The Wnt Coreceptor Ryk Regulates Wnt/Planar Cell Polarity by Modulating the Degradation of the Core Planar Cell Polarity Component Vangl2^*

Abstract

Introduction

Wnt signaling pathways control many fundamentally important processes in both embryonic morphogenesis and adult physiology. Abnormal Wnt signaling leads to birth defects and other diseases, including cancer (1, 2). Among the Wnt pathways, the Wnt/planar cell polarity (PCP)² pathway has recently emerged as an evolutionarily conserved signaling axis, which is required for embryonic morphogenesis (3–6). PCP originally refers to the polarity of epithelial cells within a plane orthogonal to their apical-basal axis and was first discovered in Drosophila melanogaster, where PCP controls polarized processes such as the directional alignment of the wing hair cells. A group of core PCP regulatory genes are identified in Drosophila, and these core PCP components are functionally conserved in vertebrates (7–9). Although a Wnt ligand has not been found to regulate PCP in Drosophila, Wnt5a and Wnt11 are required in PCP-regulated processes in vertebrates (3–6, 10, 11). In mammals, PCP signaling is essential for many critical developmental processes such as convergent extension movements, neural tube closure, inner ear hair cell polarity, and limb elongation (3, 12–22). However, despite the significantly expanded functional spectrum of PCP in vertebrate development, the mechanisms underlying Wnt-mediated regulation of PCP still remain to be elucidated.

Wnt5a, which signals mainly through β-catenin-independent pathways (23, 24), regulates the establishment of PCP by inducing the formation of a receptor complex containing the Wnt-binding receptor tyrosine kinase (RTK) Ror2 and Vangl2 (3), a core PCP component. Mouse mutants of Wnt5a, Ror2, and Vangl2 display shortened long bone cartilage (3, 14, 19, 20, 25). In humans, mutations in components of the Wnt/PCP pathway such as WNT5A and ROR2 cause Robinow syndrome and brachydactyly type B1, which are characterized by short limb dwarfism (26–29). Mutations in both VANGL1 and VANGL2 cause congenital neural tube defects (NTD) such as spina bifida (30, 31). Wnt5a acts through Ror2 to induce Vangl2 phosphorylation in a Wnt5a-dose-dependent manner (3). Similar to Ror2, the receptor-like tyrosine kinase (Ryk) is a member of the RTK family (32) and its Drosophila homolog derailed transduces wnt5 signals in axon pathfinding (33). The interaction of wnt5 and derailed is conserved in mammals (34), raising the possibility that Ryk may transduce Wnt5a signals in PCP. In Xenopus and zebrafish embryos, Ryk may mediate Wnt11-regulated convergent extension (3, 35, 36). However, because the mouse Ryk^−/− and Wnt11^−/− mutants do not display an obvious PCP phenotype (37, 38), it is important to test whether Ryk acts as a core component transducing Wnt signal to regulate PCP.

Here, we present evidence to show that Ryk is a core PCP regulatory component that controls PCP in multiple developmental processes. Ryk interacts with Vangl2 genetically and biochemically and this interaction is enhanced by Wnt5a. Mechanistically, Ryk may regulate PCP by binding to Vangl2 and increasing the stability of Vangl2 protein. Our findings suggest that human mutations in RYK may also be involved in NTD, Robinow syndrome, and brachydactyly.

EXPERIMENTAL PROCEDURES

RESULTS

Ryk and Vangl2 Interact Genetically

Wnt5a interacts with Ryk during axon guidance (34). To address whether Wnt5a also signals through Ryk to regulate other developmental processes, we generated compound mutants of Wnt5a and Ryk. By phenotypic comparison, we found that loss of Ryk did not enhance the phenotypes of the Wnt5a^+/− or Wnt5a^−/− mutants. The Wnt5a^−/−;Ryk^+/− embryos displayed the same phenotypes as the Wnt5a^−/− embryos (supplemental Fig. S1) (20). These results suggest that during development, Ryk may act in the same pathway of Wnt5a and may only transduce a subset of Wnt5a signals.

