Foxa2 mediates critical functions of prechordal plate in patterning and morphogenesis and is cell autonomously required for early ventral endoderm morphogenesis

Axial mesendoderm is a transient midgestation tissue located at the midline of 7.0–8.5 day post coitus (dpc) mouse embryos which is critical for patterning adjacent tissues, including neural ectoderm and paraxial mesoderm (Jessell, 2000; Wilson and Maden, 2005; Lee and Anderson, 2008). Axial mesendoderm is derived from anterior regions of primitive streak during gastrulation including the node organizer, is initially contiguous with endodermal epithelium, and is anatomically distinct due to its direct apposition with overlying neural groove ectoderm (Jurand, 1974; Beddington, 1994; Sulik et al., 1994; Kinder et al., 2001; Lewis and Tam, 2006; Yamanaka et al., 2007; Lee and Anderson, 2008).

Axial mesendoderm is comprised of notochordal plate and prechordal plate, the most anterior mesendoderm (Sulik et al., 1994; Yamanaka et al., 2007). Notochordal plate will give rise to notochord and contribute to dorsal foregut endoderm, and prechordal plate will contribute to ventral foregut endoderm, oral endoderm, and ventral cranial mesoderm (Couly et al., 1992). Tissue ablation studies have demonstrated that prechordal plate reinforces forebrain induction and regulates other aspects of anterior development (Shimamura and Rubenstein, 1997; Camus et al., 2000). Additionally, defects in anterior embryonic patterning have been associated with disruption of prechordal plate markers (Ang and Rossant, 1994; Weinstein et al., 1994; Filosa et al., 1997; Bachiller et al., 2000; Petryk et al., 2004; Nishioka et al., 2005; Warr et al., 2008). Attribution of tissue specific roles of genes expressed in prechordal plate in notochord development, however, has been limited by the lack of a prechordal plate-specific Cre line that is not active in other axial mesendodermal regions.

The forkhead transcription factor Foxa2 is expressed in critical signaling centers during development, including node, notochord, and prechordal plate, and is also expressed in both extraembryonic and embryonic endoderm. Global knockout of Foxa2 has revealed critical roles in embryonic patterning and a requirement for endoderm formation (Ang and Rossant, 1994; Weinstein et al., 1994; Dufort et al., 1998). Conditional Foxa2 deletion utilizing an endodermal-specific Foxa3-Cre line, however, did not result in defects in gut formation, perhaps owing to persistence of Foxa2 protein in mutant foregut endoderm at 9.5 dpc (Lee et al., 2005). Thus far, roles for Foxa2 in prechordal plate, or in early ventral endoderm, have not been investigated.

Dorsal and ventral foregut endoderm have unique embryonic origins, with distinct timing of emergence from the primitive streak and divergent migratory pathways that converge to comprise the entire foregut endoderm (Tremblay and Zaret, 2005; Tam et al., 2007). Gut tube formation begins in the anterior embryo at early to late head fold stages (EHF-LHF, ∼8.0–8.25 dpc) when an initially flat sheet of endodermal epithelium is transformed into a pocket structure. Ventral midline of the foregut is a discrete population derived from prechordal plate (Kirby et al., 2003; Franklin et al., 2008; Aoto et al., 2009). The foregut pocket remains open anterior to the site of fusion, termed the anterior intestinal portal.

To investigate tissue-specific roles for Foxa2 in prechordal plate and ventral foregut endoderm, we have ablated Foxa2 utilizing Isl1-Cre (Sund et al., 2000; Yang et al., 2006). Use of Isl1-Cre has allowed us to genetically isolate prechordal plate from posterior axial mesendoderm to address patterning by axial mesendoderm, and our results reveal a requirement for prechordal plate in development of notochord and dorsal foregut endoderm, both derivatives of notochordal plate. Our data also demonstrate a cell autonomous role for Foxa2 in ventral foregut endoderm for endoderm specification and foregut morphogenesis.

