Regulation of extra-embryonic endoderm stem cell differentiation by Nodal and Cripto signaling

During mammalian preimplantation development, an initial cell fate decision occurs during morula compaction, when outer cells become specified to form the trophectoderm, which will give rise to the trophoblast and extra-embryonic ectoderm, whereas inner cells will generate the pluripotent inner cell mass of the blastocyst (reviewed by Arnold and Robertson, 2009; Rossant and Tam, 2009; Zernicka-Goetz et al., 2009). A second cell fate restriction occurs prior to implantation, when the inner cell mass gives rise to the epiblast (primitive ectoderm) and the primitive endoderm. Following implantation, the primitive endoderm generates the visceral endoderm that surrounds the epiblast and extra-embryonic ectoderm, as well as the parietal endoderm that interacts with trophoblast cells.

Many studies have now demonstrated the crucial role of the visceral endoderm in epiblast patterning and differentiation. In particular, a central role in anterior-posterior (A-P) axis specification is played by the anterior visceral endoderm (AVE), a specialized population of cells within the visceral endoderm that overlies and patterns the adjacent epiblast at pregastrulation and early gastrulation stages (Arnold and Robertson, 2009; Rossant and Tam, 2009). In early postimplantation stages, at ∼5.0 days post-coitum (dpc), the visceral endoderm overlying the epiblast (the embryonic visceral endoderm, EmVE) becomes morphologically and molecularly distinct from the visceral endoderm overlying the extra-embryonic ectoderm (the extra-embryonic visceral endoderm, ExVE) (Mesnard et al., 2006). Following the differentiation of EmVE, the distal visceral endoderm (DVE) forms in the most distal portion of the postimplantation egg cylinder at 5.0-5.25 dpc, and then translocates to the prospective anterior side through active directional cell migration by 6.0 dpc (Migeotte et al., 2010; Rivera-Perez et al., 2003; Srinivas et al., 2004; Stuckey et al., 2011; Takaoka et al., 2011; Thomas et al., 1998; Torres-Padilla et al., 2007; Trichas et al., 2011).

The proper differentiation of the EmVE as well as the formation and movement of the DVE require Nodal, a transforming growth factor beta (TGFβ) ligand that plays a central role in early vertebrate embryogenesis (Schier, 2009; Shen, 2007). Notably, the activity of Nodal requires epidermal growth factor-Cripto/FRL-1/Cryptic (EGF-CFC) co-receptors such as Cripto and Cryptic (Tdgf1 and Cfc1 – Mouse Genome Informatics) (Reissmann et al., 2001; Yan et al., 2002; Yeo and Whitman, 2001). In particular, membrane-bound Cripto appears to recruit Nodal to an activin receptor complex composed of a dimer of the type I serine-threonine receptor Alk4 (Acvr1b) and a dimeric type II activin receptor, either ActRII or ActRIIB (Acvr2a or Acvr2b – Mouse Genome Informatics). Following receptor activation, Smad2 and/or Smad3 are phosphorylated and accumulate together with Smad4 in the nucleus to mediate transcriptional responses (Yan et al., 2002; Yeo and Whitman, 2001). The activated Smad2-Smad4 complex interacts with nuclear transcription factors that include the FoxH1 (Fast2) winged-helix transcription factor (Chen et al., 1996; Chen et al., 1997). In the absence of EGF-CFC proteins, Nodal lacks signaling activity through activin receptors, in contrast to activin, which does not seem to require a co-receptor for its signaling activity (Gritsman et al., 1999; Kumar et al., 2001; Yan et al., 2002). Interestingly, soluble Cripto protein can also display signaling activity in cell culture and in vivo, indicating that EGF-CFC proteins can display trans-acting activities (Chu et al., 2005; Yan et al., 2002).

Numerous studies have shown that Nodal and its EGF-CFC co-receptors are required for multiple aspects of EmVE differentiation and formation of the DVE. In the absence of Nodal, the EmVE retains an extra-embryonic morphology and pattern of marker expression, indicating a role for Nodal in the differentiation of the EmVE (Mesnard et al., 2006). Moreover, Nodal signaling is essential for DVE specification, as no evidence of a DVE or a molecular or morphological A-P axis is apparent in Nodal null mutants or in Cripto; Cryptic double mutants (Brennan et al., 2001; Chu and Shen, 2010; Norris et al., 2002).

