Gata3 regulates trophoblast development downstream of Tead4 and in parallel to Cdx2

The first developmental decisions during mouse development lead to the establishment of the embryonic and extraembryonic tissue lineages. Stem cell lines have been isolated from these early lineages, including embryonic stem (ES) and trophoblast stem (TS) cells (Evans and Kaufman, 1981; Martin, 1981; Tanaka et al., 1998). Both stem cell types are self-renewing and capable of lineage-appropriate differentiation. For example, ES cells can differentiate into a wide range of fetal cell types, but fail to form trophoblast (Beddington and Robertson, 1989). Conversely, TS cells differentiate along the trophoblast/placenta lineage, and fail to form fetal cell types (Tanaka et al., 1998). To create a placenta, the trophoblast lineage must achieve several distinct goals simultaneously at the blastocyst stage. Trophoblast cells must override the pluripotency program of the embryonic lineage, they must establish the ability to self-renew, and they must maintain the ability to differentiate into mature trophoblast cell types. Cdx2 and Eomes are required for trophoblast survival and maturation starting around the blastocyst stage (Russ et al., 2000; Strumpf et al., 2005). These genes are also important for TS cell establishment (Strumpf et al., 2005), suggesting roles in proliferation. However, not all cells of the trophectoderm are proliferative, as some trophectoderm cells visibly differentiate as early as implantation. This suggests that programs that promote proliferation and differentiation might coexist at the blastocyst stage.

Besides Cdx2 and Eomes, genetic evidence suggests that other genes participate in trophoblast formation in the blastocyst. For example, loss of Tead4, which is required for expression of Cdx2 in the trophectoderm (Yagi et al., 2007; Nishioka et al., 2008), leads to a more severe phenotype than loss of Cdx2. Thus, Tead4 must have multiple trophoblast targets acting at the blastocyst stage to regulate trophoblast development. Consistent with this proposal, constitutively active Tead4 is sufficient to induce trophoblast formation even in the absence of Cdx2 in ES cells (Nishioka et al., 2009). Other factors capable of overriding the pluripotency pathway and promoting trophoblast fate must therefore exist.

To identify new factors involved in early lineage decisions in the mouse, we used a bioinformatic strategy to compare expression profiles of stem cells from the blastocyst. Transcripts encoding the transcription factor Gata3 were specifically enriched in TS cells and in the trophoblast lineage, consistent with recent reports (Home et al., 2009; Ray et al., 2009). Although Gata3 expression is restricted to the trophectoderm at the blastocyst stage, we found that this expression does not depend on Cdx2. Rather, expression of Gata3, like that of Cdx2, depends on Tead4. We show that Oct4 (Pou5f1) is not initially involved in restricting expression of either Cdx2 or Gata3 to the trophectoderm, but Oct4 maintains repression of these genes in the epiblast. In ES cells, Gata3 is capable of overriding pluripotency and directing the expression of a multitude of Cdx2-independent trophoblast genes, whereas in TS cells Gata3 promotes differentiation.

For comparison of ES, TS and XEN cell expression profiles, MGU74v2A microarray.CEL files for ES (GSE3766), XEN (GSE2204) and TS (GSE3766) cells were downloaded from the GEO website (http://www.ncbi.nlm.nih.gov/projects/geo/) and processed using GCOS software (Affymetrix) with the ‘statistical method’ algorithm to generate signal intensities and absent/present calls. Log₂ ratios and fold-change calls for probe sets in TS and XEN cell samples versus ES cells were determined. Probe sets with more than a 2-fold difference in expression in TS cells (P<0.003) versus ES cells and with GO annotation as a transcription factor (GO term ID: 000635) were selected for hierarchical clustering. GO annotation was obtained from the Affymetrix annotation file for the gene chip. Hierarchical clustering was performed with Cluster 3.0 (de Hoon et al., 2004) using the ‘uncentered correlation similarity’ metric with average linkage, and resulting clusters were visualized with Java TreeView.

