Murine gastrulation requires HNF-4 regulated gene expression in the visceral endoderm: tetraploid rescue of Hnf-4^−/− embryos

The visceral endoderm (VE) derives from the primitive endoderm, which itself differentiates from the inner cell mass cells of the blastocyst at around 4.0-4.5 days post coitum (E4.0) in the mouse (Gardner, 1983). The transcription factor HNF-4 is expressed in the primitive endoderm as soon as a morphologically distinct endoderm can be identi?ed, suggesting that it could be important for differentiation of this tissue (Duncan et al., 1994). During gastrulation the VE joins with the extraembryonic mesoderm to form the visceral wall of the yolk sac. Its function as part of the yolk sac has been well characterized in postgastrulation rodent embryos where it has a critical role in the maternofetal exchange of nutrients prior to formation of a functioning placenta (Jollie, 1990). That a functioning yolk sac is critical for development of the postgastrulation mammalian embryo is illustrated by the observation that antisera which recognize proteins in the yolk sac are teratogenic (for review see Brent et al., 1990). During this period the VE is also responsible for the synthesis and secretion of several serum factors including α -fetoprotein (AFP), transthyretin (TTR), and several apolipoproteins (Apo) (Meehan et al., 1984) and, in addition, is the site of embryonic hematopoeisis. Although preceding and during gastrulation little is known about VE function, gene targeting studies have suggested that the VE may have important roles during early embryogenesis (Spyropoulos et al., 1994; Ang et al., 1994; Weinstein et al., 1994; Chen et al., 1994).

Hepatocyte nuclear factor 4 (HNF-4) is a transcription factor identi?ed in liver extracts as a DNA binding protein which binds to the promoters of the transthyretin (TTR) and apolipoprotein CIII (ApoCIII) genes (Sladek et al., 1990). A member of the steroid hormone receptor family which lacks a known ligand (Sladek et al., 1990), HNF-4 is evolutionarily conserved, with homologues found in Xenopus laevis, Drosophila melongaster, and humans (Zhong et al., 1993; Drewes et al., 1996). Targeted disruption of the Hnf-4 gene in mice revealed that it was critical for completion of gastrulation (Chen et al., 1994). Although Hnf-4^{− /−} embryos initiate gastrulation, as evidenced by production of cells expressing markers of nascent mesoderm, eg. Brachyury, they fail to produce cells expressing late mesoderm markers such as mox-1, suggesting that HNF-4 has a critical role in supporting the progression of gastrulation (Chen et al., 1994). Analyses of Hnf-4 expression by in situ hybridization revealed that during this stage of development Hnf-4 mRNA was restricted to the extraembryonic visceral endoderm (VE), with no Hnf-4 mRNA detected in embryonic tissues (Duncan et al., 1994). The ?rst expression of Hnf-4 in embryonic tissues was not detected until induction of the liver diverticulum at about E9.0 (Duncan et al., 1994; Taraviras et al., 1994). This implied that the VE has a critical function in maintaining gastrulation of the murine epiblast.

To determine whether HNF-4 is required for differentiation of the VE and, furthermore, to de?ne the role of the VE during murine gastrulation we took advantage of the unique properties of embryonic stem (ES) cells to form VE in vitro and embryos in vivo. We show here that in the absence of HNF-4 differentiation of the VE is incomplete, resulting in a functionally defective tissue which fails to express several serum proteins including AFP, Apo-AI, Apo-AIV, Apo-B, TTR, retinol binding protein (RBP), and transferrin (TFN). Furthermore, Hnf-4^{− /−} embryos are rendered gastrulation competent when they are complemented with Hnf-4^+/+ VE, con?rming that Hnf-4^{− /−} embryonic ectoderm cells can complete gastrulation and that the VE has a critical function in de?ning an environment supportive of gastrulation in the mouse.

