Foxa2 regulates polarity and epithelialization in the endoderm germ layer of the mouse embryo Free

During gastrulation the multilayered body plan of the mouse embryo is established through differentiation and highly coordinated morphogenetic events. By the start of gastrulation, at embryonic day (E) 6.5, the embryonic cup-shaped epiblast is surrounded by a single-layered epithelium of visceral endoderm (VE) that will give rise to the endodermal component of the yolk sac(Wells and Melton, 1999). Pluripotent epiblast cells constitute the progenitor cells for all cell lineages in the embryo proper and differentiate to form the three principal germ layers: endoderm, mesoderm and ectoderm(Beddington and Robertson,1999; Tam and Loebel,2007). Clonal analysis of epiblast cell fate revealed that in the early-streak embryo at E6.5, the proximal one-third of the posterior epiblast contains the precursors of the extra-embryonic mesoderm and the primordial germ cells. By contrast, the distal region of the epiblast contains the precursors of the entire neural ectoderm, and the intermediate posterior epiblast contains the precursors for the anterior mesoderm and definitive endoderm (Lawson et al., 1991; Lawson and Pedersen, 1992; Tam and Beddington, 1992; Lawson and Hage, 1994). Clonal descendants were not necessarily confined to a single germ layer, indicating that these lineages are not separated at the beginning of gastrulation. In support of this notion are embryonic stem (ES) cell differentiation experiments (Kubo et al.,2004), as well as conditional gene targeting results, indicating that bipotential mesendodermal progenitor cell populations exist(Lickert et al., 2002) (for a review, see Rodaway and Patient,2001). At various stages of gastrulation, the primitive streak(PS) has been shown to contain precursor cells of different mesodermal and endodermal lineages that are destined for different parts of the body(Kinder et al., 1999; Kinder et al., 2001). Therefore, allocation of mesoderm and endoderm in the embryo takes place in an anteroposterior (AP) manner determined by the timing and order of recruitment through the PS. The majority of definitive endodermal (DE) cells ingress through the anterior end of the primitive streak (APS) at the mid-streak (MS)stage and intercalate into the overlying VE to give rise to the foregut(Kwon et al., 2008); however,a small population of DE cells might directly delaminate into the VE from the epiblast (Tam and Beddington,1992). Taken together, these studies clearly indicate that mesoderm and endoderm are specified in a spatiotemporal manner during gastrulation; however, it is not clear if these cells become specified in the epiblast or PS region and when these cells differentiate into morphological and molecular distinct cell populations.

The T-box transcription factor brachyury (T) was shown to mark progenitor cells for mesoderm and endoderm in ES cell differentiation cultures,suggesting that these cells originate from a common progenitor(Kubo et al., 2004). In the mouse embryo, T protein is localized in the posterior epiblast at the early-streak stage and is detected in nascent mesoderm in the PS region during gastrulation, as well as in the node and notochord from the late-streak (LS)stage onwards (Inman and Downs,2006). T localization in the mesoderm and notochord suggests that abnormalities in these cell populations are responsible for the homozygous mutant phenotype (Wilkinson et al.,1990). By contrast, Foxa2 is also expressed in the posterior epiblast from the early stage onwards and is then confined to anterior definitive endoderm (ADE) and axial mesoderm, which consists of the head process, prechordal plate, notochord and node(Sasaki and Hogan, 1993; Monaghan et al., 1993). Foxa2 is a member of the Forkhead transcription factor family, which includes three related transcription factors: Foxa1, Foxa2 and Foxa3, first identified by their ability to regulate liver-specific gene expression(Lai et al., 1990; Lai et al., 1991). A null mutation of the Foxa2 gene leads to absence of ADE and axial mesoderm(Ang and Rossant, 1994; Weinstein et al., 1994). Foxa1 and Foxa3 are expressed from E7.5 onwards in the definitive endoderm and can compensate for the loss of Foxa2 in the null mutants, which allows hindgut,but not fore- and midgut formation (Sasaki and Hogan, 1993; Monaghan et al., 1993; Ang and Rossant,1994; Weinstein et al.,1994; Dufort et al.,1998). These results collectively demonstrate that T and Foxa2 are functionally important for mesoderm and endoderm development; however, it is not clear how these transcription factors regulate a molecular and cellular program for the differentiation of these cell populations.

