Activation of the PTHRP/adenylate cyclase pathway promotes differentiation of rat XEN cells into parietal endoderm, whereas Wnt/β-catenin signaling promotes differentiation into visceral endoderm

The blastocyst of rodents consists of three cell types: the trophectoderm (TE), the epiblast and primitive endoderm (PrE), the latter two derived from the inner cell mass (ICM). During the implantation into the uterus, the PrE splits into two layers, the parietal endoderm (PE) contacting the TE, and the visceral endoderm (VE) covering the developing epiblast and extraembryonic ectoderm (Gardner, 1983; Bielinska et al., 1999; Beddington and Robertson, 1999).

PE cells are dispersed over the inner surface of the trophoblast giant cell layer. There is a thick multilayered basement membrane, so called Reichert's membrane between the PE cells and the trophoblast cells, which is composed of extracellular matrix proteins, in particular, laminins, collagen 4, nidogen and perlecan (Hogan et al., 1980; Gersdorff et al., 2005). PE cells have a morphology characteristic of motile cells with minimal cell–cell adhesion and express the mesenchymal markers, snail and vimentin (Gardner, 1983; Nieto et al., 1992; Smith et al., 1992). VE cells surround the developing epiblast as an epithelial layer and implement the protective barrier and trophic function for the developing embryo (Gardner, 1983; Bielinska et al., 1999). Similar to gut epithelium, the VE consists of polarized cells with microvilli and E-cadherin based adhesion junctions, forming a continuous layer of cells (Gardner, 1983; Ninomiya et al., 2005; Kimura-Yoshida et al., 2005). In addition to metabolic function, the VE plays a crucial role in patterning of the embryo (Beddington and Robertson, 1999; Arnold and Robertson, 2009; Rossant and Tam, 2009).

Previous studies showed that the VE cells derived from early postimplantation embryos were able to contribute to both PE and VE formation when they were transplanted to recipient blastocysts (Hogan and Tilly, 1981; Gardner, 1982). In vitro data demonstrated that VE cells explanted from the postimplantation mouse embryos were able to trans-differentiate into PE cells (Ninomiya et al., 2005). These data suggest that some degree of flexibility between VE and PE lineages exists, but the signaling events keeping the cells in either of these states or mediating the transition between them remain largely unknown.

Extraembryonic endoderm (XEN) cells were originally derived from mouse preimplantation embryos (Kunath et al., 2005). The role of several signaling pathways required for the establishment of the PrE lineage was elucidated using preimplantation embryos and XEN cells. It was shown that the specification of PrE from the ICM was mediated by the FGF4/Grb2/MAP kinase pathway (Chazaud et al., 2006; Nichols et al., 2009; Yamanaka et al., 2010). PDGF signaling was critical for the early expansion of PrE and the derivation of XEN cells (Artus et al., 2010). Several transcription factors, such as GATA4 and GATA6 acted in genetic pathways that directed the differentiation of embryonic stem (ES) cells to XEN cell lineage (Fujikura et al., 2002; Yamanaka et al., 2006). Sox17 was found to activate a number of genes involved in sorting of the cells within the ICM, including GATA4 and GATA6, and was also required to establish XEN cells from the blastocyst (Niakan et al., 2010).

Microarray studies revealed that mouse XEN cells expressed markers of PrE as well as PE and VE lineages (Kunath et al., 2005; Brown et al., 2010), although they contributed mainly to the PE lineage in chimeras and are therefore considered as PE-like cells (Kunath et al., 2005). Nevertheless, differentiation to VE cells, specifically anterior VE cells, has been achieved with mouse XEN cells using the ligands of Nodal and Cripto pathways (Kruithof-de Julio et al., 2011). More recently, mouse XEN cells were differentiated into extraembryonic VE cells after treatment with BMP ligands (Artus et al., 2012; Paca et al., 2012).

ES and embryonic carcinoma (EC) cells produced extraembryonic endoderm during differentiation in vitro. Previous studies showed that ES and EC cells after treatment with retinoic acid were able to differentiate into PrE cells, which further differentiated into PE cells after activation of the cyclic adenosine monophosphate (cAMP) signaling pathway (Strickland, 1981; van de Stolpe et al., 1993; Verheijen and Defize, 1999). Stable analogs of cAMP, such as db-cAMP and forskolin, triggered the trans-differentiation of VE to PE cells in embryoid bodies (EBs) differentiated from mouse ES cells (Maye et al., 2000). In vivo, the cAMP signaling was activated by parathyroid hormone related peptide (PTHRP), a ligand produced by giant trophoblast cells (Beck et al., 1993). This mechanism was proposed to play a main role in the formation of PE from PrE (van de Stolpe et al., 1993; Behrendtsen et al., 1995; Verheijen and Defize, 1999).

