Head formation requires Dishevelled degradation that is mediated by March2 in concert with Dapper1

Wnt signaling governs various cellular processes, including cell proliferation, differentiation, migration, establishment of polarity and stem cell self-renewal. Dysregulation of Wnt signaling causes developmental defects and human diseases, including cancer (Barker and Clevers, 2006; Clevers, 2006; Logan and Nusse, 2004; MacDonald et al., 2009). Wnt signaling is divided into the β-catenin-dependent canonical pathway and β-catenin-independent non-canonical pathway. In the canonical Wnt pathway, binding of Wnt ligands to the LDL receptor-related protein 6 (LRP6) co-receptor and Frizzled (Fz) receptor induces aggregation and phosphorylation of LRP6 through Dsh-mediated scaffold formation (Bilic et al., 2007; Gao and Chen, 2010; Sheng et al., 2014). Upon ternary LRP6-Wnt-Fz complex formation, Dsh binds to Axin, which recruits glycogen synthase kinase 3β (GSK3β) to phosphorylate LRP6 (Bilic et al., 2007). Inhibition of GSK3β by the phospho-LRP6-containing activated receptor complex protects β-catenin from proteasomal degradation (Logan and Nusse, 2004; MacDonald et al., 2009). Consequently, β-catenin accumulates and interacts with T-cell factor/lymphoid-enhancer factor (TCF/LEF) in the nucleus, thereby initiating transcription.

During vertebrate development, the canonical Wnt pathway plays a crucial role in head formation (Glinka et al., 1997; Kiecker and Niehrs, 2001; Yamaguchi, 2001). Proper head formation requires inhibition of the canonical Wnt pathway by various Wnt antagonists, including Dickopf-1 (Dkk1) and secreted Fz-related proteins (sFRPs), which are secreted from the head organizer of gastrula embryos and thereby maintain a low level of Wnt signaling at the putative anterior region of embryos (De Robertis and Kuroda, 2004; Niehrs, 2004). A recent study revealed that the transmembrane protein Tiki is required for head formation via cleavage and inactivation of Wnt ligands (Zhang et al., 2012). These well-identified Wnt antagonists involved in head formation collectively exert their roles by regulating perimembranous or extracellular events in Wnt activation, including Wnt ligand binding to receptors or modification of the ligand itself. However, whether antagonism of the Wnt signal necessary for head formation can be achieved through regulation of Dsh protein stability has not been extensively investigated.

Membrane-associated RING-CH (March) proteins are recently identified members of the mammalian E3 ubiquitin ligase family (Bartee et al., 2004). It is known that one member of this family, March2, is localized to endosomal vesicles as well as to the plasma membrane, and binds to syntaxin 6, which is involved in endosomal trafficking (Nakamura et al., 2005). In addition, March2 is involved in downregulation of discs large 1 (Dlg-1), a polarity-determining protein containing three PDZ domains (Cao et al., 2008). Recent research also suggests that March2 regulates lysosomal sorting of carvedilol-bound β2-adrenergic receptors via ubiquitylation (Han et al., 2012). However, the role of March2 during development and its targeted developmental signal have not yet been elucidated.

In this study, we have defined March2 as a novel ubiquitin ligase that targets Xenopus Dsh for ubiquitin-mediated degradation, thereby antagonizing Wnt signaling and regulating head formation. We have also addressed the molecular details of the mechanism by which March2 inhibits Wnt, discovering that March2 interacts with and degrades Dsh in concert with another Dsh scaffold protein, Dapper1 (Dpr1/DACT1), in Xenopus embryos. Dpr1 has been identified as a Dsh-interacting protein that inhibits Wnt signaling by mediating Dsh degradation (Cheyette et al., 2002; Zhang et al., 2006). Interestingly, however, how Dpr1, a scaffold protein that lacks intrinsic enzymatic activity, regulates the stability of Dsh protein remains unclear. Although several E3 ubiquitin ligases have been suggested to regulate Dsh stability (Angers et al., 2006; Chan et al., 2006; Ganner et al., 2009; Miyazaki et al., 2004; Zhang et al., 2014), none seems to be directly linked with Dpr1, suggesting the possible involvement of an unidentified, genuine E3 ubiquitin ligase that downregulates Dsh protein in concert with Dpr1. Our new findings suggest that March2 is this novel E3 ubiquitin ligase and exerts this role in Xenopus head formation.