In addition to cellular differentiation, the gastrulating embryo also undergoes dramatic morphological changes to form the three principal germ layers and the basic body plan. One of the first morphogenetic events is the formation of the PS when signals and factors trigger epithelial-mesenchymal transition (EMT) of epiblast cells to give rise to mesoderm and endoderm(Thiery and Sleeman, 2006). During this process, epiblast cells lose their apical-basal (AB) epithelial polarity, downregulate the cell-cell adhesion molecule E-cadherin (cadherin 1– Mouse Genome Informatics) and break through the basement membrane (BM)to invade into the PS region. The interstitial mesodermal cells acquire a mesenchymal cellular fate and migrate over long distances between the endoderm and the ectoderm germ layer before they re-aggregate to form distinct organs such as the heart or kidney. By contrast, cells that are fate-specified to become DE appear in the APS region from MS to LS stage(Lawson et al., 1991; Tam et al., 1997; Kinder et al., 2001; Tam and Beddington, 1992). These cells acquire an epithelial fate and intercalate into the outside epithelium, but it is not clear if these cells undergo EMT followed by mesenchymal-epithelial transition or alternatively maintain epithelial polarity and just transiently downregulate cell-cell adhesion molecules to leave the epiblast epithelium. By the end of gastrulation the germ layers have formed and have already acquired AP, dorsoventral (DV) and left-right (LR)patterning information through signals from the embryonic organizer tissues,which include anterior VE, ADE, axial mesoderm, node, notochord and floorplate(Tam and Loebel, 2007).

Functional analysis of genes in mouse has greatly contributed to the understanding of germ-layer formation in the mouse embryo; however, the phenotypic analysis has been hampered by static techniques that often only describe end points, as well as the fact that embryogenesis in all placental mammals occurs in utero and is not easily amenable to ex vivo observation. The establishment of static embryo culture systems and the genetic introduction of fluorescent marker proteins in transgenic animals has now allowed for direct imaging of mouse embryogenesis (Yamanaka et al., 2007; Kwon et al.,2008). In this study, we established an ex utero static embryo culture system to continuously monitor the cellular processes occurring during germ-layer formation. The generation of genetic mosaics using aggregation chimera allowed us to distinguish embryonic and extra-embryonic lineages using fluorescent labels in order to follow mesoderm and endoderm formation at cellular resolutions. We present evidence for a specific role of Foxa2 in the formation of polarized and epithelialized cell types, namely the definitive endoderm and axial mesoderm (node and notochord).

Genes encoding fluorescent proteins (td-Tomato, YFP) were amplified by PCR using the following primers: Tomato fwd(5′-NotI-Kozak-XbaI),5′-GCGGCCGCAGCCACCATGTCTAGAATGGTGAGCAAGGGCGAGGAG;Tomato rev (5′-SpeI),3′-NNNACTAGTTTACTTGTACAGCTCGTCCATGCCG; YFP fwd,5′-GCGGCCGCATCTAGAATGGTGAGCAAGGGCGAGGAGCTGTTC; YFP rev,3′-ACTAGTTTACTTGTACAGCTCGTCCATGCCGAGAG. NotI/SpeI-digested PCR products were cloned into the pBKS vector.

For generation of Lyn-Tomato, an oligonucleotide was subcloned between the NotI and XbaI sites in the pBKS vector in front of the td-Tomato: Lyn-Oligo fwd,5′-GGCCGCATAACTTCGTATAGCATACATTATACGAAGTTATGCCACCATGGGATGTATTAAATCAAAAAGGAAAGACGGGGCCCGGTACT;Lyn-Oligo rev,5′-CTAGAGTACCGGGCCCCGTCTTTCCTTTTTGATTTAATACATCCCATGGTGGCATAACTTCGTATAATGTATGCTATACGAAGTTATGCTTATGC. The NotI/SpeI-digested fluorescent markers were subcloned into the NotI/NheI sites of the eukaryotic expression vector pCAGGS (Niwa et al.,1991).

The fluorescent ES cell and mouse lines used in this study were generated by electroporation of ScaI-linearized pCAGGS vector DNA containing dsRed, YFP or Lyn-Tomato into wild-type IDG3.2 ES cells(Hitz et al., 2007) or Foxa2^–/– R1 ES cells (Ang et al., 1994). Cells were selected with 1 μg/ml puromycin, and resistant clones were screened for uniform and ubiquitous reporter expression in cell culture and in vivo using embryos derived from ES cells.