XEN cells from rat preimplantation embryos were derived recently. Interestingly, although rat XEN cells exhibited similarities to the respective mouse lineages, they were found to be positive for Oct4, Rex1 and SSEA1, which are markers for undifferentiated cells (Debeb et al., 2009; Chuykin et al., 2010). After injection into the blastocyst, rat XEN cells contributed to both PE and VE formation, demonstrating a broader developmental potential compared to mouse XEN cells (Debeb et al., 2009).

In this study we evaluated the role of Wnt, BMP4 and PTHRP/cAMP signaling pathways in the transition of rat XEN cells into VE-like versus PE-like cells. We found that activation of Wnt or BMP4 pathways promoted the formation of tight junctions between the cells and the acquisition of ZO1/E-cadherin/cortical actin junctional complexes, consistent with the transition to the tight epithelium of the VE. In contrast, activation of the cAMP-dependent signaling pathway promoted the formation of the PE and an apparent epithelial mesenchymal transition (EMT). Microarray and real-time PCR analyses confirmed the differentiation of rat XEN cells into VE- versus PE-like cells. Furthermore, application of PTHRP promoted the upregulation of the EMT marker snai1, and a loss of membrane-associated E-cadherin, and caused EMT in rat XEN cells.

We have previously observed that the morphology of rat XEN cells varied depending on feeder versus plastic cultivation (Chuykin et al., 2010). When we cultured rat XEN cells at low density, we found that the majority of cells were rounded (Fig. 1A, arrows) or fibroblast-like (Fig. 1A, arrowheads), and highly motile, and made transitions between round and fibroblast-like shapes (supplementary material Movie 1). When the cells were cultured at higher density, they tended to make cell–cell contacts and adopted flattened and epithelioid morphology (Fig. 1A, asterisk; supplementary material Movie 2).

Wnt signaling was shown to participate in PrE differentiation (Krawetz and Kelly, 2008) and was found to be active in VE cells of the developing mouse embryos (Kimura-Yoshida et al., 2005; Ferrer-Vaquer et al., 2010). Therefore, we investigated whether activation of the Wnt pathway would result in morphology changes of rat XEN cells. We cultured the cells in the presence of a GSK3 inhibitor, CHIR99021 (CHIR), for two days and observed the formation of epithelial colonies (Fig. 1B). The live imaging showed that rat XEN cells cultured with CHIR formed and maintained cell contacts while proliferation (supplementary material Movie 3).

We then tested whether the presence of CHIR was required to sustain the epithelial phenotype. XEN cells were first cultured with CHIR for 2 days, and thereafter the agent was removed. We found that cells detached from each other 1 day after the removal of CHIR (Fig. 1C).

Activation of the Wnt pathway through stabilization of β-catenin could account for the formation of cell–cell contacts since β-catenin was implicated in cell–cell adhesion (Verheyen and Gottardi, 2010). To test it we performed a gain of function for β-catenin in rat XEN cells by infecting them with lentiviruses bearing T7-tagged β-catenin. The transduced rat XEN cells formed cell contacts (Fig. 1D). Our data demonstrate that activation of the Wnt signaling pathway either by GSK3 inhibition or by the elevation of β-catenin expression promotes the formation of cell contacts in rat XEN cells.

Previous studies showed that PE formation was achieved by the activation of the cAMP-dependent signaling pathway during differentiation of EC and ES cells (Strickland, 1981; van de Stolpe et al., 1993; Verheijen and Defize, 1999; Maye et al., 2000). We studied whether a similar process could be triggered in rat XEN cells. When rat XEN cells were cultured with forskolin, a stable analog of cAMP for 1 day, cells formed filopodia (Fig. 2A, arrowheads). A decreased formation of filopodia was observed 1 day after the removal of forskolin (Fig. 2A, lower panel). Furthermore, when we cultured rat XEN cells with CHIR for 1 day and then treated with forskolin, we observed a dramatic change in morphology, displaying disruption of cell contacts and acquisition of a fibroblast-like shape (Fig. 2B, lower photo; supplementary material Movie 4), which is contrasted by the epithelializing effect of CHIR alone (Fig. 2B, upper photo).

Time-lapse recordings revealed that the motility of rat XEN cells cultured with CHIR alone was significantly lower compared to controls and also to the cells cultured in the presence of both CHIR and forskolin (supplementary material Fig. S1; Movies 1–4).

Our data indicate that activation of Wnt or cAMP pathways results in the transition of rat XEN cells to different cell types. In addition, the morphology changes of rat XEN cells resulted from treatments with CHIR or forskolin are reversible.

