β-catenin–mediated signaling and cell adhesion in postgastrulation mouse embryos

Authors


Abstract

β-Catenin plays two major roles during the development of multicellular organisms. It is the downstream effector of the canonical Wnt signaling cascade, which is involved in many developmental processes and in tumor formation. Additionally, it is linked to classic cadherins and is required for the correct assembly and function of adherens junctions. β-Catenin loss of function mutants show early gastrulation lethality. To address the role of β-catenin in postgastrulation stages and to overcome the early embryonic lethality, we performed conditional gene targeting, using Cdx1::Cre, a newly established mouse line. By this approach, β-catenin was depleted in the entire posterior embryo after the gastrulation process at embryonic day 8.0, when the three germ layers were established. We observed defects in signaling and adhesion which are temporarily separated. At an early event, known targets of Wnt/β-catenin are down-regulated in the paraxial mesoderm. Moreover, Fgf8 and Wnt3a, the key players of the segmentation process, are down-regulated in the neural ectoderm (NE). Wnt3a expression was rescued in mutant embryos by exogenous Fgf and inhibition of Fgf signaling in wild-type embryos resulted in Wnt3a down-regulation. Based on these results, we assume the existence of an autoregulatory feedback loop in the NE where Fgf8 regulates Wnt3a, which in turn, by means of β-catenin, maintains Fgf8 expression. In later stages, the lack of β-catenin caused a progressive posterior disintegration. We found that β-catenin is required for the correct localization of N-cadherin at the membrane of neural ectodermal cells and that its absence causes a disintegration of the neural tube. Developmental Dynamics 239:191–199, 2009. © 2009 Wiley-Liss, Inc.

INTRODUCTION

The canonical Wnt signaling pathway controls cell fate decision in many developmental processes in metazoa (for review, see Nusse,2005; Barker,2008). In the adult organism, it is required for maintaining tissue homoestasis, whereas its disregulation has been implicated in various types of cancer and other disorders (Clevers,2006). Upon binding of Wnt ligands to their correspondent receptor complex of the Frizzled (Fz) and low-density lipoprotein receptor-related protein (LRP) family and several intermediate steps, hypophosporylated β-catenin accumulates in the cytoplasm and translocates to the nucleus where it acts as a transcriptional coactivator of the Tcf/Lef transcription factor family (Hecht and Kemler,2000; Willert and Jones,2006). In addition to its central role in transducing the Wnt signal from the cytoplasm to the nucleus, β-catenin is also an integral component of classic cadherin complexes at intercellular adherens junctions (Vestweber and Kemler,1984; Stemmler,2008). Wnt signaling and E-cadherin regulation are closely related in the process of epithelial to mesenchymal transition (EMT) during gastrulation (Huber et al.,1996; Kemler et al.,2004). Here, Wnt/β-catenin signaling induces mesodermal gene expression such as T-brachyury and Lef1 (Rivera-Perez and Magnuson,2005; Tam and Loebel,2007), and down-regulation of E-cadherin in the prospective mesodermal cells is ensured by the transcription factor Snail, which in turn depends on active Wnt signaling (Yook et al.,2005,2006). The zygotic depletion of β-catenin in mouse embryos results in early gastrulation lethality, with no mesoderm formation and a blockage in the anterior–posterior axis formation and head development (Haegel et al.,1995; Huelsken et al.,2000). We were interested in the role of β-catenin in the posterior embryo after the gastrulation process, when the three germ layers have been established at embryonic day (E) 8.0. BAT-Gal reporter embryos (Maretto et al.,2003) show at this stage Wnt activity in all three germ layers of the posterior embryo. With the use of Cdx1::Cre, a recently established transgenic Cre line (Hierholzer and Kemler,2009), we deleted β-catenin in all three germ layers of the posterior embryo. In this line, the Cre recombinase is driven by a promoter fragment of the Caudal-related homeobox gene (Cdx) 1, which recapitulates the endogenous Cdx1 expression in gastrulating embryos (Lickert and Kemler,2002). Cre mRNA is detected during gastrulation from E7.5 onward. Recombination activity can be observed after the establishing of the three germ layers at E8.0 as monitored in ROSA reporter embryos. This allowed us to overcome the β-catenin null phenotype and to study β-catenin function in signaling and adhesion in the emerging mesoderm and neural ectoderm (NE).