Diploid or tetraploid chimeras were generated according to standard protocols (Nagy, 2003). Embryos were collected from dsRed-(Vintersten et al., 2004) and YFP- (Hadjantonakis et al.,2002) expressing mouse lines, both maintained on mixed genetic backgrounds (CD1/129Sv/C57/Bl6). T-GFP targeting construct was used to generate ES cells and a mouse line as previously described(Fehling et al., 2003).

Embryos were dissected in DMEM containing 10% FCS and 20 mM HEPES. Embryos were cultured on glass-bottom dishes using 200 μl embryo culture medium(50% rat serum, 40% DMEM without Phenol Red, 2 mM glutamine, 100 μM 2-mercaptoethanol and 1 mM sodium pyruvate in a 37°C incubator with 5%CO₂ and 5% O₂). To avoid evaporation the medium was covered with mineral oil. Image acquisition was performed on a Leica DMI 6000 confocal microscope and image analysis was carried out using Leica LAS AF software.

Cell measurements were carried out using Leica LAS AF software. Average and standard deviation are shown in the graphs. P-values were determined using a two-tailed Student's t-test with unequal variance with the number of cells and embryos stated in the figure legends.

Whole-mount in situ hybridization was performed as previously described(Lickert at al., 2002). The following probes were used: Eomes(Ciruna and Rossant, 1999), Hex (Hhex – Mouse Genome Informatics)(Thomas et al., 1998) and claudin 4 (RZPDp981G04226D). Embryos were photographed using a Zeiss Stereo Lumar V12 microscope.

Immunofluorescence whole-mount stainings were performed as previously described (Nakaya et al.,2005). Briefly, embryos were isolated, fixed for 20 minutes in 2%PFA in PBS, and then permeabilized in 0.1% Triton X-100 in 0.1 M glycine pH 8.0. After blocking in 10% FCS, 3% goat serum, 0.1% BSA, 0.1% Tween 20 for 2 hours, embryos were incubated with the primary antibody o/n at 4°C in blocking solution. After several washes in PBS containing 0.1% Tween-20 (PBST)embryos were incubated with secondary antibodies (donkey anti-mouse 594,donkey anti-rabbit 488, donkey anti-goat 594 Alexa fluor, Molecular Probes) in blocking solution for 3 hours. During the final washes with PBST, embryos were stained with 4′,6-diamidino-2-phenylindole, dihydrochloride (DAPI),transferred into 40% glycerol and embedded between two coverslips using 120μm Secure-Seal spacers (Invitrogen, S24737) and ProLong Gold antifade reagent (Invitrogen, P36930). Antibodies: Foxa2 (Abcam, Ab408749), brachyury(N-19, Santa Cruz), GFP (A11122, Invitrogen), E-cadherin (610181, BD), ZO-1(Tjp1 – Mouse Genome Informatics) (33-9100, Zymed).

