Xenopus ADAM19 regulates Wnt signaling and neural crest specification by stabilizing ADAM13

The vertebrate neural crest (NC) cells are multipotent embryonic cells that give rise to the craniofacial cartilages and bones, sensory neurons and glia, pigment cells, cardiac tissues, and many other cells and tissues. Deficiencies in NC development are a major cause of human birth defects, ranging from relatively subtle craniofacial abnormalities to more severe diseases such as CHARGE and DiGeorge syndromes (Bronner and Simões-Costa, 2016; Mayor and Theveneau, 2013). NC-based cell therapies are also being developed to repair tissues that are damaged by injuries or deteriorate with age (Kunisada et al., 2014; Zhu et al., 2016). Understanding the basic mechanisms that control the specification of NC lineage is crucial for developing new methods to prevent NC-related birth defects and to regenerate NC-derived tissues for clinical applications.

In vertebrate embryos, the NC cells are induced at the neural plate border (NPB) during gastrulation. NC induction requires coordinated actions of multiple signaling pathways, which activate the expression of transcription factors such as Pax3/7, Zic1 and Msx1 that induce NPB formation (i.e. the NPB specifiers). The NPB specifiers in turn activate the expression of another set of transcription factors (i.e. the NC specifiers), such as Snail2, FoxD3 and Sox9, which specify NC lineage in certain NPB cells. Once specified, the NC cells undergo proliferation while maintaining their multipotency. Around the time of neural tube closure, the NC cells delaminate and migrate to specific destinations, where they differentiate into various types of cells and tissues (Bronner and Simões-Costa, 2016; Simões-Costa and Bronner, 2015; Stuhlmiller and García-Castro, 2012).

A major signaling pathway that is essential for NC induction is the canonical Wnt pathway. In Xenopus embryos, Wnt seems to be important for both the initial formation of the NPB and the subsequent specification of NC lineage. During NPB formation, Wnt is required for inducing the expression of Pax3 and Msx1, which cooperate with Zic1 to set up the NPB (Li et al., 2009; Monsoro-Burq et al., 2005). Gain-of-function studies suggest that the combination of Pax3 and Zic1 is sufficient to specify the NC lineage ex vivo in the animal cap or ectopically in the ventral ectoderm, and this effect is again Wnt-dependent (Milet et al., 2013; Monsoro-Burq et al., 2005; Sato et al., 2005). The direct Wnt target in NC specification is probably Snail2, as snail2 is the only NC specifier gene whose enhancer is known to contain a Wnt-responsive LEF/TCF-binding site (Simões-Costa and Bronner, 2015; Vallin et al., 2001). However, it remains unclear whether the specification of endogenous NC lineage also requires Wnt activity.

The disintegrin metalloproteinases (ADAMs) are multi-domain transmembrane proteins that play important roles in development and disease. Most of the functional domains of ADAMs are extracellular, including a metalloproteinase domain that is catalytically active in some but not all ADAMs, as well as disintegrin and cysteine-rich domains that can mediate protein-protein interactions (Wei, 2013). About half of the ADAMs are expressed predominantly in the testis and are mainly involved in reproduction, whereas the others have significant presence in somatic tissues. Several somatic ADAMs are known to control cell signaling through their protease activities. For example, proteolytic processing of the Notch receptors by ADAM10 is necessary for Notch signal activation, and shedding of the ligands for epithelial growth factor receptors (EGFRs) by multiple ADAMs is an essential regulatory event in EGFR signaling (Edwards et al., 2008; Reiss and Saftig, 2009). We have also shown that another ADAM, ADAM13, induces NC in Xenopus embryos by regulating canonical Wnt signaling. Knockdown (KD) of ADAM13 leads to reduced expression of markers for pre-migratory NC and inhibition of Wnt8-induced axis duplication. We further identified ephrin B1 and ephrin B2, two cell-surface ligands for EphB receptor tyrosine kinases, as the in vitro and in vivo substrates for ADAM13 and antagonists of Wnt signaling. By cleaving these ephrins, ADAM13 permits Wnt activation, which in turn induces the NC (Wei et al., 2010b). Although these results clearly demonstrate the importance of ADAM13 in Wnt signaling and NC induction, direct evidence showing the regulation of endogenous Wnt signaling by ADAM13 in one or both stages of NC induction was still missing.

In the current study, we have investigated further the roles of ADAM13 and its close paralog ADAM19 in Wnt signaling and NC induction in X. tropicalis embryos. Using a transgenic Wnt reporter line, we detected high endogenous Wnt activity at the NPB at late gastrula stages. KD of either ADAM13 or ADAM19 caused strong inhibition of Wnt activity at the NPB and reduced expression of NC markers; however, the expression of NPB markers remained unaffected. Thus, ADAM13 and ADAM19 are both required for a particular wave of Wnt signaling that directly specifies the NC lineage but is not required for NPB formation. We further show that the function of ADAM19 in NC specification is independent of its protease activity but via protection of ADAM13. ADAM13 and ADAM19 colocalized in the ER and physically interacted with each other, and KD of ADAM19 resulted in reduction of ADAM13 protein levels that was caused by proteasomal degradation. Such proteasomal degradation was mediated by a conserved C-terminal lysine residue of ADAM13, as mutation of this lysine residue to arginine or deletion of the whole cytoplasmic tail generated stabilized ADAM13 mutants that were resistant to ADAM19 KD. Ectopic expression of these stabilized ADAM13 mutants could not only rescue the NC specification defects caused by ADAM19, but also induce robust expression of snail2 in the future epidermis in a Wnt-dependent manner. Together, our data suggest a novel non-proteolytic function of ADAM19 in regulating the turnover of ADAM13 and a distinct wave of Wnt signaling that specifies the NC lineage.