RESULTS

Ablation of β-Catenin in the Cdx1::Cre Expression Domain Leads to Progressive Loss of Posterior Structures

To verify and to visualize active Wnt signaling in a spatial manner in early postgastrulation embryos, we made use of the Wnt reporter line BAT-Gal (Maretto et al.,2003) and compared the expression with the Cdx1::Cre recombination activity (Fig. 1). Both transgenic lines exhibited rather superposable expression patterns, indicating that Cdx1::Cre may delete β-catenin–mediated Wnt signaling in all three germ layers of the posterior embryo. Interestingly, the posterior NE exhibited both, active Wnt and Cre recombinase activity. When Cdx1::Cre and floxed β-catenin mice were crossed, mutant embryos look phenotypically quite normal at E8.0–E8.5. Occasionally, mutant embryos were shortened in their anterior–posterior axis at E8.5 (compare Fig. 3A and B). Clear morphological defects became evident at E9.5 (Fig. 2). As expected from the Cre expression domain (Fig. 1B), in mutant embryos the trunk was shortened lacking a tail bud (Fig. 2D; wild-type [WT], 2A) and the whole trunk posterior to the heart became progressively disintegrated between E10.5 to E11.5 (Fig. 2E,F; WT, 2B,C). After E11.5, mutant embryos died possibly due to a detachment of the umbilical cord. The mutant embryos phenocopied to a large extent Wnt3a loss of function mutants (Takada et al.,1994) providing further in vivo evidence that Wnt3a signals through β-catenin in patterning posterior structures.

Figure 1.

Comparison of Wnt activity and the expression pattern of Cdx1::Cre at early headfold stage embryos (embryonic day [E] 8.0). A: Wnt activity was monitored by using the BAT-Gal reporter line. B: Cre activity was detected by crossing Cdx1::Cre to ROSA26 reporter mice. Recombination is observed in the primitive streak region (PS) in all three germlayers where also high Wnt activity is detected.

Figure 2.

Ablation of β-catenin by Cdx1::Cre leads to a progressive disintegration of the posterior embryo. DF: Mutant embryos at embryonic day (E) 9.5 (D), E10.5 (E), and E11.5 (F) loose progressively posterior structures caudal to the heart. AC: Corresponding WT embryos.

Figure 3.

Immunohistochemical analysis shows presence of membrane bound β-catenin in the mutant posterior embryo. A,B: No morphological differences are apparent between wild-type (WT; A) and mutant (Mut; B) embryos at embryonic day (E) 8.5. CF: A reduction of β-catenin levels is observed in the posterior embryo of mutants (D,F) in comparison to WT (C,E), but β-catenin is still present at the membrane in cells of the three germ layers. G,H: T-brachyury is detected in the presomitic mesoderm of WT embryos (G) and is abolished in mutant embryos (H) except of notocordal cells where Cre is not expressed. C–H: Consecutive sagittal sections of the embryos shown in (A) and (B). NE, neural ectoderm; PM, presomitic mesoderm; NC, notocord; Gut, hindgut diverticulum.

Lack of β-Catenin in the Posterior Embryo Affects Genes for Mesoderm Specification