To investigate overlap between Isl1-Cre and Foxa2 expression, Isl1-Cre; Rosa^+/26R-lacZ embryos were analyzed beginning at early headfold (EHF) stages. Whole mount X-gal staining of EHF Isl1-Cre; Rosa+/26R embryos demonstrated low levels of β-galactosidase staining in the allantois. At these stages, β-galactosidase-positive domains were not detected in the embryo proper or in yolk sac endoderm anterior to the emergent head folds (supplementary material Fig. S1). At 3–5 som, β-galactosidase staining was not detected in extraembryonic tissues outside of the allantois in Isl1-Cre; Rosa^+/26R-lacZ embryos (supplementary material Fig. S1). Isl1-Cre activity is first evident in embryonic endoderm at 2–3 som (Park et al., 2006). At this stage, histological section analyses demonstrated that Isl1-Cre-derived cells were present in prospective ventral foregut caudal to the foregut pocket, and in ventral foregut endoderm extending to the rostral pocket tip (Fig. 1A–F). Rostralmost Isl1-Cre labeled cells included prechordal plate as indicated by direct apposition to overlying neural ectoderm, consistent with previous studies demonstrating Isl1 expression in prechordal plate in chick embryos (Yuan and Schoenwolf, 2000). Isl1-Cre-derived cells were also evident in head and heart mesoderm (Tzahor and Evans, 2011), and in yolk sac endoderm underlying the heart.

Examination of Foxa2 expression by immunohistochemistry demonstrated that Foxa2 and Isl1-Cre overlapped in prechordal plate, foregut endoderm, and yolk sac endoderm underlying cardiogenic mesoderm (Fig. 1G–I). As expected, ablation of Foxa2 with Isl1-Cre resulted in reduction of Foxa2 expressing cells in overlapping regions including prechordal plate (Fig. 1J), ventral foregut endoderm (Fig. 1K,L), and yolk sac endoderm underlying the heart (Fig. 1J). Foxa2 expression was also diminished in the midline of the neural groove (Fig. 1J,K), indicating a non-cell autonomous loss of Foxa2 in this expression domain.

Further analysis confirmed that Isl1-Cre lineages did not overlap with Foxa2 expression in node, notochordal plate or its derivatives, including dorsal foregut endoderm and notochord, or floor plate of the neural tube (Fig. 2).

Previous studies have demonstrated a role for Foxa2 in axial mesendoderm for forebrain and anterior embryonic patterning. However, owing to lack of suitable Cre drivers, it has been difficult to dissect the requirement for Foxa2 in prechordal plate from its requirement in other axial patterning tissues. Our analysis of Isl1-Cre lineages demonstrated that Isl1-Cre would allow us to selectively examine the cell autonomous requirement for Foxa2 in prechordal plate in forebrain and anterior patterning apart from its requirement in notochord or floorplate. To investigate the potential role of Foxa2 in tissues which overlapped with those labeled by Isl1-Cre, including prechordal plate and foregut endoderm, we utilized Isl1-Cre to generate tissue-specific deletion of Foxa2.

Heterozygous Foxa2^+/loxP; Isl1-Cre mice were viable, fertile, and displayed no obvious abnormalities. Most homozygous Foxa2^loxP/loxP; Isl1-Cre mutants were dead by 12.5–13.5 dpc (n = 9/11, supplementary material Table S1), were morphologically distinct at 8.5 dpc (n = 31/41), and displayed a range of phenotypes (supplementary material Table S2; Fig. 3). Mutant embryos displayed defects in anterior patterning, including abnormal head fold morphology, and abnormal neural groove (Fig. 3C,E). Laterality defects were evidenced by reversed or no cardiac looping in some mutants. Lack of foregut closure and cardia bifida were also seen (Fig. 3E, F; supplementary material Fig. S2). Previous studies have demonstrated that defects in endoderm which result in lack of foregut closure can result in cardia bifida (Maretto et al., 2008; Alexander et al., 1999). Holoproencephaly, a disorder resulting from deficient ventral forebrain patterning (Muenke and Beachy, 2000), was also observed (supplementary material Fig. S1).

To determine whether ectopic Isl1-Cre activity contributed to the range of phenotypes observed at 9.5–10.5 dpc, Foxa2^loxP/loxP; Isl1-Cre embryos were generated on a Rosa^26R-lacZ lineage reporter background (Soriano, 1999; Yang et al., 2006). Results showed specific Isl1-Cre expression in mutants displaying the full spectrum of abnormalities (supplementary material Fig. S2C–E').