For comparison of Gata3 and Cdx2 activity in wild-type and Cdx2 null cells, Affymetrix MOE430 2.0 microarrays were performed on four independent Gata3-expressing lines, two Cdx2-expressing lines, and three Gata3-expressing Cdx2 null lines. Data for these and TS cells were batch processed using Expression Console software (Affymetrix) to normalize arrays. Probe intensities were calculated using the PLIER algorithm. To reduce sample complexity, redundant probes matching to a single gene were filtered for a representative probe that had the largest number of signals greater than three times the global median signal across all samples. Ratios of gene expression for Gata3-expressing and Cdx2-expressing cells were calculated relative to tamoxifen-treated R1 cells and were log₂ transformed. Ratios for Gata3-expressing Cdx2 null cells were calculated relative to tamoxifen-treated Cdx2 null cells. Expression ratios for TS cells and their differentiated samples were calculated relative to the median expression level of the probe set across all the TS samples. To facilitate mining of the data set, all expression and probe set annotation was placed into a custom database using MySQL and queried using command line scripts. Raw data are available at GEO with accession numbers GSE12985 and GSE12986.

The ES cell lines R1 and dKO23-5 (Niwa et al., 2005) were maintained on gelatin using standard culture conditions. To create stably transformed lines, 1-3×10⁷ ES cells were electroporated with 20-30 μg plasmid (pCAG-hCdx2ERT2-ires-puro^r, or pCAG-hGata3ERT2-ires-puro^r), and were then seeded on two 10-cm gelatinized plates. Cells were fed with medium to select for expression of the plasmid (ES medium + 1.2 μg/ml puromycin) and individual colonies expanded. To induce transgene activity, cells were treated with induction medium [TS medium + 1 μg/ml tamoxifen (Sigma)]. For microarray analysis, 2×10⁵ cells were seeded on gelatinized 35-mm wells in ES medium, and then switched to TS cell medium with tamoxifen the following day, and fed daily for 6 days. To derive TS-like cell lines, cells were treated with tamoxifen on mouse embryonic feeder layers for 6 days in TS cell medium, and then passaged onto fresh feeders in TS cell medium.

TS cell lines included TS3.5 and TS6.5, which were isolated from blastocyst and E6.5 embryos, respectively, and were maintained or differentiated as described (Tanaka et al., 1998), and TS_WT, which was isolated from ICR blastocysts. For overexpression of Gata3, TS_WT cells were electroporated with Gata3ER, followed by selection for plasmid expression (with 1.2 μg/ml puromycin in TS cell medium) and tamoxifen treatment (1 μg/ml) for 5 days.

RNA was harvested from plated cells using Trizol (Invitrogen). For real-time PCR analysis, cDNA was synthesized using the Quantitect Kit (Qiagen). Real-time PCR analyses were performed using SYBR Green and a LightCycler 480 (Roche). All reactions (12 μl) were performed in triplicate, with 100-200 ng cDNA and 300 nM primers (shown 5′ to 3′) per reaction: Hprt1, AAACAATGCAAACTTTGCTTTCC and GGTCCTTTTCACCAGCAAGCT; Gata3, GGGTTCGGATGTAAGTCGAG and CCACAGTGGGGTAGAGGTTG; Cdx2, AGACAAATACCGGGTGGTGTA and CCAGCTCACTTTTCCTCCTGA; Prl2c2, AGCCCCATGAGATGCAATAC and CATCCAAAATCATGGCTCCT; Bmp4, AGGAGGAGGAGGAAGAGCAG and ACTGGTCCCTGGGATGTTCT; Pdgfra, ACGTTCAAGACCAGCGAGTT and CGATCGTTTCTCCTGCCTTA; Ascl2, TTTTCGAGGACGCAATAAGC and CACTGCTGCAGGACTCCCTA; Eomes, GTGACAGAGACGGTGTGGAGG and AGAGGAGGCCGTTGGTCTGTGG; Elf5, TGCCTTTGAGCATCAGACAG and TACTGGTCGCAGCAGAATTG; Tead4, ACGGAGGAAGGCAAGATGTA and CTGGAGACCTGCTTCCTTGT. A standard primer efficiency curve for each primer pair was generated using TS3.5 cDNA. Levels of Hprt1 served as an internal reference for all reactions. Amplification of a single PCR product for each reaction was confirmed by melting curve analysis, and all primers spanned exon junctions. RNA samples used for microarray hybridization were collected using Trizol, and then further purified using the RNeasy Mini Kit (Qiagen). Samples were analyzed by Affymetrix mouse MOE 430 2.0 cDNA microarrays (The Center for Applied Genomics, Toronto, Canada).