Foxa2 is a Forkhead transcription factor required for anterior axial mesoderm and definitive endoderm formation(Ang and Rossant, 1994; Weinstein et al., 1994),whereas the T-box transcription factor brachyury (T) is necessary for posterior, but not anterior, mesoderm formation(Wilkinson et al., 1990). As previously reported, the mRNAs for both genes are expressed during gastrulation in the posterior epiblast, but it was not clear whether the proteins are synthesized in the same cells of the epiblast and epiblast-derived mesoderm and endoderm descendants(Sasaki and Hogan, 1993; Monaghan et al., 1993; Herrmann, 1991). To investigate the cellular distribution of these two transcription factors during gastrulation, we used whole-mount immunohistochemistry (IHC) with antibodies to T and Foxa2 and laser scanning microscopy (LSM) of fixed embryos. Surprisingly, in pre-streak-stage embryos the double immunofluorescent antibody staining revealed that T and Foxa2 protein was synthesized in two intermingled, but mutually exclusive, cell populations in the posterior epiblast (Fig. 1A; see Fig. S1A in the supplementary material). At MS stage,these two cell populations segregated into proximal and distal domains of the posterior epiblast (Fig. 1B;see Fig. S2 in the supplementary material). Proximal epiblast cells upregulate T protein after EMT and show a round, mesenchymal cellular phenotype in the PS region (Inman and Downs,2006), whereas distal epiblast cells upregulate Foxa2 protein after EMT and show a flattened cell morphology (see Fig. S2 in the supplementary material). Interestingly, Foxa2-expressing flattened cells appeared anterior to the anatomical end of the PS, forming a two-cell-diameter row of polarized cells (see Fig. S2D, white arrows, in the supplementary material). From fate maps (Tam and Beddington, 1992) and our imaging results (see Fig. 2 below), this seems to be the region in which the first DE cells appear at the surface, suggesting that a small population of DE cells might directly delaminate into the outside VE. At the LS stage, only epiblast cells underlying the APS were Foxa2-positive and showed signs of EMT (Fig. 1C; see Fig. S1B in the supplementary material). We could clearly distinguish three cell populations in the PS region: T-positive posterior mesoderm; Foxa2- and T-double-positive axial mesoderm; and Foxa2-positive VE and DE populations. At the LS stage, we could rarely detect single Foxa2-positive cells in the APS region. This implies that Foxa2⁺epiblast cells quickly upregulate T protein after EMT, which suggests that cells for the axial mesoderm are recruited from the APS region(Kinder et al., 2001). These results collectively indicate that endoderm and mesoderm is specified in the epiblast and differentiates after EMT, which can be distinguished by morphology and marker gene expression. Interestingly, Foxa2 epiblast precursor cells gave rise to polarized and epithelialized endoderm and axial mesoderm,including the polarized and epithelialized cells of the node and notochord.

To gain further insight into the morphogenetic mechanisms underlying mesoderm and endoderm formation during gastrulation, we developed a static embryo culture system using time-lapse confocal imaging(Fig. 2A-D). One major difficulty in the analysis of endoderm development is the lack of appropriate marker genes that can distinguish the embryonic DE from the extra-embryonic VE(Lewis and Tam, 2006). To this end, we analyzed germ-layer formation using aggregation chimera, which allowed us to label the embryonic and extra-embryonic lineages by means of different fluorescent marker genes (Fig. 2A). In tetraploid (4n) or diploid (2n) embryo ↔ wild-type 9(wt) ES cell aggregation chimera (hereafter called 2n/4n ↔ wt chimera),the ES cells can only contribute to the embryonic epiblast, whereas the extra-embryonic lineages are always formed by the cells of the 2n or 4n embryo(Tam and Rossant, 2003). Therefore, using 2n and 4n chimera allowed us to distinguish between embryonic lineages, namely ectoderm, mesoderm and DE, and extra-embryonic lineages,specifically trophectoderm and VE. Generating 2n chimera allowed us additionally to generate genetic mosaics in the epiblast to study mutant cells in an otherwise wild-type environment. For the time-lapse imaging we used an inverse confocal microscope in combination with a static embryo culture system(Fig. 2B). Analyzing single optical sections using LSM at the mid-sagittal and surface level of fixed 4n chimeras revealed that the epiblast was always completely derived from the ES cells at the pre-streak stage (E6.5) and was covered by a single-layered epithelium of VE from the 4n embryo (Fig. 2C) (n>100). As predicted from fate map studies of the mouse embryo (Lawson et al.,1991; Lawson and Pedersen,1992; Tam and Beddington,1992), the first DE cells were recruited from the epiblast and intercalated into the surface VE in the APS region at MS stage(Fig. 2C). By the LS stage, the recruitment of DE was almost finished and the VE was mostly displaced by DE(Fig. 2C). From these results we concluded that fluorescent-lineage tagging using aggregation chimeras generates useful genetic mosaics to monitor cellular processes and lineage allocation in the pre- to LS-stage embryo. Next we performed time-lapse live imaging analysis using LSM of static immobilized 4n dsRed ↔ wt YFP chimera during gastrulation (Fig. 2D). As already indicated by our analysis of fixed MS chimera(Fig. 2C), DE cells formed in the APS region and intercalated into the overlying VE at this developmental stage (Fig. 2D; see Movie 1 in the supplementary material). The time-lapse analysis revealed that the DE and the mesoderm populations are morphologically distinct cell populations in the PS region, even before the DE cells intercalate into the outside VE(Fig. 2D) (time 0:00–0:39). The DE cells showed flat morphology and had an average length-width ratio of 4:1 (l=13.1±1.7 μm; w=3.24±0.67 μm;l/w=4.21±0.9; n=50), whereas the mesoderm cells showed a characteristic round morphology with an approximate length-width ratio of 1.4:1 (l=7.73±1.54 μm; w=5.63±1.16 μm;l/w=1.41±0.36; n=50) at LS stage. Furthermore, T-positive mesoderm cells and Foxa2-positive endoderm cells showed distinct morphology at the MS stage (see Fig. S2 in the supplementary material), indicating that mesoderm and endoderm can be distinguished by marker gene expression and morphology. This observation is consistent with results previously obtained in zebrafish (Warga and Nüsslein-Volhard, 1999), demonstrating that mesoderm and endoderm cell populations can also be distinguished by morphological criteria in higher vertebrates and that these cell populations are specified in the epiblast (Fig. 1) and differentiate and segregate in the PS region(Fig. 2).