Mesoderm-specific gene expression was studied by whole-mount in situ hybridization experiments at E8.5 when mutant embryos looked phenotypically still quite normal compared with WT littermates (Fig. 3A,B). Lef1 and T-Bra are typically expressed in mesodermal cells and both depend on β-catenin–mediated transcription (Yamaguchi et al.,1999; Filali et al.,2002). A clear down-regulation of Lef1 and T-brachyury became apparent in mutant embryos indicating an efficient depletion of β-catenin in the posterior region (Fig. 4A–D). T-bra expression in the notochord was not affected (Figs. 4C,D, 3H) because Cre is not expressed in notochord cells and their ancestors (unpublished data). An important factor for mesoderm specification and maintenance is Snail, which was found to be expressed in a oscillating manner throughout the mesoderm (Dale et al.,2006). We found that Snail was greatly reduced in the embryonic mesoderm of mutant embryos (Fig. 4E,F). These results demonstrate that mesoderm-specific gene expression is altered in mutant embryos, which raised the possibility that E-cadherin expression may be deregulated due to incomplete epithelial-mesenchymal transition. However, no difference between mutant and WT embryos was observed for E-cadherin mRNA which localizes to cells of the surface ectoderm and the gut primordium (Fig. 4G,H). This indicates that there are other factors not related to β-catenin which maintain mesodermal fate and E-cadherin repression in postgastrulation embryos. Taken together, these results demonstrate that the down-regulation of Wnt/β-catenin target genes is an early event of the ablation of β-catenin. Interestingly, in immunohistochemistry β-catenin can still be detected at the cell membrane of mesodermal and neural ectodermal cells at this stage of development when T-brachyury protein was greatly reduced (Fig. 3C–H).

Figure 4.

Marker gene analysis by in situ hybridization in wild-type (WT) and mutant (Mut) embryos at embryonic day (E) 8.5. AF: Genes specific for mesodermal cells, such as Lef1 (A), T-Brachyury (C), and Snail (E) are down-regulated in mutant embryos (B,D,F). Shown is the dissected caudal part of the embryo anterior to the left. G,H: No change in E-cadherin expression is observed between WT (G) and Mut (H) shown by transversal section through the caudal part of the embryo.

Fibroblast Growth Factor 8 and Wnt3a Expression in the Neural Ectoderm Is Down-regulated in a Temporally Distinct Manner

BAT-Gal reporter embryos reveal an active Wnt domain in the posterior NE, which overlaps with Cdx1::Cre–mediated depletion of β-catenin. Because the role of canonical Wnt signaling in early NE specification is less well understood, we examined genes which are expressed in the NE and found that the key player of the oscillation clock, Wnt3a and Fgf8 are down-regulated upon ablation of β-catenin (Fig. 5A–J). However, the Fgf8 mRNA level was already reduced at E8.0 (Fig. 5A–D). Changes in Wnt3a levels were first detectable later in development, at E8.5 (Fig. 5E–J). This may indicate that Fgf8 is directly regulated by β-catenin, whereas the delayed decline of Wnt3a may be the consequence of the reduced Fgf8 expression. Of interest, however, neither Fgf8 nor Wnt3a were required for neural ectoderm specification indicated by Sox2 mRNA, which did not show differences between WT and mutant embryos (Fig. 5K,L). Hence, canonical Wnt signaling is involved in the regulation of Wnt3a and Fgf8 expression in the NE.

Figure 5.

Genes expressed specifically in the neural ectoderm detected by in situ hybridization in wild-type (WT) and mutant (Mut) embryos. A,B,E,F: Fgf8 is down-regulated at embryonic day (E) 8.0 in mutant embryos (B) compared with WT (A), while no major changes are observed in Wnt3a expression at this stage (E,F). G,H: However, Wnt3a down-regulation becomes apparent at E8.5. K,L: No changes in Sox2 expression between WT (K) and Mut (L) is observed. Shown is the dissected caudal part of the embryo anterior to the left. C,D,I,J: Fgf8 (C,D) and Wnt3a (I,J) are mainly expressed in the neural ectoderm as shown by transversal sections.

Recombinant Fgf Rescues Wnt3a Expression in the Neural Ectoderm of Mutant Embryos

Considering the results described above and the coexpression of Wnt3a and Fgf8 in the NE, we assumed that Fgf8 expression may directly depend on β-catenin activity. In turn, Fgf8 activates Wnt3a and by this, maintains active Wnt signaling in the NE. To investigate this possibility, we performed embryo cultures using recombinant Fgf2 to activate Fgf signaling in the embryo. We used KSOM as a minimal medium to avoid side effects by serum containing media. Upon Fgf2 stimulation, Wnt3a expression was restored in the tail bud region of mutant embryos after 16 (Fig. 6D, insert in D) and 24 hr of incubation (Fig. 6F, arrow), WT embryos are shown in Figure 6C and 6E, respectively. Of interest, the sole presence of Fgf2 in a minimal medium was sufficient to boost growth and development of both mutant and WT embryo, compared with KSOM (Fig. 6A,B) which underlined the importance of the role of Fgf signaling in these developmental stages.