For single-blastocyst qPCR, total RNA was extracted from individual blastocysts using the PicoPure RNA Isolation Kit (Arcturus Bioscience), and cDNA synthesized at 37°C for 2 hours using the high-capacity cDNA Archive Kit (Applied Biosystems). One eighth of each cDNA preparation was preamplified for 16 cycles (95°C for 15 seconds and 60°C for 4 minutes) using the TaqMan PreAmp Master Mix Kit (Applied Biosystems) and gene-specific primers. Products were then diluted 5-fold for PCR (Applied Biosystems) in 48.48 Dynamic Arrays on a BioMark System (Fluidigm). Threshold cycle (Ct) values were calculated using the system's software (BioMark Real-time PCR Analysis) and were normalized to Actb Ct values.

Preimplantation embryos were harvested, stained and examined by confocal microscopy as described previously (Ralston and Rossant, 2008). Primary antibodies included mouse anti-Cdx2 (1:200, Biogenex CDX2-88), rabbit anti-Cdx2 (1:200) (Chawengsaksophak et al., 1997), mouse anti-Gata3 (1:20, Santa Cruz H-48) and mouse anti-Oct4 (1:10, Santa Cruz C-10). Secondary antibodies included Alexa488- or Alexa546-conjugated goat anti-mouse, rabbit or rat IgG (Molecular Probes) and Cy3-conjugated anti-mouse, rabbit or rat IgG (Jackson). Secondary antibody-only controls were performed in parallel (not shown). Whole-mount embryo in situ hybridization was performed as described (Yamanaka et al., 2007).

Mouse strains used in this study included wild-type (ICR) mice and mice heterozygous for null alleles of Cdx2 (Cdx2^tm1Fbe) (Chawengsaksophak et al., 1997), Oct4 (Pou5f1^tm1Scho) (Kehler et al., 2004) and Tead4 (Tead4^tm1Hssk) (Nishioka et al., 2008). All mice were treated in accordance with institutional guidelines. For genotyping, blastocysts were individually recovered following confocal microscopy and lysed using the Extract-N-Amp Tissue PCR Kit (Sigma) in a total of 10 μl per embryo, of which 2 μl was used for 10 μl PCR genotyping reactions, with 5 μl PCR Red mix and 0.5 μl each 10 μM primer (Kehler et al., 2004; Strumpf et al., 2005; Nishioka et al., 2008).

To identify potential trophoblast-inducing factors, we compared microarray expression profiles of stem cells derived from the blastocyst lineages: ES, TS and extraembryonic endoderm stem (XEN) cells (Kunath et al., 2005) (see Materials and methods). Probe sets that were significantly increased in TS and XEN cells relative to ES cells (P<0.003) were retained, and analysis of corresponding gene ontology (GO) terms yielded 122 transcription factors specifically enriched in TS cells (see Fig. S1 in the supplementary material), as represented by 138 probe sets. Importantly, this list included genes known to be essential for early trophoblast development, such as Cdx2 (Strumpf et al., 2005), Eomes (Russ et al., 2000; Strumpf et al., 2005) and Tead4 (Yagi et al., 2007; Nishioka et al., 2008). We therefore reasoned that other transcription factors enriched in this list could likewise be important for trophoblast development.

We focused on the zinc-finger transcription factor Gata3, the expression of which was specifically highly enriched in TS cells. Consistent with the microarray data, quantitative RT-PCR (qPCR) analysis of Gata3 levels indicated a greater than 100-fold enrichment of Gata3 in TS cells compared with ES cells (Fig. 1A). Gata3 is known to be expressed in, and required for the function of, the trophoblast lineage at later stages of development of the placenta (Ma et al., 1997). However, its high level of expression in TS cells suggested a previously unrecognized role for Gata3 in regulating stem cells of the trophoblast lineage. Examination of differentiating TS cells revealed that Gata3 levels increase during TS cell differentiation (Fig. 1B). These observations suggested that Gata3 might promote TS cell differentiation, consistent with its requirement in directing the formation of giant cells in the placenta (Ma et al., 1997).