In principle, the analysis of primitive endoderm formation and its subsequent differentiation can be facilitated by the isolation of stem cell lines with primitive endoderm characteristics. Such XEN cell lines can be isolated from the mouse primitive endoderm and display many of the expected morphological and molecular properties of primitive endoderm cells (Artus et al., 2010; Brown et al., 2010; Kunath et al., 2005). Curiously, however, these XEN cells can only contribute efficiently to parietal endoderm in chimeras and only very rarely can contribute to visceral endoderm (Kunath et al., 2005).

We have investigated the possibility that Nodal signaling regulates the differentiation of visceral endoderm and AVE from XEN cells. We show that treatment of XEN cells with recombinant Nodal or Cripto proteins leads to visceral endoderm and AVE differentiation in culture and to their contribution to these tissues in chimeric embryos. Unexpectedly, the effects of Cripto treatment are distinct from those of Nodal as they are not inhibited by treatment with the Alk4/Alk5/Alk7 kinase inhibitor SB431542. In combination with bioinformatic analyses of global gene expression patterns, we conclude that Nodal and Cripto act through distinct signaling pathways to mediate visceral endoderm differentiation by XEN cells.

Most of the experiments shown were performed using XEN cells from an enhanced yellow fluorescent protein (YFP)-expressing cell line (passage 12), which was generously provided by Janet Rossant (Hospital for Sick Children Research Institute, Toronto, Canada). We have also derived six additional wild-type XEN cell lines: three from GFP-expressing Tg(CAG-EGFP)B5Nagy/J mice (Hadjantonakis et al., 1998) and three from Swiss-Webster mice. All experiments with the XEN-YFP line, except for bioinformatic analyses, have been replicated using at least one of the newly derived wild-type cell lines. In addition, we derived one Cryptic homozygous mutant XEN cell line and three heterozygous lines from intercrosses of Cryptic heterozygous mutant mice (Yan et al., 1999).

XEN cells were derived and cultured as previously described (Kunath et al., 2005). Cryptic homozygous and heterozygous mutant XEN cells were derived in hanging drops in 30% XEN cell medium with 15% FCS/70% conditioned medium (CM) from mouse embryonic fibroblasts supplemented with 25 ng/ml FGF4 (R&D Systems) and 1 μg/ml heparin (Sigma). After 10 days of culture, the resulting cell aggregates were collected and cultured in four-well plates coated with 0.1% gelatin (Sigma). At day 20, cells were passaged and replated into standard XEN cell medium to establish lines.

Differentiation of XEN cells was performed by addition of recombinant Nodal (50 ng/ml) or Cripto (100 ng/ml) (R&D Systems) to XEN cell cultures for a total of 4 days. Alternatively, we used CM from Nodal- or Cripto-overexpressing HEK 293T cell lines, which were generated as described (Yan et al., 2002). In some experiments, the Alk4/Alk5/Alk7 kinase inhibitor SB435142 (1 μM, Sigma), or DMSO (Sigma) as a control, was also added.

XEN cells were transfected with Lipofectamine LTX (Invitrogen) according to the manufacturer's instructions. Luciferase assays for Nodal signaling activity were performed as previously described (Yan et al., 2002). Relative luciferase activities represent the average of the results of at least three independent experiments performed in duplicate.

Total RNA was isolated from XEN cells using the RNeasy mini RNA isolation kit (Qiagen). First-strand cDNA was prepared by random priming of Superscript III reverse transcriptase (Invitrogen) and analyzed by RT-PCR or quantitative real-time PCR (see Table S1 in the supplementary material) using a LightCycler instrument (Roche) and analyzed using LinRegPCR software (Ramakers et al., 2003).

For western blot experiments, 2×10⁵ XEN cells were seeded 1 day prior to treatment with recombinant Nodal (50 ng/ml), Cripto (100 ng/ml), SB435142 (1 μM) and/or DMSO for 1 hour. Western blotting was performed with phospho-Smad2 (Ser465/467; 3108, Cell Signaling; 1:1000), Smad2 (511300, Invitrogen; 1:200) and β-actin (sc-47778, Santa Cruz; 1:500) antibodies as described (Yan et al., 2002).