Our phylogenetic and syntenic (gene linkage) analyses suggest that ADAM12, ADAM13 and ADAM19 form a subfamily of proteolytically active ADAMs that is conserved among vertebrates (note that the mammalian ortholog of Xenopus ADAM13 is ADAM33) (Bahudhanapati et al., 2015; Wei et al., 2010a). Similar to its X. laevis ortholog (Neuner et al., 2009), X. tropicalis adam19 was found to be expressed in the dorsal mesoderm during mid-gastrulation, and broadly in the dorsal ectoderm at the end of gastrulation (Fig. S1A,B). At the same stages, transcripts of adam13 were detected in the mesoderm and pre-migratory NC, respectively (Wei et al., 2010b). By contrast, adam12 did not display any specific expression patterns (data not shown).

We reported previously that KD of ADAM13 causes defects in craniofacial and eye development in X. tropicalis embryos (Wei et al., 2012, 2010b). To understand the functions of ADAM12 and ADAM19, antisense MOs were designed to block the translation of individual ADAMs (Fig. 1A) (Wei et al., 2010b). Morphology of the craniofacial structures was characterized using a transgenic X. tropicalis reporter line that expresses enhanced green fluorescent protein (eGFP) driven by a ∼4 kb fragment of the snail2 enhancer/promoter, which allows direct imaging of NC derivatives in live embryos (J.L., M.P. and S.W., unpublished). We injected individual MOs to target the anterodorsal region of the snail2 reporter embryos, and raised these embryos (‘morphants’) to tadpole stages. Although ADAM12 morphants displayed no apparent phenotype (data not shown), injection of either ADAM19 MO (MO 19-1 or 19-2) led to reduction in head cartilage structure on the injected side, a phenotype similar to that caused by ADAM13 MOs (MOs 13-3 and 13-1; Fig. 1B) (Wei et al., 2010b). Because the head cartilage structures are NC derivatives and ADAM13 is known to induce the NC, we tested whether loss of ADAM19 also affects NC induction. Indeed, in situ hybridization results show that KD of either ADAM13 or ADAM19 drastically reduced the NC marker/specifier snail2 at stage ∼12.5 (Fig. 1C), when NC specification has just started, suggesting that both ADAMs are essential for NC induction.

Another phenotype caused by ADAM13 KD is the impaired eye structures (Wei et al., 2012), which was also manifested in ADAM19 morphants (Fig. S2A). Examination of the eye-field markers pax6 and rx1 reveals that both markers were downregulated by stage ∼12.5 (Fig. S2B,C) (Wei et al., 2012), indicating that depletion of either ADAM13 or 19 interferes with early eye field induction. Therefore, KD of ADAM19 phenocopies ADAM13 loss of function by inhibiting the induction of both the NC and eye field.

Because NC induction occurs in two steps, the initial formation of the NPB and the subsequent specification of NC lineage, we asked which step(s) are controlled by ADAM13 and ADAM19? In addition to snail2, sox9 and foxd3, two other NC markers/specifiers, were both downregulated at stage ∼12.5 by KD of ADAM13 or ADAM19; however, little or no effect was detected for NPB markers/specifiers pax3, zic1 or msx1 (Fig. 1D). We conclude from these results that both ADAM13 and ADAM19 are required for NC specification but not for earlier NPB formation.

The pivotal role of ADAM13 in NC specification has been attributed to its ability to regulate the activation of canonical Wnt signaling, as KD of ADAM13 inhibits Wnt8-induced axis duplication, and restoration of Wnt signaling rescues the NC defects caused by ADAM13 KD (Wei et al., 2010b). Wnt is thought to be essential for both the earlier NPB formation and the subsequent NC specification (Abu-Elmagd et al., 2006; Simões-Costa and Bronner, 2015; Stuhlmiller and García-Castro, 2012). The inhibition of NC specification but not NPB formation by KD of ADAM13 or ADAM19 suggests that these ADAM proteins may be required for Wnt activation at the established NPB to specify the NC lineage. To test this hypothesis, we took advantage of another transgenic X. tropicalis eGFP reporter line, which has been shown to faithfully report the endogenous Wnt activity (Tran et al., 2010). At stage ∼12.5, we detected strong eGFP signal at the NPB in the Wnt reporter embryos (Fig. 1E, white arrowheads), confirming that Wnt signaling is highly active at the NPB. Although Wnt signaling is also required for NPB formation, this earlier step (at stage ∼10.5) occurs roughly 3.5 h prior to stage ∼12.5 in X. tropicalis embryos at our experimental temperature (22°C). Because the destabilized eGFP protein used in the Wnt reporter line has a half-life of around 2 h (Tran et al., 2010), it is unlikely that the strong Wnt signal we observed here was the remnant of NPB formation. Depletion of ADAM13 drastically downregulated eGFP expression at the NPB in all injected embryos, and this effect was again mimicked by ADAM19 KD (Fig. 1E). Thus, both ADAM13 and ADAM19 are required in the right place and at the right time for Wnt activation to specify the NC lineage.