To analyze how the Foxa2 transcription factor regulates definitive endoderm development on the cellular level, we took advantage of the Foxa2 knockout ES cell line (Ang and Rossant,1994) and analyzed genetic mosaics using LSM. In 4n ↔ wt chimeras, DE cells intercalated into the outside VE in the APS region from MS stage onwards and displaced and dispersed the VE by the LS stage(Fig. 3A) (n>20). In striking contrast, all 4n ↔ Foxa2^–/–chimeras showed no sign of DE intercalation and failed to form an anatomical node at the distal tip of the embryo even at the end of LS stage, indicating that the node and definitive endoderm cells are either not formed or that these cells do not reach the surface epithelial layer(Fig. 3A) (n>30). We noticed that cells accumulated in the APS region and frequently led to an indentation of posterior epiblast epithelium into the amniotic cavity from E7.5 onwards (Fig. 3A; see Movie 3 in the supplementary material; data not shown). We next performed time-lapse imaging using LSM of Foxa2^–/– ES 2n chimeras to analyze the behavior of Foxa2 mutant cells in an otherwise wild-type environment (Fig. 3B; see Movie 2 in the supplementary material) (n=9). As shown earlier in this study, Foxa2-positive epiblast cells reside in the APS region (Fig. 1), leave the epiblast epithelium and form DE, which intercalates into the overlying VE(Fig. 2). Imaging Foxa2 null cells from MS stage onwards clearly revealed that APS cells leave the epiblast and ingress into the APS region(Fig. 3B) (t=0:00-1:45 h, white asterisk). In contrast to wild-type DE cells, Foxa2^–/– `endoderm-like' cells showed endoderm morphology (Fig. 3B) (time 0:00; l=13.3±2.6 μm; w=3.8±0.7 μm; l/w: 3.6±0.9; n=50), made contact and partially integrated into the outside VE, but failed to epithelialize (Fig. 3B) (time 0:00-1:15, black asterisks). We wondered whether in 2n↔ wt chimeras wild-type cells had a competitive advantage and substituted or rescued DE formation; thus this might have been the reason that Foxa2^–/– cells were not integrated in the outside epithelium. Therefore we analyzed the cellular behavior of mutant cells in 4n ↔ Foxa2^–/– chimeras(Fig. 3C; see Movie 3 in the supplementary material). We clearly observed cells, which were intercalated but left the outside epithelium (Fig. 3C) (time 0:00-0:36). This indicates that Foxa2 is necessary for functional integration of DE cells into the VE epithelium.