Chimeras were generated by injection of XEN cells into 3.5-dpc blastocysts, or by morula aggregation with 2.5-dpc embryos from ICR mice (Nagy et al., 2003). For blastocyst injection, 10-15 XEN cells were injected per embryo, followed by transfer into the uterus of 2.5-dpc pseudopregnant Swiss-Webster females. The resulting embryos were dissected at 6.5 dpc and analyzed by direct imaging or by whole-mount immunofluorescence using a Leica TCS SP5 confocal microscope. Immunofluorescent staining of whole-mount embryos was performed as described (Kwon et al., 2008) using Cerl primary antibody (R&D Systems; 1:300) and Alexa 555 secondary antibody (Invitrogen; 1:250). Immunofluorescent staining of cells was performed using Cerl (R&D Systems) or Afp (R&D Systems) primary antibodies, both at 1:600, and Alexa 555 secondary antibody (Invitrogen; 1:250).

For gene expression profiling, total RNA was isolated using the MagMAX-96 total RNA isolation kit (Ambion). Total RNA (200 ng) was reverse transcribed and biotin labeled by in vitro transcription using the Illumina TotalPrep RNA amplification kit (Ambion). Then 1.5 μg of the resulting cRNA was hybridized on mouse WG-6 v2 BeadArrays (Illumina). Hybridization data were obtained with an iScan BeadArray scanner (Illumina) and pre-processed by variance stabilization and robust spline normalization implemented in the lumi package under the R-System (Du et al., 2008). Expression data are available from the Gene Expression Omnibus database under GSE23675.

Prior to unsupervised cluster analysis, the expression data were normalized between slides to remove potential slide-dependent systematic bias. The signal of each probe k in array i of slide j, y_kji, was normalized by subtracting the difference between the intra- and inter-slide means:

where n and N are the number of arrays per slide and the total number of arrays, respectively. In this way, the normalized intra-slide mean of probe k is equal to the inter-slide mean. Hierarchical cluster analysis was performed with a complete linkage agglomerative algorithm (hclust function of R-System v2.8.0) using 1–r, where r is the Pearson's correlation coefficient, as distance metric. Principal components analysis (PCA) was performed with a singular value decomposition (SVD) algorithm (prcomp function of R-System v2.8.0).

For differential expression analysis, we did not apply any inter-slide normalization. We modeled the potential systematic bias by including a random factor for the experimental batch in a lineal model y_k=μ_k+α_k+β_k+ε_k, where α is the fixed treatment effect and β is the random experimental batch effect. We fitted this model to the data and estimated the treatment P-value using the limma package implemented in R-System (Smyth, 2004). Redundant probes were eliminated by keeping the one showing the highest dynamic range (coefficient of variation) among samples. False discovery rate (FDR) was estimated as described (Benjamini and Hochberg, 1995). Enrichment of Gene Ontology Biological Process (GO-BP) gene sets on Nodal and Cripto gene expression signatures were assessed by gene set enrichment analysis [GSEA (Subramanian et al., 2005)].

Statistical significance for the overlap between signatures was computed by Fisher's exact test and by a modified version of GSEA [two-tails GSEA (Lim et al., 2009)]. Briefly, we divided the query gene set into two subsets: a positive subset containing the upregulated part of the query signature, and a negative subset encompassing the downregulated part of the query signature. The target signature was then sorted from the most upregulated to the most downregulated gene (signature A) and the rank positions for the positive query subset were computed. The rank positions for the negative subset were computed from the target signature, but this time sorted from the most downregulated to the most upregulated gene (signature B). The enrichment score (ES) was computed as described (Subramanian et al., 2005), using the computed rank positions for the positive and negative subsets, but taking the score values only from signature A. The normalized enrichment score (NES) and nominal P-values were estimated by uniformly permuting the query signature ranks 10,000 times. Genes in common between two signatures were obtained from the union of the leading-edge gene sets (Subramanian et al., 2005).