Embryonic stem (ES) cells were maintained in ES cell medium supplemented with 1000 u/ml LIF on a primary embryonic ?broblast feeder layer as described by Robertson (1987). ES cells were induced to differentiate to form EBs in vitro according to the method of Robertson (1987). 3× 10⁶ ES cells were plated on gelatin-coated tissue culture dishes and grown for 3 days in the absence of feeder ?broblasts and LIF. After addition of trypsin/EDTA, ES cell clumps were collected, divided between bacterial grade Petri dishes (Fisher) containing ES cell medium and grown in suspension for the speci?ed duration. To generate ES cells homozygous for a targeted mutation in the HNF-4 gene (Hnf-4^{− /−} ) the method of Mortensen et al. (1992) was followed. Speci?cally, 5.0× 10⁶ HNF-4 heterozygote ES cells (Hnf4^+/− ) (clone 2-69; Chen et al., 1994) were plated in medium containing 1.25 mg/ml G418 and 1000 units/ml LIF. After 9 days in culture surviving colonies were expanded and their genotype ascertained by Southern blot.

DNA was prepared from ES cells, mouse tail biopses and mouse embryos as described by Plump et al. (1992). In Southern blots of NcoI digested genomic DNA a wild-type Hnf-4 allele generated a 4.5 kb restriction fragment which hybridized to an Hnf-4 speci?c probe, while the targeted allele generated a 3.8 kb fragment, as previously reported (Chen et al., 1994). The genotype of ES cell/teraploidderived mouse embryos was determined by polymerase chain reaction (PCR). Embryonic DNA served as templates for PCRs which contained all four deoxyribonucleotides (dNTPs), [α -³²P]dATP and taq DNA polymerase and utilized primers speci?c to either HPRT (agcgcaagttgaatctgc, agcgacaatctaccagag), the neomycin resistance gene (Neo^r) (gccaacgctatgtcctgatagcggt, agccggtcttgtcgatcaggatgat), or Hnf-4 (ccccatctgaaggtgccaacctc, ggttcttcctcacgctcctcctgaa).

Conditions used for RT-PCR followed the method of Wilson and Melton (Wilson et al., 1994) with minimal modi?cations. Total RNA was extracted from ES cells or mouse embryos using TRIzol reagent and following the manufacturers instructions (Gibco-BRL) and contaminating genomic DNA was removed using 1 μl of RNase-free DNase-I (Boehringer)/10 μ g RNA. cDNA was synthesized using MMLV-RT (Gibco-BRL) with dNTPs and random hexamer primers (Gibco-BRL). These cDNAs provided templates for PCRs using speci?c primers at an annealing temperature of 65ºC in the presence of dNTPs, [α -³²P] dATP and taq DNA polymerase. The following forward and reverse primers were used for speci?c ampli?cation: HPRT; agcgcaagttgaatctgc, agcgacaatctaccagag, GATA-4; ctaagctgtccccacaaggctatgca, cagagctccacctggaaaggtgtttg, vHNF-1; gaaagcaacgggagatcctccgac, cctccactaaggcctccctctcttcc, Apo-E; aggatgcctagccgagggagagc, tagatcctccatgtcggctccgagt, Hnf-4; cttccttcttcatgccag, acacgtccccatctgaag, AFP; tcgtattccaacaggagg, aggcttttgcttcaccag, TFN; tggcacaggaacactttg, tcctgctgattccgaatg, Apo-AI; acacacgtagactctctg, ctgggctttggtcttaag, Apo-AIV; agccaaggaaactgagag, tctccttgatcgtggtct, Apo-B; cttcagggaacaaagcag, tcaagggtgagctgattg, TTR; ctcaccacagatgagaag, ggctgagtctctcaattc, RBP; atccagtggtcatcgtttcctcgct, gaacttcgacaaggctcgtttctctgg, HNF-1; ttctaagctgagccagctgcagacg, gctgaggttctccggctctttcaga.

In situ hybridizations were performed as described previously (Duncan et al., 1994). Sense and anti-sense Hnf-4, [³³P]UTP labeled RNA probes were synthesized in vitro from the plasmid p4-is. The speci?city of this probe for Hnf-4 has previously been described (Chen et al., 1994). Probes were hybridized to paraffin sections of day 14 J1 ES cell aggregates (Hooper et al., 1987), exposed to photographic emulsion for 14 days, before developing and counterstaining with hematoxylin and eosin. Flourescein-isothiocyanate (FITC)labeled Sophora japonica (SJA) (Sigma) was used to label the VE of either Hnf-4^+/+ or ^{− /−} day-11 ES cells aggregates in whole mount as described (Soudais et al., 1995; Wu et al., 1983).