The striking similarity between the loss-of-function phenotypes of ADAM13 and ADAM19 implies a connection between the in vivo roles of these two metalloproteinases. Because ADAM19 is one of the closest paralogs of ADAM13 (Bahudhanapati et al., 2015; Wei et al., 2010a), the simplest explanation is that these two ADAMs have redundant functions. To examine whether this is the case, we first compared the in vivo roles of the protease activities of ADAM13 and ADAM19.

As shown in Fig. 2A, ectopic expression of wild-type ADAM13 caused anterior expansion of the expression domains of snail2 and sox9, effects that are similar to those caused by ectopically expressed β-catenin or a dominant-negative EphB receptor (Hong et al., 2008; Li et al., 2009; Wei et al., 2010b). These results provide further support for our model that ADAM13 induces the NC by inhibiting forward ephrin B signaling and promoting Wnt signaling (Wei et al., 2010b). Conversely, overexpression of the protease-dead E340A mutant of ADAM13 greatly reduced both snail2 and sox9 (Fig. 2A), suggesting that this mutant dominant-negatively inhibited NC specification. The dominant-negative effect of the E/A mutants has been reported for several ADAMs, including ADAM13 (Alfandari et al., 2001), and is presumably caused by the abilities of the mutants to bind and sequester substrates from the endogenous ADAM proteases. The expansion and inhibition of the NC caused by wild-type ADAM13 and the E340A mutant, respectively, confirm the importance of ADAM13 protease activity in NC specification. Surprisingly, ectopic expression of wild-type ADAM19, as well as the protease-dead E345A mutant, slightly expanded the expression domains of snail2 and sox9 (Fig. 2A), indicating that ADAM19 may function independently of its protease activity. We further tested whether the E345A mutant can rescue the NC specification defects caused by ADAM19 KD. Indeed, co-injection of the mRNA encoding either the wild-type ADAM19 or the E345 mutant with MO 19-1 rescued the reduction in snail2 and sox9 expression (Fig. 2B). This is in stark contrast to our previous observations that wild-type ADAM13, but not the E340A mutant, rescued the NC-specification defects caused by ADAM13 KD (Wei et al., 2010b). These results indicate a fundamental difference in the mechanism of action for the two ADAMs: unlike ADAM13, the protease activity of ADAM19 is dispensable for its function in NC specification.

The ability of wild-type ADAM19 and the E345A mutant to rescue the NC specification phenotypes caused by MO 19-1 confirms that these phenotypes are specific for ADAM19 KD. To further examine the efficacy and specificity of ADAM13 and 19 MOs, we assessed the abilities of these MOs to inhibit the translation of corresponding mRNA targets, as well as their closest paralogs. Western blot analyses show that each MO knocked down its target protein (Fig. 3A, lanes 4, 11 and 12). However, both MOs 19-1 and 19-2, two ADAM19 MOs with no overlap in their sequence (Fig. 1A), drastically downregulated the exogenously expressed ADAM13 (Fig. 3A, lanes 5 and 6). By contrast, the ADAM12 MO (MO 12-1) did not have a significant effect on ADAM13 (lane 3), and neither MO 12-1 nor 13-1 downregulated ADAM19 (lanes 9 and 10). Moreover, ectopically expressed ADAM10 and ADAM12 were also unaffected by MO 19-1 (data not shown), suggesting that the loss of ADAM13 caused by MO 19-1 or 19-2 was specific. Western blot analyses using an anti-ADAM13 antibody further show that whole-embryo injections of MO 19-1 or 19-2 resulted in over 60% reduction of endogenous ADAM13 in stage ∼12.5 X. tropicalis embryos; a similar effect was observed in X. laevis embryos (Fig. 3B and data not shown). It is unlikely that the loss of endogenous and exogenous ADAM13 in embryos injected with MO 19-1 or 19-2 was caused by off-target effects, as the transcripts of adam13 and adam19 do not share any significant sequence similarity despite the moderate homology between the two proteins (∼43% identities); we were also unable to detect any potential binding sites for either MO 19-1 or 19-2 on ADAM13 mRNA. Rather, these results imply that ADAM19 somehow post-transcriptionally regulates ADAM13 levels in Xenopus embryos. This is further supported by our observation that exogenous ADAM19 increased the levels of co-expressed ADAM13 in X. tropicalis embryos (Fig. 3C). Notably, a comparable increase in ADAM13 was observed with both wild-type ADAM19 and the protease-dead E345A mutant (Fig. 3C), suggesting that the regulation of ADAM13 is independent of ADAM19 protease activity.

To further investigate the interaction between ADAM13 and ADAM19, we determined the subcellular localization of these proteins in Xenopus XTC cells. Owing to the low expression levels of ADAM19 in XTC cells, we overexpressed this protein. Immunocytochemistry results reveal that the exogenous ADAM19 was predominantly localized to the ER, as shown by the overlapping patterns of ADAM19 and the ER marker calnexin (Fig. 4A). Our bioinformatic analyses also suggest that there are multiple possible ER-localization signals in the cytoplasmic tail of ADAM19 (R. Neuner, PhD Thesis, University of Massachusetts-Amherst, MA, USA, 2011). The endogenous ADAM13 colocalized with the exogenous ADAM19, but there was also a significant part of ADAM13 that was found in the lamellipodia (Fig. 4B). The localization of ADAM13 to the lamellipodia is consistent with the function of this protease in cleaving cell-surface substrates, such as cadherin 11, to promote migration (McCusker et al., 2009). Similar colocalization of ADAM13 and ADAM19 was also detected in 293T cells (data not shown).