Figure 6.

Wnt3a expression depends on Fgf signaling. CF: Wnt3a expression is rescued in mutant embryos after 16 hr (D) or 24 hr (F) using recombinant Fgf2 in embryo culture; wild-type (WT; C,E). A,B: A control experiment without Fgf2 wild-type (WT; A) and mutant (B) embryos is shown. G,H: Treatment for 6 hr with the Fgf inhibitor SU5402 of WT embryos results in Wnt3a down-regulation (H) in comparison to the control (G). I,J: Fgf8, which may depend on Wnt3a is slightly down-regulated (J) compared with control (I). Arrows, Wnt3a expression domain at the caudal part of the embryo; arrowheads, Wnt3a expression in the headfold region. D: Insert shows a magnification of the tail bud region.

To further substantiate the Wnt3a rescue experiments, a reverse approach was undertaken by blocking Fgf signaling in WT embryos. Embryo cultures were treated with a pharmacological Fgf inhibitor, SU5402, which efficiently blocks receptor tyrosine kinase activity of the Fgf receptor (Mohammadi et al.,1997). After 6 hr of treatment, a clear decrease of Wnt3a mRNA in the posterior embryo was observed providing further evidence that Fgf signaling regulates Wnt3a expression (Fig. 6G,H). Moreover, Wnt3a expression was reduced in the headfold region as well, where Fgf8 and Wnt3a are also coexpressed (Fig. 6G,I), suggesting a similar interplay of both signaling pathways as well as in the posterior embryo. In addition, Fgf8 levels were slightly decreased as compared with control embryos (Fig. 6I,J). This decrease in Fgf8 could be due to the absence of Wnt3a which suggests that Wnt signaling is required for Fgf8 expression. However, other explanations could be possible.

Loss of β-Catenin Affects Cadherin-Mediated Cell Adhesion

The results described so far clearly demonstrated that the early events of the conditional inactivation of β-catenin in the posterior embryo is due to the lack of the nuclear functions of β-catenin. However, from E9.5 on, mutant embryos lost their posterior tissue architecture (Fig. 2), which suggested that the lack of β-catenin affected additionally cadherin-mediated cell adhesion. Sagittal sections through E9.5 mutant embryos revealed a collapsed structure of the neural tube and an enlarged dorsal aorta, in addition to the lack of tail bud structures. (Fig. 7B). Most of the cells in the posterior embryo had lost β-catenin when compared with wild-type embryos (Fig. 7B,B′,B″) The presence or absence of β-catenin was compared with the membrane localization of E- and N-cadherin. In the gut tube, E-cadherin and β-catenin colocalized at the epithelial cell membrane (Fig. 7B′, B‴,D,D′), which indicates that β-catenin was not ablated here. In contrast, in cells of the mutant neural tube N-cadherin exhibited only weak and irregular membrane localization and was mainly distributed in the cytoplasm, which correlated with the absence of β-catenin in these cells (Fig. 7B″,C,C′). From these results, we conclude that the lack of β-catenin reduces N-cadherin–mediated cell adhesion, which alters the integrity of the neural epithelial cell layer. In line with such a view, mutant cells were detached from the neural epithelial cell layer and the epithelial structure became progressively disintegrated. Delaminated cells from the posterior neural tube were found trapped in the anterior, still intact side of the neural tube at E10.5 (Fig. 8A). Detached cells were negative for β-catenin (Fig. 8D) but N-cadherin protein (Fig. 8E) and mRNA could be detected (Fig. 8F). Without β-catenin, these cells finally undergo apoptosis, which is clearly detectable from E10.5 onward (Fig. 8C). However, no significant increase of apoptosis was found at E9.5 (data not shown). A slightly different picture emerges from the analysis of the surface ectodermal cell layer in the posterior region. Although β-catenin was efficiently deleted (Fig. 8G,G′), E-cadherin was well localized to the cell membrane (Fig. 8H,H′). It is likely, that plakoglobin can substitute the lack of β-catenin in these cells, although other explanations are possible.