Next we analyzed the identity of endoderm-like cells, which formed and intercalated into the outside VE in the absence of Foxa2. We performed whole-mount in situ hybridization to detect endoderm-specific genes that regulate EMT and mesendoderm formation [Eomes(Arnold et al., 2008)],transcription and endoderm formation [Hex(Thomas et al., 1998; Martinez Barbera et al.,2000)], as well as cell-matrix adhesion [integrin alpha3(Tamplin et al., 2008)] and tight junction formation (claudin 4). In completely Foxa2^–/– ES-cell-derived MS-stage embryos, Eomes was expressed at normal levels in the PS, confirming that Foxa2 mutant cells underwent EMT and formed mesendoderm(Fig. 4A; Fig. 3). Moreover, ADE formation was clearly induced in Foxa2 mutant cells, as indicated by the expression of the endoderm marker gene Hex(Fig. 4B). We previously used gene expression profiling of 4n ↔ wt and Foxa2^–/– chimeras to identify differentially expressed genes at the gastrulation stage(Tamplin et al., 2008). This analysis revealed that the tight junction markers claudin 4 and the cell-matrix adhesion molecule integrin alpha3 (Itga3) are potential target genes for Foxa2 in the DE. Strikingly, we found that whereas Itga3 was strongly expressed in the APS region of 4n ↔ Foxa2^–/– chimeras at the head-fold stage(Tamplin et al., 2008)(Fig. 2C,F), the tight junction marker claudin 4 was not detectable in the anterior endoderm region of 4n↔ Foxa2^–/– chimera(Fig. 6C). To further characterize the identity of Foxa2^–/– cells on a cellular level, we performed whole-mount IHC to detect the mesoderm marker protein T. As expected, Foxa2^–/– endoderm-like cells, which where partially integrated into the outside VE, were negative for T protein (Fig. 4C), indicating that Foxa2^–/– endoderm cells did not switch to a mesodermal fate, but still remained endoderm-like, expressing Eomes,Hex and Itga3, but not the tight-junction marker claudin 4. These results clearly indicate that endoderm-like cells are formed in Foxa2 mutants, but accumulate in the APS region, fail to induce claudin 4 and do not functionally integrate into the outside VE (compare with Fig. 3).

To better understand the cellular and molecular defects of Foxa2mutant endoderm cells, we analyzed the process of endoderm intercalation in greater detail. For this purpose, we made use of a ubiquitous Lyn-Tomato-expressing ES cell line (Fig. 5). The 10 N-terminal amino acids of the Lyn-kinase containing a consensus N-myristoylation and S-palmitoylation sites were fused to the N-terminus of the Tomato protein to target the fusion protein to the plasma membrane. Surprisingly, the ubiquitously expressed Lyn-Tomato protein accumulated on the apical membrane surface of the epiblast and DE epithelium(Fig. 3A, red asterisks). In the PS region, where epiblast cells lose polarity and undergo EMT, the continuous apical localization of the Lyn-Tomato protein was disrupted. Using Lyn-Tomato as a tool to analyze cell polarity, we performed time-lapse live imaging of intercalating DE cells in 4n YFP ↔ wt Lyn-Tomato chimera(Fig. 5A; see Movie 4 in the supplementary material). By the beginning of intercalation, DE cells in contact with the outside VE were not polarized, but extended filopodia processes into the outside epithelium (Fig. 5A) (time 0:00). During intercalation, DE cells became more and more polarized (Fig. 5A) (time 0:15–0:30) and by the end of the process clearly showed AB cell polarity by the means of Lyn-Tomato localization(Fig. 5A) (time 0:45).

Using Lyn-Tomato as an apical membrane marker, we compared cellular polarization in 4n ↔ wt or Foxa2^–/–chimeras (Fig. 5B). Analyzing MS- to LS-stage embryos revealed that Foxa2 mutant cells were able to localize Lyn-Tomato to the apical membrane(Fig. 5B, white arrowheads). However, the analysis also showed a statistically significant difference in the cellular polarization between wt and Foxa2 mutant cells. To investigate the cause of the cell polarity defects, we analyzed the formation of adherens and/or tight junctions using whole-mount immunolocalization studies to detect E-cadherin and ZO-1 in the endoderm epithelium. Comparing MS to LS stage 4n ↔ wt and Foxa2^–/– chimeras clearly demonstrated that the adherens junction protein E-cadherin was not localized at junctions between adjacent mutant cells, but surprisingly mutant-wt cell junctions showed a similar extent of basolateral localization as wt-wt adherens junctions (Fig. 6A). We speculate that the correct positioning of E-cadherin in mutant-wt cell junctions is due to homotypic molecular interactions of the E-cadherin protein in mutant cells with those correctly localized to the basolateral domain in wt cells. However, these interactions may be transient,as the mutant cells failed to functionally integrate into the outside epithelium. Due to the fact that claudin 4 is not expressed in Foxa2mutants (Fig. 6C), we investigated the localization of the tight-junction protein ZO-1. Comparing MS to LS stage 2n ↔ wt and Foxa2^–/– chimeras revealed that wt cells localized ZO-1 to the basolateral junctions in a punctate manner, whereas most mutant cells ectopically localized ZO-1 to the apical surface (Fig. 6B). It is well known that Claudins are the major cell-adhesion molecules of tight junctions (Tsukita et al.,2001; Furuse and Tsukita,2006) and bind specifically to ZO-1, ZO-2 (Tjp2 – Mouse Genome Informatics) and ZO-3 (Tjp3 – Mouse Genome Informatics) via an intracellular PDZ domain (Itoh et al.,1999). Therefore, failure to induce claudin 4 or other Claudins expressed in the endoderm (Sousa-Nunes et al., 2003; Hou et al.,2007) might explain the ectopic localization of ZO-1 at the apical membrane of Foxa2 mutant endoderm-like cells. Alternatively, Foxa2 might regulate a molecular program of cell polarity important to establish functional tight and adherens junctions.