To fully understand the regulatory mechanism and function of Dsh during Xenopus embryogenesis, we sought to identify new binding partners of Dsh protein that might regulate Wnt signaling in Xenopus. To achieve this we performed a computational screening analysis based on the predicted ligand selectivity of the PDZ domain (Kim et al., 2012) of Dsh with Xenopus proteins that contain PDZ-binding motifs. Among the putative Dsh binding proteins, Xenopus March2, the RING-type E3 ubiquitin ligase, was highly scored but its function during vertebrate development has not previously been reported (Table S2). Thus, we identified and cloned the cDNA for the Xenopus orthologue of March2, which consists of 741 nucleotides encoding a protein of 246 amino acids (Gene ID: 495994) that contains a RING (really important new gene) E3-ligase domain, two transmembrane domains and a C-terminal PDZ-binding motif (Fig. S1A). The encoded protein showed high similarity to other vertebrate orthologs (human, 86%; mouse, 79%; chick, 93%; zebrafish, 89%).

Before verifying the potential binding of March2 to Dsh, we analyzed the spatiotemporal expression of March2 transcripts and protein. Both maternal and zygotic expression of endogenous March2 transcripts and protein were assessed by RT-PCR and western blotting throughout Xenopus development (Fig. S1B,U). In situ hybridization showed that March2 transcripts were strongly enriched at the animal hemisphere in blastula stages (st. 3 and st. 9) (Fig. S1C,D). As gastrulation proceeded (st. 10 and st. 11) (Fig. S1E-J), March2 transcripts became ubiquitously expressed, but also showed noteworthy expression in the dorsal marginal zone (DMZ), including the head organizer (Fig. S1G,J,T). During neurulation (st. 12 to st. 20), March2 transcripts were detectable in the anterior neural plate and neural plate border region. At later times, March2 transcripts were exclusively found in the head region and heart field (st. 27 and st. 36) (Fig. S1Q-S). These data suggest that March2 is consistently expressed during Xenopus embryogenesis and may play a role in head development.

To validate the potential binding of Xenopus March2 with Dsh, which was suggested by our computational analysis, we performed co-immunoprecipitation (co-IP) analyses. Our results clearly showed that March2 interacts with Dsh ectopically (Fig. 1A) and endogenously (Fig. 1B) in HEK293FT cells (Fig. 1B). In addition, confocal imaging revealed that cytosolic puncta of March2 protein colocalized with Dsh in Xenopus animal caps (Fig. 1C) and human cells (Fig. 1D), providing additional evidence that these proteins can associate with each other. To further characterize the binding of March2 with Dsh, we compared binding of Dsh deletion-mutant forms lacking each conserved domain (DIX, PDZ and DEP) with full-length March2 (Fig. 1E). Our results showed that Dsh binding to March2 was completely lost in a mutant lacking the DEP domain, and was dramatically weakened in a mutant lacking the PDZ domain. In contrast, deletion of the DIX domain had no effect on the binding ability or affinity of Dsh for March2 (Fig. 1E). These results support the interpretation that both DEP and PDZ domains of Dsh are required for the interaction with March2, with the DEP domain playing a more prominent role (Fig. 1E).

To identify the March2 domain responsible for Dsh binding, we next constructed two deletion-mutant forms of March2, one lacking the C-terminal region containing the PDZ-binding motif (Mar2ΔC) and a second lacking the N-terminal region containing the RING domain (Mar2ΔN), and analyzed the mutants for their ability to bind Dsh (Fig. 1F). Our result revealed that Dsh bound to Mar2ΔN but not to Mar2ΔC, indicating that the C-terminal region of March2 interacts with Dsh.

Since our screening analysis was based solely on the possibility that the PDZ-binding motif consists of four amino acids at the C terminus of March2 that specifically interact with Dsh, we next directly tested whether this motif itself is indeed responsible for Dsh binding. Unexpectedly, co-IP analyses revealed that deletion of the PDZ-binding motif from March2 did not prevent March2 from binding Dsh (data not shown). This unanticipated result implies that the C-terminal region of March2 that includes amino acids 200-246, rather than a single PDZ-binding motif, is important for Dsh binding. Nonetheless, taken together, these results suggest that March2 is a new Dsh-interacting partner that binds through association of the C-terminal region of March2 with both PDZ and DEP domains of Dsh.

As March2 is known as an E3-ubiquitin ligase that mediates the ubiquitylation and subsequent degradation of its targets (Cao et al., 2008; Cheng and Guggino, 2013; Han et al., 2012), we hypothesized that March2 binding to Dsh serves to regulate the stability of Dsh protein in Xenopus. To test this hypothesis, we analyzed the effects of dose-dependent expression of March2 on the protein level of Dsh in Xenopus (Fig. 2A). Our result showed increasing amounts of March2 remarkably decreased Dsh protein levels (Fig. 2A). This ability of March2 to reduce Dsh protein was maintained from late blastula to early gastrula stages (Fig. S2A). Conversely, knockdown of March2 by injecting three respective March2 antisense morpholino oligonucleotides (Mar2MO, Mar2MO2 and Mar2MO3), which efficiently depleted March2 protein expression in Xenopus (Figs S6B,C and S7B), increased Dsh protein level in a dose-dependent manner, indicating that decreasing March2 function stabilizes Dsh protein. Mar2MO-induced increase in Dsh protein was attenuated by co-expression of the MO-resistant form of March2 mRNA, which lacks 5′UTR region to which the MOs bind (Fig. 2B, Figs S6D and S7C). We then analyzed transcript levels of Dsh following injection with increasing amounts of either March2 mRNA or Mar2MO to eliminate the possibility that the effect of March2 overexpression or knockdown on Dsh stability occurred through regulation of Dsh transcription. RT-PCR analyses showed that neither overexpression nor knockdown of March2 affected Dsh transcription (Fig. S2B,C).