Figure 7.

Lack of β-catenin reduced cell surface localization of N-cadherin in the neural ectoderm. Immunohistochemistry performed on paraffin sections from embryos at embryonic day (E) 9.5. AD: Sagittal sections from wild-type (WT) and mutant (Mut) embryos were stained with antibodies against β-catenin (A,B), N-cadherin (C) and E-cadherin (D). Details of A, B, C, and D are highlighted in additional images as indicated. C,C′: N-cadherin in neural tube cells of mutant embryos shows reduced membrane staining. B,B′,B″: β-Catenin is highly reduced in these cells. D,D′,B, ′B‴: In cells of the gut diverticulum, E-cadherin (D,D′) and β-catenin (B′,B‴) exhibit normal membrane localization.

Figure 8.

AH: Progressive disintegration of the posterior embryo lacking β-catenin. A–F: Delaminated cells from the posterior neural tube are found trapped in the anterior side of the neural tube that is still intact. A: Transverse sections of embryonic day (E) 10.5 mutant embryos shown in Figure 2 were stained with hematoxylin and eosin (H&E). A–C: Trapped cells in the neural tube (designated in A) undergo apoptosis detected by TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling) analysis (C, brightfield image in B). D,E,G: Immunohistochemistry performed with antibodies against β-catenin (D,G) and N-cadherin (E), reveals that trapped cells do not have any β-catenin but still express N-cadherin. F: In situ hybridization with N-cadherin riboprobe. D′ and E′ highlight membrane localization of β-catenin and N-cadherin, respectively, in the intact part of the neural tube. G,H: Immunohistochemistry of the posterior part of the embryo (designated in A). G,G′: Staining against β-catenin shows residual protein in the gut and surrounding tissue (G), whereas the surface ectoderm in the posterior region is completely devoid of beta-catenin (highlighted in G′). H,H′: E-cadherin is still present in the gut diverticulum (H) and in the surface ectoderm (highlighted in H′). DA, dorsal aorta.

In summary, the ablation of β-catenin from E9.5 on results in a loss of adhesive property mainly seen in the neural tube due to an improper localization of the N-cadherin molecule at the membrane. This leads to cell shedding and finally to the collapse and apoptosis of the neural epithelium.

DISCUSSION

Here, we show the importance of β-catenin in patterning the posterior embryo after gastrulation when the segmentation process starts. Our results highlight the dual role of β-catenin in Wnt signaling and in cadherin-mediated cell adhesion during these developmental processes. Notably, we have observed a clear temporal order of β-catenin function in mutant embryos, with signaling and the expression of target genes being affected initially, followed by the reduction of cadherin-mediated cell adhesion. Cre transcription and thus, ablation of β-catenin is turned on in the primitive streak at E7.5 (Hierholzer and Kemler,2009). This is followed by the reduced expression of Wnt/β-catenin target genes in mesodermal cells while cell adhesion remains unaffected. These results suggest different β-catenin pools in the cell for either signaling or cell adhesion. The cytoplasmic pool which is subjected to high protein turnover and which is tightly regulated (Aberle et al.,1997; Seidensticker and Behrens,2000) can become exhausted rather fast, after the Cre-mediated inactivation of β-catenin. In comparison, the cadherin-associated and membrane-bound pool of β-catenin is more stable (Clevers,2006), thus persisting longer in the mutants. Our experiments support the notion that these two functions are separately performed by different pools of β-catenin which is in coherence with previous results (Fagotto et al.,1996; Tolwinski and Wieschaus,2004).

