Phosphorylation of Sox9 is required for neural crest delamination and is regulated downstream of BMP and canonical Wnt signaling

Abstract

Coordination of neural crest cell (NCC) induction and delamination is orchestrated by several transcription factors. Among these, Sry-related HMG box-9 (Sox9) and Snail2 have been implicated in both the induction of NCC identity and, together with phoshorylation, NCC delamination. How phosphorylation effects this function has not been clear. Here we show, in the developing chick neural tube, that phosphorylation of Sox9 on S64 and S181 facilitates its SUMOylation, and the phosphorylated forms of Sox9 are essential for trunk neural crest delamination. Both phosphorylation and to a lesser extent SUMOylation, of Sox9 are required to cooperate with Snail2 to promote delamination. Moreover, bone morphogenetic protein and canonical Wnt signaling induce phosphorylation of Sox9, thereby connecting extracellular signals with the delamination of NCCs. Together the data suggest a model in which extracellular signals initiate phosphorylation of Sox9 and its cooperation with Snail2 to induce NCC delamination.

Neural crest cells (NCCs) belong to a transient population of embryonic multipotent stem cells that are induced at the interface between the neuroectoderm and the prospective epidermis. These cells delaminate from the epithelium and adopt a characteristic migratory pathway into the periphery, where they contribute to several tissues, notably cranial facial structures and peripheral nervous system (1). During delamination, NCCs undergo an epithelial–mesenchyme transition (EMT), characterized by loss of cell–cell contacts and polarity, as well as acquisition of migratory capabilities (2). Successful EMT and delamination are essential steps for the subsequent migration and differentiation of NCCs. Consequently, mutations in genes involved in NCCs EMT are often associated with congenital defects (3).

Previous studies showed that Sry-related HMG box-9 (Sox9), a member of the SoxE family of transcription factors, plays several roles in NC formation, survival, EMT, and differentiation (4–6). In mice harboring a null mutation in Sox9, NCCs were still generated but died by apoptosis before or shortly after NC delamination, suggesting that Sox9 is required for NC survival (4). In chick, misexpression of Sox9 in the neural tube was sufficient to induce NC identity at the expense of neural fate but only inefficiently induced delamination and EMT (5). By contrast, coexpression of Sox9 with the zinc finger–type transcription factor, Snail2, promoted systematic ectopic EMT and NC delamination along the dorsal–ventral extent of the neural tube (4). In addition, in vitro studies showed that Sox9 physically interacts with Snail2 (6). Altogether, these data suggest that the partnership between Sox9 and Snail2 may underlie the ability of Sox9 to induce NC delamination and EMT. However, how the combination of Sox9 and Snail2 achieves this is unclear.

The ability of Sox9 to initiate NC delamination appears to be regulated by protein kinase A (PKA)-mediated phosphorylation at two serine residues (S64 and S181) (7). Mutations of these two sites into alanine abolished the ability of Sox9 to initiate NC delamination in cranial neural explants, indicating that Sox9 phosphorylation is essential for NC delamination (6). In addition, previous studies showed that SoxE factors (Sox9 and Sox10) can be SUMOylated in early Xenopus embryos, and this modification has distinct effects on the ability of SoxE factors to induce and inhibit NC precursor or otic placode formation, respectively (8). SUMOylation may, therefore, play a regulatory role conferring a context-dependent function, perhaps by determining the choice of partner interactions. Indeed, a recent report showed that the small ubiquitin-like modifier (SUMO) moiety converts SoxE protein from a transactivator to a repressor that recruits the corepressor, Groucho 4 (Grg4) (9). Collectively, these findings suggest a critical role for both phosphorylation and SUMOylation in modulating Sox9 functions during NC development. Moreover, previous evidence demonstrated the involvement of bone morphogenetic protein (BMP) and canonical Wnt signaling in controlling the onset of trunk NC delamination (10–12), and PKA activity is required for the induction of EMT by BMP4 and Sox9 (6). However, these studies did not address the epistatic relationship between phosphorylation and SUMOylation of Sox9, nor whether they are subjected to the regulation of BMP and/or canonical Wnt signaling. Moreover, how posttranslational modifications confer the ability of Sox9 to induce NC formation and delamination is unclear.