ES cells are normally restricted in developmental potential to embryonic fates, having lost or suppressed the ability to generate trophoblast cell types (Beddington and Robertson, 1989). Overexpression of key trophoblast factors has been shown to lead to an increase in trophoblast gene expression in ES cells within a 6-day time frame (Niwa et al., 2005; Lu et al., 2008; Ng et al., 2008; Nishioka et al., 2009; Nishiyama et al., 2009). Subsequent passage of these cells in TS cell medium can lead to the establishment of self-renewing TS-like cells in some cases (Niwa et al., 2005; Lu et al., 2008; Nishioka et al., 2009). We examined the ability of Gata3 to induce the formation of TS-like cells by overexpressing Gata3ER, which encodes a fusion between Gata3 and the ligand-binding domain of the estrogen receptor (ER). Gata3ER was activated by addition of tamoxifen and, under TS cell derivation conditions, TS-like colonies were detected among cultures within 6 days (Fig. 1C; 4/5 lines examined), but not in control ES cells grown under the same conditions (Fig. 1E). However, endoderm-like cells were also present in all Gata3 cultures (Fig. 1C′), and these were not present in ES cells overexpressing Cdx2 (Fig. 1D). Continued passage of Cdx2-overexpressing ES cells led to the establishment of TS-like cell lines (4/5 lines examined). TS-like colonies were continuously detected among Gata3-overexpressing cells. However, cultures were consistently heterogeneous, and the TS cell phenotype could not be enriched under the conditions examined (5/5 lines examined). Thus, although both genes appear capable of inducing trophoblast differentiation in ES cells, only Cdx2 produced stable TS cell lines when overexpressed in ES cells.

We next compared the ability of Gata3 and Cdx2 to induce trophoblast at the gene expression level, comparing global gene expression profiles of ES cells overexpressing either gene. To restrict our analysis to trophoblast-specific genes, we began by defining a set of ~1800 core trophoblast genes, using TS cells as a reference (Fig. 2A; see Table S1 in the supplementary material; see Materials and methods). We then used this set to filter data sets from Cdx2-expressing and Gata3-expressing ES cell lines. This led to lists of genes induced by Gata3 (449/1794 core trophoblast genes) or Cdx2 (326/1794 core trophoblast genes) (see Table S2 in the supplementary material). Gata3 was therefore capable of inducing more trophoblast genes than Cdx2.

To examine qualitative similarities and differences in trophoblast genes induced by the overexpression of these two genes, we examined the overlap between the two lists. This revealed trophoblast genes induced by Gata3 (225 genes) or Cdx2 (102 genes) alone, as well as common genes induced by either factor (224 genes) (Fig. 2B; see Table S2 in the supplementary material). Thus, although around half of the trophoblast genes induced by Gata3 were also induced by Cdx2, each factor also induced the expression of a unique set of trophoblast genes. These differences were validated by qPCR for a subset of the genes (see Fig. S2 in the supplementary material). This analysis suggested that Gata3 expression is induced by ectopic Cdx2, although Gata3 targets were not in turn detected in the Cdx2-overexpression array. This disparity could be due to differences in the levels of overexpressed Gata3 in these two conditions. Alternatively, Cdx2 might directly or indirectly repress the expression of a subset of Gata3 target genes. This possibility is further addressed below.

These observations predict that Gata3 and Cdx2 will have both shared and distinct roles during trophoblast development. Among the genes induced by either factor, an examination of phenotypes for those that have been knocked out (Fig. 2C) revealed defects in multiple trophoblast subtypes and at multiple developmental stages (see Table S3 in the supplementary material). No single trophoblast phenotype was predominant in any of the lists, suggesting that Cdx2 and Gata3 targets are likely to play diverse, and possibly overlapping, roles in trophoblast development.

This analysis, which was designed to focus on the trophoblast roles of these genes, excluded genes that were not included among the core trophoblast gene list. We noted 347 ‘non-trophoblast’ genes induced by Gata3 and 72 induced by Cdx2 (see Table S4A,B in the supplementary material). Interestingly, among the genes induced by Gata3 were many known endodermal genes, including Foxa2, Sox17 and Sox7. Thus, both Cdx2 and Gata3 are capable of inducing non-trophoblast targets in ES cells, consistent with the diverse developmental roles played by these genes and the plasticity of ES cells to respond to inductive cues.

In ES cells, downregulation of Oct4 leads to upregulation of Cdx2 and the adoption of trophoblast fate (Niwa et al., 2000), raising the possibility that Gata3 overexpression could induce trophoblast gene expression by simply altering Oct4/Cdx2 levels. To address this possibility, we examined trophoblast gene expression following Gata3 overexpression in the dKO23-5 ES cell line that is Cdx2 null and expresses Oct4 constitutively (Niwa et al., 2005). Gata3 overexpression in dKO23-5 cells led to changes in cell morphology, and TS cell lines could not be established in this genetic background, as expected (5/5 lines examined). A microarray comparison of differences in the induction of trophoblast genes following overexpression of Gata3 in wild-type and dKO23-5 ES cells revealed that Gata3 was still able to induce a large number of core trophoblast genes (284/1794, compared with 449/1794 in wild-type cells). Gata3 is therefore sufficient to induce trophoblast gene expression in a Cdx2-independent manner.