We further examined whether ADAM13 and ADAM19 physically interact with each other by overexpressing both ADAM proteins in 293T cells. Immunoprecipitation of either ADAM13 or ADAM19 pulled down the other ADAM that was co-expressed, indicating that these two ADAMs were associated with each other (Fig. 4C). To identify the domain(s) of ADAM13 that are responsible for interacting with ADAM19, we co-expressed ADAM19 with full-length ADAM13, or the disintegrin and cysteine-rich domains individually and combined, in 293T cells. Results of co-immunoprecipitation experiments show a strong association of ADAM19 with the cysteine-rich domain of ADAM13 (Fig. 4D), suggesting that this domain plays a major role in the interaction between ADAM13 and ADAM19.

We next asked how the interaction between ADAM13 and ADAM19 results in the increase of ADAM13 levels, as seen in Fig. 3. Because ADAM13 is known to cleave itself (Gaultier et al., 2002), it is possible that binding of ADAM19 to ADAM13 prevents the autocleavage of ADAM13. If this were the case, KD of ADAM19 should have no effect on the protease-dead E340A mutant of ADAM13. As shown in Fig. 5A, ectopic expression of wild-type ADAM13 with a C-terminal myc₆ tag gave rise to a smaller fragment (indicated by an asterisk) in addition to the pro- and mature forms of ADAM13. This fragment was not present in embryos expressing the E340A mutant, indicating that it was the C-terminal autocleavage product. Both the E340A mutant and the autocleavage product of wild-type ADAM13 were effectively downregulated by MO 19-1, arguing against the possibility that ADAM19 inhibits the autocleavage of ADAM13. By contrast, KD of ADAM13 resulted in an increase of the autocleavage product of ADAM19 (Fig. S3). Together, these data suggest that the interaction between these two ADAMs inhibits the autocleavage of ADAM19 but not ADAM13. Interestingly, MO 19-1 had little effect on the levels of an ADAM13 mutant with the cytoplasmic tail deleted (ΔC; Fig. 5A). Thus, the cytoplasmic tail of ADAM13, but not the protease activity, is crucial for the regulation by ADAM19.

We have previously found that deletion of the cytoplasmic tail results in an ∼8-fold increase in cell surface levels of X. laevis ADAM13 (Alfandari et al., 2001). A similar effect was observed for X. tropicalis ADAM13 (Fig. 5F), raising the possibility that the cytoplasmic tail may regulate the turnover of ADAM13, and that such turnover may be blocked by ADAM19. One of the most common mechanisms for membrane protein turnover is the ubiquitylation of the cytoplasmic tail, which targets the proteins for proteasome-mediated degradation (MacGurn et al., 2012). Owing to difficulties in detecting endogenously ubiquitylated ADAM13, we co-expressed HA- or myc₆-tagged ADAM13 with FLAG-tagged ubiquitin in 293T cells. Co-immunoprecipitation assays detected FLAG-tagged ubiquitin that was associated with ADAM13 (Fig. 5B,E), suggesting that ADAM13 can be ubiquitylated. To test whether ADAM13 undergoes proteasome-mediated degradation in vivo, we dissociated X. tropicalis embryos expressing ectopic ADAM13, and cultured the cells with and without the proteasome inhibitor MG132. Western blot analyses show significantly higher levels of ADAM13 upon MG132 treatment (Fig. 5C). Notably, MG132 rescued the effects of ADAM19 KD on the levels of ADAM13 (Fig. 5C), indicating that ADAM19 regulates the proteasomal degradation of ADAM13.

There are 20 lysine residues in the cytoplasmic tail of X. tropicalis ADAM13, and our prediction using the BDM-PUB algorithm suggests that the most C-terminal K911 residue has a very high probability of being ubiquitylated (Table 1). This lysine residue is conserved in ADAM13 from fish to opossums, but not in the orthologs from higher mammals, such as mice or humans. Some other ADAMs, including the closely related ADAM12 from all animal species examined, also contain a lysine residue at the C terminus that is likely ubiquitylated (Table 1 and data not shown), suggesting a possible common mechanism of regulation. To examine the role of the K911 residue of X. tropicalis ADAM13, we mutated this residue to arginine. When the same amount of expression plasmid was transfected into XTC or the mammalian 293T or 3T3 cells, the K911R mutant was expressed consistently at higher levels than wild-type ADAM13 (Fig. 5D and data not shown). Importantly, when compared with wild-type ADAM13, the K911R mutant had reduced ubiquitylation when co-expressed with FLAG-tagged ubiquitin in 293T cells (Fig. 5E).

To assess the effects of ubiquitin on ADAM13 function in NC specification, we injected the same amount of transcripts encoding different variants of ADAM13, with and without the plasmid encoding ubiquitin, into X. tropicalis embryos. Similar to wild-type ADAM13, both the ectopically expressed K911R and ΔC mutants were able to induce an expansion of the sox9 expression domain. Although co-expressed ubiquitin completely inhibited the expansion of sox9 expression induced by wild-type ADAM13 and even caused a reduction of endogenous sox9, the effect was much weaker on the K911R mutant and nearly nonexistent on the ΔC mutant (Fig. S4). Finally, we determined the responses of different ADAM13 variants to ADAM19 MO in vivo. As shown in Fig. 5F, both the K911R and ΔC mutants were expressed at higher levels than wild-type ADAM13; unlike wild-type ADAM13, neither the K911R nor the ΔC mutant was reduced upon ADAM19 KD. Therefore, the K911 residue of ADAM13 is likely ubiquitylated and is essential for the regulation by ADAM19.