In this study, we establish that, in the developing chick neural tube, Sox9 is phosphorylated on S64 and S181, and this facilitates its SUMOylation. Although ectopic Sox9 protein lacking posttranslational modifications is able to initiate NC induction, nonphosphorylatable versions of Sox9 fail to initiate efficient trunk NC delamination. Both phosphorylation, and to a lesser extent SUMOylation, of Sox9 results in a physical interaction between Snail2 and Sox9, and Sox9 phosphorylation is necessary to cooperate with Snail2 to trigger NC delamination. Moreover, BMP and canonical Wnt signaling promote phosphorylation of Sox9, thereby connecting extracellular signals with the delamination of NCC cells. Together these data indicate that extracellular signals initiate NCC delamination by inducing the phosphorylation of Sox9 and its cooperation with Snail2.

Results

Sox9 Protein Can Be SUMOylated and Phosphorylated In Vivo.

Although it is known that amino acid residues for phosphorylation and SUMOylation of Sox9 are conserved between mouse, Xenopus, and chicken (Fig. 1A) (13), whether Sox9 is phosphorylated and/or SUMOylated in chick NCCs remained to be determined. To address this, we first examined the expression of phosphorylated Sox9 protein by performing immunofluoresence on transverse sections of a Hamburger and Hamilton (HH) stage 12 chicken embryo with an antibody specifically recognizing phosphorylated Sox9 at the S181 residue (anti-Sox9p^S181). Expression of phosphorylated Sox9 was not observed in the dorsal neural folds but became detectable in Sox9 expressing premigratory and emigrating NCCs (Fig. 1B). These data suggested that endogenous Sox9 protein is phosphorylated in prospective and delaminating NCCs.

Fig. 1.

Detection of phosphorylated and SUMOylated forms of Sox9 in the developing chick neural tube. (A) Schematic diagram showing chick Sox9 protein with a high mobility group (HMG) DNA binding domain (red) and transactivating domain (TA) (black). SUMO conjugation sites (K61, K254, and K376) are labeled in blue and phosphorylation residues (S64 and S181) in green. Phosphorylation-dependent SUMOylation motif (PDSM) is indicated with a bracket. (B) Transverse sections of a HH12 chicken embryo along the anterior-posterior (A-P) axis showing the expression of phosphorylated Sox9 at serine 181 is detected in emigrating neural crest cells (white arrows) overlapping with parental Sox9 expression (yellow arrows). (Scale bar, 100 μm.) (C) Phosphorylated (Sox9-P) and SUMOylated forms of Sox9 protein are detected from Sox9 immunopreciptated (IP) protein lysates extracted from HH12-13 chick neural tubes. (D) Western blots showing detection of SUMOylated Sox9 protein in embryos transfected with WT-Sox9 and Sox9^K61R but not with Sox9^K376R and Sox9^{K61R,K254R,K376R}. (E) Ectopic expression of WT-Sox9 but not Sox9^S64A,S181A is detected by anti-Sox9p^S181. (Scale bars for A and E, 50 μm.)

To examine whether endogenous Sox9 can be SUMOylated in the trunk NCCs, we immunoprecipited Sox9 from the thoracic neural tube of HH stage 12–13 embryos and performed Western blots with anti-Sox9 or anti–SUMO-1. In addition to the native form of Sox9 protein at ∼68 kDa, we detected two slower migrating bands with molecular weights of ∼72 and ∼98 kDa, the latter being consistent with the addition of a SUMO moiety (Fig. 1C). These data confirm that endogenous Sox9 protein can be phosphorylated and SUMOylated in the developing chick neural tube.

Because there are three SUMOylation sites (K61, K254, and K376) in chick Sox9 protein, we sought to identify which was SUMOylated. To this end, we generated a Flag-tagged version of Sox9 expression plasmid with or without point mutations that convert the lysines in each SUMOylation site to arginine (K to R). Following in ovo electroporation for 24 h, each construct was immunoprecipitated (IP) with anti-Flag, and blots were probed for SUMO-1. We detected a SUMOylated form of Sox9 at ∼98 kDa in protein lysates from chicken embryos electroporated with Sox9 but not in embryos transfected with pCIG empty vector control, suggesting that ectopic Sox9 protein can be SUMOylated in vivo (Fig. 1D). Moreover, SUMOylation was observed in K61R but not in K376R and K61R/K254R/K376R (Fig. 1D), suggesting that K376 is the predominant SUMOylated residue in vivo, consistent with the SUMOylation site in Sox9 from Xenopus embryos (8).

