammalian HNF3β (Ruiz i Altaba and Jessell, 1992). Dorsal injection of pintallavis RNA generates a relatively modest effect on anterior structures, including a loss of eyes (27% affected, n=44; Fig. 2E); equivalent concentrations of HNF3β were sufficient to disrupt gastrulation (Fig. 2E). Similar effects of ectopic pintallavis have been reported previously (Ruiz i Altaba and Jessell, 1992). Our data thus suggest some overlap in the effects seen with exogenous HNF3β and Pintallavis, although HNF3β RNA elicits a significantly more potent response.

We next examined the effects of HNF3β RNA on dorsal marginal zone explants. Consistent with the block to tadpole-stage somite development, HNF3β inhibits the expression of the somitic marker muscle actin (Mohun et al., 1984); collagen type II expression in these explants is also diminished, indicating a block to notochord formation (Fig. 3A) (Su et al., 1991). The accompanying increase in the expression of α-globin, a marker of blood, itself a ventral mesodermal derivative, suggests that these explants are ventralized, although ventral mesoderm may simply be less sensitive to inhibition by HNF3β (Banville and Williams, 1985). Higher concentrations of HNF3β RNA were lethal, however, and we were not able to identify a dose of HNF3β RNA that blocked all mesoderm formation (data not shown).

Fig. 3.

Effects of HNF3β on marginal zone explants. (A) Injection of HNF3β RNA inhibits development of dorsal mesoderm. RT-PCR analysis of dorsal marginal zone (DMZ) explants harvested at late neurula stages. (B) Ventral injection of HNF3β RNA dorsalizes mesoderm, but does not induce secondary axis formation. RT-PCR analysis of ventral marginal zone (VMZ) explants harvested at late neurula stages (top); lateral views of stage 35 embryos, anterior is to right (bottom). (C) Ventral injection of pintallavis RNA dorsalizes mesoderm. RT-PCR analysis of ventral marginal zone (VMZ) explants harvested at late neurula stages. (D) Injection of HNF3β RNA inhibits mesoderm induction and dorsoanterior development in early gastrulae. RT-PCR analysis of marginal zone explants harvested immediately after dissection at early gastrula stages. 250 pg (A, B, C) or 500 pg (D) HNF3β or pintallavis RNA was injected, as listed, into 4–8 cell stage embryos.

Although ectopic dorsal expression of HNF3β inhibits dorsal mesoderm formation, ventral expression of HNF3β had the surprising effect of dorsalizing marginal zone explants: globin expression was decreased, and both notochord and somite marker expression was increased, in ventral marginal zone explants derived from embryos injected with HNF3β RNA (Fig. 3B); similar effects were seen following ventral injection of pintallavis RNA (Fig. 3C). Ventral injection of these RNAs, however, rarely gave rise to secondary axes in intact embryos (3% of HNF3β RNA-injected embryos formed secondary axes; n=64), although we observed concentrations of pigment in lateral and/or ventral aspects of some embryos at tadpole stages (Fig. 3B, bottom panel, and data not shown). Infrequent axis duplication has also been reported following ventral injection of pintallavis RNA (Ruiz i Altaba and Jessell, 1992). These results indicate that ectopic expression of forkhead proteins, including HNF3β and Pintallavis, can dorsalize ventral mesoderm, but that this tissue is only infrequently organized into axial structures.

In order to get a better understanding of the effects of ectopic HNF3β, we analyzed its effects on the expression of molecular markers during gastrula stages. HNF3β inhibits the expression of the panmesodermal marker Xbrachyury (Xbra) in both dorsal and ventral explants (Smith et al., 1991); in addition, HNF3β inhibits the expression of all dorsal endomesodermal markers examined, including chordin, goosecoid, dikkopf (dkk), and otx2 (Fig. 3D, compare lanes 1 and 3, 2 and 4) (Blitz and Cho, 1995; Cho et al., 1991; Glinka et al., 1998; Pannese et al., 1995; Sasai et al., 1994). Expression of the anterior endoderm markers Xhex and cerberus are also suppressed by HNF3β, although the panendodermal marker Sox17β is unaffected (Fig. 3D) (Hudson et al., 1997). Finally, ventral expression of the ventrolateral markers vent1 and Xwnt8 are both inhibited by HNF3β (Fig. 3D, compare lanes 2 and 4) (Christian et al., 1991; Gawantka et al., 1995; Smith and Harland, 1991); dorsal Xwnt8 expression is, however, modestly elevated in some experiments (Fig. 3D, compare lanes 1 and 3). These experiments demonstrate that ectopic HNF3β antagonizes mesodermal and anterior endodermal development during gastrula stages. We found no evidence of gastrula stage dorsalization in ventral explants injected with HNF3β RNA, which suggests that the dorsalization seen in late neurula-stage ventral explants may reflect a later and/or secondary consequence of HNF3β activity (see Discussion).