To gain further insight whether the observed changes in morphology could be characterized as transition of rat XEN cells to VE cells in the presence of CHIR and to PE cells in the presence of forskolin, we applied Affymetrix rat gene 1.0 ST arrays and performed a global gene expression analysis of rat XEN cells cultured in four different conditions: (i) without treatments (control), (ii) 2 days treated with CHIR alone, (iii) 1 day treated with CHIR and 1 day with both CHIR and forskolin, and (iv) 1 day treated with forskolin alone.

Fig. 3 shows a heatmap plot consisting of 19 clusters that represent all changes in transcription between the four studied conditions. Clusters 4, 7, 8, 11, 12, 13, and 18 (Fig. 3, green numbers) contain genes which were upregulated in XEN cells cultured in the presence of CHIR alone (ii) compared with control (i). In these clusters we found previously known VE genes, such as epcam, furin, hnf4a, foxq1, ihh, tmprss2, npas2, irf6, krt19, lgals2 and cited1 (Maye et al., 2000; Kunath et al., 2005; Hou et al., 2007; Sherwood et al., 2007; Brown et al., 2010) (Table 1). The VE is characterized as an epithelium bearing extensive cell–cell contacts. Accordingly, a panel of genes involved in cell adhesion and formation of cell–cell contacts, such as zo1, svil, cldn6, ocln, st14, cdh1, cldn4, cldn7 and llgl2 (Pestonjamasp et al., 1997; Verheijen and Defize, 1999; Balda and Matter, 2000; Chalmers et al., 2005; Kimura-Yoshida et al., 2005; Sherwood et al., 2007), were also found in these clusters (Fig. 3; Table 1). All of these genes, except for cited1 and furin, were downregulated in rat XEN cells cultured in the presence of both CHIR and forskolin (iii) or forskolin alone (iv) (Fig. 3; Table 1).

However, we did not identify expression changes of afp and vil1, two previously established VE markers (Kunath et al., 2005; Ninomiya et al., 2005). We also did not find expression changes of cerl, dkk1, hhex, otx2, fzd8 and nodal, which are associated with the expansion of the VE to distal and anterior parts of the embryo (Bielinska et al., 1999; Kimura-Yoshida et al., 2005; Brown et al., 2010; Paca et al., 2012). The full list of genes identified by microarray was included in supplementary material Table S1.

The genes in clusters 1, 2, 5, 6, 15 and 19 were upregulated in the presence of forskolin alone (iv) (Fig. 3, brown numbers), whereas they were downregulated in the presence of CHIR alone, except for the genes in cluster 15 (Fig. 3). In these clusters we found a panel of previously established PE markers including genes encoding proteins of extracellular matrix, nid1, col4a1, col4a2 and lamb1 (Gardner, 1983; Verheijen and Defize, 1999; Kunath et al., 2005), genes encoding proteins involved in remodeling of extracellular matrix, thbd and plat (Kunath et al., 2005), and other genes, such as pth1r and snai1 (Veltmaat et al., 2000; Kunath et al., 2005). Another PE marker, sparc (Kunath et al., 2005), which was found in cluster 18, was also upregulated in the presence of forskolin alone (Table 1).

Interestingly, we found that a panel of EMT associated genes, such as vim encoding vimentin, a regulator of actin cytoskeleton, snai1 encoding snail, an inducer of EMT (Thiery et al., 2009), adam8 encoding an extra-cellular matrix proteinase (White, 2003), l1cam (Shtutman et al., 2006) and plau (Irigoyen et al., 1999) were downregulated in the presence of CHIR alone (ii), but upregulated in the presence of both CHIR and forskolin (iii) (Table 1). Except for vim, all of these EMT-associated genes were also upregulated in the presence of forskolin alone (iv) (Table 1).

Next, we confirmed the expression of several VE and PE genes identified in the microarray by real-time PCR analysis. In rat XEN cells treated with CHIR alone (ii) the expression of VE genes ihh, hnf4a, furin and cited1 was significantly upregulated as compared to the control (i). The expression of VE genes cdh1, alcam, ocln, zo1, llgl2, cldn4, cldn7 and svil involved in cell–cell contact formation was upregulated only slightly, but not significantly (Fig. 4A). In contrast, a number of EMT and PE genes l1cam, adam8, vim, col4a2, lamb1, thbd and plat were significantly downregulated in rat XEN cells cultured with CHIR alone as compared to the control (Fig. 4B). In condition CHIR plus forskolin (iii) mRNA levels for most of VE genes, except for furin, were lower in comparison to the condition CHIR alone (ii) (Fig. 4A). The downregulation in condition CHIR plus forskolin (iii) versus CHIR (ii) was significant for genes alcam, ihh, hnf4a and cited1 (Fig. 4A). In the same time PE genes vim and snai1 were significantly upregulated in condition CHIR plus forskolin (iii) compared to CHIR alone (ii) (Fig. 4B). Treatment with forskolin alone (iv) promoted a significant downregulation of VE genes cdh1, ocln, cldn4, svil and hnf4a (Fig. 4A), and a significant upregulation of PE genes, such as snai1 and pth1r (Fig. 4B).