ES cell-derived embryos were produced essentially following the method described by Nagy and Rossant (1993) with minor modi?cations. 2-cell stage embryos isolated from naturally mated CD-1 mice were collected in M2 medium (Specialty Media Inc.), equilibrated in fusion buffer (0.3 M mannitol, 0.1 mm MgSO₄, 50 μM CaCl₂, 3% BSA) (McLaughlin, 1993) and placed between the electrodes of a CF150 cell fusion instrument (Biochemical Laboratory Service LTD., Hungary). Embryos were aligned in a 0.7 volt AC ?eld and fused with three 100 volt DC pulses of 45 μ seconds each.

Embryos were washed in M2 medium, transferred to KSOM medium (Specialty Media Inc.) and incubated for 1 hour at 37°C/5% CO₂ at which time successfully fused embryos were recovered and cultured overnight under the same conditions. ES cell-tetraploid chimaeric embryos were produced by aggregation as described elsewhere (Nagy and Rossant, 1993).

During the earliest stages of implantation, inner cell mass cells of the blastocyst which juxtapose the blastocoel cavity are speci?ed to form the primitive endoderm lineage from which all extraembryonic endoderm is derived. The embryo at this time is relatively resistant to a molecular investigation and so, to determine whether HNF-4 has any role in differentiation of the extraembryonic endoderm lineages, we decided to use an ES cell in vitro differentiation assay. When ES cells are grown in suspension culture in the absence of LIF they differentiate to form embryoid bodies (EBs) (Evans et al., 1981; Martin, 1981). These EBs, which resemble mouse embryos at early developmental stages, contain differentiated VE (Doetschman et al., 1985). To ascertain whether HNF-4 expression was induced during EB formation in vitro, we assayed for the presence of Hnf-4 mRNA by in situ hybridization (Fig. 1A). 5 μ m paraffin sections of day-14 postaggregation EBs were hybridized to a [³³P]UTP labeled RNA probe which had previously been shown to recognize only Hnf-4 (Duncan et al., 1994; Chen et al., 1994). While a sense-strand RNA probe showed no hybridization above background (data not shown), Fig. 1A shows that an anti-sense probe identi?ed Hnf-4 transcripts which were restricted to the VE layer of the EB. From these data we conclude that Hnf-4 mRNA is expressed in, and restricted to, the VE of ES cell EBs.

The expression of HNF-4 in the VE of ES cell EBs allowed us to adopt an in vitro genetic approach to ask whether HNF-4 was central to VE differentiation and/or function. We have previously described the generation of HNF-4 heterozygote (Hnf4^+/− ) ES cells in which the DNA binding domain of one Hnf4 allele was deleted by homologous recombination (Chen et al., 1994). To construct ES cell lines which were homozygous for this targeted mutation (Hnf-4^{− /−} ), Hnf-4^+/− cells were passaged in high concentrations of G418 (high [G418]). This procedure has previously been shown to efficiently produce ES cell lines which are homozygous for a targeted allele (Mortensen et al., 1992). ES cell clones which survived 9 days in culture under 1.5 mg/ml G418 were assayed for loss of the wild-type Hnf-4 allele by Southern blot analysis. An Hnf-4 speci?c DNA probe identi?ed a 4.5 kb NcoI fragment in wild-type genomic ES cell DNA (Fig. 1B; lane 1) while the targeted allele generated a 3.8 kb NcoI fragment (Fig. 1B; lanes 2-6) (Chen et al., 1994). Of 44 high [G418] resistant clones analyzed, 12 were Hnf-4^{− /−} and the remainder Hnf-4^+/− . Three Hnf-4^{− /−} lines (A8, B9 and B13) and one Hnf-4^+/− line (B16, referred to as Hnf-4^+/−N), were selected for subsequent experiments.