Furthermore, electroporation of WT Sox9 (WT-Sox9) but not Sox9^S64A,S181A can be detected by anti-Sox9p^S181 (Fig. 1E), indicating that phosphorylation of ectopic Sox9 protein occurs in vivo. Collectively, both ectopic and endogenous Sox9 protein can be SUMOylated and phosphorylated in vivo, suggesting that components of SUMOylation and phosphorylation machinery present in the developing chick neural tube are sufficient to carry out the modification processes.

Phosphorylation of S64 and S181 Is Required for Sox9 SUMOylation.

The observation that Sox9 protein is SUMOylated and phosphorylated prompted us to test whether phosphorylation regulates Sox9 SUMOylation or vice versa. We found that WT-Sox9 and single phospho-mutant Sox9 were efficiently SUMOylated, but SUMOylation was completely abolished in Sox9^S64A,S181A and not restored even in the presence of a constitutive active protein kinase A (CA-PKA) (Fig. 2A). By contrast, the non-SUMOylatable isoform of Sox9 remained detectable with anti-Sox9^S181 (Fig. 2B). These data indicate that phosphorylation of either S64 or S181 is required for Sox9 SUMOylation, but phosphorylation is not dependent on Sox9 SUMOylation. Moreover, we also found that there is another serine residue (S381) downstream of K376 forming a composite phosphorylation-dependent SUMOylation motif (PDSM), IKTEQLSP (Fig. 1A). This motif has been shown to regulate transcription activities of many proteins through sequential phosphorylation and SUMOylation (14). To examine whether similar sequential regulation occurs for Sox9, we substituted S381 with alanine (Sox9-SA). Strikingly, SUMOylation still occurs in Sox9-SA mutant (Fig. 2C), suggesting that the phospho-SUMO switch does not occur in the PDSM of Sox9.

Fig. 2.

SUMOylation of Sox9 depends on phosphorylation. (A) SUMO conjugation occurs in WT-Sox9, Sox9^S64A, and Sox9^S181A but not in Sox9^S64A,S181A mutants. Addition of CA-PKA does not rescue SUMOylation of the Sox9^S64A,S181 mutant. (B) Ectopic expression of the non-SUMOylatable form of Sox9 in the developing chick neural tube is detected by anti-Sox9p^S181. (Scale bar, 100 μm.) (C) SUMOylated Sox9 protein is detected in embryos transfected with WT-Sox9 or Sox9-SA mutant. (D) SUMOylation of Sox9 is enhanced by CA-PKA and markedly diminished by DN-PKA without significantly altering the levels of phosphorylated form of Sox9 compared with WT-Sox9. (E) Western blots showing similar levels of SUMOylation in both WT-Sox9 and phosphomimetic mutant. (F) Reduced amount of UBC9 is observed in IP fraction from Sox9^S64A,S181A compared with the WT-Sox9. (G) V5-Snail2 and V5-UBC9 proteins are detected in the IP fraction of Sox9^S64D,S181D transfected embryos. Input represents 1% of IP.

We then asked whether PKA-mediated phosphorylation could facilitate Sox9 SUMOylation. We examined the phosphorylation and SUMOylation levels of WT-Sox9 in the presence of CA-PKA or a dominant negative form of PKA (DN-PKA). Sox9 SUMOylation was enhanced by CA-PKA and substantially diminished by DN-PKA without significantly altering the levels of Sox9 phosphorylation (Fig. 2D). These data suggest that, although PKA is not required for Sox9 phosphorylation, it does enhance SUMO conjugation, perhaps by regulating another protein in the SUMOylation machinery.

To investigate further whether phosphorylation of Sox9 is sufficient to promote Sox9 SUMOylation, we generated a phosphomimetic mutant by replacing serine with aspartate residues (Sox9^S64D,S181D) and assayed for SUMOylation. The levels of SUMO conjugation for Sox9^S64D,S181D was similar to that of WT-Sox9 (Fig. 2E).

An interaction of SoxE proteins with SUMO-conjugation enzyme UBC9 appears to be a crucial step for their SUMOylation (8), raising a possibility that the lack of SUMO conjugation in Sox9^S64A,S181A could be caused by the inability to interact with UBC9. To examine this, we coexpressed WT-Sox9 or Sox9^S64A,S181A with a expression vector encoding a chick full-length UBC9 tagged with V5 and tested their ability to interact by co-IP. As shown in Fig. 2F, Sox9^S64A,S181A appeared to interact with UBC9 less efficiently than WT-Sox9, suggesting that the lack of SUMOylation of Sox9^S64A,S181A might be caused by a reduced ability to associate with UBC9.