To investigate whether defects in anterior patterning were separable from defects in foregut closure and cardia bifida, mutants were analyzed for mutual exclusivity of these two broad classes of phenotypes (supplementary material Table S3). Results of this analysis demonstrated that these abnormalities in foregut and cardiac morphogenesis could occur in the absence of defects in anterior patterning, suggesting distinct spatial and/or temporal requirements for Foxa2 for these two aspects of the Foxa2^loxP/loxP; Isl1-Cre mutant phenotype.

Specific defects in anterior patterning and laterality defects as evidenced by reversed or no cardiac looping in Foxa2^loxP/loxP; Isl1-Cre mutants suggested perturbations in axial mesendoderm. In situ hybridization was used to assay axial mesendoderm marker gene expression in 8.0–8.5 dpc Foxa2^loxP/loxP; Isl1-Cre mutants. At these stages, Goosecoid (Gsc), a homeobox transcription factor, is a marker of prechordal plate (Filosa et al., 1997). Foxa2^loxP/loxP; Isl1-Cre mutants exhibited markedly reduced Gsc expression at E8.0 (6 som) (Fig. 4B). Fibroblast growth factor 8 (Fgf8) marks anterior neural ridge, a signaling center that interacts with prechordal plate to regulate anterior forebrain patterning and expansion (Crossley and Martin, 1995; Shimamura and Rubenstein, 1997). Fgf8 is also expressed at the midbrain-hindbrain boundary (Fig. 4C). Expression of Fgf8 in anterior neural ridge was reduced in mutant embryos, while its expression at the midbrain-hindbrain junction was unaffected (Fig. 4D).

Sonic hedgehog (Shh) is a secreted factor expressed throughout axial mesendoderm which is required for survival of prechordal plate cells and dorsal ventral patterning of the neural tube (Echelard et al., 1993; Chiang et al., 1996; Ding et al., 1998; Aoto et al., 2009). At E8.0, Shh mRNA expression was absent in anterior regions of mutant embryos. Trunk regions exhibited patchy expression of Shh which did not extend to posterior regions as observed in somite-matched controls (Fig. 4E–F). Immunostaining confirmed that Shh protein was also absent from anterior regions of Foxa2^loxP/loxP; Isl1-Cre mutants, and was abnormally positioned to the left of the midline in more posterior regions (supplementary material Fig. S3).

Brachyury is a T-box transcription factor expressed in notochord at 8.5 dpc (Fig. 4G,I). Notochord cells marked by Brachyury expression were disorganized and abnormally positioned to the left of the midline in mutant embryos (7 som) (Fig. 4H), while other more anterior and posterior regions showed complete absence of Brachyury expression (Fig. 4J).

TUNEL and immunohistochemistry to detect Brachyury were used to investigate notochord survival in 8.5 dpc Foxa2^loxP/loxP; Isl1-Cre mutants. In control littermates, notochord cells were detected in a continuous anterior-posterior domain, and were not associated with apoptosis (Fig. 5A–D). Again, Brachyury was absent from the anterior region of mutant embryos (Fig. 5E–F). The most anterior limit of Brachyury expression was detected posterior to the anterior intestinal portal and positive cells were again observed to be somewhat disorganized and abnormally positioned leftward of the neural groove (Fig. 5H). In mutant embryos, apoptosis was observed in midline regions rostral to Brachyury expression (Fig. 5).

Our data demonstrated that loss of Foxa2 in regions which overlap with Isl1-cre lineages resulted in defects in notochord formation and morphogenesis. Isl1-cre lineages overlap with Foxa2 in both prechordal plate and ventral foregut endoderm. In individual mutant embryos, we observed loss of notochord markers even when Foxa2 was present in ventral foregut endoderm, suggesting that loss of notochord markers is not consequent to loss of Foxa2 in ventral foregut endoderm (Fig. 6A–H). Notochord defects were scored by absence of a morphologically recognizable notochord structure and lack of characteristic floor plate morphology, lack of Foxa2 protein expression in neural plate midline, and lack of detectable Brachyury or Shh protein expression. Hlxb9 is expressed in notochord and dorsal endoderm, and was also found to be greatly reduced in both tissues in these mutants (Fig. 6F). In other mutants, expression of axial tissue and dorsal foregut endoderm markers was maintained despite loss of Foxa2 from ventral foregut endoderm (Fig. 6I–J). Markers utilized were Foxa2 in presumptive floor plate, Brachyury in notochordal plate, and Hlxb9 in notochord and dorsal foregut.