However, the expression of many trophoblast genes was lost in this genetic background. Examining the intersection between the lists of core trophoblast genes induced by Gata3 in either wild-type or dKO23-5 cells (Fig. 3A) revealed that the expression of 172/449 Gata3 targets was unchanged, whereas 277/449 targets were no longer induced by Gata3 in dKO23-5 cells. Therefore, the expression of some Gata3 targets relied on the Oct4/Cdx2 pathway, whereas the expression of others, such as Eomes and Ascl2, did not (Fig. 3B). The genes that were dependent on Cdx2, however, did not necessarily overlap with those induced by Cdx2, suggesting differences in the necessity and sufficiency of Cdx2 for trophoblast gene expression.

This analysis also identified 112 genes that were induced by Gata3 in dKO23-5 and not wild-type ES cells (Fig. 3A; see Table S5 in the supplementary material). This suggests that Cdx2 might repress the Gata3-mediated induction of some trophoblast genes. Taken together, these observations suggest that Gata3 can act via Cdx2, and in parallel to Cdx2, to induce trophoblast gene expression. In addition, these observations suggest that Gata3 might play a unique role in regulating trophoblast development independently of Cdx2.

The findings that Gata3 is enriched in TS cells and is sufficient to induce trophoblast gene expression in ES cells suggested that Gata3 might be expressed in the trophoblast during lineage establishment in vivo. We examined the expression of Gata3 during trophoblast development at preimplantation stages (Fig. 4A-E). Gata3 protein was detectable within the nuclei of the trophectoderm at the blastocyst stage, where it colocalized with Cdx2 (Fig. 4D) (n=10). In fact, Gata3 colocalized with Cdx2 at earlier preimplantation stages as well (Fig. 4A-C) (n=31 embryos, 8- to 32-cell stages). Prior to becoming restricted to outside cells of the nascent trophectoderm, Cdx2 is expressed in an unpatterned, mosaic manner beginning around the late 8-cell stage (Dietrich and Hiiragi, 2007; Ralston and Rossant, 2008). Gata3 colocalized with Cdx2 in nuclei on a cell-by-cell basis (723/730 cells) in embryos examined at the 8- to 32-cell stages (31 embryos). Among embryos in which Gata3 and Cdx2 expression did not perfectly correlate (5/31 embryos), Gata3-positive/Cdx2-negative and Cdx2-positive/Gata3-negative nuclei were detected at equivalent frequency (four and three nuclei, respectively). Thus, Gata3 is coexpressed with Cdx2 from the earliest developmental stages.

Since TS cells have also been derived from post-implantation embryos, around the time of gastrulation (Tanaka et al., 1998; Uy et al., 2002), we next examined Gata3 expression around gastrula stages by in situ hybridization. Gata3 expression was detected throughout the trophoblast lineage from embryonic day (E) 6.5 to 8.5 (Fig. 4E) (n=12). This expression was consistent with previous reports (George et al., 1994) and included the extraembryonic ectoderm (EXE)/chorion and ectoplacental cone (EPC). However, we noted that higher levels of Gata3 were detected in the EPC than in the EXE. By contrast, Cdx2 levels appeared higher in the EXE than in the EPC (Fig. 4F). These differences were confirmed by qPCR, following microdissection of these regions (Fig. 4G). Since the EPC is thought to be more differentiated than the EXE, these observations suggest that Gata3 levels increase during trophoblast differentiation.

Notably, Gata3 and Cdx2 were also detected within the embryo proper around the gastrula stage and later, with Gata3 in a restricted anterior region (Fig. 4E and data not shown), consistent with previous reports (Manaia et al., 2000), and Cdx2 in posterior regions (Fig. 4F and data not shown) (Beck et al., 1995). This pattern is consistent with the proposal that both genes can also induce non-trophoblast targets in ES cells.