Phosphorylation of Sox9 Is Essential for Trunk NC Delamination.

To investigate whether phosphorylation or SUMOylation has a role in mediating Sox9 protein activity during trunk NC development, we examined the impact of misexpressing Sox9^S64A,S181A on NC induction and delamination. In ovo electroporation was performed in the caudal hemineural tube of HH stage 10–11 chicken embryos at the level of segmental plate mesoderm and analyzed 12 h posttransfection (hpt) for markers characteristic of NC identity before the onset of NC delamination. Similar to WT-Sox9, Sox9^S64A,S181A induced expression of migratory NC markers human natural killer-1 (HNK-1) and Sox10, whereas expression of Sox2 was decreased (n = 4/4; Fig. S1), suggesting that the ability of Sox9 protein to induce the switch from neural progenitor to NCCs is not affected by the mutation of the two phosphorylation residues. In contrast to WT-Sox9, the majority of cells expressing Sox9^S64A,S181A remained in the neuroepithelium (n = 7/7; Fig. 3A) without disrupting laminin expression on the basement membrane (Fig. 3B), and expression of FoxD3 in migrating NCCs on the transfected side was reduced (n = 4/4; Fig. 3 C and E). Importantly, misexpression of Sox9^S64A,S181A resulted in down-regulation of endogenous Sox9 expression in the premigratory NC domain but did not affect the ability of WT-Sox9 to induce HNK-1 expression and NC delamination, ruling out the dominant-negative impact of Sox9^S64A,S181A on emigration of NCCs (Fig. S2 A and B). These data suggest that phosphorylation of Sox9 is essential for trunk NC delamination. Because phosphorylation regulates Sox9 SUMOylation, we tested whether the non-SUMOylatable form of Sox9 induced NC identity and delamination. NC induction and delamination still occurred in cells expressing Sox9^{K61R,K254R,K376R}, albeit the number of emigrating NCCs was moderately reduced compared with WT-Sox9 (n = 5/5; Fig. S1, Fig. 3 A, B, and E). These data suggest that phosphorylation and to a lesser extent SUMOylation, of Sox9 contribute to the regulation of trunk NC delamination. To further substantiate this, we misexpressed Sox9^S64D,S181D in the trunk neuroepithelium and observed no dramatic alterations of its function in NC induction and delamination compared with WT-Sox9 (n = 3/3; Fig. 3 A, B, and E). Similar findings were obtained in neural explant cultures in which Sox9^S64A,S181A inhibited NC delamination (Fig. S3). Together, our gain-of-function studies indicate that both phosphorylation and to a lesser extent SUMOylation, confer the ability of Sox9 protein to initiate trunk NC delamination.

Fig. 3.

Sox9 phosphorylation is required for trunk neural crest delamination. (A) Dorsal view of the trunk neural tubes electroporated with the indicated constructs 24 hpt showing GFP-positive emigrating NCCs. (Scale bar, 50 µm.) (B) Transverse sections of electroporated embryos from A showing robust ectopic induction of HNK-1 in the transfected neural tube by each construct. WT-Sox9–, Sox9^S64D,S181D-, or Sox9^{K61R,K254R,K376R}-expressing cells are observed delaminating from the pial surface of the dorsal neural tube where laminin is lost (white arrowheads), whereas SOX9^S64A,S181A-transfected cells still remained in the neuroepithelium without disrupting basement membrane (yellow arrow). (Scale bar, 50 µm.) (C) FoxD3 expression remains unaltered in Sox9 electroporated embryo at 36 hpt, whereas reduced FoxD3 expression in emigrating neural crest cells is observed on the transfected side of the neural tube (Left) with Sox9^S64A,S181A compared with the untransfected side. (Scale bar, 100 μm.) (D) Electroporation of Sox9-MO into the right side of the embryo results in most of the transfected cells remain in the neural tube with a marked reduction of the amount of FoxD3-positive migrating NCCs compared with the untransfected side or with the Ctrl-MO–treated embryo where NC emigration is observed (white arrows). Coelectroporation of Sox9-MO plus WT-Sox9 restores NC emigration (white arrows) together with the amount of FoxD3-positive migrating NCCs, whereas coelectroporation with Sox9^S64A,S181A fails to rescue the loss-of-function phenotype with a drastic reduced amount of FoxD3-positive migrating NCCs. (E) The number of emigrating GFP+ve cells expressing HNK-1 from the transfected side of the neural tube with the indicated constructs. Each bar represents an average of emigrating GFP+ve cells expressing HNK-1 from three to five well-transfected embryos with each analyzing 10–15 sections. *P < 0.01; **P < 0.001; ***P < 0.05.