Activation of HNF3β target genes inhibits mesoderm formation

Although cell culture studies demonstrate that HNF3β can function as a transactivator, the transcriptional activity of this protein during early development has not been defined (Cereghini, 1996). Towards this end, we constructed two activated forms of HNF3β, which contain HNF3β sequences fused to a VP16 activation domain (Fig. 4) (Kessler, 1997). Both constructs behaved similarly in the functional assays described below; however, we found that VP-HDNAB, a VP16-HNF3β DNA-binding domain construct, was significantly less toxic than VP-HNF3β, a fusion construct that included additional HNF3β residues (data not shown). VP-HDNAB was thus the construct used in most of our experiments.

FIG. 4.

Chimeric HNF3β constructs used in this study. See text for details.

Dorsal injection of VP-HDNAB RNA results in embryos that are indistinguishable from those injected with wild-type HNF3β RNA, both morphologically and in the dramatic reduction or absence of somitic mesoderm (100% affected, n=14; Fig. 5A). Dorsal marginal zone explants from VP-HDNAB RNA-injected embryos showed a reduction in the expression of the dorsal markers chordin and goosecoid, similar to the affects seen following injection of HNF3β; these explants also displayed reduced expression of both Xbra and Xwnt8 (Fig. 5B, compare lanes 1 and 2). Ventral marginal zone explants expressing VP-HDNAB also show a reduction in both Xbra and Xwnt8 expression, suggesting that VP-HDNAB inhibits both dorsal and ventral mesoderm formation (Fig. 5B, compare lanes 3 and 4). Finally, we find that VP-HDNAB suppresses mesoderm formation in ectodermal (animal cap) explants treated with the Transforming Growth Factor β (TGFβ) superfamily protein Activin (Fig 5C, compare lanes 1 and 2); this suppression can be rescued by co-expression of a Drosophila Engrailed repressor-Xenopus HNF3β fusion protein, EnR-HNF3β, pointing to the specificity of suppression by VP-HDNAB (Fig. 4; 5C, compare lanes 2 and 3, and see below). These results, coupled with data on the activity of EnR-HNF3β RNA (see below) suggest that HNF3β functions as a transcriptional activator during early development, and indicates that activation of HNF3β target genes inhibits mesoderm formation during early Xenopus development.

FIG. 5.

Inhibition of mesoderm by activated HNF3β constructs. (A) Injection of VP-HDNAB RNA disrupts gastrulation and inhibits mesoderm formation in intact embryos. Embryos were injected dorsally with VP-HDNAB RNA, cultured until stage 39, and probed with the somite-specific antibody 12/101 (Kintner and Brockes, 1984). Embryos are lateral views; anterior is to left. (B) Dorsal and ventral marginal zone expression of VP-HDNAB RNA inhibits mesoderm formation. RT-PCR analysis of marginal zone explants harvested immediately after dissection at early gastrula stages. (C) VP-HDNAB-mediated inhibition of mesoderm formation is rescued by co-expression of EnR-HNF3β RNA. RT-PCR analysis of animal caps dissected at late blastula stages and cultured until midgastrula stages. 1 ng VP-HDNAB or EnR-HNF3β RNA was injected at early cleavage stages, as listed. 0.5 ng/mL Activin was added to stage 9 animal caps, as listed.

Dorsal repression of HNF3β target genes inhibits head formation

We have thus far demonstrated that activation of HNF3β target genes in the marginal zone represses mesoderm formation in Xenopus embryos and explants. We next asked whether the converse was also true; i.e., whether repression of HNF3β target genes promotes mesoderm formation. To address this issue, we examined the effects of HNF3β target gene inhibition within the normal domain of HNF3β expression. We first injected RNA generated from the Engrailed repressor-HNF3β fusion construct, EnR-HNF3β, into dorsal vegetal blastomeres; injection of EnR-HNF3β RNA dramatically inhibits head formation in a high percentage of injected embryos (87%; n=106) (Fig. 6A). In addition to anterior defects, we observed ectopic structures, resembling axial duplications, in approximately 10% of the embryos following dorsal vegetal injection of EnR-HNF3β RNA; this phenotype could be viewed most readily in late tailbud stage embryos (Fig. 6A, right panel). Somewhat surprisingly, these structures were not positive for the somite antibody 12/101 (data not shown). Lineage tracing in control embryos demonstrated that injected blastomeres tend to populate the anterior endoderm and, to a lesser extent, the axial mesoderm and anterior neural structures (Fig. 6B, top panel); HNF3β-injected blastomeres appear to adopt a similar positional fate (Fig. 6B, bottom panel). These data indicate that repression of HNF3β target genes in the anterior endoderm inhibits head formation.