As co-IP analysis (Fig. 1E) showed that PDZ and DEP domains of Dsh are important for the binding of March2, we next asked whether March2 is unable to degrade Dsh mutant proteins lacking either PDZ or DEP domains. Expression of March2 in Xenopus did not affect the Dsh mutants lacking PDZ or DEP domains but effectively reduced DIX-domain-deleted forms of Dsh protein, though not as much as the wild type (Fig. 2C).

To further elucidate whether the reduction of Dsh stability upon March2 expression was attributable to enhanced ubiquitylation of Dsh by March2 E3-ubiquitin ligase activity, we generated a mutant form of March2 (March2CS) that lacks this activity by virtue of a substitution of serine residues for four crucial cysteine residues within the RING domain of March2 (Cao et al., 2008) and analyzed its effect on the Dsh protein (Fig. 2D). Our results showed that March2CS did not reduce the Dsh protein levels (Fig. 2D).

We next carried out in vivo ubiquitylation assays (Fig. 2E) using inhibitors of proteasomal (MG132) and lysosomal degradation (bafilomycin A1 and leupeptin) simultaneously, because previous reports have shown that the stability of Dsh protein can be regulated by both proteasome- and lysosome-dependent pathways (Angers et al., 2006; Chan et al., 2006; Ganner et al., 2009; Gao et al., 2010; Miyazaki et al., 2004; Zhang et al., 2006, 2014). We found that wild-type March2, but not March2CS, dramatically induced Dsh poly-ubiquitylation. In addition, March2 depletion led to a loss of Dsh poly-ubiquitylation (Fig. 2F). Collectively, these results indicate that March2 acts through interactions with PDZ and DEP domains of Dsh to specifically target Dsh protein for poly-ubiquitination-dependent degradation during Xenopus embryogenesis.

We next asked to which pathway does March2 guide poly-ubiquitylated Dsh proteins for degradation? To address this question, we performed in vivo ubiquitylation assays in Xenopus (Fig. 3A,B) to determine the pathway responsible for March2-dependent Dsh degradation. These analyses revealed that March2 mediated poly-ubiquitylation of Dsh protein was increased by treatment with bafilomycin A1 (Fig. 3A) or leupeptin (Fig. 3B), but not by MG132 (Fig. 3A), supporting the notion that March2-mediated ubiquitylation targets Dsh to the lysosomal pathway. In addition, we monitored the subcellular localization of March2 protein in Xenopus animal cap tissues (Fig. 3C) and human culture cells (Fig. 3D). The cytosolic puncta of March2 proteins was strongly colocalized with the late endosome/lysosomal marker protein Rab7 and Lamp1 (Fig. 3C,D), but not with the early endosome marker EEA1 (Fig. 3C). Taken together, these results demonstrate that March2 regulates the stability of Dsh protein in Xenopus by mediating its poly-ubiquitylation and subsequent lysosomal degradation.

Given that the canonical Wnt pathway requires Dsh as a key scaffold protein for multiple steps in the signaling process, including Wnt-Fz-LRP5/6 receptor complex formation and aggregation (Gao and Chen, 2010; Sheng et al., 2014), LRP5/6 phosphorylation and β-catenin stabilization (Bilic et al., 2007), we next investigated the evident possibility that March2 is involved in the Wnt pathway and antagonizes various signaling steps by restricting the Dsh protein pool during Xenopus embryogenesis. To examine the effects of loss or gain of March2 function on Wnt-dependent transcriptional activity, we performed reporter assays in Xenopus embryos using a TOP-flash reporter containing TCF-binding sites. Knockdown of March2 boosted Wnt-dependent transcriptional activity (Fig. 4A), whereas ectopic expression of March2 inhibited Dsh-induced transcriptional activity (Fig. 4D). However, March2 expression could not affect the transcriptional activity induced by β-catenin, indicating that March2 antagonizes Wnt signaling and functions upstream of β-catenin (Fig. 4D). In addition, western blot analyses revealed that knockdown of March2 promoted activation of β-catenin (Fig. 4B, Figs S6E and S7D) as well as phosphorylation of LRP6 (Fig. 4B), whereas overexpression of March2, but not March2CS, inhibited both Wnt8-mediated processes, confirming the idea that March2 acts as a negative regulator of canonical Wnt signaling (Fig. 4C, lanes 4-6).