The requirement of ADAM19 for maintaining the protein levels of ADAM13, along with the similarity between the phenotypes of ADAM13 and ADAM19 KDs, strongly implies that ADAM19 may function through regulation of ADAM13. To test this hypothesis, we evaluated the abilities of both ADAMs to rescue each other's KD phenotypes. The ΔC and K911R mutants, but not wild-type ADAM13 that was depleted upon ADAM19 KD, effectively rescued the NC specification phenotypes caused by ADAM19 KD, as shown by in situ hybridization for snail2 and sox9 (Fig. 6A and Fig. S5A). We also noticed that there was ectopic snail2 expression in embryos injected with either the ΔC or the K911R mutant, even in the presence of ADAM19 MO (Fig. S5A), suggesting that both mutants are sufficient to induce snail2 expression outside of the NC territory independently of ADAM19 (to be further investigated below). To rule out the possibility that the myc₆ tag interferes with the biochemical properties of ADAM13, we generated constructs encoding untagged ADAM13 variants. The untagged ΔC and K911R mutants were indistinguishable from their myc₆-tagged counterparts in rescuing the NC specification defects caused by MO 19-1 and in inducing ectopic snail2 (Fig. S5B), confirming that the myc₆ tag does not affect ADAM13 function. By contrast, neither wild-type ADAM19 nor the E/A mutant was able to rescue the reduced sox9 expression caused by ADAM13 MO (Fig. 6B). Taken together, these data indicate that the in vivo function of ADAM19 in NC specification is dependent on ADAM13, but not vice versa. We further tested whether the eye field induction defects caused by ADAM19 KD can also be rescued by the ΔC and K911R mutants of ADAM13. Indeed, expression of the eye field markers pax6 and rx1 was restored when mRNA encoding either ADAM13 mutant was co-injected with MO 19-1 (Fig. S6), suggesting that ADAM19 also functions in eye-field induction by regulating ADAM13.

We have previously identified Snail2 as a major downstream target that mediates ADAM13 function in NC specification, as exogenous Snail2 rescued the NC defects caused by ADAM13 KD (Wei et al., 2010b). Our current finding that the ΔC and K911R mutants of ADAM13 can induce ectopic snail2 in the dorsal ectoderm (Fig. S5A,B) not only validates these results, but also suggests that the ADAM13-Wnt signaling axis may partially re-specify non-NC cells to adopt some NC-like properties. To test this possibility, we injected transcripts encoding different variants of ADAM13 into the posterior-ventral region of X. tropicalis embryos, where the cells are destined to become epidermis and do not express detectable levels of endogenous snail2. As shown in Fig. 6C, a low dose of mRNA encoding wild-type ADAM13 induced weak ectopic expression of snail2 in the ventral ectoderm, whereas the same amount of the ΔC and K911R mutant transcripts caused a robust induction. A similar induction of snail2 expression by wild-type ADAM13 was obtained in animal cap assays using X. laevis embryos (Khedgikar et al., 2017). By contrast, the E/A mutant failed to induce ectopic snail2, and co-injecting a β-catenin MO completely abolished the snail2-inducing activity of the K911R mutant (Fig. 6C). These results are consistent with our previous observations that ADAM13 protease activity and downstream Wnt signaling are required for NC specification (Wei et al., 2010b). However, we were unable to detect any ectopic expression of other NC or NPB specifiers that were tested, including sox9, foxd3, pax3, zic1 and msx1, in any of these embryos (data not shown). Thus, the ADAM13-Wnt axis is sufficient to induce snail2 in the ventral ectoderm but not to produce a complete conversion to the NC lineage.

Many extracellular metalloproteinases, including several ADAMs, have been implicated in NC development (Christian et al., 2013). The ADAMs are key regulators of cell signaling, mainly due to their abilities to cleave signaling molecules such as ligands, receptors, agonists and antagonists. However, some ADAM proteases may function through their non-proteolytic activities in certain physiological contexts, and there are also non-proteolytic ADAMs that play important roles in development and disease (Edwards et al., 2008; Reiss and Saftig, 2009). As compared with the well-characterized functions of ADAM protease activities, little is known about the non-proteolytic roles of ADAMs in vivo. Here we report that Xenopus ADAM19 specifies the NC lineage through non-proteolytic protection of ADAM13. To our knowledge, there is no published example of one protease protecting another in vivo. A possible mechanism for this novel protease-protease interaction is discussed below.