To demonstrate that no functional HNF-4 could be expressed by the Hnf-4^{− /−} ES lines, Hnf-4^+/+, Hnf-4^+/− and Hnf-4^{− /−} day14 EBs were assayed for the presence of Hnf-4 mRNA by RTPCR (Fig. 1C). To control for the relative amounts of RNA used in each RT-PCR assay we included primers which identi?ed HPRT mRNA since this gene is expressed ubiquitously at relatively constant levels. As shown in Fig. 1C, no product was ampli?ed by HPRT primers in the absence of reverse transcriptase (HPRT − RT) or DNA (lane 1), showing that all products were ampli?ed from cDNA rather than from contaminating genomic DNA. In the presence of reverse transcriptase (HPRT +RT) a speci?c 219 bp product was identi?ed at comparative levels in each sample indicating that each reaction started with a similar amount of template. While HNF-4 primers generated a speci?c PCR product from Hnf-4^+/+ (lane 2), Hnf-4^+/− (lane 3), and Hnf-4^+/−N (lane 7) EB cDNAs, no product was detectable in any of the three Hnf-4^{− /−} EB samples (lanes 4-6). These data verify that ES cell lines A8, B9 and B13 are homozygous Hnf-4^{− /−} mutants.

Because HNF-4 is detected in the primitive endoderm at the earliest stages of its differentiation (Duncan et al., 1994) we wanted to determine whether HNF-4 was required for specification of the extraembryonic endoderm lineage. We therefore measured steady state mRNA levels from the VE marker genes Gata-4 (Soudais et al., 1995), vHnf-1 (Cereghini et al., 1992) and Apo-E (Basheeruddin et al., 1987; Harrison et al., 1995) by RT-PCR in Hnf-4^+/+, Hnf-4^+/−and Hnf-4^−/−EBs (Fig. 2A). As before, ampli?cation with HPRT-speci?c primers demonstrated that each sample started with a comparable concentration of template. In several repetitions of this experiment no signi?cant difference in levels of Gata-4, Apo-E, or vHnf-1 mRNAs were detected between Hnf-4^+/+, Hnf-4^+/− or Hnf-4^{− /−} EBs, suggesting that VE tissue could be produced in the absence of HNF-4. To con?rm this, day-11 Hnf-4^+/+ or Hnf-4^{− /−} EBs were stained in whole mount with FITC-labeled Sophora japonica agglutinin (SJA) which speci?cally reacts with the VE (Sato et al., 1985). Fig 2B,D shows phase contrast micrographs in which the formation of endoderm, evident as a cuboidal epithelium, can readily be identi?ed in both Hnf-4^+/+ and Hnf-4^{− /−} EBs. The same tissue also stains with FITC-labeled SJA showing that this endoderm is VE (Fig. 2C,E). Control EBs did not label when N-acetylgalactosamine was pre-incubated with the lectin (data not shown), con?rming the speci?city of the staining (Sato et al., 1985). Cumulatively, these data demonstrate that Hnf-4 is not required for early speci?cation of the VE lineage.