As Wnt5a has been identified to control PCP in mammals (3, 5), we reasoned that Ryk may also regulate PCP if it acts in the Wnt5a signaling pathway. To test this, we crossed the Ryk^+/− mice with the Vangl2^+/− mice. The Ryk^+/−;Vangl2^+/− mice were viable and fertile and did not display any obvious defects (Fig. 1). However, the Ryk^−/−;Vangl2^+/− embryos exhibited open neural tubes; 61% (n = 17/28) of these embryos displayed a completely open neural tube (craniorachischisis), similar to that shown in the Vangl2^−/− embryos. The Ryk^+/−;Vangl2^−/− embryos appeared to be similar to the Vangl2^−/− mutant. However, the Ryk^−/−;Vangl2^−/− embryo exhibited more severe phenotypes than the Vangl2^−/− mutant. The anterior-posterior body axis and the limbs of these embryos were markedly shortened. In addition, the ventral body wall in the Ryk^−/−;Vangl2^−/− embryos failed to close, similar to what has been observed in the Vangl1^−/−;Vangl2^−/− embryo (14). Importantly, the phenotypes of the Ryk^−/−;Vangl2^−/− embryos, including shortened limbs, Anterior-posterior body axis and frontonasal processes, were similar to those of the Wnt5a^−/− embryo (20). Furthermore, we found, that the closure of the eyelid, which is controlled by PCP (39), was affected in the Ryk^−/−;Vangl2^+/− embryos. This defect was not observed in the single mutants of Ryk or Vangl2 (supplemental Fig. S2). The long bones in the Ryk^−/−;Vangl2^−/− forelimb were significantly shortened, but their widths were slightly increased, leading to a reduced length-to-width ratio (Fig. 2, A–F). A similar defect was observed in the Wnt5a^−/− mutants (19, 20). However, the digits of the Ryk^−/−;Vangl2^−/− embryo were less affected than those in the Wnt5a^−/− embryo (Fig. 2, F′ and E′). Only one of the three phalanges was absent in the Ryk^−/−;Vangl2^−/− embryo. By contrast, the rib cage in the Ryk^−/−;Vangl2^−/− embryo was more affected than that of the Wnt5a^−/− embryo (Fig. 2, I and J). The number of ribs was greatly reduced, and rib bifurcation was observed. Mild rib bifurcation was also observed in the Vangl2^−/− mutant (Fig. 2H). The sternum failed to close completely in the Ryk^−/−;Vangl2^−/− embryo, similar to that in the Wnt5a^−/− embryo (Fig. 2, I′ and J′). These results indicate that Ryk regulates PCP and that Wnt5a-mediated PCP signaling was impaired in the Ryk^−/−;Vangl2^−/− embryo. Consistent with these results, we found that Ryk was ubiquitously expressed (supplemental Fig. S3). We stained E14.5 Ryk^−/− limbs to detect β-gal expression, as these embryos express the lacZ gene under the control of the Ryk promoter (37). Ryk expression was increased in the chondrocytes and osteoblasts (supplemental Fig. S3).

Open in a separate window

FIGURE 1.

Genetic interaction of Ryk and Vangl2. Lateral views of E18.5 mouse embryos (top row) and their skeletal preparations (bottom row). The Vangl2^−/− embryo showed a complete open neural tube and shortened frontonasal process, limbs, and anterior-posterior body axis. The Vangl2^−/− phenotype was not enhanced by heterozygous loss of Ryk. However, the Ryk^−/−;Vangl2^+/− embryo phenocopied the Vangl2^−/− embryos. The Ryk^−/−;Vangl2^−/− embryo resembled the Wnt5a^−/− embryo and was more severe phenotypically compared with the Vangl2^−/− embryo.

Open in a separate window

FIGURE 2.