Additionally, we studied the transcription level of several VE genes that were not present in the microarray. We found that vil1 encoding a microvilli protein was significantly upregulated in cells treated with CHIR alone. In contrast, vil1 was significantly decreased in cells cultured with forskolin alone versus control and in CHIR with forskolin versus CHIR alone (Fig. 4C). Marker genes of anterior visceral endoderm nodal, fzd8, hhex and dkk1 were significantly decreased in forskolin alone versus control (Fig. 4C). In the presence of CHIR nodal and fzd8 were upregulated, whereas hhex levels remained the same and dkk1 was downregulated (Fig. 4C). We did not detect the mRNA of afp gene by real-time PCR analysis in all four studied conditions (data not shown).

The morphological changes of rat XEN cells cultured in the presence of CHIR were consistent with the formation of an epithelium (Fig. 1B). Therefore, we analyzed the expression of ZO1, a tight junction component by immunostaining (Fig. 4D). We found a weak staining for ZO1 in some of the control cells (Fig. 4D), but a strong staining at junctions of cells cultured with CHIR, confirming the formation of an epithelium (Fig. 4D). Treatment with forskolin resulted in an overall downregulation of ZO1, even below control levels (Fig. 4D).

Recent studies on mouse XEN cells demonstrated that Nodal/Cripto signaling induced the anterior VE formation (Kruithof-de Julio et al., 2011), whereas BMP4 promoted the extraembryonic VE differentiation (Artus et al., 2012; Paca et al., 2012). We found that treatment with Nodal did not cause any morphological transition in rat XEN cells (data not shown). However, when we treated the cells growing on laminin in N2B27 medium with BMP4 following the established protocol (Artus et al., 2012; Paca et al., 2012) (Fig. 5A), we noted the formation of an epithelium on day 2 (1 day after the BMP4 treatment) (Fig. 5B). On day 4, cells treated with BMP4 alone formed epithelial colonies (Fig. 5C), and became E-cadherin positive (Fig. 5D). In control cells, cell contacts were only occasionally formed, and only few E-cadherin-positive cells were found (Fig. 5D). In the presence of forskolin alone there were no epithelial-like and E-cadherin-positive cells (Fig. 5C,D). It is noteworthy that the addition of forskolin to rat XEN cells cultured in the presence of BMP4 completely blocked the formation of the epithelium (Fig. 5C) and E-cadherin expression (Fig. 5D).

Our data showed that rat XEN cells cultured in control medium were in a primitive state, and that activation of Wnt or BMP4 signaling promoted them to acquire VE-like properties. The additional activation of the cAMP dependent signaling pathway prevented this effect.

We further detailed the formation of cell contacts in rat XEN cells treated with CHIR by staining for actin, E-cadherin and ZO1. Images taken at low magnification revealed E-cadherin and ZO1 proteins at the cell borders, and actin at the periphery of the cells residing in the center of the epithelial structure (Fig. 6A). Images of the central cells at higher magnification revealed cortical actin characteristic of an epithelium. In the same time E-cadherin and ZO1 were distributed linearly at the border of neighboring cells (Fig. 6B). In addition, the tight junction component, ZO1, was situated above E-cadherin, as shown in reconstructed 3D image (Fig. 6C), indicating that cells growing with CHIR acquire a bottom-top polarity.

In contrast to the central part, images of the peripheral cells of the epithelial cluster at higher magnification revealed that actin stress fibers in one cell were often connected to stress fibers of the adjacent cell (Fig. 6D). Furthermore, the linear distribution of the tight junction proteins ZO1 and occludin (Ocln) was impeded in the peripheral cells of the epithelium (Fig. 6D) as compared to the central cells (Fig. 6B). Furthermore, when cells cultured with CHIR were treated with forskolin for 6, 12 and 24 hours, we observed an increase in the formation of actin stress fibers and a resulting disintegration of cells in the epithelial sheet (Fig. 6E). These data suggest that cells at the periphery of the epithelial cluster undergo EMT.