The depletion of β-catenin in the posterior embryo demonstrates its pivotal function in transducing Wnt signaling in the paraxial mesodermal cell layer. The role of canonical Wnt signaling in forming and maintaining mesodermal cell fate has been investigated in conditional knock out approaches, targeting the mesodermal cell population (Dunty et al.,2008; Marikawa et al.,2008).

In addition, Cdx1::Cre enabled us to ablate β-catenin in the neural ectoderm (NE) at the onset of the segmentation process. Wnt3a and Fgf8 play a key role in this process and both are expressed in a posterior to anterior gradient in the NE and tail bud (Wahl et al.,2007; Aulehla et al.,2008). Interestingly, conditional ablation of Fgf8 in the paraxial mesoderm (PM) using T-Cre did not result in segmentation defects (Perantoni et al.,2005), in contrast to the ablation of FgfR1 (Wahl et al.,2007), which indicates that PM cells require Fgf-mediated stimuli for segmentation. Wnt3a ligands act on the PM forming a β-catenin gradient and conditional inactivation of β-catenin in the PM results in axial truncation and abolishes the segmentation process (Aulehla et al.,2008). Upon ablation of β-catenin using Cdx1::Cre, we found that both Wnt3a and Fgf8 are down-regulated. However, Fgf8 was lost immediately and Wnt3a expression was abolished later. This suggests that Fgf8 could be regulated directly by β-catenin, whereas Wnt3a reduction was the consequence of the decrease of Fgf8. If so, these two signaling pathways can interact and regulate each other, in a reciprocal manner to maintain their expression in the NE. In accordance to this, we found that the application of an inhibitor of Fgf signaling in WT embryos leads to an early reduction of Wnt3a expression followed by the slight decrease of Fgf8 expression. Although the embryo growth was marginally reduced, a general effect on gene transcription was not observed as indicated by persisting Fgf8 mRNA levels and the embryos were still alive after the treatment. This is in line with the observation that in our mutant embryos at E8.5, proliferation in the posterior part of the embryo was not disturbed by β-catenin ablation since we found normal levels of PCNA compared with WT (data not shown).

Examples for the interaction of the Wnt and Fgf pathways have been described in other experimental systems. In Xenopus laevis animal cap explants, fibroblast growth factors are up-regulated by β-catenin or XWnt3a mRNA. Similar to this, the induction of posterior neural markers by Wnt/β-catenin requires Fgf signaling in Xenopus, indicating that Fgf acts upstream of Wnt (Domingos et al.,2001).

Adhesion defects due to a lack of β-catenin became apparent later in development when the neural ectoderm have formed the neural tube. Upon ablation of β-catenin, cells from the neural tube delaminate and a progressive disintegration of the embryo is observed. This is consistent with the observation which was made in β-catenin–deficient forebrain epithelium where N-cadherin–mediated adhesion was lost upon β-catenin ablation (Junghans et al.,2005). Whereas the membrane localization of N-cadherin is lost in the neural tube upon β-catenin ablation, E-cadherin is still detectable at the membrane of the surface ectoderm in areas where β-catenin is deleted. These apparent differences between the cell surface localization of E- and N-cadherin in the absence of β-catenin requires, however, further analysis. In conclusion, with the conditional gene inactivation approach undertaken, we show that the dual function of β-catenin in signaling and adhesion is essential for establishing and patterning the entire posterior embryo after gastrulation.

EXPERIMENTAL PROCEDURES

Mouse Strains

The generation and expression pattern of the Cdx1::Cre mouse line was described previously (Hierholzer and Kemler,2009). Cdx1::Cre was used to recombine the loxP flanked β-catenin allele (Brault et al.,2001). Following recombination activity of Cdx1-Cre, we crossed animals with the ROSA26 reporter line (Soriano,1999). The BAT-Gal reporter line was used to visualize active canonical Wnt signaling in the embryo by LacZ staining. (Maretto et al.,2003).