The first two testicular ADAMs, ADAMs 1 and 2, were discovered as a heterodimer that is involved in sperm-egg fusion (Blobel et al., 1992). However, the physiological consequence of this dimerization is unclear, and similar interaction has not been reported for somatic ADAMs. Instead, ADAMs 10 and 17, two somatic ADAMs, are known to interact with tetraspanins and non-proteolytic members of the Rhomboid family (iRhoms), respectively, which regulate the trafficking, stability and activity of these ADAMs (Adrain et al., 2012; Dornier et al., 2012; Haining et al., 2012; Matthews et al., 2017; McIlwain et al., 2012). The activity of Xenopus ADAM13 has also been shown to be regulated by other proteins, such as PACSIN2 and XFz4 (Abbruzzese et al., 2015; Cousin et al., 2000). In the current study, we observed a protective effect of ADAM19 towards ADAM13, which underlies the in vivo function of ADAM19 in NC specification. Our data suggest that, in the absence of ADAM19, ADAM13 is ubiquitylated at the C-terminal K911 residue and degraded through the proteasome pathway. When ADAM19 is present, it binds to the cysteine-rich domain of ADAM13 and possibly masks the K911 residue in the cytoplasmic tail of ADAM13, resulting in the stabilization of ADAM13 that subsequently specifies the NC lineage (Fig. 6D). Interestingly, although ADAM13 inhibits the autocleavage of ADAM19, ADAM19 does not seem to cleave ADAM13 or interfere with its protease activity (Fig. S3 and Fig. 5A). Therefore, the heterodimer is likely asymmetric: although the protease activity of ADAM19 may be blocked by ADAM13, ADAM13 remains free to cleave its substrates (Fig. 6D). This structural asymmetry is likely crucial for the difference in the mechanism of action for the two ADAMs in NC specification.

Most of the studies on the functions of ADAMs in early NC development have been carried out in two Xenopus species, X. laevis and X. tropicalis (Christian et al., 2013). Our previous work unveiled the essential roles of ADAM13 in NC induction in X. tropicalis embryos (Wei et al., 2010b), which is confirmed by the current study. However, KD of ADAM13 in X. laevis did not affect any NC markers, raising the issue of whether the function of ADAM13 in NC induction is conserved (Cousin et al., 2011). A major difference between the two frog species is that X. tropicalis is diploid, whereas X. laevis is allotetraploid (Session et al., 2016). The X. laevis genome contains a pseudoallelle of adam13 (also known as xmdc13), which encodes a protein that is ∼88% and 87% identical to X. laevis and X. tropicalis ADAM13, respectively (Cai et al., 1998). KD of both ADAM13 and xMDC13 did result in reduction of snail2 expression prior to NC migration in X. laevis embryos (D.A., unpublished), indicating that ADAM13 function in NC specification is conserved between the two frog species. Unlike adam13, there is only one copy of adam19 in X. laevis, and KD of this gene also causes reduction of pre-migratory NC markers at stage ∼15 (Neuner et al., 2009). However, it was unclear whether this effect was caused by inhibition of induction or post-induction maintenance of the NC. Combined with the current study, these results suggest that ADAM19 is required for NC specification in both Xenopus species. At later stages, ADAM13 is required for NC migration through its protease activity, as well as the cytoplasmic tail, which is cleaved and translocalizes into the nucleus to regulate target gene transcription (Cousin et al., 2011). By contrast, we show here that the function of ADAM13 in NC specification is independent of its cytoplasmic tail, as the ΔC mutant can effectively rescue the early NC defects caused by ADAM19 KD and induce ectopic snail2 (Fig. 6A,C, Fig. S5A,B). Thus, the mechanisms of action for ADAM13 in NC specification and migration appear to be different.

Previous research has implicated more than one wave of canonical Wnt signaling in NC induction. In Xenopus embryos, several Wnt ligands and receptors have been reported to have NC-inducing capacity, and loss-of-function studies have confirmed that at least two canonical Wnt ligands, Wnt3a and Wnt8, as well as two Wnt receptors, Xfz3 and Xfz7, are required for endogenous NC induction (Abu-Elmagd et al., 2006; Deardorff et al., 2001; Elkouby et al., 2010; Hong et al., 2008). These Wnt ligands and receptors are expressed in different regions at the time of NC induction (Abu-Elmagd et al., 2006; Steventon et al., 2009), suggesting that, instead of functioning redundantly, they may form separate waves of Wnt signaling to regulate different aspects of NC induction. Indeed, NC induction can be divided into two steps, the formation of NPB and the specification of NC lineage, both of which are thought to depend on Wnt signaling (Simões-Costa and Bronner, 2015). However, although the regulation of NPB markers/specifiers by Wnt signaling has been well studied, whether and how Wnt controls endogenous NC specification remained elusive. This is probably because NC specification immediately follows, and is difficult to be separated from, NPB formation. In particular, it was unclear whether the two steps of NC induction are controlled by the same wave of Wnt signaling or separate ones. In this study, we provide three important lines of evidence suggesting that a distinct wave of Wnt signaling, which is controlled by ADAM13 and ADAM19, is required for endogenous NC specification but not NPB formation in X. tropicalis embryos. First, by using a transgenic Wnt reporter line, we found that canonical Wnt signaling is highly active within the NPB at late gastrula stages (Fig. 1E). To our knowledge, this is the first direct evidence indicating that there is strong Wnt signal at the NPB during NC specification, although previous studies indicate that Wnt activity in NC explants increases between gastrula and neurula stages (Steventon et al., 2009). This spatiotemporal pattern of Wnt activity is consistent with its function in NC specification. Second, KD of either ADAM13 or ADAM19 inhibits Wnt signaling and NC specification at the NPB, but the formation of NPB itself is unaffected (Fig. 1C-E). These results differ from previous reports that Wnt inhibition causes loss of both NPB and NC markers (Li et al., 2009; Monsoro-Burq et al., 2005). Thus unlike the ‘core’ Wnt components that are required ubiquitously for Wnt signal activation, ADAM13 and ADAM19 are needed in only a subset of tissues, probably where the Wnt signal is cross-antagonized by ephrin B signaling. Such differential requirement allows us to separate the wave of Wnt signaling that specifies NC lineage from the one that establishes the NPB. Last, we show that ectopic expression of ADAM13 induces the NC specifier snail2 in the ventral ectoderm in a Wnt-dependent manner (Fig. 6C). This gain-of-function phenotype provides additional evidence for the specification of NC lineage by the ADAM13-Wnt signaling axis. Among the NC specifier genes, snail2 is the only one that is known to contain a cis-regulatory element that binds LEF/TCF transcription factors, and has been speculated to be the direct Wnt target during NC specification (Simões-Costa and Bronner, 2015). Interestingly, ADAM13 cannot induce ectopic expression of any of the other NC or NPB markers that were examined, suggesting that snail2 is indeed the direct target of the ADAM13/19-Wnt axis. This is further supported by our previous finding that exogenous Snail2 can rescue the reduction of other NC markers, such as sox9 and twist, caused by ADAM13 KD (Wei et al., 2010b).