To gain insight into the spatial-temporal requirement of Sox9 activity for NC delamination, we ablated Sox9 function using a fluorescein-tagged Sox9 morpholino (Sox9-MO) electroporated into the caudal hemineural tube (right side) of HH stage 11 embryos, a time at which NCCs are specified but delamination is yet to be initiated. This approach avoids an early impact of Sox9 down-regulation on NC survival (4). At 24 hpt, Sox9 protein was diminished in the premigratory NC domain expressing Sox9-MO (n = 7/7) compared with the untransfected side or embryos transfected with control-MO (Ctrl-MO; n = 0/6; Fig. S4 A–C). Importantly, the depletion of Sox9 expression did not cause apoptosis, and expression of NC specifier gene, Snail2, was retained (Fig. S4 C and D), suggesting that NC survival and specification were unaffected. However, the majority of Sox9-MO expressing cells remained in the neural tube and did not delaminate, whereas NC emigration initiated in Ctrl-MO–treated embryos. In agreement with this, a reduced amount of FoxD3-positive migrating NCCs was observed in Sox9-MO–treated embryos (n = 7/8) compared with the untransfected side or the Ctrl-MO–treated embryos (n = 1/8; Fig. 3D). Similar lack of NC emigration was observed in Sox9-MO–treated neural tube explants compared with the Ctrl-MO–treated explants (Fig. S5A). These results are consistent with the involvement of Sox9 in NC delamination.

To test the specificity of the morpholino knockdown, we performed rescue experiments in which Sox9-MO was electroporated with a pCIG vector encoding full-length Sox9 or Sox9^S64A,S181A protein. After electroporation, WT-Sox9 or Sox9^S64A,S181A plus Sox9-MO was able to induce HNK-1 expression 24 hpt (Fig. S5B). In addition, embryos transfected with WT-Sox9 plus Sox9-MO restored migratory FoxD3-positive NCCs (n = 6/7; Fig. 3D). By contrast, cells expressing both Sox9^S64A,S181A and Sox9-MO failed to delaminate, and there was a marked reduction in FoxD3-positive migrating NCCs compared with the untransfected side (n = 7/7; Fig. 3D). Altogether, these results further confirm that phosphorylation of Sox9 is required for efficient trunk NC delamination.

Phosphorylation of Sox9 Is Necessary to Cooperate with Snail2 to Trigger NC Delamination.

We previously demonstrated that coexpression of Sox9 and Snail2 is sufficient to trigger NC delamination throughout the dorsal–ventral neural tube with EMT-like characteristics (4). To investigate whether phosphorylation and SUMOylation are essential for delamination induced by the combination of Sox9 and Snail2, we coexpressed Sox9^S64A,S181A with Snail2. The data revealed that most transfected cells remained in the neuroepithelium, and there was little if any disruption in the expression of the apical marker atypical protein kinase C (aPKC) or basement membrane–localized laminin (Figs. 4A and 3E; n = 5/5). By contrast, forced expression of WT-Sox9 or Sox9^S64D,S181D and Snail2 induced robust NC delamination with EMT features (n = 3/3). We also observed a decrease in the number of delaminating NCCs from the dorsal neural tube expressing Sox9^{K61R,K254R,K376R} and Snail2, but the degree of reduction was less dramatic than with Sox9^S64A,S181A (Figs. 4A and 3E; n = 4/4). These data provide evidence that phosphorylation of Sox9 is necessary for Sox9-Snail2–induced NC delamination, whereas SUMOylation plays a less significant role. To further substantiate this, we generated two tethered constructs that express WT-Sox9 or Sox9^S64A,S181A linked in frame to Snail2 as a fusion protein and examined their effect on NC delamination by in ovo electroporation. After 24 hpt, both fusion proteins are capable of inducing HNK-1 expression, but only Sox9-Snail2–expressing cells undergo ectopic delamination in the medial and ventral extent of the neural tube, downregulating expression of laminin and aPKC (Fig. 4C; n = 6/6). In contrast, the majority of cells expressing Sox9^S64A,S181A-Snail2 remain in the neuroepithelium as demonstrated by the intact laminin and aPKC expression (Fig. 4C; n = 6/7). Taken together, these results further demonstrate that phosphorylation of Sox9 is essential for cooperating with Snail2 to trigger NC delamination.