FIG. 6.

Effects of HNF3β target gene inhibition in the anterior endoderm. (A) Dorsal vegetal injection of 250 pg EnR-HNF3β RNA inhibits head formation. Left panel: lateral view of stage 40 embryos; anterior is to left. Embryo at top is an uninjected control. Right panel: EnR-HNF3β RNA induces ectopic structures, resembling secondary axes, in a subset (approximately 10%) of injected embryos; these structures do not contain somitic mesoderm (data not shown). Lateral views of stage 28 embryos; anterior is to left. Embryo at top is an uninjected control. (B) Dosal vegetal cells expressing EnR-HNF3β populate the head and anterior gut. Embryos injected in dorsal vegetal blastomeres with 200 pg β-Gal RNA (top) or both 250 pg EnR-HNF3β and 200 pg β-Gal RNA (bottom), and stained with X-gal as a substrate. Lateral views of stage 32 embryos; anterior is to left; embryo in bottom panel, viewed at higher magnification, was cleared in 2:1 benzyl benzoate/benzyl alcohol. (C) EnR-HNF3β-MT does not affect dorsal development. Top panels: gel shift analysis of EnR-HNF3β and EnR-HNF3β-MT; in vitro translation products are shown at left. TTR, specific competitor; AP2, non-specific competitior; control, no RNA included in in vitro translation. Bottom panel: lateral views of stage 35 embryos; anterior is to left. 8-cell stage embryos were injected dorsovegetally with 250 pg of EnR-HNF3β-MT RNA. (D) Dorsal suppression of HNF3β target genes alters mesendodermal gene expression. RT-PCR analysis of dorsal explants dissected and immediately harvested at early gastrula stages; explants used for lane 1 were isolated from embryos injected with 250 pg EnR-HNF3β RNA in both dorsal vegetal blastomeres at the 8-cell stage. (E) Pintallavis-EnR does not affect dorsal development. Top panel: RT-PCR analysis of dorsal explants dissected and immediately harvested at early gastrula stages; explants were isolated from embryos injected with 250 pg of either EnR-HNF3β RNA or pintallavis-EnR RNA in both dorsal vegetal blastomeres at the 8-cell stage. Bottom panel: Lateral view of stage 28–30 embryos injected with 250 pg pintallavis-EnR RNA in both dorsal vegetal blastomeres at the 8-cell stage; anterior is to right.

To verify that EnR-HNF3β function is dependent on DNA binding, we constructed EnR-HNF3β-MT, in which we introduced the identical mutations described above for HNF3β-MT (Fig. 4). Unlike EnR-HNF3β, EnR-HNF3β-MT does not bind a consensus HNF3 binding site in mobility shift assays (Fig. 6C, top panel). 78% of embryos injected with EnR-HNF3β-MT RNA are indistinguishable from uninjected controls (Fig. 6C, bottom panel). Affected embryos have smaller, but still recognizable, head structures (data not shown). Thus, as expected, EnR-HNF3β activity, like that of wild-type HNF3β, is primarily mediated through interaction with DNA. The residual activity seen after injection of either EnR-HNF3β-MT or HNF3β-MT RNA (see above), however, suggests that these mutants retain some level of DNA-binding activity, via additional forkhead box residues, that is below the level of detection of our in vitro assay. Alternatively, we cannot rule out the possibility that EnR-HNF3β and HNF3β also act through mechanisms independent of DNA binding.