In this study we analyzed germ-layer formation in wild-type and Foxa2 mutant embryos and chimera by immunohistochemistry and time-lapse live imaging. We showed that T-positive mesoderm and Foxa2-positive axial mesoderm and endoderm cell populations are already specified in the epiblast. These cells undergo EMT and ingress into the PS region, where they differentiate and segregate into molecularly and morphologically distinct populations of mesoderm and endoderm. Flat endoderm cells polarize and integrate into the overlying epithelium by formation of adherens and tight junctions. We showed that Foxa2 is translated in epiblast precursor cells of polarized and epithelialized cell types: namely endoderm and axial mesoderm(node and notochord). Axial mesodermal cells upregulate T protein after EMT,which suggests that Foxa2 is upstream of T in this cell population. In Foxa2 mutants, an anatomically characteristic node structure is not formed at the distal tip of the embryo, and although endoderm-like cells are formed and accumulate in the anterior PS region, they do not functionally integrate into the outside epithelium. These cells fail to polarize and epithelialize, implicating that Foxa2 regulates a molecular program important for these processes.

An important question in embryology and stem cell biology is when and how precursor cells are specified and differentiate. To our surprise, the T-box transcription factor brachyury (T) and the Forkhead box transcription factor Foxa2 are specifically synthesized in specified mesoderm and endoderm precursor cells in the posterior epiblast during gastrulation. Using time-lapse imaging and immunohistochemistry, we have shown that mesodermal and endodermal cells quickly segregate and differentiate after EMT. T-positive epiblast cells differentiate into T-positive mesenchymal cells in the PS,whereas Foxa2-positive epiblast cells differentiate into Foxa2-positive epithelial endodermal cells that integrate into the overlying epithelium and Foxa2-positive, T-positive axial mesodermal cells. Fate map analyses have revealed that the cells in the anterior end of the PS of the MS-stage embryo,which we have shown are Foxa2-positive, will give rise to anterior mesoderm and endoderm (Kinder et al.,2001), whereas cells in the posterior region of the PS, which we have shown are T-positive, will give rise to posterior as well as extra-embryonic mesoderm (Kinder et al.,1999). This is consistent with the gene functional analysis of either of these genes. T null mutants lack posterior mesoderm and notochord(Wilkinson et al., 1990; Kispert and Herrmann, 1994),whereas Foxa2 null mutants lack anterior mesoderm and endoderm, as well as the node and notochord (Ang and Rossant, 1994; Weinstein et al., 1994). Using a T-Cre and Foxa2-Cre genetic lineage tracing approach, we and others have recently shown that Foxa2 epiblast precursor cells give rise to anterior mesoderm and endoderm, whereas T epiblast precursors give rise to posterior mesoderm and endoderm(Uetzmann et al., 2008; Park et al., 2008; Kumar et al., 2007; Verheyden et al., 2005). These results are consistent with the idea that a bipotential mesendodermal progenitor cell population exists in mammals(Rodaway and Patient, 2001; Lickert et al., 2002; Kubo et al., 2004). Taken together, these results suggest that the posterior epiblast can be divided into a distal Foxa2-positive and proximal T-positive precursor cell population, giving rise to anterior and posterior mesendodermal cell populations, respectively.