Ventral endoderm forms a highly organized epithelium with tight junctions. Recent studies have demonstrated a role for Foxa2 in maintenance of epithelial integrity in the early embryo (Burtscher, 2009). Previous studies in mouse liver tissue have shown that Foxa2 directly binds to regions within the tight junction protein 1 (Tjp1) locus encoding ZO-1 suggesting ZO-1 is a direct downstream target of Foxa2 (Wederell et al., 2008). Therefore, we performed immunostaining for ZO-1 and ZO-2 in Foxa2loxP/loxP; Isl1-Cre embryos at 4–5 som. Apico-lateral localization of ZO-1 and ZO-2 was observed in epithelium of both formed and prospective foregut endoderm of control embryos (Fig. 7A). Foxa2loxP/loxP; Isl1-Cre endodermal segments composed of Foxa2- negative cells exhibited reduced to absent expression of both ZO-1 and ZO-2 (Fig. 7B). Isolated groups of Foxa2-expressing cells embedded in Foxa2-negative epithelium displayed persistent expression of ZO-1 (data not shown) and ZO-2 (Fig. 7C,D), demonstrating that the observed effect is specific and cell autonomous.

To investigate whether Foxa2 is required to maintain endodermal identity in foregut, immunohistochemistry was used to investigate expression of the transcription factor Sox17, a marker of endodermal identity (Kanai-Azuma et al., 2002). Control embryos with 4–5 som displayed several sites of specific positive signal for Sox17 in addition to foregut endoderm, including endocardium and cells throughout the mesenchyme that are often in close association with the foregut (Fig. 7E). Low levels of Sox17 expression reported in vascular and endocardial endothelium could account for some of the positive signal detected outside of the foregut (Sakamoto et al., 2007). Co-immunohistochemistry with an anti-E-cadherin antibody was performed to positively identify Sox17-expressing foregut endoderm. Foxa2loxP/loxP; Isl1-Cre embryos with 5 som displayed patches of E-cadherin-positive endoderm lacking Sox17 expression in the ventral foregut (Fig. 7E) that correlated with regions of reduced-to-absent Foxa2 expression (data not shown). Sox17 expression was unchanged in dorsal foregut of mutant embryos.

Shortened foregut and defective midline cardiac fusion in Foxa2loxP/loxP; Isl1-Cre mutants indicated a role for Foxa2 during ventral morphogenesis. Embryos homozygous null for the transcription factor Gata4 similarly exhibit cardia bifida and lack of foregut closure (Kuo et al., 1997). Additionally, Foxa2 directly binds to the Gata4 locus in adult mouse liver, supporting a role for Foxa2 as a direct upstream regulator of Gata4 expression (Wederell et al., 2008). Gata4 protein was detected in endoderm associated with cardiogenic mesoderm near the anterior intestinal portal of control embryos with 4–5 som (Fig. 7G). Endodermal Gata4 expression was greatly reduced in this region of Foxa2loxP/loxP; Isl1-Cre mutants and correlated with areas of reduced-to-absent Foxa2 protein levels (Fig. 7H,I).

To quantify shortened foregut pockets in Foxa2^loxP/loxP; Isl1-Cre mutants, foregut pocket length was measured in 8.5 dpc control and mutant embryos at 5 som. Analysis revealed that Foxa2^loxP/loxP; Isl1-Cre mutants exhibited significantly shorter rostral-caudal foregut pocket length when compared to somite-matched controls (supplementary material Fig. S4).

