Wnt5a and Wnt11 regulate mammalian anterior-posterior axis elongation

In vertebrate embryos, the basic body plan is established during gastrulation through the formation of three germ layers and a major rearrangement of cells. Gastrulation is initiated with primitive streak (PS) formation at the posterior of the embryo (Tam and Behringer, 1997). The PS elongates anteriorly during gastrulation. Within the PS, cells of the primitive ectoderm delaminate and undergo an epithelial-mesenchymal transition (EMT) (Arnold and Robertson, 2009) to give rise to mesodermal and endodermal structures (Beddington and Robertson, 1999; Tam and Loebel, 2007). Defective EMT results in severe phenotypes, as cells fail to migrate away from the PS and instead accumulate (Sun et al., 1999; Ciruna and Rossant, 2001). Besides its essential role in gastrulation, EMT is also crucial for other developmental processes, wound healing, fibrosis and cancer metastasis (Thiery et al., 2009). However, the molecular mechanisms that regulate EMT remain to be fully elucidated.

Anterior-posterior (A-P) body axis elongation requires continuous generation and proper organization of axial and paraxial mesoderm. Fate-mapping studies have revealed that cells emerging from the anterior PS give rise to the axial mesoderm that forms the notochord (Lawson et al., 1991). The notochord serves as an important signaling center that patterns the overlying neuroectoderm and adjacent somites through secretion of signaling molecules such as sonic hedgehog (Shh) (Echelard et al., 1993; Chiang et al., 1996; reviewed by Cleaver and Krieg, 2001). Anterior to the PS, an indentation called the posterior notochord (PNC, also referred to as the node or ventral node) forms at mid-gastrula stages (Blum et al., 2007). The PNC provides instructive signals during mouse left-right determination and is continuous with the notochord (Sulik et al., 1994; Nonaka et al., 2002). Cells of the trunk and tail notochord are derived from the PNC (Sulik et al., 1994; Yamanaka et al., 2007; Ukita et al., 2009). However, since PNC cells are ciliated and proliferate at a very low rate, a region consisting of the anteriormost streak and the posteriormost part of the node/PNC, called the node/streak border (NSB), is considered the origin of notochord precursors (Sulik et al., 1994; Bellomo et al., 1996; Cambray and Wilson, 2002, 2007; Ukita et al., 2009). PNC cells express Noto and the epithelial marker E-cadherin (Abdelkhalek et al., 2004; Plouhinec et al., 2004; Yamanaka et al., 2007). Genetic fate-mapping studies revealed that the Noto-expressing cells in the PNC contribute to notochord precursor cells (NPCs) that form the trunk and tail notochord as well as some of the paraxial mesoderm (Yamanaka et al., 2007; Ukita et al., 2009). In addition, paraxial mesodermal cells are derived from more posterior regions of the PS and migrate anteriorly to give rise to the somites (Tam and Beddington, 1987; Tam et al., 2000).

The trunk notochord is derived from the PNC and forms by convergent extension (CE) (Sausedo and Schoenwolf, 1994; Ybot-Gonzalez et al., 2007), whereas the tail notochord (posterior notochord extension) is generated by peripheral PNC cells actively migrating posteriorly (Sulik et al., 1994; Yamanaka et al., 2007). Although continuous production of notochord cells by the NPCs is crucial to establish the correct body plan (Ukita et al., 2009), the molecular and cellular regulation of NPC generation remains unknown. The Wnt signaling pathways regulate many fundamentally important developmental processes, including gastrulation. Although Wnts can signal through the canonical pathway mediated by β-catenin, the β-catenin-independent Wnt/planar cell polarity (PCP) pathway has recently emerged as a regulator of many key processes (Song et al., 2010; Gao et al., 2011). The PCP pathway was initially identified in Drosophila melanogaster, where it controls the establishment of a polarity within an epithelial sheet perpendicular to apical-basal polarity. A group of core PCP components has been identified in Drosophila, and these are functionally conserved in vertebrates (Tree et al., 2002; Seifert and Mlodzik, 2007; McNeill, 2010). In vertebrates, PCP regulates CE of axial and paraxial mesodermal cells during axis elongation, and defects in PCP signaling result in a shortened and widened A-P axis (Greene et al., 1998; Keller, 2002; Wallingford et al., 2002; Ybot-Gonzalez et al., 2007; Roszko et al., 2009; Song et al., 2010; Mahaffey et al., 2013). CE movements are characterized by polarized cell behavior as cells converge along one axis, intercalate and finally extend along the axis perpendicular to the initial cell movements (Keller, 2002). In mammals, PCP signaling is essential for many morphological processes, including closure of the neural tube, inner ear hair cell polarity, left-right asymmetry and elongation of the A-P axis (Kibar et al., 2001; Curtin et al., 2003; Montcouquiol et al., 2003; Wang et al., 2006; Ybot-Gonzalez et al., 2007; Borovina et al., 2010; Hashimoto et al., 2010; Song et al., 2010; Gao et al., 2011). Mutations in human core PCP components cause congenital neural tube defects (NTDs) such as spina bifida (Kibar et al., 2007; Lei et al., 2010). However, despite the importance of PCP signaling, the mechanism underlying Wnt regulation of PCP signaling is poorly understood.

Wnt5a and Wnt11 have been shown to regulate vertebrate PCP signaling. Loss of Wnt5a results in a severe shortening of the A-P axis and limb truncations (Yamaguchi et al., 1999a). Wnt5a regulates PCP establishment by inducing phosphorylation of Vangl2, a core PCP protein (Gao et al., 2011). In Xenopus embryos, expression of a dominant-negative Wnt11 results in NTDs and CE defects (Tada and Smith, 2000) and the zebrafish wnt11 mutant silberblick exhibits CE defects in the developing notochord that result in a shortened A-P axis (Heisenberg et al., 2000). However, mouse Wnt11^−/− embryos do not show PCP defects, in contrast to the silberblick mutant (Majumdar et al., 2003), raising the question of whether Wnt11 regulates PCP signaling in mammals.