We also tested whether the severity of phenotypes caused by forced Wnt signaling is increased by knockdown of March2. The ability of Xenopus embryos to form a secondary axis by Wnt8 in ventral regions was enhanced by March2 depletion (Fig. 4E). This antagonistic role of March2 was specific for canonical Wnt signaling, because March2 inhibited Wnt- and Dsh-induced Xnr3 expression (Fig. 4F,G), but did not affect FGF-, BMP- or nodal (Xnr1)-induced Xbra expression (Fig. 4I-K). Notably, March2 did not inhibit β-catenin-induced Xnr3 expression, consistent with reporter assay results showing that March2 functions upstream of β-catenin (Fig. 4D,H).

Our results clearly show that March2 antagonizes the canonical Wnt pathway by ubiquitylating Dsh protein and targeting it for lysosomal degradation. To elucidate the endogenous function of March2 in Xenopus, we injected Mar2MOs into the two dorsal blastomeres of four-to eight-cell stage embryos and analyzed their phenotypes at tadpole stages (Fig. 5A, Figs S6F and 7E). Intriguingly, almost 90% of March2 morphants exhibited defects in anterior head formation, including shrinkage or loss of forebrain structures, and in the cement gland and eyes, whereas embryos injected with control MO (CoMO) were normal. The phenotypes of March2 morphants were rescued by co-injection of the MOs-resistant form of mRNA or March2 DNA, confirming the specificity of the effects of the MOs (Fig. 5A,B, Fig. S6F,G, Fig. S7E,F). Meanwhile, gain-of-function of wild-type March2, but not March2CS, caused gastrulation defects (Fig. S3) that phenocopy of Dsh morphants (Sheng et al., 2014), supporting our results suggesting that March2 downregulates the stability of Dsh protein (Fig. 2).

As March2 knockdown promotes excessive Wnt signaling by preserving the cytosolic pool of Dsh protein (Fig. 4), we asked whether the head defects caused by March2MO could be rescued by Axin1, an inhibitor of Wnt signaling that acts downstream of Dsh, and whether the head-enlargement defects caused by expression of Dkk1, a Wnt inhibitor upstream of Dsh, could be moderated by depletion of March2 (Fig. S4A,B). The head malformation caused by Mar2MO was significantly rescued by Axin1 (Fig. S4A,D), whereas that caused by Dkk1 was significantly rescued by Mar2MO (Fig. S4B). These results support the notion that the defects caused by March2 depletion are attributable to excessive Wnt signaling by stabilized Dsh protein.

Next, we asked whether March2 knockdown leads to impaired dorsal head organizer formation. Importantly, our results revealed that March2-depleted embryos showed decreased expression of the dorsal head organizer marker genes Chd, Gsc and Otx2 at mid-to-late gastrula stages (st. 11-12), and these reductions were restored by expression of the MO-resistant form of March2 mRNA; in contrast, March2 depletion did not affect expression of the notochord marker Xnot, which is dispensable for head organizer formation (Fig. 5C and Figs S5A, S6H,I, S7G,H). Notably, prior to stage 11, Mar2MO did not inhibit expressions of Gsc and Chd (data not shown), which are regulated by maternal Wnt signaling (Cho et al., 1991; Christian and Moon, 1993; Zhang et al., 2012). Moreover, March2 mRNA rescued head defects caused by Wnt8 DNA, which is capable of activating zygotic Wnt signaling (Fig. S4C). Collectively, these results suggest that March2 is required for head organizer formation by inhibiting zygotic Wnt signaling in Xenopus.

It has been reported that maintenance of the lowest levels of Wnt signaling at the head organizer during gastrulation results in precise anterior-posterior (AP) neural patterning by keeping anterior Wnt signaling low (De Robertis and Kuroda, 2004; Harland and Gerhart, 1997; Niehrs, 2004; Zhang et al., 2012). To further explore this concept, we next investigated the possibility that the effect of March2 depletion on the head organizer is translated into AP patterning. Injection of Mar2MO into one dorso-animal blastomere at the eight-cell stage reduced the expression of a subset of genes, including the forebrain markers BF1 and Otx2, the midbrain marker En2, and the hindbrain marker Krox20; notably, these reductions were rescued by the MO-resistant form of March2 mRNA (Fig. 5D and Fig. S5B). Given that Wnt inhibition is also required for neural induction (Zhang et al., 2015), we examined the possibility that March2 may be involved in this process by analyzing the effects of March2 depletion on expression of the pan-neural marker Sox2. We found that Mar2MO did not affect Sox2 expression, implying that March2 function is dispensable for neural induction (Fig. 5D). Interestingly, moderate depletion of endogenous March2 using a sub-optimal dose of Mar2MO that does not efficiently induce head defects, combined with a small increase in Dsh expression, synergistically induced head defects (Fig. 5E). This result can be interpreted to mean that the aberrant amount of Dsh protein arising from both Mar2MO-stabilized endogenous Dsh protein and ectopically expressed Dsh mRNA elevates Wnt signaling sufficiently to mediate head defects. Furthermore, the severity of head defects in March2 morphants caused by sufficient amounts of Mar2MO was ameliorated by moderate knockdown of Dsh with a sub-optimal dose of DshMO (Fig. 5F). These results imply that translation blockade of newly synthesized Dsh mRNA by DshMO counteracted the preservation of the existing pool of Dsh protein by Mar2MO (Fig. 5F), thereby rescuing the head defects in March2 morphants caused by the elevation in Wnt signaling attributable to increased Dsh protein. Collectively, these results underscore the requirement of March2 for anterior head formation through its Dsh-degradation-dependent antagonism of Wnt signaling.