The potential application of NC cells in tissue regeneration for the treatment of related birth defects or damages has drawn a great deal of research interest. Methods are being actively developed to differentiate pluripotent stem cells or reprogram non-NC lineages to generate NC cells (Kunisada et al., 2014; Zhu et al., 2016). In this study, we found that the stabilized ADAM13 mutants that can function independently of ADAM19 induce robust expression of the key NC specifier snail2 in the ventral ectoderm; however, the lack of other NC or NPB markers/specifiers indicates an incomplete conversion to the NC lineage (Fig. 6C). By contrast, overexpression of ADAM13 on the dorsal ectoderm causes expansion of both snail2 and other NC markers, such as sox9 along the NPB (Fig. 2A), suggesting that additional molecules, which are normally present at the NPB, are needed to cooperate with ADAM13 to induce a complete conversion of the ventral ectoderm cells to NC lineage. Potential candidates include inhibitors of bone morphogenetic protein (BMP) signaling, such as Chordin and Noggin, because high BMP signal on the ventral side of the embryos is known to inhibit NC induction (LaBonne and Bronner-Fraser, 1998; Steventon et al., 2009). Our results also call for caution in interpreting some of the published data on ectopic NC induction, where snail2 is used as the only marker for the NC lineage. Finally, Wnt signaling and Snail2 are known to be key regulators of the epithelial-to-mesenchymal transition (EMT) in development, wound repair and cancer. In the traditional models of EMT, Wnt and Snail2 control the downstream expression of metalloproteinases, which in turn modify cell-cell and cell-matrix adhesion through their proteolytic activities to promote cell migration and invasion (Lim and Thiery, 2012; Powell et al., 2013). Here, we show that both the proteolytic and nonproteolytic activities of ADAMs play important roles upstream to regulate Wnt signaling and snail2 expression. It would be of interest to examine whether similar mechanisms are also involved in developmental and/or pathological EMT.

X. tropicalis cDNA clones encoding adam12, adam13 and adam19 were obtained previously (Wei et al., 2010b). Constructs encoding C-terminally myc₆- or HA-tagged wild-type ADAMs were generated by subcloning into pCS2+ vectors modified to append corresponding epitopes, and mutations were introduced using a QuikChange mutagenesis kit (Stratagene). Rescue constructs were made by introducing sense mutations downstream of the translational start codon in corresponding myc₆-tagged expression constructs. Constructs for preparing the in situ hybridization probes for snail2, sox9, foxd3, pax3, zic1, msx1, pax6 and rx1 were obtained previously (Li et al., 2011; Wei et al., 2012, 2010b). The FLAG-ubiquitin construct was a gift from Dr Carole Labonne (Northwestern University, Evanston, IL, USA). In vitro transcription was carried out as described previously to generate mRNA transcripts and in situ hybridization probes (Sive et al., 2000). Commercial primary antibodies that were used in this study include: mouse anti-myc (DSHB 9E10, 1:500), mouse anti-HA (Sigma-Aldrich H9658, 1:20,000), mouse anti-FLAG (Sigma-Aldrich F1804, 1:1000) and rabbit anti-calnexin (Enzo Life Sciences ADI-SPA-860, 1:200). The rabbit anti-ADAM13 (6615F, 1:200) and anti-ADAM19 (2C5 E1, 1:200) antibodies were generated previously (Alfandari et al., 1997; Neuner et al., 2009). Commercial secondary antibodies that were used include HRP-conjugated rabbit anti-mouse (Sigma-Aldrich A9044, 1:80,000) and goat anti-rabbit (Sigma-Aldrich A0545, 1:80,000), as well as Alexa Fluor 488 donkey anti-mouse (Jackson Laboratories 715-545-150, 1:200) and Alexa Fluor 594 donkey anti-rabbit (Jackson Laboratories 711-585-152, 1:200). For loading controls, unconjugated mouse anti-β1-integrin (DSHB 8C8, 1:10), anti-GAPDH (Millipore MAB374, 1:10,000) or HRP-conjugated mouse anti-β-actin (Sigma-Aldrich A5316, 1:5000) were used.