Severe length reduction in the Ryk^−/−;Vangl2^−/− humerus. A–F, Skeletal preparation of the forelimbs of the indicated genotypes at E18.5. Bone length was only mildly reduced in the Vangl2^−/− (C) and Ryk^−/−(B) mutant, and a more severe reduction in limb elongation was found in the Ryk^−/−;Vangl2^+/− limb (D). Reduction in humerus length in Ryk^−/−;Vangl2^−/− embryos (E) was very severe and similar to that in the Wnt5a^−/− limb (F). A′–F′, enlarged views of the anterior paws shown in A–F. One phalange in the Ryk^−/−;Vangl2^−/− embryo was missing (E′). Distal elements were completely missing in the Wnt5a^−/− limb (F′). G–J, boxed rib regions of E18.5 embryos are shown at higher magnification. G, normal development of the rib cage in the Ryk^+/−;Vangl2^+/− embryo. H, single bifurcation in the Vangl2^−/− rib cage. I, reduced number of ribs accompanied by severe bifurcation in the Ryk^−/−;Vangl2^−/− embryo. J, bifurcation of a subset of the ribs in the Wnt5a^−/− embryo. G′–J′, ventral views of sternum in embryos shown in G–H. The sternum in the Ryk^−/−;Vangl2^−/− (I′) and the Wnt5a^−/− (J′) embryos were closed incompletely (arrows).

Ryk Regulates PCP Signaling in the Inner Ear

To further investigate the role of Ryk as a PCP regulator, we examined the polarity of sensory hair cells in the inner ear. Expression of Ryk was recently reported by Macheda et al. (36). The hair cell polarity was mostly normal in the Ryk^+/−;Vangl2^+/− and Ryk^−/− embryos (Fig. 3, A and B). However, they were misaligned in the Ryk^−/−;Vangl2^+/− inner ears at E18.5 (Fig. 3D). This phenotype was most severe in the outer hair cell row 3 (Fig. 3I). Furthermore, we found an additional row of hair cells in the middle cochlea (arrow in Fig. 3D), and a similar defect was found in the inner ear of the Wnt5a^−/− mutant (5). The misalignment of hair cells in the Ryk^−/−;Vangl2^−/− cochlea (Fig. 3E) was similar to that observed in the Vangl2^−/− embryos (Fig. 3C) (14). However, in addition to the misalignment, we found up to two additional rows of hair cells in the middle cochlea of Ryk^−/−;Vangl2^−/− embryos (arrows in Fig. 3E), a defect that was not observed in the Vangl2^−/− cochlea (Fig. 3C). The hair cells of these additional rows were mostly misoriented. Interestingly, the combination of defective hair cell orientation and abnormal convergent extension in the Ryk^−/−;Vangl2^−/− cochlea was also reported in the Wnt5a^−/−;Vangl2^Lp/+ mutant (5). These results further support the notion that Ryk mediates Wnt5a signaling and regulates PCP in the mouse inner ear.

Open in a separate window

FIGURE 3.

Misalignment of sensory hair cells in the Ryk and Vangl2 mutant cochleae. A–E, projection of a z-stack of confocal sections of sensory hair cells of E18.5 cochleae stained for acetylated tubulin (green) and actin (red). F–I, quantification of hair cell alignment. Number of hair cells in relation to their deviation from correct alignment is displayed. Normal alignment of hair cells in Ryk^+/−;Vangl2^+/− (A) and Ryk^−/− (B) inner ears. C, loss of orientation in the Vangl2^−/− cochlea. D, in the Ryk^−/−;Vangl2^+/− cochlea, the orientation of the hair cells was impaired, most severely in the outer hair cell row 3 (I, p < 0.0001, two-tailed Student's t test against Ryk^+/−;Vangl2^+/−) and to a minor degree in the remaining hair cell rows (F and G). E and I, misalignment of hair cells was similar in the Ryk^−/−;Vangl2^−/− cochlea compared with Vangl2^−/−. One and up to two additional rows of hair cells were detected in the middle but not the proximal cochlea of Ryk^−/−;Vangl2^+/− and Ryk^−/−;Vangl2^−/− embryos, respectively (marked by arrows in E and F). IHC, inner hair cell; OHC, outer hair cell.