To investigate potential causes for shortened foregut pocket length in mutants, foregut endoderm cell proliferation and apoptosis were compared between 8.5 dpc Foxa2^loxP/loxP; Isl1-Cre and control embryos (4–6 som) (Fig. 8). To quantify proliferation, cells positive for the mitotic marker phosphohistone H3 (P-H3) were counted and calculated as a percentage of total foregut pocket cell number. Total foregut pocket P-H3 levels were significantly reduced in mutant versus control populations and correlated with significantly reduced proliferation in ventral foregut (Fig. 8 A–C). TUNEL assays were performed to assay apoptosis. TUNEL-positive cells were counted and calculated as a percentage of total foregut pocket cell number. Total foregut pocket TUNEL levels were significantly increased in mutant versus control populations and correlated with significantly increased levels in dorsal and ventral foregut (Fig. 8 D–F).

Embryological approaches in mouse have shown that prechordal plate is required to support identity and growth in overlying forebrain ectoderm (Shimamura and Rubenstein, 1997; Camus et al., 2000) with previous work suggesting involvement of Foxa2 in these roles. Anterior embryo patterning phenotypes in mouse, such as holoprosencephaly, are correlated with disrupted expression of prechordal plate marker genes including Foxa2 (Ang and Rossant, 1994; Weinstein et al., 1994; Filosa et al., 1997; Bachiller et al., 2000; Petryk et al., 2004; Nishioka et al., 2005; Warr et al., 2008). A role for Foxa2 in prechordal plate during anterior embryo morphogenesis has also been supported by the phenotype of double Gsc null-Foxa2 heterozygous mutants which exhibited abnormal forebrain morphology similar to that observed in Foxa2^loxP/loxP; Isl1-Cre mutants (Filosa et al., 1997). Until now, genetic approaches to directly investigate patterning roles of prechordal plate, and roles for Foxa2 in this tissue, have been encumbered by lack of a prechordal plate-specific Cre line that is exclusive of other axial mesendoderm components such as notochord. Isl1Cre ablates in prechordal plate and in ventral foregut endoderm, but not other axial mesendoderm. Isl1-Cre-mediated ablation of Foxa2 results in anterior embryo patterning phenotypes previously seen consequent to perturbations of axial mesendoderm. These anterior patterning phenotypes were observed in mutant embryos which had preserved expression of Foxa2 in ventral foregut endoderm. Conversely, notochord and dorsal foregut endoderm markers were present in embryos in which Foxa2 was absent from ventral foregut endoderm. Taken together, our data suggest a primary role for Foxa2 in prechordal plate for notochord development and/or maintenance. Ideally, it would be preferable to complement our studies with a Cre recombinase which would be selectively active in ventral foregut endoderm, and not in prechordal plate. However, our attempts to utilize Nkx2.5cre to this effect were unsuccessful, owing to perdurance of Foxa2 protein in ventral foregut endoderm, despite successful recombination as indicated by lineage indicators.

Previous studies of dorsal foregut endoderm have focused on elucidating pathways that stimulate budding of organ primordia, such as pancreas (Zorn and Wells, 2009). Little is known as to factors which pattern dorsal foregut endoderm at earlier stages. Foxa2^loxP/loxP; Isl1-Cre mutants exhibit reduced expression of the dorsal foregut marker Hlxb9. Lack of Isl1-Cre activity in dorsal foregut indicates a non-autonomous requirement for Foxa2 in patterning of dorsal foregut endoderm. We observed a lack of correlation between loss of Foxa2 in ventral foregut endoderm and downregulation of Hlxb9, suggesting a role for prechordal plate in maintenance of dorsal foregut endoderm identity. It is interesting to speculate that mutual signaling between contiguous dorsal endoderm and notochord may be required for specification or maintenance of both, and that this crosstalk may be disrupted in Foxa2^loxP/loxP; Isl1-Cre mutants.

Previous work supports a model in which Shh secreted from axial mesendoderm induces Foxa2 expression in overlying prospective floor plate; Foxa2 in turn activates expression of Shh in this tissue. Shh secreted from floor plate is critical for dorsal-ventral neural patterning, such that homozygous null Shh mutants exhibit holoprosencephaly (Chiang et al., 1996; Sasaki et al., 1997; Ding et al., 1998; Matise et al., 1998; Epstein et al., 1999; Jeong and Epstein, 2003; Wilson and Maden, 2005). Floor plate secretes factors that function as a molecular boundary preventing asymmetric left-sided signaling networks from diffusing across the embryonic midline. Laterality phenotypes have been associated with disruption of this function in mice (Meno et al., 1998; Tsukui et al., 1999). Reduced or absent expression of Shh in the anterior embryo and Foxa2 in neural groove indicated that this pathway was disrupted in Foxa2^loxP/loxP; Isl1-Cre mutants which exhibited holoprosencephaly and laterality defects.