Embryo Genotyping and Imaging

Embryonic stages were counting on the day of vaginal plugs as 0.5 days postcoitum. Genotyping was performed by polymerase chain reaction using DNA, isopropanol precipitated from yolk sac or embryonic tissue which was digested in Proteinase K containing buffer. Primer sequences were taken from original publications. A mutant embryo was identified by Cre-positive and homozygous β-catenin floxed alleles. WT indicates Cre positive and a flox/+ genotype. Pictures were taken using a Nikon DXM1200F digital camera.

LacZ Whole-Mount Staining

Detection of β-galactosidase activity was performed as described (Hierholzer and Kemler,2009).

Histology and Immunohistochemistry

For immunohistological analysis, embryos were fixed in 4% paraformaldehyde/phosphate buffered saline (PBS), dehydrated, embedded in paraffin and sectioned at 7 μm. Sections were dewaxed, rehydrated, and treated with hydrogen peroxide to avoid endogenous peroxidase activity. Antigen retrieval was carried out by boiling the sections in Tris/ethylenediaminetetraacetic acid (EDTA; pH 9.0). After blocking in 1% bovine serum albumin, we stained with antibodies against E-cadherin (1:200 dilution), β-catenin (1:200) and N-cadherin (1:200). All antibodies were purchased from BD Transduction Labs. The T-Brachyury antibody (1:200) was a gift from B. Hermann. The EnVision Plus System (Dako) with diaminobenzidine peroxidase substrate (Sigma) was used to detect and amplify the signals. For histological analysis sections were stained with Hematoxylin and Eosin according standard protocols. Specimens were analyzed using a microscope (Axioskop2, Zeiss) equipped with a digital camera (Axiocam, Zeiss).

In Situ Hybridization

Whole-mount in situ hybridization (ISH) was carried out as previously described (Kemler et al.,2004). Digoxigenin-labeled riboprobes were used from plasmids encoding Lef1, T-Bra, Snail (a gift from A. Nieto), E-cadherin, Wnt3a, Fgf8 and Sox2. For ISH on tissue section, embryos were dissected in ice-cold PBS, fixed over night in 4% paraformaldehyde (PFA)/PBS, dehydrated, embedded in paraffin, and sectioned at 7 μm. Sections were acetylated (0.1 M triethanolamine/HCl, pH 8.0, 0.25% acetic anhydride, 15 min at room temperature) and hybridized with digoxigenin labeled riboprobe for N-cadherin overnight at 68°C. The sections were washed (1× standard saline citrate, 50% formamide, 0.1% Tween-20) at 65°C, and blocked in the presence of 10% inactivated goat serum and 1% blocking reagent (Roche) before incubation with alkaline phosphatase (AP) -conjugated anti-digoxigenin-Fab fragments (Roche, 1:2,000). After extensive washing, hybridized riboprobes were revealed using an AP substrate from Roche (BM Purple). After washing and postfixation in PFA, sections were embedded in Kaiser′s glycerol gelatin (Sigma).

Embryo Culture

For the rescue experiments, E8.0 embryos were isolated in M2 medium (Sigma). Subsequently, the embryos were transferred in pre-equilibrated KSOM medium (Chemicon) supplemented either with or without 100 ng/ml recombinant human Fgf2 (R&D Systems). After 16 or 24 hr of incubation at 37°C and 10% CO2, the embryos were washed briefly in PBS, fixed in 4% PFA/PBS and stored in methanol at −20°C until use.

For application of the Fgf inhibitor SU5402 (Calbiochem), E8.5 embryos were isolated in M2 medium and transferred in preequilibrated RPMI1640 medium (Invitrogen) either supplemented with 40 μM SU5402 dissolved in dimethyl sulfoxide (DMSO) or the correspondent amount of DMSO as a control. After 6 hr of incubation at 37°C and 10% CO2 embryos were treated as described above.

TUNEL Analysis

TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling) analysis on sections were performed using a detection kit (Roche) according to the manufacturer's instructions.

Acknowledgements

We thank Dr. Marc Stemmler and Rose Black for helpful discussions and critical reading of the manuscript. We appreciate the excellent technical assistance from Lei Yang and Katie Hansen. A.H. is a PhD student of the Faculty of Biology, University of Freiburg.