As our in situ hybridization and RT-PCR results suggested the maternal and ubiquitous expression of Xenopus March2 transcripts during gastrulation (Fig. S1), we initially expected March2 to be involved in Wnt signaling in both dorsal and ventral regions of the embryo, and accordingly addressed the effects of March2 knockdown in each region. Intriguingly, however, Mar2MOs injection into the dorso-animal side caused head malformation, which is characteristic of genes that antagonize Wnt signaling (Fig. 5A, Figs S6F and S7E) (Shimomura et al., 2010; Zhang et al., 2012, 2015), whereas ventral injection did not cause any defects (data not shown). Consistent with this unanticipated physiological requirement, we found that March2 mediated degradation of Dsh protein in only dorsal and animal explants of the gastrula (stage 10.5) embryo (Fig. S8A). Notably, Dsh protein was stabilized to a greater extent in ventral explants.

These unexpected findings raise a question regarding how March2 specifically mediates degradation of Dsh protein in the dorso-animal region to promote precise head formation. One possible hypothesis is that an unidentified regulator of March2 that is enriched in the dorso-animal region ensures the region-specific function of March2. To test this idea, we postulated that the Dsh-interacting protein Dpr1/DACT1 is an unidentified regulator of March2, based on reports suggesting that: (1) Dpr1, which lacks enzymatic activity, interacts with PDZ and DEP domains of Dvl to mediate Dvl degradation via the lysosomal pathway (Kivimäe et al., 2011; Zhang et al., 2006); (2) Dpr1 transcripts are strongly expressed in the dorso-animal region of the Xenopus gastrula (Cheyette et al., 2002) (Fig. S8B); and (3) gain and loss of Dpr1 function phenocopy March2 gain and loss of function (Cheyette et al., 2002) (Fig. S3).

To corroborate the involvement of Dpr1 in March2 function in Wnt signaling and head formation, we first examined whether depletion of Dpr1 affected the ability of March2 to mediate Dsh degradation. Surprisingly, Dsh protein degradation induced by ectopically expressed March2 was attenuated by increasing amounts of co-injected Dpr1MO, which efficiently depleted Dpr1 protein expression in Xenopus (Fig. S8D), and additional co-injection of the MO-resistant form of Dpr1 mRNA restored the ability of March2 to degrade Dsh (Fig. 6A). Conversely, ectopically expressed Dpr1 decreased Dsh protein, and this effect was impaired by Mar2MOs and restored by co-injection of March2 mRNA (Fig. 6B and Fig. S6J). Moreover, March2 did not decrease Dsh protein when co-expressed with the dominant-negative form (DN) of Dpr1 (Zhang et al., 2006) (Fig. 6C). In vivo ubiquitylation assays showed that, in the presence of bafilomycin A1, Dpr1MO reduced March2-mediated poly-ubiquitylation of Dsh (Fig. 6D). Collectively, these results suggest that March2 requires Dpr1 to mediate poly-ubiquitylation and lysosomal degradation of Dsh, and further imply that the previously reported function of Dpr1 in promoting Dsh degradation (Zhang et al., 2006) depends on March2.

It is known that Dpr1 is a non-enzymatic Dsh-binding scaffold protein that lacks domains responsible for protein ubiquitylation and degradation. These observations, together with our findings, suggest that Dpr1 binds March2, serving as a bridge between March2 and Dsh to support the E3 ubiquitin ligase function of March2. To provide evidence supporting this hypothesis, we analyzed whether Dpr1 interacts with March2. Co-IP analyses showed that Dpr1 bound March2 (Fig. 6E). Consistent with this, confocal imaging (Figs 6F-J and 3D) showed that Dsh, Dpr1 and March2 proteins are co-localized in cytosolic vesicles (Fig. 6I,J) that our data suggest are late endosomes/lysosomes (Fig. 3C,D). Importantly, knockdown of Dpr1 both eliminated Dsh from lysosomes where March2 is localized (Fig. S9) and dramatically decreased the binding affinity of March2 for Dsh (Fig. 6K), suggesting that March2 requires Dpr1 to ensure stable interactions with Dsh in lysosomes and subsequent mediation of Dsh degradation.