In vivo, cAMP signaling could be activated by PTHRP, a ligand produced by giant trophoblast cells (Behrendtsen et al., 1995; Verheijen and Defize, 1999). On our gene expression list we identified pth1r gene (Table 1), which encodes the receptor of PTHRP (Kunath et al., 2005). It had been reported that pth1r mRNA was induced during in vitro differentiation of EC and ES cells, and PTHRP substituted db-cAMP in inducing PE differentiation (van de Stolpe et al., 1993). To define whether PTHRP would promote PE formation, rat XEN cells cultured with CHIR were treated with recombinant PTHRP. As revealed by real-time PCR analysis, the treatment with PTHRP resulted in a significant upregulation of snai1 gene, a master regulator of EMT, and a decreased expression of cdh1 gene encoding E-cadherin (Fig. 7A). Thus, PTHRP induced the same transcriptional changes of genes associated with EMT in rat XEN cells as forskolin (Fig. 4). Furthermore, we observed that the 1-day PTHRP treatment caused a dramatic change in morphology of rat XEN cells: disruption of E-cadherin based cell–cell contacts and acquisition of a motile phenotype (Fig. 7B).

To characterize the effect of PTHRP on rat XEN cells in further detail, we cultured the cells in suspension similar to EB differentiation protocol, which was previously used for VE differentiation from EC cells (LaMonica et al., 2009). Rat XEN cells formed spheres after 4-day suspension culture with CHIR (supplementary material Fig. S2). After plating on laminin, the spheres attached to the culture dish, and cells formed cell–cell contact by E-cadherin dependent junctions (supplementary material Fig. S2). By contrast, when cells were grown with PTHRP, they were detached from each other and lost E-cadherin expression (supplementary material Fig. S2). All together our data showed that PTHRP, similar to forskolin, induced EMT in rat XEN cells.

Here we demonstrated that rat XEN cells acquired VE or PE characteristics in response to activation of Wnt/BMP4 or cAMP signaling pathways, respectively. The transitions to VE or PE cells resulting from the treatments with CHIR or forskolin were reversible. In addition, we showed that both forskolin and PTHRP induced EMT in rat XEN cells.

Rat XEN cells were derived as primary cultures from preimplantation embryos of different genetic backgrounds (Debeb et al., 2009; Chuykin et al., 2010). Previous studies showed that rat XEN cells expressed Oct4 and SSEA1 (Debeb et al., 2009; Chuykin et al., 2010) and contributed efficiently to both PE and VE development after blastocyst injections (Debeb et al., 2009). Therefore, rat XEN cells were proposed to be at the earliest known stage observed in in vitro culture among the extraembryonic endoderm lineage (Debeb et al., 2009). Together with these previous studies, our data demonstrate that rat XEN cells are PrE-like cells, and possess the potential to differentiate into both PE and VE lineages.

Cultivation of mouse XEN cells was established in 2005, and has since become a popular model to study lineage specification in extraembryonic endoderm. However, there were no reports concerning the effects of db-cAMP/forskolin and PTHRP on mouse XEN cells (Kunath et al., 2005; Artus et al., 2010; Brown et al., 2010; Kruithof-de Julio et al., 2011; Artus et al., 2012; Paca et al., 2012). Published data indicated that mouse XEN cell lines were restricted to the PE lineage. After blastocyst injections, mouse XEN cells contributed efficiently to PE, but not to VE development (Kunath et al., 2005). Moreover, mouse XEN cells expressed low E-cadherin levels unless when treated with Nodal/Cripto or BMP4 (Kruithof-de Julio et al., 2011; Artus et al., 2012; Paca et al., 2012).

We showed here that activation of the Wnt pathway in rat XEN cells by the GSK3 inhibitor CHIR or by gain of function of β-catenin induced the formation of an epithelium characteristic of VE cells (Fig. 1). A previous study showed that Wnt signaling was active in VE cells as revealed by the activity of TCF/LEF:H2B-GFP reporter gene (Ferrer-Vaquer et al., 2010). In addition, the active form of β-catenin was localized in cytoplasm of the entire VE cells and even in the nucleus of the proximal and distal aspects of VE (Kimura-Yoshida et al., 2005). Previous data also suggested that the Wnt pathway might be involved in PrE formation during in vitro differentiation of EC cells. Wnt6 gene encoding a canonical Wnt ligand was induced by retinoic acid in EC cells. Overexpression of Wnt6 or Wnt6-conditioned medium promoted EC cells to form PrE cells in vitro (Krawetz and Kelly, 2008). In line with that, wnt6 was expressed at blastocyst stage (Kemp et al., 2005). Thus, Wnt6 might be involved in the specification of extraembryonic endoderm in the early embryo. In addition, several transcription factors including Sox17 were required to establish XEN cells from the blastocyst (Niakan et al., 2010).