Fig. 4.

Sox9 phosphorylation is required for cooperating with Snail2 to induce an EMT. (A) Embryos transfected with Sox9/Snail2; SOX9^S64D,S181D/Snail2; SOX9^{K61R,K254R,K376R}/Snail2 result in NC delamination in the intermediate and ventral extent of the neural tube where expression of basal laminin and apical PKC are lost (white arrows). However, most of the Sox9^S64A,S181A/Snail2-transfected cells remain in the neuroepithelium without disrupting basal laminin and apical PKC expression. (B) The amount of Snail2 protein presence in the IP fraction of Sox9^{K61R,K254R,K376R} transfected embryos is less than that of WT-Sox9. No Snail2 protein is detected in the IP fraction of Sox9^S64A,S181A-transfected embryos. (C) Schematic of WT-Sox9 or Sox9^S64A,S181A-Snail2 tethered construct. Horizontal white line in the Insets indicate the plane of sectioning of the transfected neural tube with WT-Sox9-Snail2 or Sox9^S64A,S181A-Snail2. Both tethered constructs can induce ectopic HNK-1 expression. Cells expressing Sox9-Snail2 emigrate from the dorsal, medial, and ventral extent of the neural tube (yellow arrows) with loss of basal laminin and aPKC expression. In contrast, the majority of Sox9^S64A,S181A-Snail2-expressing cells remain in the neuroepithelium without disrupting laminin and aPKC expression. (Scale bars, 100 µm.)

Phosphorylation Is Required for Sox9 to Interact with Snail2.

Previous in vitro co-IP analysis revealed a physical interaction of Sox9 and Snail2 proteins (6). To test whether this interaction was promoted by phosphorylation and/or SUMOylation of Sox9, we performed co-IP of protein extracts from chicken embryos coexpressing Snail2 with WT-Sox9, nonphosphorylatable or non-SUMOylatable Sox9 mutants. Consistent with previous in vitro studies, IP of ectopic WT-Sox9 from embryos extracts showed an association with the Snail2 protein (Fig. 4B). A similar association was observed between Sox9^S64D,S181D and Snail2 (Fig. 2G). The interaction was abolished with Sox9^S64A,S181A and somewhat reduced with Sox9^{K61R,K254R,K376R} (Fig. 4B). These data suggest that phosphorylation is required for Sox9 to interact with Snail2.

BMP and Canonical Wnt Signaling Regulates Sox9 Phosphorylation and SUMOylation.

Given the essential role for BMP and canonical Wnt signaling to induce NC delamination (10, 11), we analyzed whether phosphorylation and SUMOylation of Sox9 are regulated by these pathways. Consistent with previous studies (10), we found that inhibition of BMP and canonical Wnt signaling by Smad6 and β-catenin–engrailed repressor (β-cat-EnR), respectively, had no effect on the ability of Sox9 to confer NC identity, but both reduced NC delamination (Fig. 5A). Strikingly, phosphorylation and SUMOylation of Sox9 was abolished in the presence of Smad6 or β-cat-EnR (Fig. 5 B and D), whereas activation of BMP or canonical Wnt signaling pathway using a constitutively active BMP receptor IA (CA-BMPR-IA) or TCF-VP16 increased the phosphorylation and SUMOylation of Sox9 (Fig. 5 C–E). These results support a model in which the transcriptional targets of BMP and canonical Wnt signaling are indirectly required for Sox9 phosphorylation and SUMO conjugation. To confirm that phosphorylation of Sox9 is required for BMP-induced NC delamination, we coexpressed Sox9^S64A,S181A with CA-BMPR-1A and examined the impact on NC delamination. Although misexpression of CA-BMPR-1A alone resulted in increased NC delamination from the dorsal neural tube, delamination is inhibited in the presence of Sox9^S64A,S181A (Fig. 5F). Collectively, these results provide evidence that BMP or its downstream canonical Wnt signaling induces Sox9 phosphorylation followed by SUMO conjugation to initiate NC delamination events.

Fig. 5.