Here we show that, upon loss of both Wnt5a and Wnt11, the phenotype of Wnt5a^−/− embryos is exacerbated as the A-P axis is further shortened, indicating functional redundancy of these two signaling molecules during axis formation in the murine embryo. We further show that Wnt5a and Wnt11 regulate CE, EMT and cell migration, disruption of which results in defects in notochord formation and in patterning of the neural tube and somites.

wnt11 is required to regulate axis elongation through PCP in zebrafish (Heisenberg et al., 2000). The lack of similar defects in mouse Wnt11 mutants (Majumdar et al., 2003) suggests that Wnt11 might play redundant roles with Wnt5a during mouse gastrulation. We first examined the expression of Wnt5a and Wnt11 in early mouse embryos and confirmed that Wnt5a is expressed in a caudal-to-rostral gradient in the PS (supplementary material Fig. S1A-D) (Yamaguchi et al., 1999a), whereas Wnt11 expression is more restricted (supplementary material Fig. S1E-H) (Kispert et al., 1996). Wnt11 was expressed in the PNC and in the forming heart, as previously reported (Kispert et al., 1996). To investigate a possible redundancy between Wnt5a and Wnt11 during early embryonic development, we generated Wnt5a; Wnt11 double-mutant mouse embryos. Wnt5a^−/−; Wnt11^−/− embryos were found at the expected Mendelian ratio between E8.5 and E10.5 (Fig. 1A-L), and died between E10.5 and E11.5. The phenotype of Wnt5a^−/−; Wnt11^−/− embryos became apparent at ∼E8.5 (Fig. 1A-D) and was much more severe than that of the Wnt5a single mutant, as the A-P axis was further shortened (Fig. 1G,H,K,L). Therefore, Wnt11 plays redundant roles with Wnt5a in regulating early mouse embryonic development. There was no difference between Wnt5a^−/−; Wnt11^+/− and Wnt5a^−/− embryos in terms of morphology and marker gene expression (data not shown).

A-P axis elongation is driven by PCP-mediated CE movements within the notochord and paraxial mesoderm. Defects in PCP signaling result in a shortened and widened A-P axis (Ybot-Gonzalez et al., 2007; Song et al., 2010). To test whether similar defects were caused by loss of Wnt5a and Wnt11, we examined the length/width ratio at E8.5. Wnt5a^−/−; Wnt11^+/− embryos displayed a decrease in the length/width ratio, which was enhanced in Wnt5a^−/−; Wnt11^−/− mutants (supplementary material Fig. S2A-D). Next, we investigated the expression of Shh, which marks the notochord and floor plate at E9.5 (Echelard et al., 1993), to assess notochord formation (Fig. 1M-P). Surprisingly, the notochord was not only reduced in length but also in width in Wnt5a^−/−; Wnt11^−/− embryos (Fig. 1M′-P′). To further understand the observed notochord malformation, we investigated expression of the transcription factor brachyury (T), a marker of the nascent mesoderm and the notochord (Wilson et al., 1995). At E8.0, we found a sparse and irregular T expression pattern, and fewer cells expressed T in Wnt5a^−/−; Wnt11^−/− than in control embryos (Fig. 2A,B).

To test whether the reduction in notochord length in Wnt5a^−/−; Wnt11^−/− embryos was due to defective CE, we performed in vitro DiI labeling experiments to trace cells of the PNC. After 12 h of in vitro culture, most labeled PNC cells had migrated anteriorly along the midline in the control (Fig. 2C,C′; n=27/33). However, in Wnt5a^−/−; Wnt11^−/− embryos, the labeled PNC cells lacked directionality and migrated in more random directions (Fig. 2D,D′; n=5/6). Thus, the distance of migration away from the initial labeling position (PNC) was shorter. To confirm that notochord cells had been successfully labeled by DiI injections, we sectioned the labeled embryos and found that T-positive cells were labeled by DiI (supplementary material Fig. S3A-D). Additionally, we observed labeled cells in the paraxial mesoderm (Fig. 2C,D; supplementary material Fig. S3C,D). Taken together, these results show that Wnt5a and Wnt11 are required for proper CE of notochord and paraxial mesoderm cells.

During lengthening of the notochord along the A-P axis, the cells in the axial midline elongate along the mediolateral axis and intercalate to form two adjacent rows of notochord cells. This is typical of highly conserved CE movements. The PCP pathway has been shown to be instructive for the correct alignment of notochord cells during CE and is regulated by Wnt5a (Qian et al., 2007; Ybot-Gonzalez et al., 2007; Gao et al., 2011). Therefore, we tested whether notochord cells exhibit asymmetric localization of the core PCP protein Vangl1, as the asymmetric localization of core PCP proteins is a hallmark of PCP signaling (Klein and Mlodzik, 2005; Song et al., 2010). In control embryos, in many notochord cells, as marked by T expression, Vangl1 was localized asymmetrically with an increase on cell membranes perpendicular to the A-P axis as compared with the parallel axis. Additionally, the notochord cells were elongated mediolaterally (Fig. 2E,E′). By contrast, in Wnt5a^−/−; Wnt11^−/− embryos, Vangl1 protein levels were much reduced. Very few cells showed asymmetrically localized Vangl1 and the cells were less elongated, with irregular morphology (Fig. 2F,F′). These results indicate that PCP signaling, and hence CE regulated by PCP, is reduced in the absence of Wnt5a and Wnt11.