Differentiation of the VE requires the orderly formation of an epithelial layer upon a basement membrane with the subsequent expression of characteristic late marker genes (Grover et al., 1983a,b). Many of these genes encode secreted serum proteins that are also expressed in hepatocytes (Meehan et al., 1984), where HNF-4 is believed to be important in regulating their expression (Sladek, 1994). We therefore determined whether HNF-4 was required for late phase VE differentiation by using RT-PCR to measure the steady state levels of mRNAs expressed from such genes. Primers were designed to detect AFP, TFN, Apo -AI, -AIV, and -B, TTR and RBP mRNAs by RT-PCR in day-14 Hnf-4^+/+, Hnf-4^+/− and Hnf-4^{− /−} EBs. In addition, since HNF-4 has been implicated in the transcriptional regulation of the transcription factor HNF-1 (Tian et al., 1991; Kuo et al., 1992), primers were included which could detect Hnf-1 mRNA. As before, no product was detected in the absence of DNA or reverse transcriptase showing that products were ampli?ed from cDNAs (Fig. 3A; HPRT – RT, and lane 1). A similar amount of starting material was used in each reaction, as shown by an equivalent amount of product generated by HPRT primers (Fig. 3A; HPRT +RT). GATA-4 was expressed at comparable levels between samples indicating that similar amounts of VE had been formed by the different EBs and, as expected, while HNF-4 was expressed in Hnf-4^+/+ and Hnf-4^+/− EBs none could be detected in the Hnf-4^{− /−} EBs (Fig. 3A: GATA-4 and HNF-4). Analysis of serum protein gene expression gave the striking result presented in Fig. 3A. While expression of AFP, TFN, Apo-AI, Apo-AIV, and Apo-B was easily detected in Hnf-4^+/+ or Hnf-4^+/− EBs, expression was virtually undetectable in Hnf-4^{− /−} EBs. Expression of TTR and RBP mRNAs was also grossly reduced in the Hnf-4^{− /−} EBs and, although less striking, levels of Hnf1 mRNA were also down. These data demonstrate that HNF4 is a key regulator of VE gene expression and is essential for the complete differentiation of VE in vitro.

We next determined whether this disruption to the expression of serum protein genes in the absence of HNF-4 in EBs also held true in vivo. E8.5 embryos were collected from crosses of Hnf-4^{− /−} embryos and pooled according to phenotype; E8.5 embryos which express HNF-4 have formed a distinct headfold and allantois, contain somites and exhibit clear organization of germ layers, while Hnf-4^{− /−} embryos show no obvious morphological signs of gastrulation (Chen et al., 1994). The genotype of the pooled embryos was con?rmed by RT-PCR using HNF-4 speci?c primers (Fig. 3B; HNF-4). Expression of mRNAs for the same genes described above were, once again, assayed by RT-PCR (Fig. 3B). HPRT – RT primers did not generate a product con?rming the absence of contaminating genomic DNA, while the HPRT +RT reaction shows that equivalent amounts of starting material were used in each sample (Fig. 3B; HPRT – RT, HPRT +RT). Fig 3B also shows that higher levels of Gata-4 mRNA are found in Hnf4^{− /−} embryos than in embryos expressing HNF-4, re?ecting the fact that the ratio of VE cells to cells of embryonic lineage is greater in Hnf-4^{− /−} embryos because they fail to undergo normal gastrulation (Chen et al., 1994). Fig. 3B shows that the transcripts of all genes assayed were detected in normal embryos; however, as is the case in vitro, expression of AFP, TFN, Apo-AI, Apo-AIV, and Apo-B mRNAs was almost undetectable in Hnf-4^{− /−} embryos and, as before, expression of TTR, RBP and Hnf-1 was reduced. In sum, from both in vitro and in vivo data, we conclude that ablation of HNF-4 results in a striking dysregulation of gene expression in the VE that severely compromises its paracrine activity.

In situ hybridization anlayses demonstrated that expression of Hnf-4 mRNA was restricted to the VE prior to formation of the hepatic diverticulum at around E9.0. Since no expression could be detected in the embryonic tissues before this stage, we postulated that the disruption to gastrulation was due to dysfunction of the VE (Duncan et al., 1994; Chen et al., 1994). If this model is correct, speci?c complementation of Hnf-4^{− /−} embryos with Hnf-4^+/+ VE should allow Hnf-4^{− /−} embryos to complete gastrulation. As illustrated in Fig 4A, when chimaeras are formed between tetraploid (4n) morulae and diploid (2n) ES cells, the resulting fetuses are derived entirely from ES cells, whereas tetraploid cells contribute to the extraembryonic tissues including the VE (Nagy et al., 1990). This procedure has previously been used to rescue Mash-2^{− /−} embryos, which die due to de?ciencies in development of extraembryonic tissues (Guillemot et al., 1994). Using this system we produced Hnf-4^{− /−} fetuses containing Hnf-4^+/+ VE (Fig. 4).