In our study, rat XEN cells treated with CHIR demonstrated an increased expression of VE marker genes, whereas the expression of PE genes was decreased (Fig. 4). One of the previously established VE markers is ihh gene encoding the ligand Indian hedgehog (Ihh) of the Hedgehog signaling pathway. Ihh produced by VE cells was critically important for blood island formation from extraembryonic mesoderm (Maye et al., 2000; Byrd et al., 2002). Using real-time PCR analysis, we found that the VE marker gene vil1 (Ninomiya et al., 2005) encoding the microvilli protein villin1 was strongly upregulated in rat XEN cells treated with CHIR. The expression of two anterior VE markers nodal and fzd8 (Kimura-Yoshida et al., 2005; Brown et al., 2010; Paca et al., 2012) was upregulated in the presence of CHIR, whereas hhex was not changed and dkk1 was downregulated (Fig. 4). These data indicated that treatment with CHIR did not cause a complete differentiation to the VE lineage.

Furthermore, we showed that CHIR induced the formation of cell–cell adhesion in rat XEN cells (Fig. 1). This was consistent with the formation of VE-like epithelium as which were shown to form cell contacts (Gardner, 1983; Bielinska et al., 1999; Sherwood et al., 2007). We demonstrated that in the presence of CHIR the cells formed tight junctions containing occludin, ZO1, E-cadherin and cortical actin (Fig. 6). However, at the edges of the epithelial clusters the distribution of tight junction components was different compared to central areas and there were actin stress fibers (Fig. 6D). This observation suggests that some cells maintain characteristics of motile cells in the presence of CHIR, indicating an incomplete transition to the epithelium.

Studies in mouse XEN cells established the role of the Nodal/Cripto and BMP pathways in VE lineage specification. Nodal induced the formation of anterior VE (Kruithof-de Julio et al., 2011), whereas combination of BMP4 and laminin promoted extraembryonic VE formation (Artus et al., 2012; Paca et al., 2012). Treatment of mouse XEN cells with BMP4 resulted in the formation of an epithelium. This morphological transition was reversible (Artus et al., 2012; Paca et al., 2012). In our study, we did not observe that Nodal cause any morphological transition. However, we found that BMP4 induced the formation of an epithelium as revealed by the expression of E-cadherin, indicating that the BMP4 signaling pathway was involved in VE formation in rat XEN cells (Fig. 5). In addition, the contribution of the BMP signaling to formation of the epithelium has been reported in different experimental systems including rat renal fibroblasts, human melanoma cells and mouse fibroblasts during reprogramming (Na et al., 2009; Samavarchi-Tehrani et al., 2010; Zeisberg et al., 2005).

Future studies are necessary to establish the culture conditions for complete differentiation of the VE lineage. Presumably this would need a culture with one or a combination of established ligands, such as activators of Wnt, Nodal and BMP4 signaling pathways, together with a chemically defined medium and plating substrates, such as laminin.

A question raised from the study is whether and how the effects of BMP4 and CHIR on rat XEN cells are related. Previous studies showed that Wnt and BMP pathways were functionally integrated in a variety of different systems, and often regulated similar biological processes, although the functional and mechanistic interaction between these two pathways relied on tissue-specific mechanisms. Wnt and BMP pathways were able to function independently from each other. However, recent studies have revealed many cases where these two pathways cooperated or attenuated each other (Guo and Wang, 2009; Itasaki and Hoppler, 2010; Hikasa and Sokol, 2013 for reviews). As mentioned above, previous studies suggested that the both Wnt and BMP pathways might be involved in PrE and VE formation. However, the cross-talk between them is not known in PrE formation and PE versus VE determination. Further studies are necessary to uncover whether BMP4 and Wnt signalings collaborate or act independently to induce VE differentiation in rat XEN cells and in the early embryo.

Our study showed that activation of the cAMP pathway in rat XEN cells by forskolin induced the formation of filopodia (Fig. 2A) and even surpassed the VE formation induced by CHIR and BMP4 (Figs 2, 4–6). Several previously established PE marker genes were upregulated in rat XEN cells after treatment with forskolin (Table 1; Fig. 4B). Interestingly, we found that a panel of EMT associated genes was downregulated in the presence of CHIR alone, but upregulated in the presence of both CHIR and forskolin. These data are in line with the previous studies showing that activation of the cAMP pathway induced PE differentiation from PrE formed during in vitro differentiation of EC and ES cells, in both monolayer (Strickland, 1981) and EB cultures (Maye et al., 2000).