To investigate the potential role of Nodal pathway activity in primitive endoderm, we examined the ability of XEN cells to differentiate into visceral endoderm following Nodal or Cripto treatment, using a YFP-expressing mouse XEN cell line. In our studies, we have found that treatment of these XEN-YFP cells with recombinant Nodal or Cripto protein over 4 days of culture leads to upregulation of the visceral endoderm marker alpha fetoprotein (Afp) (Fig. 1A-F,J). We obtained similar results using conditioned medium produced from Nodal- or Cripto-overexpressing HEK 293T cell lines (Yan et al., 2002) and have confirmed these observations using six different wild-type XEN cell lines produced in our laboratory (see Fig. S1A in the supplementary material; data not shown). Similarly, a significant percentage of treated XEN cells display upregulation of the AVE marker Cerberus-like (Cerl; Cer1 – Mouse Genome informatics) (Fig. 1G-J; see Fig. S1A in the supplementary material), as 17.8±1.8% of Nodal-treated and 17.7±1.1% of Cripto-treated cells displayed Cerl immunostaining. We also observed upregulation of E-cadherin, which is expressed by visceral endoderm and not parietal endoderm (see Fig. S2 in the supplementary material). By contrast, no effects were observed on the expression of the visceral endoderm marker Hnf4a or the primitive endoderm markers Sox7, Gata4 and Gata6; as expected, the epiblast marker Oct4 (Pou5f1 – Mouse Genome informatics) and the trophectoderm marker Cdx2 were not expressed (Fig. 1J; see Fig. S1 in the supplementary material). Quantitative real-time PCR analyses confirmed significant upregulation for both Afp and Cerl following treatment with Nodal or Cripto, whereas little or no effect was observed on the expression of Gata6 and Hnf4a (Fig. 1K). Despite these alterations in gene expression, we did not observe significant changes in the morphology of the Nodal- or Cripto-treated XEN cells after 4 days of culture (Fig. 1A-C).

Given the response of XEN cells to Nodal and Cripto treatment, we next examined the expression of genes encoding components of the Nodal signaling pathway. Previous studies have shown that this pathway can be activated by TGFβ ligands in addition to Nodal, including mammalian growth differentiation factor 1 (Gdf1) and Gdf3 (Andersson et al., 2007; Andersson et al., 2008; Andersson et al., 2006; Chen et al., 2006; Cheng et al., 2003). Although Gdf1 may act as a co-ligand for Nodal (Tanaka et al., 2007), Gdf3 appears to have intrinsic signaling activity, utilizing either the Alk4 or Alk7 (Acvr1c) type I receptors (Andersson et al., 2007; Andersson et al., 2008; Andersson et al., 2006; Chen et al., 2006; Cheng et al., 2003). We found that Alk4, Alk5 (Tgfbr1), ActRIIB and Foxh1 were expressed by XEN cells in all treatment conditions, without significant changes in their levels in response to Nodal or Cripto (see Fig. S1B in the supplementary material). By contrast, expression of Nodal, Gdf1, Gdf3 and Alk7 was not observed (see Fig. S1B in the supplementary material). Notably, XEN cells do not express the EGF-CFC co-receptor Cripto, but do express the related co-receptor Cryptic (see Fig. S1B in the supplementary material).

To determine whether the EGF-CFC co-receptor Cryptic mediates the response to Nodal in XEN cells, we derived a new XEN cell line, XEN-TIC1, which is homozygous for a Cryptic null mutation as derived from intercrosses of Cryptic heterozygous mutant mice. We examined the activity of Nodal, Cripto and Cryptic in this Cryptic mutant XEN cell line using a modification of a luciferase reporter assay that we previously developed to investigate Nodal signaling in HEK 293T cells (Yan et al., 2002). This assay for Nodal relies upon the co-expression of an EGF-CFC co-receptor and FoxH1 in transfected cells and the measurement of the activity of luciferase expressed by the activin/Nodal-responsive reporter A3-luc. Since XEN cells already express Foxh1, we measured luciferase activity following transfection of Nodal, Cripto and/or Cryptic expression vectors into the XEN-TIC1 cells. As expected, we found that expression of Nodal or Cryptic alone did not result in luciferase activity in the XEN-TIC1 cells, whereas Cripto alone stimulated luciferase activity 4.5-fold relative to the control (Fig. 1L). However, luciferase activity was strongly stimulated by co-transfection of Nodal with Cryptic, indicating that Nodal signaling remains EGF-CFC-dependent in XEN-TIC1 cells (Fig. 1L). These results were confirmed by analyses of AVE and visceral endoderm marker expression in the XEN-TIC1 cells (Fig. 1J). Thus, these findings indicate that the response to Nodal signaling in XEN cells is Cryptic dependent.