Next, we investigated whether March2 requires Dpr1 to antagonize the Wnt pathway in Xenopus. Our results showed that increasing amount of Dpr1MO impaired the ability of March2 to inhibit Wnt8-mediated β-catenin activation and LRP6 phosphorylation, whereas co-expression of Dpr1 mRNA restored the effects of March2 (Fig. 7A). Furthermore, co-expression of DN Dpr1 blocked the effect of March2 on Wnt8-mediated β-catenin activation (Fig. 7B), confirming our previous results showing that March2 is a negative regulator of the Wnt pathway and that Dpr1 is required for this process.

Last, we analyzed the involvement of Dpr1 in March2-mediated regulation of head formation. Consistent with our previous data (Fig. S4C), we found that ectopic expression of March2 efficiently suppressed the severity of head defects induced by Wnt8. However, co-injection of Dpr1MO abrogated this rescue effect of March2 (Fig. 7C). Furthermore, moderate depletion of both endogenous March2 and Dpr1 by injection of sub-optimal doses of MOs synergistically induced head defects (Fig. 7D). This result can be interpreted as indicating that the combination of moderate blockade of endogenous March2 mRNA translation by Mar2MO and inhibition of endogenous March2 protein function by depletion of Dpr1 causes a degree of March2 malfunction sufficient to stabilize aberrant amounts of Dsh protein and cause head defects.

This study reveals a previously unknown function of March2 during head formation in Xenopus. March proteins are recently identified members of the mammalian E3 ubiquitin ligase family. Compared with other classes of E3 ubiquitin ligases, evidence relating to the roles of March family proteins in vivo or in vitro is limited. A few studies in cultured cells have reported functions of March family proteins in keeping with a role in endosomal trafficking and lysosomal sorting of target proteins (Cao et al., 2008; Han et al., 2012; Nakamura et al., 2005). To date, however, the physiological functions of March family proteins have yet to be elucidated.

We investigated regulatory mechanisms of March2 that underlie head formation, and found that March2 acts as a negative regulator of canonical Wnt signaling by promoting the lysosomal degradation of Dsh. We showed that March2 is maternally and ubiquitously expressed throughout Xenopus embryogenesis, but is notable for its expression during gastrulation in the dorsal organizer and at neurula stages in anterior neural plate regions, which are responsible for head formation and are sites where Wnt signaling is actively antagonized (Figs S1 and S8). Of note, the March2 expression pattern is similar to that of Dsh in Xenopus gastrula (Gray et al., 2009), supporting our data suggesting that March2 targets Dsh for degradation in the dorsal head organizer to regulate precise head formation.

As first suggested by a bioinformatics analysis and further corroborated by biochemical analyses, our results clearly showed that Xenopus March2 is a novel Dsh-binding protein (Fig. 1). Using co-IP analyses with deletion-mutant forms of Dsh and March2, we further elucidated the mode of interaction between them. Intriguingly, we found that the C-terminal region of March2 interacts with PDZ and DEP domains of Dsh, and further showed that the conserved PDZ-binding motif at the tip of the March2 C terminus is dispensable for the interaction. Further examination to narrow down the specific amino acids of the C-terminal region of March2 will resolve this inconsistency.

According to the recently proposed model of canonical Wnt signaling, in the absence of Wnt stimulation, Dsh protein is located cytoplasmically (Barker and Clevers, 2006; Clevers, 2006; Gao and Chen, 2010). Wnt stimulation recruits Dsh to the Fz receptor in the cell membrane, where it functions as a scaffolding protein that serves to release β-catenin from the destruction complex by conveying it to the membrane. It has been reported that March2 is localized in the endoplasmic reticulum (ER), early and late endosomes, lysosome, multivesicular bodies (MVB), and Golgi structure (Han et al., 2012; Nakamura et al., 2005). However, our confocal images revealed that most March2 protein is localized to Rab7-positive late endosomes/lysosomes in Xenopus (Fig. 3C). Our results, taken together with the domain structure of March2, suggest that both the N terminus containing the RING domain and C-terminal region responsible for Dsh interaction may be exposed to the cytosol. From this vantage point, March2 captures excess cytosolic Dsh and clears it. This notion is distinguishable from the report demonstrating that Wnt activation sequesters negative regulators of Wnt signaling into the multivesicular endosomes (Taelman et al., 2010), because our examination showed that Wnt stimulation did not affect the ability of March2 to produce Dsh degradation (data not shown).