Surgical ablation of endodermal regions corresponding to prechordal plate in both chick and mouse models results in defective forebrain patterning, cardiac fusion, and foregut morphogenesis, supporting a role for prechordal plate in development of anterior-ventral tissues also affected in Foxa2^loxP/loxP; Isl1-Cre mutants (Camus et al., 2000; Withington et al., 2001). Morphological abnormalities in Foxa2^loxP/loxP; Isl1-Cre mutants also resemble what has been reported for Nodal loss-of-function mouse mutants in which defects have been attributed to deficient specification and/or formation of anterior axial mesendoderm populations, including prechordal plate (Roebroek et al., 1998; Lowe et al., 2001; Vincent et al., 2003; Dunn et al., 2004; Liu et al., 2004). While Shh is required to promote survival within a subpopulation of cells comprising prechordal plate (Aoto et al., 2009), Shh null mutants exhibit heart and foregut developmental anomalies that occurred later and are less severe than those we have observed in Foxa2^loxP/loxP; Isl1-Cre embryos (Litingtung et al., 1998; Smoak et al., 2005), identifying a Shh-independent Foxa2-mediated role for prechordal plate in ventral embryo morphogenesis.

We observed downregulation of a number of proteins which could, at least in part, account for observed defects in morphogenesis, proliferation, and survival of ventral endoderm in Foxa2^+/loxP; Isl1-Cre mutants. Isl1-Cre mutants showed corresponding loss of Foxa2 and Gata4 expression in prospective ventral foregut endoderm cells near the site of closure. Homozygous null Gata4 mutants exhibit ventral morphogenesis defects similar to those of Foxa2^+/loxP; Isl1-Cre mutants, including cardia bifida (Kuo et al., 1997; Molkentin et al., 1997). Chimeric rescue experiments demonstrated a specific requirement for Gata4 in endoderm for foregut closure (Kikugawa et al., 1997). Previous work showing that Foxa2 directly binds to the Gata4 locus support a role for Foxa2 as a direct upstream regulator of Gata4 expression (Wederell et al., 2008). Cardiac morphogenesis defects arising from ablation of Foxa2 in prechordal plate and endoderm highlight an important caveat when using Islet1Cre to investigate the role of a gene in heart development. If the gene of interest overlaps with Islet1Cre in prechordal plate and/or ventral foregut endoderm, as well as the second heart field, any observed requirement for cardiogenesis consequent to ablation with Islet1Cre cannot be attributed to a requirement within the second heart field.

Foregut endoderm of Foxa2loxP/loxP; Isl1-Cre mutants also showed reduced expression of ZO-1 and ZO-2 tight junction scaffolding components. Foxa2 has been shown to bind the Tjp1 locus, encoding ZO-1, supporting a role for Foxa2 as a direct upstream regulator of ZO-1 expression (Wederell et al., 2008). Homozygous ZO-1 null mutants die at 11.5 dpc, due to defects in yolk sac vascular development. Homozygous null ZO-2 mutants die at 9.5 dpc with disrupted tight junctions detected by 6.5–7.5 dpc. Both ZO-1 and ZO-2 null mutants displayed elevated levels of apoptosis throughout the embryo (Katsuno et al., 2008; Xu et al., 2008). In a mammary epithelial cell line, tight junction formation was only eliminated when both ZO-1 and ZO-2 function was downregulated, identifying a requirement the expression of both factors for tight junction assembly. Barrier function was compromised, although adherens-mediated cell adhesion appeared intact (Umeda et al., 2006; Umeda et al., 2004). The foregoing suggest that diminished expression of both ZO-1 and ZO-2 observed in the foregut of Foxa2loxP/loxP; Isl1-Cre embryos is likely to disrupt tight junction assembly and identifies a role for Foxa2 in endoderm for maintenance of the foregut epithelium tight junction network. The in vivo requirement for ZO-1 and ZO-2 to prevent apoptosis suggests that loss of these factors in Foxa2loxP/loxP; Isl1-Cre embryos may contribute to increased cell death observed in mutant foregut.