Hnf-4^{− /−} ES cells were aggregated with Hnf-4^+/+ tetraploid morulas, cultured to blastocyst stage and transferred to surrogate mothers. Post-gastrulation ES cell-derived embryos were collected after 9.5 days of gestation (E9.5) and their genotype ascertained by PCR. As shown in Fig. 4A, primers were designed which could detect either ES cell-derived DNA alone (Neo^r), tetraploid morula-derived DNA alone (HNF-4), or both (HPRT). Embryos derived from all three Hnf-4^{− /−} ES cell lines (A8, B9 and B13) were recovered (n=60) and shown to be devoid of Hnf-4^+/+ cells (Fig. 4B;E), whereas the yolk sacs from these embryos were strongly HNF-4 positive (Fig. 4B;Y). The phenotypes of tetraploid-A8 and -B13 (Hnf-4^{− /−} ) embryos as well as wild-type and Hnf-4^{− /−} embryos are shown in Fig. 4C-F. As described previously, Hnf-4^{− /−} embryos are grossly abnormal by E8.5 (Fig. 4D), and by E9.5 the majority are being resorbed (Chen et al., 1994). As with wild-type control embryos (Fig. 4C), tetraploid-Hnf-4^{− /−} ES cell-derived embryos (Fig. 4E,F) had undergone gastrulation, as illustrated by distinct postgastrula features including de?ned anterior-posterior and dorsal-ventral axes, segmental patterning (somites), neural tube formation and onset of oganogenesis. All ES-cell-derived embryos, including those from wild-type (Hnf- 4^+/+) ES cells (not shown), exhibited an exencephaly that was presumably inherent to the parental ES cell line and not a re?ection of a speci?c defect attributed to loss of HNF-4. These data demonstrate that Hnf-4^{− /−} embryos are capable of com- pleting gastrulation in the presence of Hnf-4^+/+ VE and, futher- more, demonstrate that a fully differentiated VE is required to support murine gastrulation.

Mouse embryos lacking a functional Hnf-4 gene are unable to support gastrulation (Chen et al., 1994). The Hnf-4^{− /−} embryos ?rst show evidence of an abnormal phenotype as early as E6.5, around the onset of gastrulation, at which time the Hnf-4^{− /−} embryonic ectoderm exhibits an increase in apoptotic cell death relative to normal littermates (Chen et al., 1994). By E7.5, when normal embryos are at the late primitive streak stage, Hnf-4^{− /−} embryos show no morphological evidence of gastrulation and by E8.5 they are grossly abnormal. Further investigation of the Hnf-4^{− /−} embryos found that although gas- trulation did initiate it was delayed and failed to progress beyond the expression of early primitive streak stage marker genes (Chen et al., 1994). The tissue distribution of Hnf-4 mRNA during early development was shown by in situ hybridization to be restricted to the visceral endoderm with no expression found in the fetus prior to E9.0 (Duncan et al., 1994; Taraviras et al., 1994). From these data we postulated that HNF-4 regulated the expression of VE secretory proteins which were required to support gastrulation (Duncan et al., 1994; Chen et al., 1994). For this model to be correct we predicted that two criteria should be satis?ed: (i) that expression of secreted protein genes should be dysregulated in Hnf-4^{− /−} VE, and (ii) that speci?c complementation of Hnf4^{− /−} embryos with Hnf-4^+/+ VE should allow Hnf-4^{− /−} embryos to complete gastrulation.

Expression of VE secreted proteins could be blocked if either the VE failed to differentiate or if gene expression was directly affected by the absence of HNF-4. To test this we made Hnf-4^{− /−} ES cells and asked whether they were capable of producing fully differentiated VE in vitro. We found that both Hnf-4^+/+ and Hnf-4^{− /−} EBs could form a morphological endoderm whose identity was con?rmed as VE by positive staining with the diagnostic lectin SJA. In addition, both Hnf4^+/+ and Hnf-4^{− /−} EBs expressed similar levels of the VE marker genes Gata-4, Apo-E, and vHnf-1, con?rming that the initial stages of VE differentiation do not require HNF-4. It has been proposed that one of the VE markers tested, the transcription factor GATA-4, is essential for the earliest stages of VE differentiation in ES cell EBs (Soudais et al., 1995). This would suggest that HNF-4 acts downstream of GATA-4 and is consistent with the proposal that differentiation of the VE is a multistep process (Grover et al., 1983a,b). Whether there is any direct regulation of HNF-4 by GATA-4, resulting in a transcriptional cascade during VE differentiation, is currently under investigation.