Ryk Promotes Vangl2 Stability but Not Vangl2 Phosphorylation

To understand the mechanism underlying the regulation of PCP by Ryk, we first tested whether Ryk and Vangl2 form a Wnt5a receptor complex. Indeed, we were able to detect Ryk and Vangl2 in the same immunoprecipitated complex (Fig. 4, A and B). As a negative control, we could not detect Ryk and another membrane protein Smoothened (Smo) in the same complex (supplemental Fig. S4). Co-expression of Wnt5a enhanced the formation of the Ryk-Vangl2 complex (Fig. 4, A and B). Previously, we have shown that Wnt5a acts through Ror2 to induce phosphorylation of Vangl2 (3). To test whether Ryk regulates PCP in a similar fashion, we investigated Vangl2 phosphorylation. However, unlike Ror2, Ryk did not induce any obvious protein mobility shift of Vangl2, suggesting that Vangl2 phosphorylation was not altered by Ryk (Fig. 5, A and B). Instead, Vangl2 protein levels were increased by Ryk expression suggesting a reduction of Vangl2 degradation in the presence of Ryk. We then tested whether Ryk regulates Vangl2 protein levels in vivo and whether such regulation contributes to the observed genetic interaction between Ryk and Vangl2 by examining Vangl2 protein levels in the E9.5 whole embryo lysates. Indeed, a decrease in Vangl2 protein levels was detected in the Ryk^−/−;Vangl2^+/− embryos compared with their Ryk^+/+;Vangl2^+/− littermate (Fig. 6A). Furthermore, we observed a similar regulation of Vangl2 protein levels by Ryk in the mouse embryonic fibroblasts, as Ryk^−/− cells showed less Vangl2 protein compared with cells isolated from a wild type littermate (Fig. 6B).

Open in a separate window

FIGURE 4.

Ryk and Vangl2 interaction is enhanced by Wnt5. A and B, co-immunoprecipitation of Ryk and Vangl2 expressed in the HEK 293T cells. Wnt5a enhanced Ryk and Vangl2 interaction. The ratio of co-immunoprecipitated Vangl2 to Ryk was calculated and compared. The results of four independent experiments were analyzed using the Student's t test and shown on the right.

Open in a separate window

FIGURE 5.

Phosphorylation of Vangl2 is not induced by Ryk. A, in transfected CHO cells, unphosphorylated (lower arrow) and basal phosphorylated (middle arrow) Vangl2 was detected. In the presence of Wnt5a, phosphorylation of Vangl2 (upper arrows) was induced by Ror2 but not by Ryk. B, phospho-specific Vangl2 (pVangl2) antibody that only recognizes phosphorylated founder sites of Vangl2, Ser-82 and Ser-84 (3), showed a mobility shift of Vangl2 (upper arrow) only in the presence of Wnt5a and Ror2 but not Ryk.

Open in a separate window

FIGURE 6.

Ryk increases the stability of Vangl2. A, in the E9.5 whole embryo lysates, Vangl2 protein levels in the Ryk^−/−;Vangl2^+/− embryos were decreased compared with those in the Ryk^+/+;Vangl2^+/− littermate. B, decrease of Vangl2 protein levels in the Ryk^−/− MEF cells compared with wild type MEF cells. The ratio of Vangl2 to actin protein levels was calculated and compared between Ryk^−/− and wild type cells. The results of three independent experiments were analyzed using Student's t test and shown on the right. C, dose-dependent increase in levels of Vangl2 by Ryk or the lysosome inhibitor bafilomycin A1 (baf A1, 400 nm). Increasing amounts of Ryk DNA in transfection resulted in higher levels of Vangl2 protein. The ratio of Vangl2 to actin protein levels was calculated and compared. The results of four independent experiments were analyzed using Student's t test and shown on the right. D, phospho-deficient Vangl2 (ΔpVangl2), lacking all Ser and Thr phosphorylation sites (3), was stabilized by Ryk. The ratio of ΔpVangl2 to actin protein levels was calculated and compared. The results of four independent experiments were analyzed using Student's t test. E, Vangl2 degradation was analyzed using the protein synthesis inhibitor cycloheximide. In the presence of Ryk, the degradation of Vangl2 is reduced.

As a membrane protein, Vangl2 can be degraded by lysosome. Indeed, treating cells with bafilomycin A1, a vacuolar-type H⁺-ATPases inhibitor which prevents acidification of endosomes and lysosomes that degrade membrane proteins (40), led to increased Vangl2 protein levels (Fig. 6C). Ryk overexpression, similar to bafilomycin A1 treatment, resulted in increased Vangl2 protein levels (Fig. 6C). Furthermore, the effect of Ryk on Vangl2 protein levels was dose-dependent; higher doses of Ryk in transfection resulted in higher levels of Vangl2 protein (Fig. 6C). As Ryk did not seem to affect Vangl2 phosphorylation, we hypothesized that regulation of Vangl2 levels by Ryk may be independent of Vangl2 phosphorylation. To test this, we employed a Vangl2 construct in which all Wnt5a-dependent phosphorylation sites (Ser and Thr residues) have been mutated (3). Ryk also increased the levels of this phospho-deficient Vangl2, indicating that this activity of Ryk is independent of Vangl2 phosphorylation (Fig. 6D).