We were intrigued to discover an extracellular signal inducing EMT and formation of PE in rat XEN cells. By real-time PCR we found that pth1r gene was upregulated in forskolin-treated cells (Fig. 4B). Pth1r has been assigned as PE marker (Verheijen and Defize, 1999; Kunath et al., 2005). Moreover, according to the microarray analysis, pth1r mRNA was present at high levels in rat XEN cells (expression level in log2 scale was 11.9 in control, supplementary material Table S1). Expression of pth1r was induced during differentiation of EC and ES cells, correlating with the appearance of functional adenylate cyclase-coupled PTH1R (van de Stolpe et al., 1993). PTHRP, a PTH1R ligand, was found to induce PE differentiation during in vitro differentiation of EC and ES cells (van de Stolpe et al., 1993). Furthermore, TE cells of rat blastocysts started to transcribe the mRNA of PTHRP at the stage of implantation at day E5.5 (Beck et al., 1993). Given the proximity of TE and PE cells in the embryo, PTHRP was considered as the most probable ligand inducing EMT and promoting the formation of PE in the embryo.

Since PE formation required EMT (Veltmaat et al., 2000; Krawetz and Kelly, 2008) we studied the effect of forskolin and PTHRP on morphology of rat XEN cells. Treatment with forskolin or PTHRP caused the disruption of cell–cell contacts, the ablation of ZO1 and E-cadherin at the cell membranes and the formation of stress fibers (Figs 4D, 5D, 6E, 7B). We also found that snai1 was upregulated after treatment with forskolin or PTHRP (Figs 4B, 7A). Snail is a transcriptional repressor of cdh1, and was shown to be an inducer of EMT (Peinado et al., 2007; Thiery et al., 2009). Therefore, our data allow us to conclude that activation of the PTHRP/cAMP pathway in rat XEN cells induces EMT and the formation of PE.

Treatments of rat XEN cells with CHIR, BMP4 or forskolin in our study caused reversible morphology changes. This is consistent with a previous study showing that the BMP4-induced differentiation of mouse XEN cells to extraembryonic VE was reversible (Artus et al., 2012). The observed flexibility of XEN cells is likely to be an innate property since VE and PE cells derived from the embryo were also able to undergo changes (Ninomiya et al., 2005; Paca et al., 2012).

Our study show that rat XEN cells represent a valuable cell culture model to study signaling cues leading to the specification of cells in the extraembryonic endoderm lineage, molecules potentially involved in patterning of the developing embryo, and fundamental processes, such as the formation of epithelial contacts and EMT.

Rat XEN cell lines derived from different strains have been characterized previously (Chuykin et al., 2010). Experiments were performed on the rat XEN cell line F2 from Rattus norvegicus (strain Fisher 344) at passage 20–30. Cells were cultured in DMEM/F12 (Sigma-Aldrich, St Louis, USA) supplemented with 10% fetal bovine serum (Sigma-Aldrich), 100 U/ml penicillin, 100 µg/ml streptomycin, 50 µM β-mercaptoethanol, 500 U/ml LIF (PAA, Pasching, Austria). When indicated, media were supplemented with 3 µM GSK3 inhibitor CHIR99021 (Axon Medchem, Groningen, The Netherlands), 10 µM forskolin (Sigma-Aldrich), 50 ng/ml human PTHRP (Sigma-Aldrich) or 50 ng/ml BMP4 (R&D Systems, USA). Rat XEN cells were routinely maintained on cell culture dishes without feeder cells and gelatin. Confluent XEN cells were passaged every 4–5 days in 1∶40–1∶50 ratio using TrypLE Express (Invitrogen, New York, USA). For treatment with BMP4, cells were plated on laminin and cultured in N2B27 medium: 1∶1 mixture of DMEM/F12 supplemented with N2 and neurobasal medium supplemented with B27 without vitamin A (Invitrogen). Laminin-coated slides were prepared by coating with 10 µg/cm² poly-L-ornithine (Sigma) for 30 min at room temperature, followed by coating with 0.15 µg/cm² laminin (Sigma) overnight at room temperature. Formation of spheres was achieved by cultivation of rat XEN cells in suspension on non-adhesive plastic dishes.

Lentiviral vector LVTHM encoding T7-tagged β-catenin designed in the group of Rolf Kemler (MPI, Freiburg) was provided by Alexey Tomilin (Institute of Cytology, Russian Academy of Sciences, St. Petersburg). 293T cells were transfected with envelope-encoding plasmid pMD2G (5 mg), packaging pCMV-dR8.74PAX2 (5 mg) and T7-β-catenin encoding LVTHM-based plasmid (20 mg) by using the calcium-phosphate method. Lentiviral particles in cell culture supernatant were collected and processed as described elsewhere (Wiznerowicz and Trono, 2003).