All mouse lines have been previously described: Foxa2^loxP (Sund et al., 2000); Isl1-Cre (Yang et al., 2006); Rosa^26R-lacZ (Soriano, 1999). Foxa2^loxP/loxP or Foxa2^loxP/loxP; Rosa^26R/26R females were crossed with Foxa2^+/loxP; Isl1-Cre males to generate embryos for morphological analysis, X-gal staining, in situ hybridization, immunohistochemistry. Matings were timed with noon on day of vaginal plug detection defined as 0.5 dpc.

Embryos for whole-mount X-gal staining were collected between 7.5–10.5 dpc, fixed in 4% paraformaldehyde on ice 20–45 minutes depending on embryo size, stained according to described protocol (Nagy et al., 2003).

Embryos for in situ hybridization were collected between 8.0–8.5 dpc, fixed overnight at 4°C in 4% paraformaldehyde, dehydrated through a methanol series and stored in 100% methanol at −20°C. Whole-mount in situ hybridization was performed according to described protocol (Nagy et al., 2003) using the following probes: mouse Brachyury, Fgf8, Foxa2, Gsc, Shh. Samples were post-fixed overnight at 4°C in 4% paraformaldehyde. Post-fixed embryos processed for histology were embedded in paraffin wax and transverse sections cut at 8 µm.

Embryos for cryosectioning and immunohistochemistry were collected between 8.0–8.5 dpc, fixed in 4% paraformaldehyde on ice 20–30 minutes, washed twice in phosphate-buffered saline (PBS) containing 0.1% bovine serum albumin, and incubated through a PBS-sucrose series (5%, 12.5%, 20%). Incubations were performed on ice for 15–60 minutes each. Embryos were embedded in 1:1 mixture of 20% sucrose and Tissue-Tek O.C.T. Embedding Compound (Sakura Finetek). Sections were cut at 10 µm unless otherwise noted.

Immunohistochemistry was performed according to described protocol (Nagy et al., 2003). 4′,6-diamidino-2-phenylindole (DAPI, Invitrogen, D3571) was used to counterstain DNA. Primary antibodies used: β-galactosidase (1:400, MP Biomedicals, 56028); Brachyury (1:400, Santa Cruz Biotechnology, 17743); Foxa2 (1:100, Santa Cruz Biotechnology, 6554); Isl1 (1:1000, Abcam, 20670); phosphohistone H3 (1:200, Upstate Biotechnology, 06-570); Shh (1:150, Santa Cruz Biotechnology, 9024). Cleaved Caspase-3 (1:100, Cell Signaling Technology, cat. 9661); E-cadherin (1:300, Sigma-Aldrich, cat. U3254); Gata4 (1:50, Santa Cruz Biotechnology, cat. 1237); Hlxb9 (1:8000, generously provided by Sam Pfaff, Salk Institute, La Jolla, CA); SOX17 (1:300, R&D Systems, cat. AF1924); ZO-1 (1:100, Invitrogen, cat. 40-2200); ZO-2 (1:200, Invitrogen, cat. 71-1400). Secondary antibodies were from the Alexafluor series (Invitrogen Molecular Probes).

Sections for TUNEL-mediated detection of apoptosis were prepared as for immunohistochemistry and stained with In Situ Cell Death Detection Kit, TMR red according to manufacturer's instructions (Roche Applied Science, 12156792910).

We thank the Evans lab, Carl de Luca, Michel Puceat, Ju Chen, Kunfu Ouyang and Deborah Yelon for helpful discussion and critical reading of the manuscript. We thank Maretoshi Hirai, Xingqun Liang, Jody Martin, David Stachura for technical guidance and Paul Chen for technical assistance. Probes provided by Richard Behringer, Bernhard Herrmann, Gail Martin, Deepak Srivastava, Kenneth Zaret. This work supported by NIH 5 R01 HL074066-03 (S.M.E.), NIH 1 F32 HL090291-01 (Z.H.), and AHA 0825280F (Z.H.).