March2 acts through the mode of interaction described above to mediate poly-ubiquitylation and proteolysis of Dsh in Xenopus (Fig. 2). In fact, several E3 ubiquitin ligases and components that are responsible for Dsh degradation have been extensively investigated. These studies have revealed that Dsh protein turnover is regulated by NEDL-1, the KLHL12-Cullin-3 ubiquitin ligase complex, Prickle-1, ANAPC2 and NEDD4L (Angers et al., 2006; Chan et al., 2006; Ganner et al., 2009; Miyazaki et al., 2004; Zhang et al., 2014). Two novel aspects of March2-mediated Dsh degradation differentiate our study from these previous reports. First, we found that March2 is a novel E3-ubiquitin ligase that specifically regulates head formation via Dsh degradation to antagonize canonical Wnt signaling. It is well known that Dsh is a common denominator in both the canonical Wnt and the non-canonical Wnt/PCP pathway. Each pathway governs different processes in Xenopus. Canonical Wnt signaling is crucial for head formation, but non-canonical Wnt signaling is required for convergent extension (CE) movements. It has been suggested that ANAPC2 plays a central role in regulating cilia polarity in Xenopus epidermis via mediating Dsh degradation (Ganner et al., 2009). More recently, NEDD4L has also been suggested to be a Dsh-specific E3 ligase that mediates Dsh degradation and is required for CE movements (Zhang et al., 2014). Both reports described them as non-canonical Wnt signaling antagonists. Intriguingly, however, although March2 mediates degradation of Dsh, which serves as a branch point in both pathways, and March2 transcripts are also expressed in the region that undergoes CE movements, our results showed that March2 morphants only resulted in head defects that are specifically induced by excessive canonical Wnt signaling (Figs 4 and 5). They were unable to induce defective gastrulation (data not shown), which is a hallmark of mis-regulated non-canonical Wnt signaling. We suggest two regulatory mechanisms that allow Dsh to undergo canonical Wnt-specific degradation by March2: unknown components that help Dsh avoid capture by March2 in the lysosome under non-canonical Wnt stimulation; and the marking of Dsh with canonical Wnt-specific modification(s), such as phosphorylation, required for recognition by March2. Indeed, the latter idea is supported by data shown in Fig. 6B,C. We found that March2 expression significantly degraded the phosphorylated form of Dsh, which is lower in electrophoretic mobility. Although these ideas require more experimental evidence from further study, our study is the first to provide evidence of an E3 ubiquitin ligase for Dsh degradation that exerts roles in the vertebrate head formation. Furthermore, we suggest March2 as being the first identified E3 ubiquitin ligase responsible for lysosomal degradation of Dsh (Fig. 3). Although previous studies have reported that Dsh undergoes lysosomal degradation, the E3 ubiquitin ligase responsible for this pathway has remained unidentified. The Dsh-interacting protein Dpr1, though not itself an E3 ubiquitin ligase, was the only well-identified protein implicated in lysosome-dependent Dsh degradation, as evidenced by the ability of lysosomal inhibitors to recover the Dpr1-mediated reduction in Dsh protein. We thus posit March2 as a genuine ubiquitin ligase that requires Dpr1 for Dsh degradation. Indeed, our results confirm that Dpr1 aids March2-mediated degradation of Dsh by allowing stable, strong interactions between March2 and Dsh through its scaffolding function, thereby producing the antagonism of the canonical Wnt signaling necessary for head formation (Figs 6 and 7 and Fig. S8). Moreover, these results explain how ubiquitously expressed March2 can mediate the specific degradation of Dsh protein at the dorso-animal region, which is needed for precise head formation. It is still not known how Dpr1 and March2 effectively cooperate to target Dsh for degradation in the lysosomes – a mechanism that awaits further study.

On the basis of our results, we propose the following model in which March2 in the dorsal head organizer antagonizes canonical Wnt signaling by targeting Dsh for lysosomal degradation (Fig. 7E): (1) cytosolic vesicle-localized March2 binds through its C-terminal region to DEP and PDZ domains of Dsh in concert with Dpr1, which acts via its scaffolding function to ensure stable interactions between March2 and Dsh; (2) March2 mediates the poly-ubiquitylation of Dsh, and targets Dsh to the lysosomal degradation pathway; and (3) the reduced pool of Dsh protein in the cytosol ensures the turned-off status of Wnt signaling, which is necessary for head formation. Further studies will be required to elucidate in detail how Dpr1 is involved in the mechanism by which March2 regulates Dsh degradation during head formation. Nevertheless, the present study provides new insights into the mechanisms that regulate the canonical Wnt pathway as well as the process of head formation during vertebrate development.