Increasing levels of Gata3 during trophoblast differentiation, both in TS cells and in the post-implantation embryo, suggested that Gata3 promotes differentiation. We therefore examined whether Gata3 is sufficient to induce differentiation in TS cells. We introduced the Gata3ER fusion construct into a TS cell line and examined changes in morphology and gene expression following treatment with tamoxifen for 5 days. As in previous experiments, cells expressing the Gata3ER fusion protein were selected by drug resistance. Control TS cells treated with tamoxifen maintained a generally undifferentiated state (Fig. 5A). However, Gata3-overexpressing TS cells appeared largely differentiated, with numerous giant cells present throughout the culture (Fig. 5B), despite the continued presence of TS cell medium. Plasmid-electroporated cells treated with lower doses of tamoxifen did not appear differentiated, arguing that the differentiation effect was Gata3 dependent. We therefore conclude that Gata3 overexpression is sufficient to induce differentiation of TS cells. This proposal was confirmed by an examination of TS cell and giant cell markers by qPCR (Fig. 5C).

Coexpression of Cdx2 and Gata3 at the blastocyst stage led us to investigate whether these genes are regulated by a common mechanism in vivo. The transcription factor Tead4 is required for Cdx2 expression prior to the blastocyst stage (Yagi et al., 2007; Nishioka et al., 2008). In Tead4 mutants, Cdx2 is initially detected around the 16-cell stage (Nishioka et al., 2008), but this expression is lost and embryos die prior to blastocyst formation around the 32-cell stage (Yagi et al., 2007; Nishioka et al., 2008). We hypothesized that Tead4 could play a role in the regulation of Gata3 prior to blastocyst formation, and examined expression of Gata3 in Tead4 mutants at E3.5. Nuclear levels of Gata3 were greatly reduced in Tead4 mutants (n=5) compared with non-mutants (n=17) (Fig. 6A,B). Similar to Cdx2, however, low levels of Gata3 could be detected in the nuclei of some cells (not shown), consistent with Tead4 regulating the maintenance, rather than initiation, of Gata3 expression.

Cdx2 expression is also lost in Tead4 mutants (Yagi et al., 2007; Nishioka et al., 2008), suggesting that Cdx2 could be required for Gata3 expression during preimplantation. We therefore examined the requirement for Cdx2 in Gata3 expression in the trophectoderm by examining Gata3 expression in Cdx2 null embryos at E3.5. By confocal analysis, Gata3 expression was unaffected by loss of Cdx2 (Fig. 6C,D) (n=6) at the blastocyst stage. This was validated at the mRNA level by qPCR (see Fig. S3 in the supplementary material). We conclude that Cdx2 is not required for the expression of Gata3 during trophectoderm formation, consistent with the similar timing of their expression at earlier stages. Rather, Gata3 appears to be regulated by Tead4 in parallel to Cdx2 during blastocyst formation.

We have shown that, like Cdx2, Gata3 is initially expressed throughout the preimplantation embryo, suggesting that both genes become patterned by a process of repression within inside cells during blastocyst formation. Oct4 is required for repression of Cdx2 in ES cells (Niwa et al., 2000), consistent with a possible role for Oct4 in repressing trophectoderm fates in the embryonic lineage. However, whether Oct4 is required for repression of Cdx2 in the inner cell mass during blastocyst formation has not been examined. Likewise, the role of Oct4 in regulating Gata3 expression during blastocyst formation remains unknown.

To examine the requirement for Oct4 in repressing Cdx2 and Gata3 in vivo, we examined the expression of these markers in embryos lacking zygotic Oct4 (Kehler et al., 2004). At the blastocyst stage (E3.5), Oct4 protein was undetectable in Oct4 mutants (Fig. 7B) (n=3), whereas Oct4 was detected throughout the blastocyst at this stage in non-mutants (Fig. 7A). However, Cdx2 and Gata3 expression patterns were largely unaffected in Oct4 mutants (n=4/5 and n=2/2, respectively) (Fig. 7C). Although weak Cdx2 expression was detected in the inner cell mass of one Oct4 mutant embryo, Cdx2 expression levels in the trophectoderm of this mutant embryo were also weaker than in non-mutant littermates (not shown). This pattern is normally observed in early blastocysts (Ralston and Rossant, 2008), suggesting that Oct4 mutants can exhibit a slight developmental delay relative to non-mutant littermates. Indeed, this proposal is consistent with the previous observation that the trophectoderm marker keratin 8 (Krt8, detected by TROMA1 antibody) is detected in the inner cell mass of some, but not all, Oct4 null embryos (Nichols et al., 1998). Since Krt8 is also expressed in the inner cell mass of early blastocysts (Ralston and Rossant, 2008), a developmental delay of Oct4 mutants could explain this phenotype. Nonetheless, the majority of Oct4 mutants exhibited the normal trophoblast-restricted expression of Cdx2 at the blastocyst stage. These results therefore suggest that zygotic Oct4is not required for initial repression of trophectoderm genes Cdx2 or Gata3, indicating that other mechanisms lead to the restriction of both of these genes to the trophectoderm.