To determine the behavior of XEN cells in vivo, we generated chimeric embryos by blastocyst injection of XEN cells (Table 1), using either the original XEN-YFP line or one of our newly derived XEN lines that expresses GFP. In control chimeras generated by injection of untreated XEN cells, we observed contribution exclusively to parietal endoderm when analyzed at 6.5 dpc (n=18) (Fig. 2A), as reported previously (Kunath et al., 2005). However, consistent with our finding in cell culture, chimeras produced with either Nodal- or Cripto-treated XEN cells displayed a significant contribution to visceral endoderm (n=10 chimeras for Nodal treatment, n=16 for Cripto), with occasional contribution to parietal endoderm (Fig. 2B,C,G; see Fig. S3A,B in the supplementary material). We did not observe any obvious regional differences in the contribution of Nodal-treated or Cripto-treated XEN cells to the EmVE or ExVE (Fig. 2B,C,G; see Fig. S3A,B in the supplementary material).

We also generated chimeras by morula aggregation, obtaining integration of XEN cells (n=14/20) and Nodal-treated XEN cells into the blastocyst inner cell mass at 3.5 dpc (n=6/12 blastocysts) (Fig. 2D; data not shown). Similar to the chimeras generated by blastocyst injection, we observed contribution of Nodal-treated XEN cells to the visceral endoderm at 6.5 dpc (n=3) in morula aggregation chimeras (Fig. 2E,F; see Movie 1 in the supplementary material).

Finally, to determine whether treated XEN cells contribute to AVE in vivo, we examined the colocalization of GFP expressed by the XEN cells with Cerl, which marks the AVE at 6.5 dpc. Nodal-treated and Cripto-treated XEN cells injected into blastocysts resulted in chimeras containing cells that co-expressed Cerl and GFP, indicating the contribution of XEN cells to the AVE (Fig. 2G-J; see Fig. S3C-F in the supplementary material).

To investigate the transcriptional response to Nodal and Cripto treatment, we performed genome-wide gene expression profiling of XEN cells at 4 days after Nodal or Cripto treatment. Four biological replicates for each condition were hybridized on mouse WG-6 v2 Illumina BeadArrays. From the 20,311 profiled genes, 4693 and 4704 were differentially expressed after Nodal and Cripto treatments, respectively (FDR<0.01) (Table 2; see Tables S2 and S3 in the supplementary material). To examine the transcriptional response to these treatments, we used principal components analysis (PCA), a method that projects data variability onto a reduced number of orthogonal axes, such that the first axis captures the greatest degree of variance, and subsequent axes correspond to successively decreasing variance. Notably, this approach demonstrated a similar transcriptional response to both Nodal and Cripto treatments, as shown by the first principal component (PC-1), which captures 75.1% of the data variability (Fig. 3A). This conclusion is also supported by unsupervised hierarchical cluster analysis (see Fig. S4 in the supplementary material), statistical analyses (see Table S5 in the supplementary material), and heatmap analysis of the expression of selected markers of primitive, visceral and definitive endoderm, as well as TGFβ pathway components (see Fig. S5 in the supplementary material). Furthermore, there is a strong overlap between the Nodal gene expression signature and the Cripto signature (odds ratio=52.7, P<10^–16 by Fisher's exact test) (Fig. 3B).

We further inspected the similarity of the Nodal and Cripto gene expression signatures using gene set enrichment analysis (GSEA), which is an unsupervised methodology that is specifically designed for computing the overlap between gene expression signatures (Subramanian et al., 2005). GSEA is a useful analytical method because genes in a signature (e.g. those regulated by Cripto) might not have large fold changes, but do display correlated responses when a different treatment is applied (e.g. treatment with Nodal), indicating coordinated regulation. Indeed, GSEA is the most effective approach to test whether two processes produce similar effects in terms of differentially expressed genes (Subramanian et al., 2005). Using an enhanced version of this algorithm, termed two-tails GSEA, which takes into account the `direction' of the gene expression change (Lim et al., 2009), we found a strong enrichment of the Nodal signature on the Cripto gene expression profile (Fig. 3C) and vice versa (Fig. 3D). Consistent with these findings, we found only 64 genes that responded in opposite directions to Nodal and Cripto treatments, with just three genes displaying more than 2-fold change in gene expression in either signature (see Table S4 in the supplementary material). These reciprocal enrichment data strongly support the similar outcomes following Nodal or Cripto treatment of XEN cells and are highly consistent with our cell culture and chimera data in vivo.

To obtain insights into the biological processes modulated by Nodal and Cripto treatment, we determined the Gene Ontology-Biological Process (GO-BP) gene sets that were significantly enriched in the Nodal and Cripto gene expression signatures (see Table S6 in the supplementary material). We found that 54 GO-BP categories were enriched in the Nodal signature and 46 in the Cripto signature, with 31 categories enriched in both signatures.