To better characterize the effects of HNF3β target gene repression, we isolated, at early gastrula stages, explants of embryos injected with EnR-HNF3β RNA in dorsal vegetal blastomeres at the 8-cell stage. RT-PCR analysis of these dorsal explants revealed a strong and consistent down-regulation of otx2 as well as a decrease in the expression of both cerberus, and chordin (Fig. 6D, compare lanes 1 and 2); the range of expression of these latter two markers, however, was substantially more variable than was observed for otx2 (compare, for example, chordin expression in Figs. 6D, E). In all experiments, we also observed an increase in the expression of the ventrolateral mesodermal markers Xwnt8 and vent1; an increase in the expression of the dorsoanterior markers Xhex, dkk, and goosecoid, as well as the pan-endodermal marker sox17β, was observed in roughly 50% of experimental trials (Fig. 6D, and data not shown). Finally, dorsal expression of Xbra was largely unaffected by EnR-HNF3β expression (Fig. 6D). In summary, we find that dorsal suppression of HNF3β target genes activates ventrolateral mesodermal marker expression. This activation does not, however, coincide with a global decrease in dorsoanterior marker expression; rather, only a subset of such genes, including otx2 and, to a lesser extent, cerberus and chordin, are inhibited by EnR-HNF3β. Consistent with these findings, tailbud-stage expression of dorsal mesodermal markers is unaffected by EnR-HNF3β expression (data not shown).

To examine the specificity of EnR-HNF3β, we compared its activity to that of a Pintallavis repressor construct, Pintallavis-EnR (Saka et al., 2000). While marginal zone injection of pintallavis-EnR RNA disrupted early development as reported, giving rise to embryos with shortened axes, poorly developed heads, and occasional gastrulation defects (data not shown) (Saka et al., 2000), dorsovegetal expression of Pintallavis-EnR had minimal effects on either head development (Fig. 6E, bottom panel; 90% wild-type; n=40) or on the expression of marker genes repressed by EnR-HNF3β, including otx2 and chordin (Fig. 6E, top panel). Both repressor constructs did, however, increase the dorsal expression of Xwnt8 (Fig. 6E, top panel). These results indicate that the effects of EnR-HNF3β and Pintallavis-EnR are largely non-overlapping, and suggest that Pintallavis activity is not affected by dorsovegetal expression of EnR-HNF3β.

Ventral repression of HNF3β target genes induces ectopic axis formation

Finally, we attempted to inhibit HNF3β activity in ventral/posterior endoderm. Injection of EnR-HNF3β RNA into 8-cell stage ventral vegetal blastomeres resulted in the formation of ectopic structures, resembling secondary axes, in 46% of injected embryos (n=35; Fig. 7A); these secondary axes never contained recognizable head structures, such as eyes or cement glands (Fig. 7A, and data not shown). Lineage tracing experiments demonstrated that some injected cells contribute directly to the ectopic axes (Figs. 7A, B). These embryos were readily distinguishable from those injected dorsally, as ventral injection of EnR-HNF3β RNA had no effect on head formation (Compare Figs. 6A and 7A); furthermore, nearly all ectopic axes generated by ventral injection stained positive with the somite-specific antibody 12/101 (93%; n=15) (Fig. 7B), in contrast with those seen following dorsal injections. Consistent with the induction of secondary axes, ventral vegetal injection of EnR-HNF3β RNA dorsalizes ventral explants; explants from injected embryos express both the notochord marker collagen type II and the somitic marker muscle actin (Fig. 7C); these explants show a pronounced decrease in the normally high levels of globin expression. Finally, in marked contrast to the effects seen following HNF3β RNA injection, we find that injection of EnR-HNF3β RNA induces the expression of a number of dorsoanterior endomesodermal markers at gastrula stages including, most prominently, goosecoid (Fig. 7D, compare lanes 1 and 2); dkk, chordin, and Xhex expression are also elevated in ventral explants injected with EnR-HNF3β RNA, albeit none as dramatically as goosecoid. Notably, cerberus and otx2 expression are not affected by ventral EnR-HNF3β RNA injection; expression of the ventrolateral markers vent1 and Xwnt8 are also unaffected, while sox17β expression is increased, and Xbra expression is decreased, by EnR-HNF3β RNA injection. These results suggest that repression of HNF3β target genes converts ventral/posterior cells to a dorsoanterior endmesodermal fate.

FIG. 7.