To examine whether Oct4 is required for maintaining restricted expression of trophectoderm genes at later developmental stages, we attempted to examine the expression of Cdx2 and Gata3 in Oct4 mutants after implantation, at ~E4.25. At this stage, Oct4 mutants were extremely rare (1 mutant/27 non-mutant embryos), consistent with a requirement for Oct4 for embryo survival. However, in a rare mutant recovered at this stage, Cdx2 and Gata3 were clearly upregulated in cells occupying epiblast and primitive endoderm territories (Fig. 7E). By contrast, both Cdx2 and Gata3 were always excluded from nuclei in epiblast and primitive endoderm populations in non-mutants (Fig. 7D) (n=26). Thus, Oct4 is required for continued repression of Cdx2 and Gata3 in the late blastocyst, similar to its proposed role in repressing trophoblast fate in ES cells (Niwa et al., 2005). These observations indicate that the establishment of trophoblast and embryonic lineages proceeds by a mechanism that is distinct from the program that regulates this lineage restriction in established ES cells or the epiblast.

Here, we have used a combination of bioinformatic and functional genomic approaches to address fundamental questions about the first lineage restriction in the mouse. Specifically, what other factors act downstream of Tead4, are these sufficient or necessary to induce trophoblast fate, and are trophoblast factors themselves regulated in the embryo through mechanisms similar to those used in ES cells? Through genetic analyses performed in stem cells and in the mouse embryo, we provide evidence that Gata3 acts downstream of Tead4 and in parallel to Cdx2. A fundamental challenge in the field of stem cell biology is the paucity of truly trophoblast-specific markers. To overcome this challenge, we used TS cells as a reference tissue to define a set of core trophoblast genes. This enabled a deeper molecular comparison of trophoblast phenotypes resulting from the overexpression of Gata3 or Cdx2, and provides a reference for future studies of this type.

We have shown that Gata3 is sufficient to induce trophoblast genes in ES cells, consistent with another study (Nishiyama et al., 2009). Our analysis, however, revealed differences between Gata3 and Cdx2. First, although expression of Gata3 can induce trophoblast differentiation in ES cells, stable TS cell lines could not be maintained, unlike the situation with Cdx2. Rather, Gata3 appears to act as a pro-differentiation factor in TS cells. Second, unlike Cdx2, Gata3 is probably not required for the early lineage decision in the embryo. Whereas shRNA-mediated knockdown of Gata3 leads to developmental delay during the morula-to-blastocyst transition (Home et al., 2009), Gata3 null embryos survive until E10.5, whereupon they exhibit defects in the placenta and numerous fetal tissues (Ma et al., 1997). Thus, Gata3 is both necessary and sufficient to promote trophoblast maturation, but is not sufficient to stabilize the stem cell state. Other studies have shown that other factors, including Eomes, Elf5 and activated Ras, can also destabilize the pluripotent state of ES cells and drive trophoblast differentiation (Niwa et al., 2005; Lu et al., 2008; Ng et al., 2008; Nishiyama et al., 2009). Together, these observations suggest that there are multiple pathways capable of overriding the pluripotency program to induce trophoblast fate in ES cells.