The early heart tube of the Wnt5a^−/−; Wnt11^−/−embryo failed to undergo normal rightward looping and remained as a linear tube at E9.5 (Fig. 3A-C). This malformation may lead to embryonic lethality and confirmed the findings of a recent study (Cohen et al., 2012). Nkx2.5, a homeobox transcriptional regulator that is essential for heart looping (Lyons et al., 1995), was expressed throughout the myocardial layer of the heart tube in control embryos at E9.5 (Fig. 3A). Normal Nkx2.5 expression was maintained in the heart tube of Wnt5a^−/−; Wnt11^−/− embryos at E9.5 (Fig. 3C), raising the question of whether laterality determination is affected by loss of Wnt5a and Wnt11. We have previously shown that PCP breaks bilateral symmetry by positioning cilia to the posterior side of the PNC cells (Nonaka et al., 2005; Song et al., 2010), and in Xenopus embryos knocking down Wnt11b results in left-right defects (Walentek et al., 2013). In order to test whether Wnt5a and Wnt11 control PCP-regulated laterality determination, we examined asymmetric gene expression of Nodal and Pitx2 (Lowe et al., 1996; Logan et al., 1998). We observed normal, left-sided expression of Nodal and Pitx2 in nine out of ten Wnt5a^−/−; Wnt11^−/− embryos, with only one embryo showing an inverted, right-sided Nodal expression (supplementary material Fig. S4A-C; data not shown). Furthermore, we were able to detect asymmetric localization of Vangl1 in the PNC cells of Wnt5a^−/−; Wnt11^−/− embryos (supplementary material Fig. S4D,E). Thus, the absence of Wnt5a and Wnt11 signaling resulted in left-right patterning defects only at very low penetrance. These results suggest that other Wnt ligands expressed in the PS might compensate for the loss of Wnt5a and Wnt11 during laterality determination. Our data indicate that the observed heart-looping defect occurs further downstream in development and is not regulated by laterality determination in this case.

Our analysis of notochord formation and Shh expression in the Wnt5a^−/−; Wnt11^−/− embryos suggests that the reduction in notochord size resulted in a smaller floor plate, as Shh expressed in the notochord induces floor plate formation (Echelard et al., 1993; Placzek et al., 1993). Since Shh secreted from the notochord and the floor plate is required for repression of dorsal and induction of ventral cell fates in the neural tube (NT) (Echelard et al., 1993; Chiang et al., 1996), we tested whether the reduction of Shh expression in Wnt5a^−/−; Wnt11^−/− embryos resulted in a dorsalized NT. Indeed, expression of Pax6, which is a dorsal NT marker repressed by Shh (Ericson et al., 1997), was expanded ventrally in the Wnt5a^−/−; Wnt11^−/− embryos (supplementary material Fig. S5A-C). Conversely, expression of the ventral markers Nkx2.2, Olig2 and Nkx6.1, which depend on Shh induction (Chiang et al., 1996), was reduced in the Wnt5a^−/−; Wnt11^−/− mutants (supplementary material Fig. S5D-L). To exclude the possibility that the reduction in ventral cells was the result of a reduced number of cells in the NT, we quantified the ratio of labeled to unlabeled NT cells and were able to confirm a dorsalized NT in the Wnt5a^−/−; Wnt11^−/− embryos (supplementary material Fig. S5M).

Apart from a shortened A-P axis, we also observed smaller and irregular somites in Wnt5a^−/−; Wnt11^−/− embryos, indicating defects in somite formation. In addition, the A-P axis reduction in Wnt5a^−/−; Wnt11^−/− embryos was most severe in posterior regions, where cells of the PS and later the tailbud give rise to the paraxial mesoderm, the precursor of the somites (Tam and Beddington, 1987; Tam et al., 2000). To understand the defects in somitogenesis in the absence of Wnt5a and Wnt11, we first investigated the expression of Wnt3a and Fgf8, which are essential for the formation of paraxial mesoderm (Takada et al., 1994; Crossley and Martin, 1995). The expression levels of Wnt3a and Fgf8 were similar in all investigated specimens (Fig. 3D-I). Furthermore, Wnt/β-catenin signaling indicated by TOPGAL (DasGupta and Fuchs, 1999) and Fgf signaling indicated by Spry4 expression (Minowada et al., 1999) were normal in Wnt5a^−/−; Wnt11^−/− embryos (supplementary material Fig. S6A-F). However, the expression domains of Wnt3a and Fgf8 were significantly shortened along the A-P axis in Wnt5a^−/−; Wnt11^−/− as compared with control and Wnt5a^−/−; Wnt11^+/− embryos (Fig. 3D-I).

This shortening might result from a reduction in paraxial mesoderm formation and/or defects in mesodermal cell migration. Mesodermal movements are controlled by T, which is expressed in the PS (Wilson et al., 1995). In the absence of Wnt5a and Wnt11, we observed a reduction of T expression as compared with the Wnt5^−/− mutant, indicating defects in mesoderm formation and migration (Fig. 3J-L). To investigate paraxial mesoderm formation we compared the expression of Tbx6 at E8.5 and E9.5 (Fig. 3M-R) (Chapman and Papaioannou, 1998). In the Wnt5a^−/−; Wnt11^+/− embryos, the Tbx6 expression domain was reduced, and this was further enhanced in Wnt5a^−/−; Wnt11^−/− embryos at E8.5 (Fig. 3N,O). Additionally, there was a reduced level of Tbx6 expression at E9.5 (Fig. 3Q,R), suggesting defects in paraxial mesoderm formation in the absence of Wnt5a and Wnt11.