Our gene expression analyses revealed that XEN cells do not express detectable levels of Nodal or of Gdf1 and Gdf3, which encode potential Nodal-like signaling factors or Nodal co-ligands (see Fig. S1B in the supplementary material). Consequently, it is unclear how Cripto treatment could affect the differentiation of XEN cells, as the known EGF-CFC-dependent ligands are not expressed. One possibility is that the addition of Cripto could mediate the activity of another, as yet unidentified, TGFβ ligand, which could then signal through the canonical Nodal/activin/TGFβ pathway. To investigate this possibility, we utilized the small molecule inhibitor SB431542, which blocks the kinase activity of the type I receptors Alk4, Alk5 and Alk7 (Inman et al., 2002), and thereby inhibits Nodal pathway signaling. As expected, we found that treatment with 1 μM SB431542 inhibits visceral endoderm differentiation stimulated by recombinant Nodal, as assessed by expression of Afp, Cerl and E-cadherin, but has little effect by itself (Fig. 4A,B,D,E,G,H,J,K; see Fig. S2 in the supplementary material). No significant effects on expression of Sox7, Hnf4a, Gata4 or Gata6 were observed in response to SB431542, whereas Oct4 and Cdx2 were not expressed (Fig. 4J,K). Furthermore, the expression levels of Foxh1, Cryptic and Alk4 were unaffected in all treatment conditions, whereas Nodal, Gdf1 and Cripto were not expressed (see Fig. S1B in the supplementary material).

Unexpectedly, however, SB431542 treatment did not inhibit visceral endoderm differentiation induced by Cripto (Fig. 4C,F,I; see Fig. S1 in the supplementary material). Analyses of marker gene expression confirmed that SB431542 blocks the effects of Nodal treatment, but not those of Cripto (Fig. 4J,K); in particular, 14.4±1.4% of cells treated with Cripto and SB431542 displayed Cerl immunostaining. Furthermore, in chimera analyses, we found that XEN cells treated with both Nodal and SB431542 only contribute to the parietal endoderm (n=7) (Fig. 4L,M), as expected for inhibition of Nodal pathway activity. By contrast, XEN cells treated with both Cripto and SB431542 maintained their ability to contribute to visceral endoderm and AVE (n=7) (Fig. 4N,O; see Fig. S3G-L in the supplementary material).

Next, we performed expression profiling of Nodal- and Cripto-treated XEN cells in the presence of SB431542, and compared the resulting expression signatures with those generated in the absence of SB431542. PCA revealed that SB431542 significantly impaired the Nodal gene expression signature (Fig. 5A, Table 2), consistent with the unsupervised hierarchical clustering analysis (see Fig. S4 in the supplementary material). In particular, two-tails GSEA showed no significant enrichment of the Nodal signature on the Nodal+SB431542 gene expression profile (Fig. 5B). Conversely, SB431542 did not substantially affect the Cripto gene expression signature, as shown by the unsupervised analysis (Fig. 5A; see Fig. S4 in the supplementary material). This conclusion is further supported by the significant overlap between the Cripto and Cripto+SB431542 gene expression signatures (odds ratio=78.7, P<10^–16 by Fisher's exact test) (Fig. 5C), by the strong enrichment of the Cripto gene expression signature on the Cripto+SB431542 gene expression profile (Fig. 5D), by the unsupervised hierarchical cluster analysis (see Fig. S4 in the supplementary material) and by the heatmap analysis of selected gene expression (see Fig. S5 in the supplementary material). Consistent with this conclusion, we found only 37 genes that responded in opposite directions to Cripto versus Cripto+SB431542, with just three genes displaying a greater than 2-fold enrichment in either signature (see Table S7 in the supplementary material). In combination with the cell culture and chimera analyses, these results indicate that the effects of Cripto on XEN cells are largely independent of Alk4/Alk5/Alk7 receptor kinase activity.