How does axial mesoderm, namely the head process, prechordal plate,notochord and node, develop? It was previously suggested that Foxa2 is on top of a developmental program for axial mesoderm formation(Yamanaka et al., 2007). At the MS stage we detected an APS population, which was Foxa2-positive and was fate-mapped to give rise to the axial mesoderm and endoderm(Kinder et al., 2001). At the LS stage, the epiblast cells generated three distinct cell populations by morphology and marker gene expression: a T-positive posterior mesoderm population, a Foxa2-positive endoderm population and an anterior Foxa2-positive and T-positive axial mesoderm population. We noticed that Foxa2-positive epiblast cells at MS to LS stage upregulated T protein after EMT, indicating that Foxa2 epiblast cells give rise to axial mesoderm. From knockout studies it is known that Foxa2 mutants do not form axial mesoderm at all, whereas the T mutants initially form but fail to maintain posterior notochord. We also showed in this study that no anatomical node structure is formed at the distal tip of Foxa2 mutant chimera. This suggests that Foxa2 is on top of the axial mesoderm hierarchy(Yamanaka et al., 2007) and is consistent with loss of brachyury expression, specifically in the node and AME, but not PS, of tetraploid-derived Foxa2 null embryos at E7.5 (Dufort et al., 1998). Interestingly, axial mesodermal cells (node and notochord) did not acquire a mesenchymal fate along with the rest of the T-positive mesoderm population in the posterior PS, but rather constituted a population of cells that were highly polarized and connected through cell-cell adhesion. For example, node cells formed a characteristic anatomical structure in the surface endoderm layer at the distal tip of the embryo. The cells showed clear AB polarity,were monociliated and interconnected through E-cadherin-mediated cell-cell adhesion (Yamanaka et al.,2007). Also the notochord descendents of the node cells were highly polarized and formed a solid rod-like structure through cell-cell adhesion, between the endoderm and the ectoderm epithelium. This suggests that Foxa2 progenitor cells in general give rise to polarized, interconnected cell types and that Foxa2 promotes an epithelial fate and suppresses a mesenchymal fate.

In this study, we have shown that Foxa2 mutant progenitor cells leave the epiblast, but fail to integrate into the outside epithelium, which leads to an accumulation of mesenchymal cells in the APS region. This is consistent with the idea that Foxa2 regulates a program necessary to acquire an epithelial cellular phenotype. This is also accordant with the lack of polarized and epithelialized cell types in the Foxa2 mutant embryos,i.e. node, notochord and anterior definitive endoderm(Ang and Rossant, 1994; Weinstein et al., 1994), but how does Foxa2 regulate cell-cell polarity and epithelialization in the endoderm germ layer? In our attempts to identify novel Foxa2 target genes at the gastrulation stage (Tamplin et al.,2008), we have identified many potential target genes, including the homeobox transcription factors Hex and Otx2(Kimura-Yoshida et al., 2007),the signaling molecules Cer1 and Shh(Epstein et al., 1999; Jeong and Epstein, 2003), the SRY-related HMG box transcription factor Sox17 and the Forkhead box transcription factor Foxa1 (Duncan et al.,1998). Most of these endoderm-specific patterning factors are expressed in the endoderm germ layer, but not in Foxa2-positive epiblast precursor cells. This is consistent with the idea that Foxa2 is a pioneer factor, which opens compact chromatin and acts in higher-order gene regulation to allow mesendoderm and endoderm specific transcription factors to specify cell fate (Cirillo et al.,2002). But how does this molecular program translate into cellular changes that lead to the mesoderm or endoderm lineage decisions? In this respect it was interesting to find proteins involved in cell adhesion, such as the tight junction protein claudin 4, the homotypic cell-cell adhesion molecules Flrt2, Flrt3 and Pcdh19, as well as the cell-matrix adhesion molecule Itga3, as potential Foxa2 endoderm target genes(Tamplin et al., 2008). It was recently shown that hepatocyte nuclear factor 4a (HNF4a; Hnf1a – Mouse Genome Informatics), an important nuclear receptor for endoderm development(Lemaigre and Zaret, 2004),triggers formation of functional tight junctions and establishment of polarized epithelial morphology by specifically inducing Claudin expression(Chiba et al., 2003; Satohisa et al., 2005). ZO-1 has been proposed to be a scaffolding protein between transmembrane and cytoplasmatic proteins, and possibly forms a link between the adherens and tight junctions, e.g. formation of the adherens junction through E-cadherin is associated with the formation and localization of tight junction proteins,particularly ZO-1 (Rajasekaran et al.,1996; Siliciano and Goodenough, 1988). Taken together, we suggest that Foxa2mutant endoderm-like cells fail to initiate an endodermal molecular program regulated by Foxa2 and different endoderm-specific patterning factors, which results in a change of cellular morphology dictated by cell-cell, cell-matrix adhesion and cell polarity molecules.