Having found that HNF-4 is non-essential for early differentiation of the VE we were able to ask whether HNF-4 was required for expression of secreted protein genes in the VE. Since HNF-4 is believed to be important for the regulation of many genes expressed in hepatocytes it seemed reasonable to suggest that the most likely candidates for serum protein genes regulated by HNF-4 would be those expressed in both liver and VE (Meehan et al., 1984; Sladek, 1994). We therefore analysed the expression of AFP, TTR, Apo-AI, Apo-AIV, Apo-B, RBP, and TFN in Hnf-4^{− /−} ES cell EBs by RT-PCR. We found that while all genes were expressed in Hnf-4^+/+ and Hnf-4^+/− EBs their expression was grossly reduced in all Hnf-4^{− /−} EBs. Furthermore, we also found that expression of these genes was extremely reduced in Hnf-4^{− /−} embryos, demonstrating that HNF-4 is required for expression of several proteins secreted from the VE both in vitro and in vivo. Many of the genes analyzed above contain HNF-4 binding sites within their promoters/enhancers, and some of these sites are important for their expression, at least in tissue culture cells (reviewed by Sladek, 1994). This would suggest that these genes are direct targets of HNF-4 and that their expression is directly dependent upon HNF-4 action. It is generally believed that tissue speci?c transcriptional regulation is the result of the coordinated interplay of several trans-acting factors. Redundancy at the promoter/enhancer level is suggested by the observation that mutation of speci?c cis-acting elements does not usually abolish but, more frequently, subtly modulates the level of transcription of a given gene. Correspondingly, targeted disruption of transcription factor genes often has surprisingly little effect on target gene expression (see for example, Pontoglio et al., 1996). However, in striking contrast to this, our data show that HNF-4 has a central role in establishing the expression of many VE genes and supports the proposal that HNF-4 is a key regulator of complete differentiation of the VE. Whether this is due to the direct action of HNF-4 on target gene expression or as the result of an HNF-4 regulated transcriptional cascade, possibly involving HNF-1, is unknown. However, it is interesting to note that HNF-1 expression appears to be only moderately reduced in Hnf-4^{− /−} VE. This would suggest that while HNF-4 modulates, it is not essential for HNF-1 expression in the VE.

We have previously shown by in situ hybridization that before 9.0 days of gestation Hnf-4 mRNA is restricted to the VE (Duncan et al., 1994). The demonstration that VE lacking HNF-4 expressed greatly reduced levels of secreted serum factors further supported our hypothesis that, in the absence of HNF-4, disruption of VE paracrine activity could block normal gastrulation. If this was indeed the case, speci?c complementation of Hnf-4^{− /−} embryos with Hnf-4^+/+ VE should rescue the Hnf-4^{− /−} block to gastrulation. To address this we used the tetraploid aggregation technique to produce Hnf-4^{− /−} ES cell-derived fetuses which contained Hnf-4^+/+ tetraploid-derived extraembryonic tissues, as described by Nagy et al. (1990) and Nagy and Rossant (1993). This procedure has been used successfully to rescue Mash-2^{− /−} embryos and, more recently, to produce VEGF^{− /−} fetuses directly from ES cells (Guillemot et al., 1994; Carmeliet et al., 1996). We found that in the presence of Hnf-4^+/+ VE, Hnf-4^{− /−} embryos would complete gastrulation, exhibiting anterior-posterior and dorsal-ventral structures which were essentially indistinguishable from wild-type embryos. Cumulatively, these data show that Hnf-4^{− /−} embryonic ectoderm is competent to complete gastrulation and that the VE has a critical role in supporting gastrulation of the epiblast.