Based on the similarity of the functional response to Nodal and Cripto in XEN cells, we investigated whether they shared a similar downstream response through the phosphorylation of Smad2. Treatment of XEN cells with Nodal or Cripto resulted in increased levels of C-terminal phosphorylated Smad2 (pSmad2, Ser465/467), as detected by immunofluorescence and western blotting (Fig. 6); moreover, we observed no alterations in the overall levels of Smad2 (Fig. 6G). These results indicate that both Nodal and Cripto treatments activate the canonical Nodal/activin/TGFβ pathway through phosphorylation of Smad2. However, SB431542 blocked Nodal-dependent as well as Cripto-dependent Smad2 phosphorylation, indicating that phosphorylation of Smad2 relies upon Alk4/Alk5/Alk7 kinase activity in XEN cells. These results indicate that Cripto can elicit Smad2 phosphorylation in an Alk4/Alk5/Alk7-dependent manner, similar to Nodal. By contrast, XEN cell differentiation into visceral endoderm and AVE in response to Cripto is at least partially independent of Alk4/Alk5/Alk7 kinase activity and Smad2 phosphorylation.

Our findings reveal that Nodal signaling can alter the differentiation potential of XEN cells and provide new insights into the molecular regulation of Nodal signaling as well as the formation of visceral endoderm in vivo. We have shown that treatment of XEN cells with either Nodal or Cripto protein promotes their differentiation into visceral endoderm as well as AVE in culture and in chimeric embryos. Unexpectedly, our data also show that the activities of Nodal and Cripto are distinct, as the ability of Cripto to promote visceral endoderm and AVE differentiation is not blocked by SB431542. Furthermore, this activity of Cripto is at least partially independent of Smad2 phosphorylation because Smad2 phosphorylation in response to Cripto is inhibited by SB431542. Therefore, our findings raise the possibility that Cripto can signal, at least in part, through a downstream pathway in XEN cells that is independent of Alk4 kinase activity and Smad2 phosphorylation. Such a non-canonical signaling pathway might be significant for the molecular regulation of visceral endoderm and DVE formation in vivo.

Our results suggest a model in which Nodal and Cripto act cooperatively through distinct and possibly synergistic signaling pathways to regulate XEN cell differentiation into visceral endoderm and AVE. The addition of Nodal stimulates canonical signaling through Alk4 and Smad2/Smad3 to promote visceral endoderm and AVE differentiation, and this activity is dependent upon the EGF-CFC protein Cryptic (Fig. 7A). In the absence of Nodal, we propose that Cripto can either signal through type I receptors in a kinase-independent manner (which would not be inhibited by SB431542), or through an as yet unidentified novel receptor in XEN cells (Fig. 7B). Endogenous Cryptic activity would presumably be unable to activate either pathway in the absence of Nodal or Cripto, either because the levels of Cryptic are insufficient or because it is functionally dissimilar to Cripto.

This model contrasts with previous studies that have linked Cripto activity to the function of Nodal pathway ligands. Cripto is produced as a glycosylphosphatidylinositol (GPI)-linked protein, and can be readily detected in the culture medium of mammalian cells transfected with full-length Cripto expression constructs (Minchiotti et al., 2000; Yan et al., 2002). Soluble Cripto protein from culture medium, as well as commercially available recombinant protein, can stimulate Nodal pathway activity in a luciferase reporter assay in HEK 293T cells, where pathway activity requires Nodal ligand and can be inhibited by Lefty proteins (Chen and Shen, 2004; Yan et al., 2002). Furthermore, soluble Cripto can induce prechordal mesoderm differentiation from explants of chick notochord, and this activity is Alk4/Alk5/Alk7 dependent as it can be blocked by SB431542 (Chu et al., 2005). Although the mechanism by which soluble Cripto is produced in cell culture is unknown, a similar mechanism is likely to occur in vivo. Indeed, previous chimera analyses have demonstrated the non-cell-autonomy of Cripto function in the mesendoderm (Chu et al., 2005). Similarly, the non-overlapping expression patterns of Cripto and Cryptic imply non-cell-autonomy in the visceral endoderm (Chu and Shen, 2010), supporting the conclusion that Cripto expressed in the epiblast can behave as a trans-acting factor in vivo to promote DVE formation.