Our results indicated that the defects in somitogenesis were not due to abnormal Wnt3a and Fgf signaling. As Wnt5a has been suggested to regulate the proliferation of paraxial mesoderm cells (Yamaguchi et al., 1999a), we examined cell proliferation and survival in the paraxial mesoderm by anti-phospho-histone H3 (PHP3) antibody staining and TUNEL assay, respectively. At E8.5, no significant difference in cell proliferation or survival was observed in Wnt5a^−/−; Wnt11^−/− and control embryos (supplementary material Fig. S7A-F). At E9.5, cell death was significantly increased in the Wnt5a^−/−; Wnt11^−/− embryos, especially in the tail bud (supplementary material Fig. S7G-I), and proliferation in the paraxial mesoderm of Wnt5a^−/−; Wnt11^−/− embryos was slightly reduced (supplementary material Fig. S7J-L). At E10.5, a substantial increase in cell death and almost no cell proliferation were detected in Wnt5a^−/−; Wnt11^−/− embryos (supplementary material Fig. S7M-R). These results indicate that Wnt5a and Wnt11 promote cell proliferation and survival independently of Wnt3a and Fgf signaling in the paraxial mesoderm.

In Wnt5a^−/−; Wnt11^−/− embryos the somites were not only smaller but also irregular, suggesting that, apart from the defects in cell proliferation and survival, somite patterning was also disrupted. To test this, we examined the expression of Uncx4.1 (also known as Uncx), which marks the posterior compartment of mature somites (Mansouri et al., 1997). Uncx4.1 was only expressed in the caudal part of somites in control embryos (Fig. 4A). However, Uncx4.1 expression became continuous and sometimes asymmetric on the left and right side in the newly formed somites of Wnt5a^−/−; Wnt11^−/− embryos (n=3/6; Fig. 4C). These results suggest that the periodic and synchronized segmentation of the paraxial mesoderm and the A-P polarity of the somite were disrupted in the absence of Wnt5a and Wnt11.

As Notch signaling plays fundamental roles in the formation and patterning of somites (Conlon et al., 1995; Kageyama et al., 2007), we examined Notch signaling in Wnt5a^−/−; Wnt11^−/− embryos. Mesp2 acts downstream of Notch signaling to specify the rostral compartments of somites and its expression is restricted to the future rostral somitic half (Saga et al., 1997). Mesp2 expression was greatly reduced in Wnt5a^−/−; Wnt11^−/− but not in Wnt5a^−/−; Wnt11^+/− embryos (Fig. 4D-F). In addition, expression of Hes7 and Lfng, which are Notch signaling effectors (Bessho et al., 2001; Dale et al., 2003), was also significantly diminished in Wnt5a^−/−; Wnt11^−/− embryos (Fig. 4G-L), indicating that Notch signaling activity was reduced in the absence of Wnt5a and Wnt11. The expression of Notch1 and Notch2 was greatly reduced in the paraxial mesoderm of Wnt5a^−/−; Wnt11^−/− embryos (Fig. 4M-R). However, the expression levels of the Notch ligands Dll1 and Dll3 were normal in the paraxial mesoderm of Wnt5a^−/−; Wnt11^−/− embryos (Bettenhausen et al., 1995; Dunwoodie et al., 1997) (Fig. 4S-X). These results suggest that the reduced Notch signaling activity is likely to be due to diminished expression of Notch1 and Notch2 in the Wnt5a^−/−; Wnt11^−/− embryos.

Consistent with the asymmetric expression pattern of Uncx4.1 (Fig. 4C), we observed lateral asymmetric expression patterns of Mesp2 and Lfng in Wnt5a^−/−; Wnt11^−/− embryos (Fig. 4F,L), indicating defects in somite synchronization. As retinoic acid (RA) signaling is required for the bilateral symmetry of somite formation and establishment of the determination front (Vermot et al., 2005; Vermot and Pourquie, 2005; Sirbu and Duester, 2006), we examined the expression of Raldh2 (Aldh1a2), the RA biosynthetic enzyme, and of Cyp26b1, the RA degrading enzyme (Zhao et al., 1996; White et al., 2000). Raldh2 was expressed in the segmented region of the control embryos and its expression was reduced in the somites of Wnt5a^−/−; Wnt11^−/− embryos at E9.5 (supplementary material Fig. S8A-C), whereas Cyp26b1 expression in the tail bud remained normal (supplementary material Fig. S8D-F). Taken together, these results indicate that significantly reduced Notch and RA signaling inhibit proper somite formation in Wnt5a^−/−; Wnt11^−/− embryos.

The thinner notochord and smaller somites in Wnt5a^−/−; Wnt11^−/− embryos prompted us to further investigate the molecular and cellular mechanisms underlying the defects in notochord and somite formation, focusing on posterior mesoderm formation and migration during gastrulation. Interestingly, we noticed an ectopic accumulation of cells in the Wnt5a^−/−; Wnt11^−/− embryos at the posterior of the PNC at E8.5 (Fig. 5A-B′), suggesting that Wnt5a and Wnt11 are additionally required for the specification and/or migration of mesodermal cells. We speculated that this ectopic cell mass might be a result of accumulated precursors of midline and paraxial mesodermal cells failing to migrate to their proper destination. Therefore, impaired notochord and somite formation might be caused by a lack of normal motile precursor cells. To test this hypothesis, we characterized the accumulated cells in detail.

Cell labeling and genetic lineage-tracing experiments have revealed that NPCs are derived from relatively quiescent PNC cells (Sulik et al., 1994; Yamanaka et al., 2007; Ukita et al., 2009). In addition, the source of the axial and paraxial mesoderm precursors is believed to reside at the NSB (Sulik et al., 1994; Bellomo et al., 1996; Cambray and Wilson, 2007), which is where we observed the ectopic accumulation. A similar accumulation of cells has already been reported in Wnt5^−/− embryos (Yamaguchi et al., 1999a). However, in the Wnt5a^−/−; Wnt11^−/− embryos, it appeared to be more densely populated with cells and less organized. As in the Wnt5a mutants, the accumulated cells expressed T, indicating a mesodermal fate (Fig. 5A-B′) (Yamaguchi et al., 1999a). To further test whether these cells were already committed to a specific mesodermal fate, we investigated the expression of Noto, which is expressed exclusively in the PNC and notochord (Fig. 5C,C′) (Abdelkhalek et al., 2004). The accumulated cells did not express Noto, indicating that they had not differentiated into notochord cells in the Wnt5a^−/−; Wnt11^−/− embryos (Fig. 5D,D′). Next, we investigated whether the accumulated cells had adopted a paraxial fate by analyzing the expression of Tbx6 (Chapman and Papaioannou, 1998). Tbx6 was expressed in the dorsal portion of the ectopically accumulated cells, suggesting that these cells were PS cells and/or early paraxial mesodermal cells, but occupying the midline position (Fig. 5E,F). Thus, in the absence of Wnt5a and Wnt11 a mixed population consisting of paraxial (T-positive, Tbx6-positive) and axial (T-positive, Tbx6-negative) mesodermal cells accumulates posterior to the PNC.