Cells grown on 14 mm glasses (Menzel, Braunschweig, Germany) in 24 well plates were fixed in 4% paraformaldehyde, permeabilized in 0.5% Triton X-100, blocked in 1% BSA (Sigma-Aldrich) in phosphate buffer saline (PBS) with 0.05% Tween 20 (PBST) and stained overnight with primary antibodies. Mouse E-cadherin (BD Biosciences, New-Jersey, USA), mouse T7 tag (EMD Biosciences, Darmstadt, Germany), rabbit ZO1 or mouse occludin antibodies (both from Invitrogen) were used in 1∶500 dilution in blocking solution. Following washes in PBST, secondary antibodies conjugated to Alexa fluorophores (Invitrogen) diluted 1∶1000 in blocking solution were added for 1 hour at room temperature. For DNA staining samples were incubated with 10 µg/ml Hoechst 33342 (Sigma-Aldrich). Actin filaments were stained with 1 µg/ml Phalloidin-FITC (Sigma-Aldrich) after incubation with secondary antibodies for 15 min at 37°C. Samples were mounted with DAKO mounting solution (DAKO, Copenhagen, Denmark). Images were acquired with confocal Leica TCS SP5 microscope with 40× or 63× oil immersion objectives using Leica LAS AF software. Phase contrast and some fluorescent pictures were obtained at epifluorescent microscope Leica, DMI6000B or Zeiss Observer Z1. The image files containing Z-stacks acquired by Leica confocal microscope were deconvoluted in Huygens Professional software, and the resulting model was reconstructed with the Imaris 7.4 software. For live imaging, cells were plated (4×10³/cm²) on 8-well ibiTreat μ ibidi slides (Ibidi, Martinsried, Germany), after 1 day of treatment with experimental medium. Time-lapse phase contrast images were taken every 5 min at Olympus Cell∧R microscope. The motility of randomly chosen cells was estimated using Volocity software. Video files were assembled in Fiji software.

For microarray and real-time PCR analysis RNAs were extracted by RNA mini kit (Qiagen, Duesseldorf, Germany). Residual genomic DNA was removed by DNase I treatment (Qiagen). Three independent RNA preparations from the 4 different conditions were processed to fragmented and labeled ssDNAs with WT expression and Terminal labeling kits (Ambion, New York, USA) and hybridized on rat gene 1.0 ST Array (Affymetrix, Santa Clara CA, USA). Arrays were quantile-normalized with respect to the probe GC content using the RMA algorithm (GC content adjustment, RMA background correction and mean probe set summarization). There were no outlier and batch effects at principal component analysis (supplementary material Fig. S3A). Transcripts with low expression were removed by a maximum expression cut off <100. The data filtering resulted in 150,256 of 211,195 probe sets and 21,567 meta-probe sets. Differential expression of genes was calculated using Partek ANOVA statistic followed by FDR multiple testing corrections. Resulting p-values and FDR values indicated the probability of differential expression between the four conditions. Differences between individual pairs of conditions were calculated using the LSD post hoc test. Expression levels of probe sets with significant difference in ANOVA (FDR <0.001, n = 4,183) were clustered and visualized in a heatmap plot. The k-mean clustering using k = 19 was done after the k estimation over the normalized signal values using the Davies Bouldin procedure (supplementary material Fig. S3B). The microarray data have been deposited in the NCBI Gene Expression Omnibus and are accessible through GEO Series accession number GSE42438 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE42438).

For the confirmation of microarray data, independent RNA isolations were performed. RNA samples were reverse transcribed using random hexamers and Moloney murine leukemia virus reverse transcriptase M-MLV (Invitrogen). The real-time PCR approach used the SYBR green method. PCR reactions were performed with Taqman 7900 cycler for 384 well plates. A two-step protocol was applied to compare gene expression levels between groups, using the equation 2^−ΔΔCT (Livak and Schmittgen, 2001). Gene expression was normalized to the mRNA level of housekeeping gene, Nat1, using rat-mouse specific primers (Takahashi and Yamanaka, 2006). The primers used are listed in supplementary material Table S2. The experiments were run in sample size of n = 3 and technical triplicates.

The authors would like to thank Zoltan Cseresnyes and Anje Sporbert from the MDC microscope facility for their help with live-imaging, Volocity software, and 3D reconstruction in Huygens and Imaris software. We are grateful to all members of Michael Bader and Kaomei Guan labs for technical support and discussions and especially to Simin Chen for proofreading of the manuscript. We thank Alexey Tomilin and Mikhail Liskovykh for the lentivirus encoding stabilized β-catenin. We are really grateful to Walter Birchmeier, Daniel Besser and colleagues from Michael Kuehn, Sergei Sokol, Philippe Soriano and Michael Shen labs for their critical comments.