Effects of HNF3β target gene inhibition in the posterior/ventral endoderm. (A) Ventral vegetal blastomeres expressing EnR-HNF3β RNA (250 pg) contribute directly to ectopic axes. 8-cell stage embryos were co-injected with 250 pg EnR-HNF3β and 200 pgβ-Gal RNA, and stained with X-gal as a substrate; stain is found predominantly in the ectopic structures. Lateral view of stage 32 embryos; anterior is to right. (B) Secondary axes induced by ventral vegetal injection of EnR-HNF3β RNA (250 pg) contain dorsolateral mesoderm. Embryos were co-injected as in (A), stained with X-gal as a substrate, and probed with the somite-specific antibody 12/101 (Kintner and Brockes, 1984). Lateral view of stage 32 embryos; anterior is to right. Embryos were cleared in 2:1 benzyl benzoate/benzyl alcohol. (C) Injection of EnR-HNF3β RNA dorsalizes ventral explants. RT-PCR analysis of ventral explants harvested at mid-neurula stages; explants used for lane 1 were isolated from embryos injected with 250 pg EnR-HNF3β RNA in both ventral vegetal blastomeres at the 8-cell stage. (D) Ventral suppression of HNF3β target genes induces dorsoanterior mesendoderm. RT-PCR analysis of ventral explants dissected and immediately harvested at early gastrula stages; explants used for lane 1 were isolated from embryos injected with 250 pg EnR-HNF3β RNA in both ventral vegetal blastomeres at the 8-cell stage.

DISCUSSION

In this study, we report that Xenopus HNF3β is expressed throughout the deep endoderm at early gastrula stages, a pattern reminiscent of HNF3β expression in the mouse visceral endoderm. Ectopic expression of HNF3β in the marginal zone inhibits the expression of gastrula stage mesendodermal markers and the later development of axial and paraxial mesoderm; conversely, inhibition of HNF3β target genes in the ventral gastrula gives rise to the ectopic expression of dorsoanterior genes and, subsequently, truncated secondary axes. Our studies suggest that HNF3β plays an important role in the regulation of mesoderm formation in the Xenopus gastrula.

In the mouse, HNF3β (mHNF3β) is required in the visceral endoderm (VE) for proper morphogenesis of the approximate equivalent of the Xenopus marginal zone, the primitive streak (Ang and Rossant, 1994; Dufort et al., 1998; Weinstein et al., 1994). Loss of function of both HNF3β and the LIM homeodomain transcription factor Lim1 in the VE leads to a dramatic expansion of the primitive streak and the production of ectopic ventral mesoderm at the expense of epiblast cells (Perea-Gomez et al., 1999); our data demonstrate similarities between the effects of HNF3β target gene repression in Xenopus embryos and deletion of HNF3β and lim1 in the mouse. These studies thus point to a conserved function for HNF3β in the suppression of ectopic mesoderm in higher and lower vertebrates, and suggest that Xenopus HNF3β, expressed in the deep endoderm but excluded from the notochord, functions in part as the equivalent of mHNF3β in the VE. Other winged-helix factors expressed in the notochord, including pintallavis, likely mediate the axial mesodermal component of mHNF3β function in Xenopus.

How might Xenopus HNF3β mediate the inhibition of mesoderm? Misexpression of HNF3β phenocopies the effects seen following expression of an activated HNF3β construct, strongly suggesting that HNF3β functions as a transcriptional activator during early embryogenesis. HNF3β^−/−lim1^−/− mice show a severe reduction or absence of expression of both the Nodal antagonists lefty1 and cerberus-like (cerl) in the anterior visceral endoderm (AVE) (Perea-Gomez et al., 1999). Furthermore, combined targeted deletion of Lefty1 and Cerl gives rise to ectopic primitive streak formation (Perea-Gomez et al., 2002). Although dorsoventral differences in the mesoderm generated by these compound knockouts indicate that these sets of mutations are not equivalent, the data suggest that HNF3β and/or Lim1 stimulate the expression of nodal antagonists in the anterior visceral endoderm. As described in the Introduction, Nodal appears to function as a mesoderm inducer in both frog and mouse; thus, the increase in mesoderm seen in explants injected with EnR-HNF3β RNA could reflect a role for HNF3β in the maintenance of Nodal antagonist expression in the anterior endoderm.