BMP and canonical Wnt signaling regulates phosphorylation and SUMOylation of Sox9. (A) Sox9 expression was examined on electroporated embryos with the indicated constructs followed by transverse sections at the level indicated by horizontal black lines, and GFP immunofluorescence was performed to mark the transfected cells. (Inset) Region of transfection with β-cat-EnR in the dorsal neural fold where Sox9 transcripts are detected. Immunofluorescence on transverse sections of electroporated embryos showing expression of Sox9 and Laminin proteins remain intact. (Scale bar, 50 µm.) (B) WT-Sox9 is phosphorylated and SUMOylated, but both modifications are inhibited in the presence of Smad6. (C) Western blots showing an increased amount of phosphorylated and SUMOylated forms of Sox9 in the presence of CA-BMPR-1A compared with the WT-Sox9 alone. (D) Activation and inhibition of canonical Wnt signaling by TCF-VP16 and β-catenin-EnR promotes and abolishes Sox9 phosphorylation and SUMOylation, respectively. (E) Levels of phosphorylated and SUMOylated forms of Sox9 in the presence of TCF-VP16 or β-catenin-EnR by Western blot were quantified compared with the modified forms of WT-Sox9 alone. *P < 0.05; **P < 0.01. (F) Electroporation of CA-BMPR-1A promotes NC delamination from the dorsal neural tube. Coelectroporation of Sox9^S64A,S181A plus CA-BMPR-1A results in ectopic HNK-1 expression, but transfected cells remain in the neuroepithelium without disrupting laminin expression in the basement membrane (white arrow). (Scale bar, 100 µm.) (G) Schematic diagram showing the transcriptional outputs of BMP and/or Wnt signaling can induce Sox9 phosphorylation, which facilitates SUMOylation partly through interaction with UBC9 and both modifications cooperate with Snail2 to trigger NC delamination.

Discussion

The ability of Sox9 to associate with different partners has been demonstrated to confer its tissue-specific function in various developmental processes (5, 15–18). Previous studies suggested that posttranslational modifications of Sox9 protein can affect their choice of partner interactions (9, 19), although how these modifications are regulated is not known. In this report, we demonstrated that ectopic Sox9 protein is subjected to BMP- and canonical Wnt signaling–dependent phosphorylation and SUMOylation in the developing chick neural tube. Moreover SUMOylation is dependent on phosphorylation. The initial induction of NC identity by Sox9 was not affected in the absence of either or both modifications. Importantly, however, the nonphosphorylatable Sox9 isoform, and to a lesser extent a non-SUMOylatable form, did not interact with Snail2 and failed to initiate efficient NC delamination. Together these data indicate a molecular mechanism that explains how BMP and Wnt signaling controls the onset of NC delamination (Fig. 5G).

Detection of the phosphorylated form of Sox9 in vivo has been shown in the growth plate of endochondral bones (7). Here, we show that phosphorylated Sox9 is also observed in the prospective and delaminating NCCs in embryos. In addition, detection of SUMOylated Sox9 has been reported both in cell lines and in Xenopus embryos (8, 20, 21). These studies required addition of Ubc9 and/or SUMO ligase, Pias1, to promote SUMO conjugation of Sox9. Our data indicate that components of SUMOylation machinery present in the developing chick neural tube are sufficient to induce detectable SUMOylation of both endogenously and ectopically expressed Sox9 proteins. Consistent with this, SUMO1, Pias1, and Ubc9 are expressed in the developing chick neural tube (22). Our analysis suggests that SUMOylation of Sox9 protein is not entirely dependent on PKA-mediated phosphorylation of S64 and/or S181, because neither the phosphorylated or the SUMOylated form of Sox9 was abolished in the presence of DN-PKA. It is possible that other kinases such as Rho kinase and/or cGMP-dependent protein kinase II, previously implicated in Sox9 phosphorylation, are involved in the modification process (23, 24). In addition, our data also indicate that PKA may have a role in SUMO conjugation, the details of which remain to be elucidated, but our results suggest that the absence of SUMOylation on a nonphosphorylatable Sox9 might be partly explained by reduced interaction with UBC9, the key component for SUMO conjugation event. Whether the negatively charges introduced by phosphorylation on S64 and S181 stabilize an interaction with UBC9 and/or convert the lysine residues to a more efficient SUMO acceptor remains to be determined. Importantly, our data show that a phospho-mimicking Sox9 mutant in which the serines are replaced with negatively charged aspartate residues was able to interact with UBC9 and undergo SUMO conjugation as efficiently as the WT (Fig. 2G). It is interesting to note that Sox9 contains a phospho-SUMOyl switch that couples sequential phosphorylation and SUMOylation and is associated with transcriptional repression (25). However, our data showed that mutation of serine in the composite motif (ΨKxExxSP) did not alter the ability of Sox9 to be SUMOylated. Thus, this serine residue is either not subjected to phosphorylation in vivo or does not play a major role in regulating Sox9 SUMOylation and activity.