Regulated by signals from the PS, mesodermal cells are generated in the PS from pluripotent epiblast stem cells that also generate ectoderm and endodermal cells (Gardner and Beddington, 1988). At later developmental stages, the differentiation potential of the descendants of these early embryonic stem cells becomes progressively restricted (Loebel et al., 2003). Bipotential axial stem cells in the caudal lateral epiblast (CLE) (Wilson et al., 2009; Takemoto et al., 2011) generate caudal neural plate and paraxial mesoderm. Therefore, we hypothesized that Wnt5a and Wnt11 might regulate mesodermal cell fates by regulating the differentiation of late epiblast stem cells or CLE. To test this, we investigated the expression of Sox17 and Sox2, which are endodermal and ectodermal markers, respectively (Collignon et al., 1996; Kanai-Azuma et al., 2002). Sox17 expression was normal in the endodermal cell layer in Wnt5a^−/−; Wnt11^−/− embryos, as compared with the control (supplementary material Fig. S9A,B). However, as mentioned above, T was downregulated in the dorsal PS in Wnt5a^−/−; Wnt11^−/− mutants (Fig. 6A,B). By contrast, Sox2 expression was increased ventrally and posteriorly. The opposing expression gradients of T and Sox2 in the presumptive paraxial mesodermal in Wnt5a^−/−; Wnt11^−/− embryos indicated an increase in neural ectodermal and a decrease in mesodermal cell fate determination (Fig. 6A,B). To exclude the possibility that reduced cell proliferation in the PS region caused the notochord defect, we stained embryos for PHP3. Cell proliferation appeared normal in this region in Wnt5a^−/−; Wnt11^−/− embryos compared with the controls (supplementary material Fig. S9C,D). Therefore, we concluded that the loss of Wnt5a and Wnt11 led to reduced mesoderm formation.

As mesoderm generation from the ectodermal epiblast requires EMT (Shook and Keller, 2003) and cell accumulations in the PS have been reported in mouse mutants with defective EMT during gastrulation (Arnold et al., 2008), we hypothesized that incomplete EMT might be a cellular mechanism underlying the reduced mesoderm formation in Wnt5a^−/−; Wnt11^−/− embryos. To test this, we first examined the expression of Snail (also known as Snai1), a transcription factor that regulates the induction of EMT (Barrallo-Gimeno and Nieto, 2005). However, Snail expression was similar in Wnt5a^−/−; Wnt11^−/− and control embryos (supplementary material Fig. S9E-F′), suggesting that initiation of EMT occurred normally. Next, we tested whether EMT takes place in late gastrulating embryos by examining the integrity of the basement membrane and the expression of epithelial and mesenchymal cell markers (Hay and Zuk, 1995; Smyth et al., 1999; Mendez et al., 2010). In the region posterior to the PNC in wild-type embryos at E7.75-E8.25, the basement membrane had broken down, as indicated by fragmented laminin staining (supplementary material Fig. S10A-F′). Intact basement membrane was observed anterior to the PNC (supplementary material Fig. S10A-F′). In addition, we found that basement membrane breakdown was associated with downregulation of E-cadherin expression and upregulation of vimentin expression in this region (supplementary material Fig. S10D-F′). As EMT is characterized by a downregulation of E-cadherin (Hay and Zuk, 1995) and upregulation of vimentin (Mendez et al., 2010) expression, our results indicate that active EMT occurs posterior to the PNC in wild-type embryos as a continuous process from mid to late gastrula stages, and thus NSB and anterior PS cells both undergo EMT in late gastrulation. However, in Wnt5a^−/−; Wnt11^−/− embryos, cells within the ectopic accumulation expressed E-cadherin ectopically (Fig. 6C-D″), while vimentin expression was reduced (Fig. 6E,F). These cells also expressed T at similar levels to that in notochord cells, indicating that these cells were early progenitors of the notochord (Fig. 6D,D″, arrows). These findings indicate that EMT in this region generates notochord progenitor cells and that, in the absence of Wnt5a and Wnt11, even though EMT was induced and the cells were on their way to becoming mesoderm, they failed to fully acquire mesenchymal character. Increased cell-cell adhesion due to persistent E-cadherin expression in these cells prevents them from migration. In support of this, we found that ZO-2 (Tjp2) expression, which marks tight junctions in epithelial and endothelial cells, was increased in the ectopically accumulated cells, although normal expression was maintained in the endodermal layer in the absence of Wnt5a and Wnt11 (Fig. 6G,H).