Wild-type X. tropicalis and X. laevis adults (male and female) were purchased from NASCO. The transgenic X. tropicalis Wnt reporter frogs were provided courtesy of Dr Kris Vleminckx (Flanders Institute for Biotechnology, Belgium), and the transgenic X. tropicalis snail2-eGFP line was generated as described elsewhere (J.L., M.P. and S.W., unpublished). Methods involving live animals were carried out in accordance with the guidelines and regulations approved and enforced by the Institutional Animal Care and Use Committees at West Virginia University, University of Delaware and University of Massachusetts at Amherst. Morpholino oligos were designed and synthesized by Gene Tools. Embryo preparation, injection and culturing were carried out as described previously (Wei et al., 2010b). Briefly, embryos were obtained by natural mating and injections were carried out to target specific blastomeres (Huang et al., 1998) using a PLI-100A microinjector (Harvard Apparatus). Alexa Fluor 555 or 488 dextran (Invitrogen) was co-injected as a lineage tracer. Embryos were cultured in 0.1× MBS to desired stages, and injected embryos were sorted according to the injected side using a Zeiss Axiozoom.v16 epifluorescence microscope. Embryo lysates were prepared as described below for cell lysates. For embryo dissociation, X. tropicalis embryos were injected in both blastomeres at the two-cell stage and cultured to stage ∼12.5. The embryos were placed in 1.5 ml centrifuge tubes with 0.1× MBS and 100 μM MG132 (or DMSO as control). The embryos were then dissociated using a 28 gauge 1 ml syringe (Exel) and incubated in 0.1× MBS and 100 μM MG132 or DMSO at 23°C for 5 h.

HEK293T cells were purchased from ATCC and cultured in DMEM (ATCC) supplemented with 10% fetal bovine serum (FBS; Gibco) at 37°C with 5% CO₂. XTC cells (Pudney et al., 1973) were cultured in L-15 medium (GE Life Sciences) supplemented with 10% FBS and 1 mM sodium pyruvate at 23°C. Both cell lines were recently authenticated and tested for contamination. Cells were transfected using Lipofectamine 3000 (Invitrogen) at 50-70% confluency, and allowed to grow for another 24-48 h. For MG132 treatment, cells were cultured in the presence of 40 µM MG132 for 24 h before being harvested. To prepare whole-cell lysates, cells were lysed in ice-cold lysis buffer [20 mM Tris-HCl (pH 7.6), 100 mM NaCl, 1% Triton X-100, 50 mM NaF, 10 mM Na₄P₂O₇, 1 mM PMSF and 5 mM EDTA] supplemented with protease inhibitor cocktail (Sigma-Aldrich) and 10 mM o-phenanthroline, and centrifuged for 20 min at 12,000 g at 4°C. Supernatants were processed for SDS-PAGE and blotted with corresponding antibodies.

Embryos were fixed at desired stages, and in situ hybridization was carried out as described (Sive et al., 2000). For immunocytochemistry, cells were cultured and transfected on glass coverslips (Fisher Scientific), and fixed with 1:1 methanol/acetone at −20°C for 10 min. The cells were then rehydrated with 1× PBS for 10 min, blocked with 1% bovine serum albumin (Sigma-Aldrich) for 1 h at room temperature, and incubated with the primary antibody at 4°C overnight. After being washed with PBS containing 0.5% Tween-20, the cells were subsequently incubated with Alexa Fluor-conjugated secondary antibody for 1 h at room temperature, and labeled with DAPI (Sigma-Aldrich) to identify cell nuclei. Images were captured with a Zeiss 780 multiphoton confocal microscope.

Detection of western blots was carried out using HRP-conjugated secondary antibodies and chemiluminescence substrates (GE Healthcare). Membranes were stripped and reblotted for β-actin, β1-integrin or GAPDH as a loading control. For immunoprecipitation, cells were lysed with 20 mM Tris-HCl, (pH 7.6), 167 mM NaCl, 1% NP-40 and protease inhibitor cocktail (Sigma-Aldrich) at 4°C for 30 min. The lysates were then centrifuged for 20 min at 12,000 g at 4°C. Supernatants were collected and incubated with 2-10 μg of antibody with rotation at 4°C overnight. Protein G PLUS-Agarose beads (Santa Cruz) were added to the mixture and rotated for 1 h at 4°C. Beads were then washed five times with ice-cold PBS. The bound proteins were dissociated by incubating with SDS-PAGE loading buffer at 50°C for 10 min and then 100°C for 10 min, and subsequently subjected to western blot analysis.

For craniofacial phenotypes, injected snail2-eGFP embryos were allowed to develop to stage ∼46 and scored for defects in head cartilage structures. For eye phenotypes, injected wild-type embryos were allowed to develop to stage ∼38 and scored for defects in eye structures. Embryos were scored blinded as individuals selected randomly from the different treatment groups, by comparing the injected side with the uninjected side of the same embryos. The percentages of normal and reduced phenotypes were calculated for injected embryos obtained from multiple independent experiments, and Chi-squared tests were performed to compare the phenotypes in different treatment groups. Images of head cartilages (eGFP) or eyes (bright field) were taken with a Zeiss Axiozoom.v16 epifluorescence microscope. See also Wei et al. (2010b) and Fig. S1D for more examples and explanation of cartilage and eye phenotypes.

We thank Dr Kris Vleminckx for providing the transgenic Wnt reporter frogs, and Dr Carole Labonne for the FLAG-ubiquitin construct.