Another intriguing difference between Cdx2 and Gata3 lies in their expression patterns at later stages of trophoblast development. Whereas Cdx2 and Gata3 were coexpressed in the EXE, Gata3 was expressed at much higher levels within the EPC around the time of gastrulation. These observations suggest where Cdx2/Gata3 targets might be expressed. For instance, common trophoblast targets would be expected to be expressed in the EXE. Consistent with this proposal, many genes involved in EXE development, such as Fgfr2, Wnt7b and Bmp4 (Orr-Urtreger et al., 1993; Coucouvanis and Martin, 1999; Kemp et al., 2005), were induced by either Cdx2 or Gata3 overexpression in ES cells. In addition, EXE genes such as Eomes and Ascl2 (Guillemot et al., 1994; Ciruna and Rossant, 1999; Russ et al., 2000) were induced by Gata3 even in the absence of Cdx2, suggesting that Gata3 can reinforce trophoblast fate through a Cdx2-independent mechanism. Intriguingly, Gata3 was expressed at higher levels in the EPC than in the EXE, whereas Cdx2 was not. This provides potential biological relevance for the set of trophoblast genes that were induced by Gata3 only in the absence of Cdx2. Genes in this list included Pparg and Dnmt3l, loss of which lead to defects in trophoblast differentiation (Barak et al., 1999; Bourc'his et al., 2001). Thus, Gata3 may promote a program of trophoblast differentiation in the EPC where Cdx2 expression is low or lacking. When overexpressed in TS cells, Gata3 induced differentiation of the cells towards more differentiated cell fates, consistent with this role. In this way, Gata3 could play a dual role, either promoting stem cell (EXE) fates or differentiation (EPC/giant cell fates) depending on the presence of other factors such as Cdx2. This proposal is consistent with evidence that Gata3 is required for self-renewal of TS cells (Home et al., 2009). Moreover, Gata2 has been proposed to promote self-renewal versus differentiation of hematopoietic progenitor cells in a level-dependent manner (Heyworth et al., 1999), arguing that Gata3 might play a similar role in the trophoblast lineage.

We have also examined whether trophoblast factors are regulated in the embryo through mechanisms similar to those used in ES cells. In ES cells, Oct4 normally represses trophoblast fate (Niwa et al., 2000). However, it has not been clear whether this relationship applies to the embryo. Since Cdx2 and Gata3 are initially expressed in both inside and outside cells, Oct4 could repress the expression of these factors in inside cells during blastocyst formation. However, we show that Oct4 is not involved in the repression of trophoblast fate in the embryo until around the time of implantation. Indeed, trophoblast cells can coexpress Cdx2 and Oct4 in a variety of contexts (Niwa et al., 2005; Strumpf et al., 2005; Lu et al., 2008; Ng et al., 2008), arguing that Oct4 cannot be providing the initial patterning information along the inside/outside axis of the embryo. Rather, it was recently shown that the absence of Hippo signaling promotes Cdx2 expression in outside cells during blastocyst formation (Nishioka et al., 2009). Thus, the maintenance of ES cell fate might reflect molecular interactions that are relevant to stages of development following the initial lineage decisions. This proposal could help to explain why Cdx2 is not required for Gata3-mediated induction of Eomes in ES cells, even though Cdx2 is initially required for expression of Eomes in the blastocyst (Ralston and Rossant, 2008). Studies conducted in ES cells may therefore be viewed as reflecting a lineage maintenance, rather than establishment, program.

Finally, our study suggests that culture conditions influence cell fate changes induced by transcription factor overexpression in ES cells. Cdx2, and not Gata3, was sufficient to induce the formation of TS-like cells from ES cells. However, the isolation of stable TS-like cells is only possible following continued passage in TS cell medium after an initial period of transient transcription factor overexpression, a process that takes at least 6 days. This is reminiscent of the process involved in reprogramming mature cell types to pluripotency (Takahashi and Yamanaka, 2006), and suggests that similar mechanisms might be involved in the generation of stable TS cell lines. Given that many non-trophoblast genes were induced by the overexpression of either Cdx2 or Gata3, altering culture and selection conditions could therefore lead to an enrichment of different cell fate outcomes. For instance, the use of different cell culture medium, growth factors or small molecules might enable the enrichment of TS-like cells or endoderm from Gata3-expressing ES cells. Given that multiple pathways can override the pluripotent state, the manipulation of intrinsic and extrinsic factors could facilitate the selection of other lineage-specific stem cell types during this process.

We thank Owen Tamplin for in situ help; Jorge Cabezas, Colm Rea and Andres Nieto for animal help; Malgosia Kownacka, Jodi Garner and Alison Hirukawa for cell culture help; Dr Yojiro Yamanaka, Dr Michael A. Halbisen and members of the J.R. laboratory for helpful suggestions; Dr Alexey Tomilin for the Oct4 mutant mice; and Brian De Veale, the laboratory of Dr Derek van der Kooy and the Toronto Centre for Phenogenomics for help with mouse import. P.R.-G. was funded by a Canadian Institutes of Health Research Bisby Fellowship. J.S.D. was funded by the Biotechnology and Biological Sciences Research Council, UK. This work was funded by Grant #FRN13426 from the Canadian Institutes of Health Research.