Proper cell migration not only requires downregulation of E-cadherin but also a functional extracellular matrix (ECM). After cell fate determination and germ layer formation, different germ layers are separated by ECM containing fibronectin (FN; also known as Fn1), which plays important roles in cell adhesion and migration (Pankov and Yamada, 2002). Deposition of FN has been linked to CE movements and PCP (Goto et al., 2005). In zebrafish, loss of Vangl2 and Prickle1a led to decreased FN levels (Dohn et al., 2013). In the absence of FN, mouse embryos die during gastrulation and show defects in A-P axis formation (George et al., 1993). In control embryos, a layer of FN-containing ECM separated the endodermal or ectodermal layer from the mesodermal layer (Fig. 6I). However, in Wnt5a^−/−; Wnt11^−/− embryos the FN-containing layers were missing where mesodermal cell fate determination was incomplete (Fig. 6J). This supports the idea that reduced PCP signaling and incomplete mesodermal fate determination in the absence of Wnt5a and Wnt11 leads to the reduction of FN expression. Taken together, EMT defects in the Wnt5a^−/−; Wnt11^−/− embryos reduce axial and paraxial mesoderm formation and hamper cell migration, causing ectopic cell accumulation and a failure of axial elongation, with smaller and irregular somites.

E-cadherin downregulation during EMT requires an active p38 MAPK pathway (Zohn et al., 2006). To test whether p38 (Mapk14) is activated in the Wnt5a^−/−; Wnt11^−/− embryos, we investigated the phosphorylation of p38 (P-p38). We observed a substantial reduction of P-p38 at the posterior end of the Wnt5a^−/−; Wnt11^−/− embryo (Fig. 7A-B′). This indicates that Wnt5a and Wnt11 are necessary to activate p38 MAPK and downregulate E-cadherin in the posterior embryo. In mouse F9 teratocarcinoma cells, Wnt5a has been shown to induce p38 phosphorylation (Ma and Wang, 2007). In order to establish whether Wnt11 also activates p38 and if Wnt5a and Wnt11 act synergistically in this process, we treated primary mouse embryonic fibroblast (MEF) cells with recombinant Wnt5a and/or Wnt11 protein (Fig. 7C,D). We found that Wnt5a can induce p38 phosphorylation and that Wnt5a and Wnt11 together induce more robust p38 phosphorylation. To further investigate this interaction, we tested whether JNK signaling, which has been identified as a target of Wnt5a and Wnt11 signaling, is involved in the activation of p38 (Pandur et al., 2002; Yamanaka et al., 2002). We observed an increase of c-Jun phosphorylation (a target of JNK) induced by Wnt5a (Fig. 7C,E). To block c-Jun activity we employed SP600125, an inhibitor of the kinase JNK (Mapk8). However, as SP600125 treatment induced phosphorylation of p38 even at very low concentrations, we were unable to test whether c-Jun activity is required for Wnt5a/Wnt11-mediated p38 phosphorylation (data not shown). Next, we investigated whether Rho GTPase activation is required for the phosphorylation of p38, as Wnt5a and Wnt11 have been shown to activate Rho GTPases (Zhu et al., 2006). Wnt11 treatment increased RhoA activation significantly, which was further enhanced by Wnt5a treatment in MEF cells (Fig. 7F,G). To test whether RhoA activity is required for p38 phosphorylation, we employed the RhoA inhibitor Rhosin (Shang et al., 2012). Inhibition of RhoA by Rhosin prevented Wnt5a/Wnt11-mediated p38 phosphorylation in MEF cells (Fig. 7H,I). We therefore concluded that Wnt5a and Wnt11 act synergistically to enhance p38 phosphorylation and that RhoA signaling can mediate this activity.

Finally, we examined whether Wnt5a and Wnt11 regulate EMT through PCP signaling, as both EMT and PCP pathways regulate cytoskeletal reorganization and play important roles in cell migration. We analyzed Vangl1^−/−; Vangl2^−/− embryos and were unable to detect a similar accumulation of cells posterior to the notochord as we observed in Wnt5a^−/−; Wnt11^−/− embryos (supplementary material Fig. S11A,B). However, there was a slight increase in E-cadherin in this region, although the increase was less profound than that in the Wnt5a^−/−; Wnt11^−/− mutant. In addition, in the Vangl1^−/−; Vangl2^−/− MEF cells, Wnt5a and Wnt11 could still induce p38 phosphorylation, albeit at lower levels than in wild-type MEF cells (supplementary material Fig. S11C,D; Fig. 7C). Therefore, we conclude that PCP also regulates EMT, but Wnt5a and Wnt11 regulate EMT through multiple pathways including PCP. Taken together our results indicate that, in the absence of Wnt5a and Wnt11, the cells of the prospective midline and paraxial mesoderm fail to migrate due to incomplete EMT.

Here we have found that Wnt5a and Wnt11 play redundant roles in regulating multiple developmental processes in the late gastrulating embryo. Although Wnt11^−/− embryos do not have obvious developmental defects, the phenotypes of the early Wnt5a^−/−embryos were significantly enhanced by loss of Wnt11. Wnt11 is specifically expressed in the PNC (Kispert et al., 1996) and this expression accounts for its redundant role with Wnt5a in early embryonic development, as shown here. This explains why the defects observed are restricted to the notochord and the paraxial mesoderm that ingress in this region. Wnt5a and Wnt3a are both expressed strongly in the PS (Takada et al., 1994; Yamaguchi et al., 1999a). Wnt3a signals through the β-catenin-mediated canonical Wnt signaling pathway, whereas Wnt5a does not (Logan and Nusse, 2004; Kikuchi et al., 2012; Topol et al., 2003; Westfall et al., 2003). Therefore, even though both are required to regulate mouse gastrulation, the Wnt5a^−/− and Wnt3a^−/−embryos exhibit distinct defects (Takada et al., 1994; Yamaguchi et al., 1999a). In Wnt5a^−/−; Wnt11^−/−embryos, Wnt3a is still expressed at normal levels and the canonical Wnt signaling activity is largely normal, as judged by TOPGAL activity (DasGupta and Fuchs, 1999), confirming that both Wnt5a and Wnt11 act through β-catenin-independent pathways in early mouse embryos.