To further investigate whether Ryk regulates Vangl2 protein levels by controlling its stability, we tested whether Ryk regulates the half-life of Vangl2. CHO expressing Vangl2 were treated with cycloheximide to block protein synthesis. After 2 h, we observed a reduction of Vangl2 protein in control cells, whereas Vangl2 behaved very stable in the presence of Ryk (Fig. 6E). To test whether Wnt5a also regulates Vangl2 stability through Ryk, we asked whether loss of Wnt5a affected Vangl2 levels. In the limbs of E13.5 embryos, we observed a reduction of Vangl2 protein in the Wnt5a^−/− tissue compared with that from the control littermates (Fig. 7A). Next, we asked whether Wnt5a is able to enhance the stability of Vangl2. Therefore, we cultured CHO cells stably expressing Vangl2 in the presence of Wnt5a conditioned medium. Wnt5a increased Vangl2 protein levels (Fig. 7B). We then tested whether Wnt5a affects Vangl2 stability through Ryk in mouse embryonic fibroblast (MEF) cells. Wnt5a treatment increased Vangl2 protein levels in the wild type MEF cells, but not in the Ryk^−/− MEF cells (Fig. 7C). To further substantiate this finding, we expressed Ryk and Wnt5a in the Vangl2-expressing CHO cells. Wnt5a or Ryk expression increased the levels of Vangl2, and the effects were further increased when both Ryk and Wnt5a were co-expressed (Fig. 7D). Taken together, our data show that Wnt5a enhances Ryk and Vangl2 complex formation and that Wnt5a may signal through Ryk to regulate Vangl2 activity by increasing its stability.

Open in a separate window

FIGURE 7.

Wnt5a regulates Vangl2 stability. A, Vangl2 protein levels were decreased in the E13.5 embryonic limb lysate of the Wnt5a^−/− mutants compared with control littermates. The ratio between Vangl2 and Actin was calculated, and the results of four independent experiments were analyzed using Student's t test. B, the levels of Vangl2 were increased in CHO cells that stably express Vangl2 in the presence of Wnt5a-conditioned medium for 12 h. The ratio betweeenVangl2 and actin was calculated, and the results of four independent experiments were analyzed using Student's t test. C, Wnt5a-conditioned medium increased the levels of Vangl2 in wild type MEF cells but not in Ryk^−/− MEF cells. Cells were incubated for 12 h. D, in the Vangl2-expressing CHO cells, Wnt5a expression increased the levels of Vangl2 similar to Ryk expression or the treatment with lysosome inhibitor bafilomycin A1 (baf A1, 400 nm). Co-expression of Ryk and Wnt5a further increased the stability of Vangl2. The ratio of Vangl2 to actin protein levels was calculated and compared. The results of four independent experiments were analyzed using Student's t test and shown on the right.

DISCUSSION

In this study, we demonstrate that Ryk is a regulator of the Wnt/PCP pathway in mice. We found that loss of Ryk in a Vangl2^+/− background resulted in typical PCP phenotypes in multiple developmental processes, although the Ryk mutant itself had no obvious PCP defects. We further showed biochemically that Ryk and Vangl2 formed a complex, which was enhanced by Wnt5a, and that Ryk was necessary to maintain Vangl2 protein levels. Our results suggest that there is considerable redundancy at the level of Wnt5a signaling reception on the plasma membrane in controlling PCP. Multiple Wnt5a-binding proteins can receive Wnt5a signals at the plasma membrane and regulate Vangl2 in different ways. Collectively, all of these Wnt5a-binding proteins contribute to the regulation of PCP.