To perform a computational screening analysis, we predicted the ligand selectivity of the PDZ domain of Xenopus Dsh with the PDZ domain specificity prediction tool (http://sbi.postech.ac.kr/pdz/) and scored potential binding partners of Xenopus proteins. The prediction tool first selects pocket residues of PDZ domain from the query sequence with sequence alignment for the reference PDZ domain. Then, it predicts pocket-wise ligand selectivity as a positional weight matrix (PWM) based on the pocket residue similarity with those of a reference set of 133 PDZ domains using the 10 physicochemical properties of amino acids, i.e. hydrogen bond donors, polarity, volume, bulkiness, hydrophobicity, isoelectric point, positive charge, negative charge, electron ion interaction potential and free energy in water. Finally the PWM allows us to calculate binding scores between candidate proteins and PDZ domain of query protein and to prioritize them.

For expression in Xenopus embryos, the entire coding region of Xenopus March2 was amplified from gastrula-stage embryos by PCR and cloned into the pCS2+, pCS2+GFP and the pCS4+HA vectors, respectively. Deletion mutants of March2 (Mar2N and Mar2C) were generated with fusion PCR and cloned into the pCS2+GFP vector. For Xenopus March2CS, the 64th, 67th, 106th and 109th Cys residues were mutated into Ser residues using PCR-based mutagenesis. The entire coding region of Xenopus Dapper1 was also amplified from gastrula stage embryos by PCR and cloned into the pCST+ and pCS2+-Myc vectors. Capped mRNAs were synthesized from NotI-linearized plasmids using the SP6 mMessage mMachine kit (Ambion).

HEK293FT, HEK293T, HeLa cells were obtained from ATCC and maintained with Dulbecco's modified Eagle's medium (DMEM; Hyclone) containing 10% FBS and 1% antibiotics. Lipofectamine 2000 (Invitrogen) or Vivamagic transfection reagent (Vivagen) was used for cell transfections. All cells were maintained and transfected without any contamination.

RT-PCR analysis was performed as described previously (Cheong et al., 2006). Primers for RT-PCR analysis were as follows: March2 forward, 5′-ATGGCTTAAGGACCCAGGAC-3′; March2 reverse, 5′-AGTGGTAGCGGAATGACACC-3′; Dsh forward, 5′-ATGGCGGAGACTAAAGTGATTTA-3′; Dsh reverse, 5′-TTCAGGTCTGACCTCTGTAG-3′; Dpr1 forward, 5′-AGAGCATCTGGAGACAGAC-3′; Dpr1 reverse, 5′-TGTAGTCCAGCCATTCAAGA-3′. Primers for ODC, Xnr3, Siamois, Xbra, Msx1 and Chordin are listed in Table S4. Relative expression levels for Fig. S6B were analyzed using ImageJ.

Xenopus laevis females and males were purchased from Nasco and were used following the instructions from the POSTECH IACUC (Institutional Animal Care and Use Committees) (Korea) after they had certified us regarding ethical handling. Xenopus laevis eggs were obtained and fertilized as described previously (Sheng et al., 2014). Developmental stages of the embryos were determined according to Nieuwkoop and Faber (1994). Microinjections were performed as described previously (Sheng et al., 2014).

Antisense morpholino oligonucleotides (MO) were obtained from Gene Tools. MO sequences were as follows: March2 MO, 5′-TCATGGCAACCCTCAGTGCCTTCCT-3′; March2 MO2, 5′-CAACCCTCAGTGCCTTCCTCCTTTA-3′; March2 MO3, 5′-TGACTCTCCGCTGCCAGACTTTAC-3′; Dapper1 MO, 5′-GATCGGCTTCATCCTCTGGGAACGA-3′; XDsh MO, 5′-TCACTTTAGTCTCCGCCATTCTGCG-3′; control MO, 5′-CCTCTTACCTCAGTTACAATTTATA-3′.

Two- to four-cell stage embryos were injected into the animal pole with mRNAs. At stage 9.5, animal caps were dissected, fixed in 4% paraformaldehyde for 2 h and incubated in PBS with 0.1% Triton X-100 and 2% BSA for 1 h. The tissues were stained with primary antibodies (listed in Table S1) for 2 h, washed with PBS and incubated with fluorescent secondary antibodies for 2 h at room temperature. Mounted tissue images were obtained and analyzed by confocal microscopy (Olympus FluoView 1000 and Zeiss LSM 510).

Injected stage 8 embryos were treated with 40 μM MG132, 250 μM bafilomycin A1 or 100 μM leupeptin and were harvested at stage 12. Embryos were lysed in RIPA buffer, and then pull-down steps and SDS-PAGE were performed.

Whole-mount in situ hybridization was performed as previously described (Harland, 1991). Antisense in situ probes against March2 was generated by linearizing the pBSKII+March2 with BstXI and transcribing with T7 RNA polymerase.

We thank Drs R. Moon, C. Niehrs, E. Jho, C. Yeo and S. Hirose for generous gifts of reagents. We also appreciate other members of our laboratory for helpful comments.