Notably, there are several published reports of Nodal pathway-independent effects of Cripto in mammalian cell culture. Thus, Cripto can act as a non-competitive antagonist for activin (Adkins et al., 2003; Gray et al., 2003; Kelber et al., 2008), as well as an inhibitor for TGFβ (Gray et al., 2006; Shani et al., 2008). In other cases, Cripto has effects that appear to be independent of TGFβ pathway components, as it can signal independently of Alk4 through c-Src, resulting in downstream activation of the Ras/Raf/MAP kinase and phosphatidylinositol 3′-kinase (PI3K)-Akt pathways (Bianco et al., 2002; Bianco et al., 2003; Ebert et al., 1999). Activation of c-Src may be mediated by binding of Cripto to glypican-1, which is a GPI-linked heparan sulfate proteoglycan (Bianco et al., 2003). Furthermore, Cripto may also modulate another developmental signaling pathway, as a recent study has shown that human CRIPTO can potentiate Notch signaling by promoting enhanced proteolytic cleavage and maturation of Notch receptors (Watanabe et al., 2009). At present, it is conceivable that Cripto might signal in XEN cells through one of these previously described non-canonical pathways, which then converges on Nodal targets downstream of Smad2, or might instead modulate a novel non-canonical signaling pathway. However, the mechanisms by which Nodal and Cripto signaling lead to functional convergence with downstream components of the canonical Nodal signaling pathway remain to be investigated.

Our findings on the properties of XEN cells in culture and chimeric embryos raise the possibility that Nodal and EGF-CFC proteins play a central role during the early steps of visceral endoderm specification and/or differentiation in vivo. This interpretation is consistent with previous analyses of the EmVE defects in Nodal mutants and Cripto; Cryptic double mutants (Chu and Shen, 2010; Mesnard et al., 2006). In addition, Nodal signaling stimulates visceral endoderm formation during human ES cell differentiation in culture (Vallier et al., 2005). Furthermore, our results are consistent with the view that Nodal and/or Cripto signal from the epiblast to the visceral endoderm to promote DVE formation. Although a central role for Nodal has been supported by several studies (Brennan et al., 2001; Chu and Shen, 2010; Norris et al., 2002), direct evidence for the ability of Nodal to induce DVE has been previously lacking.

Although previous work has shown that Nodal signaling is required for DVE formation and translocation, the precise embryological origins and relationship of the DVE and AVE have been unclear. Previous studies have suggested that the expression patterns of Lefty1 and Cerl at the late blastocyst stage correspond to an early specification of the DVE at peri-implantation stages (Takaoka et al., 2006; Torres-Padilla et al., 2007). Indeed, recent lineage-tracing studies indicate that the AVE has a separate origin from the DVE and is newly derived from the EmVE at 5.5 dpc (Takaoka et al., 2011). Such an independent origin of the AVE is consistent with the delayed or absent expression of some AVE markers in Cripto mutant embryos (Chu and Shen, 2010). In this regard, we note that our study has not addressed whether DVE (as opposed to AVE) might arise as a distinct response of XEN cells, or whether DVE heterogeneity might result from differential responses to independent Nodal and Cripto signaling pathways.

Interestingly, recent studies have described the isolation of XEN-like (XEN-P) cells from rat blastocysts (Debeb et al., 2009), which can contribute to both visceral and parietal endoderm as well as to trophoblast in chimeric rat embryos in the absence of prior Nodal treatment (Galat et al., 2009). However, these rat XEN-P cells display key differences to mouse XEN cells, as they are cultured in the presence of leukemia inhibitory factor (Lif) and express the pluripotency marker Oct4 (Debeb et al., 2009). These findings suggest the existence of species-specific differences in XEN cell properties and/or possible differences in cell types that arose through isolation under distinct culture conditions. Indeed, recent gene expression profiling analyses have demonstrated considerable differences in gene expression between XEN cells and similar primitive endoderm-like lines derived from embryonal carcinoma cells (Brown et al., 2010). Consequently, it might be possible to derive a spectrum of XEN-like cell lines using distinct culture conditions.

Finally, recent work has shown that visceral endoderm cells can contribute to definitive endoderm in vivo (Kwon et al., 2008). Given the strong molecular similarity between the visceral and definitive endoderm, and the central role of Nodal signaling in definitive endoderm formation, our results raise the possibility that XEN, or XEN-like, cells might also be capable of contributing to definitive endoderm tissues.

We thank Janet Rossant for generously providing EYFP-expressing XEN cells, Vivienne Halili for assistance with microinjection, Po-Chien Chou for initial data on Smad2 phosphorylation, Parla Makinistoglu for assistance in generating XEN lines, and Gloria Kwon and Kat Hadjantonakis for advice on immunostaining. This work was supported by NIH grant HD42837 (M.M.S.). Deposited in PMC for release after 12 months.