Ror2 is required for Wnt5a signaling. However, the phenotypes of Ror2^−/− and even Ror2^−/−;Vangl2^−/− embryos were not as severe as those of the Wnt5a^−/− mutant in many developmental processes, suggesting the existence of functional redundancy in Wnt5a signal transduction. This notion is supported by the recently identified redundancy between Ror1 and Ror2 in receiving Wnt5a signaling (41). Our results uncover that Ryk is one of the other Wnt5a-binding proteins on the plasma membrane that regulates PCP. Ryk has been found to mediate Wnt5a activity in regulating axon guidance (34). However, Ryk^−/− mouse mutants or derailed fly mutants failed to show obvious PCP defects morphologically (37, 42), raising the question whether Ryk is mediating Wnt5a activity in controlling PCP in general. Here, we found that Ryk regulates Vangl2 activity by modulating its protein stability. As Vangl2 regulates PCP in a dose-dependent manner, it would not be surprising that in a wild type background, the Ryk^−/− mutant has no obvious PCP defects because the Vangl2 protein is not reduced to below threshold levels. When Vangl2 protein levels are reduced in the Vangl2^+/− mutant, further loss of Ryk decreases the Vangl2 protein levels below a critical threshold, resulting in PCP defects. As axon guidance is a process that can be regulated by PCP (43), the axon guidance defects, shown in Ryk knockdown mice (44), may also be a result of PCP deficiency. It is likely that Vangl2 protein stability is differential regulated in distinct developmental processes. It is also possible that axon guidance may be more sensitive to reduced Vangl2 protein levels than other morphogenetic processes.

It is interesting to note that the relative contribution of Ror2 and Ryk in regulating Vangl2 is tissue-dependent. In the limb, Ror2 appears to be a major coreceptor of Vangl2 as the Ror2^−/−;Vangl2^−/− limb phenotype is more severe than that of the Ryk^−/−;Vangl2^−/− mutant. However, in the neural tube, Ryk appears to be more important than Ror2 as 61% of the Ryk^−/−;Vangl2^+/− mutants showed NTD, whereas only 5% of the Ror2^−/−;Vangl2^+/− had similar defects (3). This may reflect tissue-specific expression of Ryk and Ror2 or other tissue-specific components involved in PCP signaling.

The occurrence of additional hair cell rows has previously been reported in the Wnt5a^−/− inner ears, but in contrast to the Ryk^−/−;Vangl2^+/− or Ryk^−/−;Vangl2^−/− mutants, the Wnt5a^−/− mutant does not have sensory hair cell polarity defects (5). These findings suggest that Wnt5a signaling is impaired in the absence of Ryk and Vangl2 and that loss of Wnt5a may be compensated for by other Wnts in regulating hair cell polarity (45). These results may also imply that in the inner ear, normal convergent extension requires higher Wnt doses than maintaining normal hair cell polarity. Consistent with this, the absence of additional phenotypes in the Ryk^−/−;Wnt5^+/− embryo suggests that Wnt5a and/or other Wnt ligands are expressed adequately such that reduction of one Wnt ligand is not enough to perturb the PCP signaling in morphogenesis even in the absence of Ryk. Given the fundamentally important function of PCP in morphogenesis, it is not surprising that multiple Wnts and multiple Wnt-binding receptors have evolved to ensure proper PCP signaling under different circumstances. Similar to our studies, Ryk has been found to interact with Vangl2 and together they might mediate the function of Wnt11 in the zebrafish (36). However, the Wnt11^−/− mice do not display PCP defects, and Ryk interaction with Wnt5a has been well established in axon guidance in mice; our genetic and biochemical analysis including characterization of the skeletal phenotypes of the Wnt5a and Ryk;Vangl2 double mutant embryos indicate that Wnt5a is the major Wnt ligand that regulates PCP through Ryk.

The PCP pathway is highly conserved in animals and mutations in human VANGL1 and VANGL2 have been identified to cause NTD (30, 31). Furthermore, human mutations in WNT5A can result in Robinow syndrome (29). As human and mouse Ryk are highly conserved, it is very likely that the role of Ryk in regulating PCP, identified in this work, is also conserved in human. The striking similarity between the phenotype of the long bones in the Ryk;Vangl2 mutants and the defects observed in patients diagnosed with Robinow syndrome and brachydactyly indicate that Ryk may play a role in these malformations. Therefore, our results suggest that mutations in human RYK might be involved in NTD and Robinow syndrome and provide new insight into the etiology of these diseases.

Acknowledgments

REFERENCES