Following gastrulation, the extraembryonic VE forms a bilayer structure with extraembryonic mesoderm to produce the visceral wall of the yolk sac. From this time until maturation of the placenta, the yolk sac is responsible for maternofetal transport of nutrients for the growing embryo (reviewed by Jollie, 1990). Although yolk sac function has been well studied much less is known about the role of the VE before formation of the yolk sac. Recently, however, the VE has been shown to provide a signal for apoptosis required during cavitation of the egg cylinder which occurs prior to gastrulation (Coucouvanis et al., 1995). Furthermore, other pre-gastrulation functions of the VE have been suggested by mutations in the Hβ 58 and even-skipped (evx-1) genes. Disruption of Hβ 58 by a transgene insertion causes defects in growth of the embryonic ectoderm during a time when the highest level of Hβ 58 expression is found in the extraembryonic VE (Lee et al., 1992). Mouse embryos lacking evx-1, whose expression at early times is found in the VE, fail to form extraembryonic endoderm and are resorbed before gastrulation begins (Spyropoulos et al., 1994). Further insight into the putative roles of the VE during early development has been provided by analyses of parthenogenetic embryos. Early reports showed that embryos derived entirely from maternal genomes could develop to mid-gestation stages although yolk sac development was severely impaired (Kaufman et al., 1977; Barton et al., 1984; McGrath et al., 1984; Surani et al., 1984). However, a recent detailed morphological analysis of parthenogenetic embryos during early developmental stages found that they fell into four categories exhibiting increasingly severe abnormalities (Sturm et al., 1994). While the least affected parthegenones developed to mid-gestation periods, as described previously, the remaining embryos all exhibited gross defects in the production of mesoderm and in axial patterning (Sturm et al., 1994). Furthermore, in the most severely affected parthegenones, VE was grossly abnormal or completely absent while in the less affected the VE, although not normal, was present and capable of forming a yolk sac (Sturm et al., 1994). These observations are therefore consistant with a critical role for the VE during murine gastrulation.

In sum, our data establish that Hnf-4^{− /−} ectodermal cells are competent to form postgastrulation embryos and that gastrulation requires an HNF-4+ VE. Since many of the genes whose expression is down-regulated in the absence of HNF-4 are serum factors, we propose that, during early stages of postimplantation development in the mouse, VE paracrine activity is critical for de?ning and maintaining an embryonic environment which will support gastrulation of the epiblast. Whether any of the serum factors we have shown to be down-regulated in the absence of HNF-4 are the cause of the Hnf-4^{− /−} phenotype is under investigation. However, since gene targeting studies have shown that many are dispensable for gastrulation (Williamson et al., 1992; Episkopou et al., 1993; Farese et al., 1995; Huang et al., 1995), we believe it more likely that the effect is cumulative and the result of a gross serum de?ciency. In this regard it is of interest to note that growth of PC13 cells, a murine embryonal carcinoma cell line, can be maintained in serum-free media if the media are supplemented with transferrin (TFN), high density-lipoprotein (HDL), and low-density lipoprotein (LDL) (Heath et al., 1983). We have shown that expression of TFN, as well as that of Apo-AI and Apo-B which are major components of HDLs and LDLs respectively, are almost undetectable in the absence of HNF-4.

Recent years have seen rapid advances in the application of targeted genetics to the study of mouse development and gene function. Application of the tetraploid aggregation technique has enabled us to bring a new dimension to our investigation of the early embryonic lethality seen in Hnf-4^{− /−} mice. By complementing the lethal VE defect, the competence of Hnf-4^{− /−} ES cell-derived embryos to progress beyond gastrulation was uncovered. This complementation opens the way for dissection of roles that HNF-4 may play in later stages of organogenesis.

We gratefully thank P. Traktman, R. F. Bachvarova, C. Horvath, and M. Stoffel for help and discussions; J. E. Darnell Jr for continued support and encouragement; S. Cereghini for advice on vHNF-1 and HNF-1 probes and the Rockefeller University transgenic facility for helping S. A. D. set up the tetraploid aggregation system. S.A.D. is a Naomi Judd American Liver Scholar and a recipient of an Alexandrine and Alexander Sinsheimer Scholar Award.