Previous reports revealed a differential effect of Sox9 phosphorylation on nuclear localization and transcriptional activity in chondrocytic and testicular cells, implying a context-dependent role of Sox9 phosphorylation (7, 26). In agreement with this, our results indicate that mutation of the two phosphorylation sites did not affect the level of expression or nuclear localization of Sox9 in the developing chick neural tube (Fig. S6). Consistent with this, unphosphorylated Sox9 retained its ability to induce markers characteristic of NCCs (Fig. S1). Our data contrast with previous studies that indicated a requirement for PKA activity for Sox9 to induce NC delamination in cranial explants. One possibility is that different kinases regulate delamination in cranial and trunk NCCs (6).

Importantly, morpholino knockdown of Sox9 function after NC specification indicated that Sox9 is required for NC delamination. Moreover, mutation of the key serines of Sox9 indicated that its phosphorylation was critical for the onset of NC delamination. Consistent with this, endogenous expression of phosphorylated Sox9 is predominantly localized in emigrating NCCs but not in cells before the initiation of delamination. Sox9 phosphorylation is required for its interaction with Snail2. Moreover, our data suggest that phosphorylation of Sox9 is necessary to cooperate with Snail2 to promote NC delamination, and a physical interaction between the two proteins was not sufficient to overcome the requirement for Sox9 phosphorylation.

Similarly, the functional consequences of SUMOylation of Sox9 depend on the specific cellular and promoter context (8, 20, 21). Recent studies in Xenopus embryos showed that the SUMO moiety recruits a corepressor, Grg4, to change the transactivator SoxE into repressive function that confers its ability to inhibit NC formation (8, 9). However, our data suggest that NC induction occurs regardless of the SUMOylation status of Sox9. The reason for the difference between Xenopus and chick is unclear at present. One possibility could be that there are species-specific differences in the contribution of the SUMOylated form of Sox9 to NC induction. The SUMO moiety on Sox9 protein might play a role in cooperating with Snail2 or other partner factors essential for NC delamination and EMT. Consistent with this, a recent report showed that coexpression of Sox9 and Snail2 promotes metastasis-seeding abilities of human breast cancer cells (27). Whether phosphorylation and SUMOylation are involved in this process remain to be elucidated. Nevertheless, these data suggest that Sox9 and Snail2 may play a developmentally conserved role in regulating cell motility. How this cooperation initiates NC motility and tumor metastasis requires further identification and functional characterization of the downstream target genes induced by Sox9 and Snail2 together.

Several studies have revealed a role for dorsal neural tube–derived BMP–canonical Wnt signaling cascade in controlling the onset of NC delamination (10–12). Inhibition of these pathways resulted in reduced NC delamination but had no effect on the protein expression of Sox9 and Snail2 conferring NC identity (10), suggesting that these two events (NC specification and delamination) are independently controlled. Our data suggest that transcriptional targets of BMP and canonical Wnt signaling are indirectly required to promote phosphorylation and thus SUMOylation of Sox9.

In summary, our study provides insight into the mechanism by which BMP and canonical Wnt signaling regulate phosphorylation and SUMOylation of Sox9 to cooperate with Snail2, thereby determining the onset of NC delamination.

Materials and Methods

Fertilized chick eggs were obtained from Jinan Poultry Co. (Tin Hang Technology) and incubated at 38 °C in a humidified incubator (Brinsea Octagon 250 incubator). Embryos were staged according to Hamburger and Hamilton (28). In ovo electroporation was carried out as described (5). Detailed protocols regarding the generation of Sox9 phospho-, SUMO-mutant, V5-UBC9, and V5-Snail2 expression vectors, tethered constructs, in ovo electroporation, frozen sectioning, in situ hybridization, immunofluorescence, Western blotting, and coimmunoprecipitation can be found in SI Materials and Methods.