During the formation of the trunk notochord at E8.5, cell proliferation and cell death appeared normal in Wnt5a^−/−; Wnt11^−/− embryos (supplementary material Fig. S7). However, at later stages we observed an increase in cell death and reduced proliferation. This reduction in cell survival might be secondary to the heart defects observed and the reduction in Shh expression due to earlier defects in notochord formation. In the early embryos, the PCP pathway controls CE that is required for notochord formation (Ybot-Gonzalez et al., 2007). Both wnt5a and wnt11 regulate CE in zebrafish (Rauch et al., 1997; Heisenberg et al., 2000), but in mammalian embryos the role of Wnt11 in CE had not been demonstrated. We found in this study that the Wnt11 function can be mostly fulfilled by Wnt5a and that Wnt5a and Wnt11 both regulate CE of the trunk notochord cells and paraxial mesodermal cells through PCP (Fig. 2). However, in Wnt5a^−/−; Wnt11^−/− embryos, asymmetrical localization of Vangl1 was still observed in some notochord cells and NT closure was normal, suggesting that other Wnt ligands present in early embryos (Yamaguchi, 2008) can also regulate PCP. Consistent with this, we did not detect any defects in neurulation, which requires PCP function. We also observed normal asymmetrical localization of Vangl1 in the ciliated PNC cells and low penetrance of left-right asymmetry defects in the Wnt5a^−/−; Wnt11^−/− embryos.

CE defects in PCP mutants result in shortened and broadened notochord and floor plate (Greene et al., 1998; Ybot-Gonzalez et al., 2007). However, in Wnt5a^−/−; Wnt11^−/− embryos, the notochord is shortened, but thinner, resulting in reduced Shh expression and hence dorsalization of the NT and somites. We found that this peculiar phenotype is caused by reduced migration and generation of NPCs from the PNC. The source of trunk and tail notochord progenitors is believed to reside in the border region between the posterior PNC and the anterior PS (Cambray and Wilson, 2007). Since the PNC (node)/streak border region is believed to contain a population of stem cells, it is also possible that both trunk and tail notochord formation are dependent on proper migration of the descendants of these stem cells, which is regulated by Wnt5a and Wnt11 (Cambray and Wilson, 2007). Thus, the thinner notochord is likely to be caused by defects in two processes regulated by Wnt5a and Wnt11. First, as PCP is required to provide directionality to migrating cells, disruption of PCP due to loss of Wnt5a and Wnt11 leads to reduced directed anterior and posterior migration of NPCs, such that there are fewer cells to form trunk and tail notochord. Second, as the PNC is organized as an epithelium and the migrating cells display mesenchymal properties, this process involves EMT. Indeed, in the PNC (node)/streak border region, we have observed breakdown of the basement membrane and a reduction of epithelial markers (supplementary material Fig. S10). An increase in cell adhesion molecules in the NPCs of Wnt5a^−/−; Wnt11^−/− embryos indicates that some of the precursors of the notochord are unable to fully adopt mesenchymal cell properties, resulting in fewer cells available for trunk and tail notochord formation. A similar EMT process has recently been shown to control the generation of limb mesenchyme cells during limb bud formation (Gros and Tabin, 2014).

Wnt/β-catenin signaling is a potent inducer of EMT in early gastrulation and it suppresses E-cadherin expression. Interestingly, blocking Wnt/β-catenin signaling by removing β-catenin in Noto-expressing cells led to persistent strong E-cadherin expression in NPCs and reduction of notochord elongation (Ukita et al., 2009), suggesting that Wnt/β-catenin signaling also regulates EMT in the generation of NPCs at late gastrulation stages. Consistent with this, it has been reported that Wnt3a controls paraxial mesoderm formation and Tbx6 expression (Nowotschin et al., 2012). The data in that report reveal that cells modulate the expression of Wnt3a as they traverse the PS: as they enter they upregulate Wnt3a, and then subsequently downregulate Wnt3a as they exit. The study suggests that reduced tail bud formation in Wnt3a^−/− mutants might be caused by the first steps of late paraxial mesoderm formation, such as ingression/EMT (Nowotschin et al., 2012). Compared with Wnt/β-catenin signaling, Wnt5a and Wnt11 are weak inducers of EMT. EMT occurred at early gastrulation stages in Wnt5a^−/−; Wnt11^−/− embryos and the phenotypes were only manifested at later stages, possibly due to progressive accumulation of a weak EMT defect.

In addition to reduced migration of NPCs, we propose that reduced A-P axis elongation is also caused by reduced paraxial mesoderm formation in the absence of Wnt5a and Wnt11 (Fig. 5). In the PS of the late gastrula embryo there is continuous EMT (Arnold and Robertson, 2009). Cells ingress and then delaminate from the epiblast ectoderm to form a loose mesenchyme that will later form the paraxial mesoderm. In this process, the ingressing cells turn off the Sox2 enhancer N1, whereas Tbx6 expression is upregulated (Takemoto et al., 2011). Once formed, the mesoderm is then separated from the neural ectoderm by a layer of FN-containing basement membrane. In Wnt5a^−/−; Wnt11^−/− embryos, reduced T expression and expanded Sox2 expression suggest that reduced paraxial mesoderm formation might also be caused by incomplete EMT. FN deposition requires dynamic cell rearrangements and the increase in cell adhesion that we observed in this region might thus be responsible for the lack of FN (Dzamba et al., 2009). The defective EMT therefore led secondarily to incomplete formation of the FN-containing basement membrane, reduced axial mesoderm migration causing a thinner notochord, and reduced paraxial mesoderm migration causing the formation of smaller and irregular somites. Therefore, in this study we have identified that Wnt5a and Wnt11 are required to control EMT in the late gastrula embryo.