The modest repression of cerberus expression by EnR-HNF3β expression supports the notion that inhibition of mesoderm by Xenopus HNF3β is mediated by enhanced transcription of Nodal antagonists, including cerberus. Several observations, however, complicate this view. First, the increase in dorsal expression of Xwnt8 and vent1 by EnR-HNF3β is only occasionally accompanied by a dramatic decrease in cerberus expression. Second, ectopic dorsal HNF3β downregulates cerberus expression, suggesting that this gene is not a simple target of HNF3β transcativation. Third, EnR-HNF3β induces mesoderm and secondary axis formation in the ventral/posterior endoderm, a region with little or no expression of cerberus (Bouwmeester et al., 1996). This indicates that cerberus repression not a prerequisite for mesoderm induction by this construct. While these caveats do not exclude the possibility that HNF3β regulates the transcription of other Nodal antagonists, the frog embryo itself presents us with an additional limitation of this model. Although Nodal-related proteins induce mesoderm in Xenopus, high levels of these molecules also induce endoderm; furthermore, Nodal-related molecules appear essential for endoderm formation in Xenopus (Clements et al., 1999; Gamer and Wright, 1995; Henry et al., 1996; Osada and Wright, 1999; Yasuo and Lemaire, 1999). Thus, inhibition of Nodal antagonists in the deep endoderm (by EnR-HNF3β) would not be expected to convert endoderm to a mesodermal fate. Although the increase in sox17β expression by EnR-HNF3β suggests that mesoderm induction in our explant assays may not come entirely at the expense of endoderm, lineage tracing experiments suggest that at least some of the injected, deep endodermal cells contribute directly to ectopic mesoderm. We believe, then, that mesoderm inhibition by HNF3β is mediated in part through a mechanism independent of Nodal antagonism.

One set of experiments reported here is ostensibly inconsistent with a model in which HNF3β inhibits mesoderm formation: we have shown that ventral expression of HNF3β dorsalizes marginal zone explants at tadpole stages. These data should, however, be viewed in the context of several additional findings. First, we find no evidence of dorsalization by HNF3β in ventral cells early in development; in fact, we observe a reduced expression of the ventral markers Xwnt8 and vent1 in ventral explants harvested at gastrula stages. This suggests that the late stage dorsalization by HNF3β is mediated by one or more indirect downstream events. For example, ventral inhibition of Xwnt8 has been shown to dorsalize mesoderm (Hoppler et al., 1996); thus, the inhibition of Xwnt8 by HNF3β could have the paradoxical effect of dorsalizing those ventral cells that retain mesodermal identity. Second, we find that pintallavis RNA is also quite effective in these assays; this suggests that HNF3β misexpression in the late ventral marginal zone may mimic the activity of other, related proteins, perhaps including Pintallavis, whose neurula stage expression in the notochord is, unlike that of HNF3β, consistent with a role for this factor in the establishment and/or maintenance of dorsal mesodermal fates (Ruiz i Altaba and Jessell, 1992). Third, it must be noted that the experiments utilizing HNF3β RNA injection are misexpression studies and should be viewed with the appropriate caveats; for example, ectopic HNF3β activity may linger well past the desired period of study. These assays, particularly those in which we examine the effects of HNF3β RNA on late stage explants, therefore have to be considered in the context of both the expression studies and the loss-of-function, EnR-HNF3β experiments.

Finally, our data suggest that Xenopus HNF3β also plays a role in the maintenance of anterior fate, as dorsal expression of EnR-HNF3β gives rise to headless embryos; a loss of anterior neural structures is also observed in a subset of HNF3β^−/− mouse embryos (Ang and Rossant, 1994; Weinstein et al., 1994). This phenotype could be mediated in part by the induction of Xwnt8, the zygotic expression of which acts as a ventralizing factor (Hoppler et al., 1996); however, we observed no loss of head structures following injection of Pintallavis-EnR RNA, despite evidence of a similar increase in Xwnt8 expression. Alternatively, EnR-HNF3β may suppress head formation by inhibiting expression of otx2, shown to play a role in head formation in both mouse and frog (Ang et al., 1996; Blitz and Cho, 1995; Pannese et al., 1995). Compound mutational analysis has previously suggested a genetic link between HNF3β and otx2 in mammalian head formation (Jin et al., 2001); suppression of otx2 by vegetal injection of EnR-HNF3β RNA suggests that maintenance and/or induction of otx2 expression requires endodermal HNF3β activity. In the early Xenopus gastrula, the primary expression domains of these genes appear to be distinct: HNF3β transcripts are excluded from the mesoderm (this study), while otx2 expression is largely restricted to cells fated to give rise to prechordal mesoderm (Ang et al., 1996; Blitz and Cho, 1995; Pannese et al., 1995). These expression data suggest that Xenopus HNF3β is not required cell autonomously for otx2 expression; this idea gains significant support from the demonstration that misexpression of HNF3β in dorsal mesoderm actually inhibits the expression of otx2. Our studies thus point to a non-cell autonomous requirement for endodermal HNF3β activity in the induction and/or maintenance of otx2 expression in the presumptive head mesoderm.