Our results suggest that Wnt5a and Wnt11 regulate EMT through RhoA-mediated activation of p38. However, we cannot exclude the possibility that other pathways downstream of Wnt5a and Wnt11 might also be involved in regulating EMT, as Wnt5a and Wnt11 can stimulate multiple non-canonical Wnt pathways (reviewed by Yang, 2012). We show that JNK signaling is activated by the addition of Wnt5a and Wnt11, but it is not clear whether JNK activation is required for p38 activation. As JNK and p38 signaling pathways often act in parallel (Widenmaier et al., 2009; Das et al., 2011; Chen et al., 2013; Tanaka et al., 2013) and JNK signaling has been shown to be involved in the induction of EMT in various tissues (Pallet et al., 2012; Chen et al., 2013), both p38 and JNK might regulate EMT during embryonic A-P axis elongation.

As Wnt/β-catenin signaling is required for T expression and T regulates Tbx6 expression (Yamaguchi et al., 1999b), the reduced T expression dorsally may cause ventrally extended Sox2 expression, which is normally suppressed by Tbx6. One scenario is that Wnt5a and Wnt11 can signal weakly through the canonical Wnt pathway. However, the strong Wnt3a expression and normal TOPGAL activity in Wnt5a^−/−; Wnt11^−/− embryos (Fig. 3F; supplementary material Fig. S6F) suggest that Wnt5a and Wnt11 might modulate T and Tbx6 expression independently of the canonical Wnt pathway. Further experiments are required to clarify this point.

Wnt5a, Wnt11, Vangl1 and Vangl2 mutant mouse strains and associated genotyping methods have been described previously (Yamaguchi et al., 1999a; Majumdar et al., 2003; Song et al., 2010). Animal care and experimental animal procedures were performed in accordance with the NHGRI institutional standards and were approved by the Institutional Animal Care and Use Committee, NHGRI, and by the NIH Institutional Review Board.

Whole-mount in situ hybridization was performed according to standard protocols (Topol et al., 2003).

RNA probes have been described previously: Shh (Echelard et al., 1993), T (Wilson et al., 1995), Nkx2.5 (Lyons et al., 1995), Wnt3a (Takada et al., 1994), Wnt5a (Yamaguchi et al., 1999a), Wnt11 (Kispert et al., 1996), Fgf8 (Crossley and Martin, 1995), Tbx6 (Chapman and Papaioannou, 1998), Uncx4.1 (Mansouri et al., 1997), Mesp2 (Saga et al., 1997), Hes7 (Bessho et al., 2001), Lfng (Dale et al., 2003), Noto (Abdelkhalek et al., 2004), Notch1 (Conlon et al., 1995), Notch2 (Bessho et al., 2001), Dll1 (Bettenhausen et al., 1995) and Dll3 (Dunwoodie et al., 1997). For immunofluorescence staining, embryos were fixed in 4% paraformaldehyde for 30 min at 4°C and processed for cryosectioning and immunostaining according to standard protocols (Song et al., 2010). Unless indicated otherwise, each result presented was observed in at least two individual specimens with a penetrance of 100%. Primary antibodies used: brachyury (1:150; R&D Systems, AF2085), E-cadherin (1:200; R&D Systems, AF748), fibronectin (1:500; gift from K. Yamada, Bethesda, MD, USA), Sox17 (1:50; R&D Systems, AF1924), Sox2 (1:500; Abcam, ab97959), Vangl1 [1:50 (Song et al., 2010)], ZO-2 (1:100; Cell Signaling, 2847), β-catenin (1:500; BD Biosciences, 610153), phospho-p38 (1:200; Cell Signaling, #9211), Pax6 (1:50), Pax7 (1:50), Nkx2.2 (1:100), Islet1 (1:50), Lim1/2 (1:50), Nkx6.1 (1:50) (Developmental Studies Hybridoma Bank), phospho-histone H3 (1:500; Cell Signaling, 9071) and Olig2 (1:100; Abcam, ab33427). Secondary antibodies were obtained from Life Technologies and used at 1:400.

Apoptosis was analyzed using the ApopTag Apoptosis Kit (S7111, Millipore). Images were acquired using a Zeiss LSM 510 NLO Meta system and Zeiss Imager D2.

Embryos were dissected in DMEM containing 10% FBS and 25 mM HEPES pH 7.4. DiI (Sigma) was injected at 0.2 mg/ml in 3 M sucrose using a glass micropipette into cells of the PNC. Embryos were subsequently cultured for 12 h at 37°C, 5% CO₂ in a roller culture (Rotator Genie, Scientific Industries) in 25% rat serum and 75% culture medium (DMEM with 10% FBS) (Rivera-Pérez et al., 2010). After culture, the embryos were analyzed using a Zeiss V20 stereomicroscope.

Mouse embryonic fibroblasts (MEFs) were grown to confluence and serum starved for 2 h. Cells were then incubated in human recombinant WNT5A (125 ng/ml; R&D Systems) and WNT11 (200 ng/ml; R&D Systems) for the indicated time period. Cells were lysed in lysis buffer (20 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.5% Nonidet P-40) with Halt Protease Inhibitor Cocktail (Thermo Scientific) and Halt Phosphatase Inhibitor Cocktail (Thermo Scientific), and incubated with antibodies against p38 (1:1000; Cell Signaling, #9212), phospho-p38 (1:1000; Cell Signaling, #9211), c-Jun (1:1000; Cell Signaling, 9165) or phospho-c-Jun (1:1000; Cell Signaling, 9261).

RhoA activity was determined using a RhoA Activity Kit (New East Biosciences, 80601). RhoA activity was inhibited by incubating the cells for 2 h in 1 µM Rhosin (EMD Millipore). JNK activity was blocked using 500 nM-15 µM SP600125 (Sigma).

We thank the Y.Y. laboratory for stimulating discussions, Kenneth Yamada for providing the fibronectin antibody, and Shelley Hoogstraten-Miller and Irene Ginty for providing rat serum.