Inhibition of neural crest formation by Kctd15 involves regulation of transcription factor AP-2

Abstract

The neural crest develops in vertebrate embryos within a discrete domain at the neural plate boundary and eventually gives rise to a migrating population of cells that differentiate into a multitude of derivatives. We have shown that the broad-complex, tramtrack and bric a brac (BTB) domain-containing factor potassium channel tetramerization domain containing 15 (Kctd15) inhibits neural crest formation, and we proposed that its function is to delimit the neural crest domain. Here we report that Kctd15 is a highly effective inhibitor of transcription factor activating enhancer binding protein 2 (AP-2) in zebrafish embryos and in human cells; AP-2 is known to be critical for several steps of neural crest development. Kctd15 interacts with AP-2α but does not interfere with its nuclear localization or binding to cognate sites in the genome. Kctd15 binds specifically to the activation domain of AP-2α and efficiently inhibits transcriptional activation by a hybrid protein composed of the regulatory protein Gal4 DNA binding and AP-2α activation domains. Mutation of one proline residue in the activation domain to an alanine (P59A) yields a protein that is highly active but largely insensitive to Kctd15. These results indicate that Kctd15 acts in the embryo at least in part by specifically binding to the activation domain of AP-2α, thereby blocking the function of this critical factor in the neural crest induction hierarchy.

The neural crest (NC) is a uniquely vertebrate migratory cell population that gives rise to multiple derivatives, including craniofacial skeleton, peripheral nervous system, and melanocytes, whereas the neural plate border is a broad competence domain that contains progenitors for the NC and placodes (1, 2). When exposed to appropriate stimuli, progenitors acquire NC fate and express multiple factors that activate downstream genes (3–8), and several signaling cascades, including bone morphogenetic protein (BMP), wingless-type MMTV integration site family (Wnt), FGF, retinoic acid, and Notch are vital in NC induction (9–12). Deficits in NC development are responsible for many birth defects (13–15), broadening the interest in this system.

An important regulatory component in NC development is transcription factor AP-2, a target of Wnt signaling (16–21). The AP-2 family contains five members (22), among which AP-2α, -β, and -γ act in NC formation. After dimerization, AP-2 binds to DNA at sites with the consensus sequence 5′-GCCNNNGGC-3′, affecting genes in a broad range of biological processes (22–27). AP-2 can be divided into two broad regions: the N-terminal part containing the transactivation domain (AD) and the C-terminal region harboring the DNA-binding domain (DBD) that also mediates dimerization (22, 24). In zebrafish and Xenopus embryos, AP-2 is expressed in the NC anlage and has a role in NC formation (20, 23, 28, 29). In zebrafish AP-2α is encoded by tfap2a, and lockjaw and montblanc mutations in this gene exhibit defects in NC derivatives (28, 30), but induction of the NC is not abolished owing to partial functional redundancy between AP-2α and AP-2γ (30). In contrast, knockdown of AP-2α blocks NC development in Xenopus (20, 29).

We have shown that the BTB domain-containing factor Kctd15 regulates NC specification (31). Zebrafish kctd15a and -b are expressed at the neural plate border immediately lateral to the forming NC domain. We proposed that the inhibitory function of Kctd15 on NC formation delimits the NC domain, preventing its lateral expansion into the placodal domain (31). Although Kctd15 inhibits the canonical Wnt pathway, a key signal in NC induction, the level of Wnt inhibition seemed insufficient to explain its effectiveness in blocking NC formation, and thus we pursued the lead that the related protein KCTD1 can interact with AP-2 in human cells (32, 33). Here, we demonstrate that Kctd15 interacts with AP-2α and is highly effective in repressing the transcriptional activity of three AP-2 family members. Kctd15 binds to the activation domain of AP-2α, preventing its function in a reconstituted model system. We propose that inhibition of AP-2 is a key mechanism in the role of Kctd15 in regulating NC formation during embryonic development.

Results

The expression of several genes required during NC formation is suppressed in kctd15 mRNA injected zebrafish embryos (31). We pursued the relationship between AP-2 and Kctd15 as Tfap2 family genes are expressed in the NC and epidermis in the embryo and are required for NC formation (16–21, 28, 30). To compare the kctd15 and tfap2a expression domains we related both to the well-defined NC marker foxD3. The expression domains of kctd15 and tfap2a extended laterally beyond the foxD3 domain, but the overlap region was different: tfap2a overlapped foxD3 throughout its domain, whereas kctd15 overlapped just barely at the outer edge (Fig. 1A). This pattern is consistent with our model in that Kctd15 is in a position to inhibit AP-2 function in the region just lateral to the NC domain. Consistently, overexpression of kctd15 inhibits expression of tfap2a (Fig. S1A), just as it inhibits the expression of other transcription factors in this region (31), emphasizing the requirement for balanced expression between these two factors in the neural plate boundary.

Fig. 1.

Kctd15 and tfap2 expression overlaps in zebrafish embryos, and the protein products interact physically. (A) Expression domains of tfap2a and kctd15 (both blue) are compared with foxD3 (red). Anterior to the left, midline to bottom. (B) KCTD15 and AP-2α interact in mammalian cells. KCTD15 tagged with FOS and TFAP2a were cotransfected into HEK293T cells. Immunoblotting (IB) was carried out with Flag, AP-2α, and tubulin antibodies. (C) Co-IP of endogenous AP-2α and Kctd15 in one-somite-stage embryo extracts. Lanes are as follows: 1: AP-2α marker from co-IP of zAP-2α and Kctd15 expressed in HEK293T cells; 2: beads plus embryo extract, used as control; 3: beads plus embryo extract and anti-Kctd15 antibody; 4: beads plus buffer; 5: empty; 6: input, embryo lysate. Lanes 2 and 3 contain extract from 100 embryos, whereas lane 6 could be loaded with 3 embryos only because of high protein levels; therefore no band is seen. Background bands indicated by asterisks.

Physical Interaction Between Kctd15 and AP-2α.

To test for physical interaction between Kctd15a and AP-2α we overexpressed human untagged AP-2α and human Flag-ONE-Strep (FOS)-tagged KCTD15 in HEK293T cells. AP-2α bound to KCTD15, whereas no signal was recovered from empty vector-transfected cells (Fig. 1B). We tested for confounding effects of the tag on zebrafish Kctd15 behavior and found that FOS-tagged kctd15 mRNA is as effective as the untagged form in inhibiting NC formation in zebrafish embryos (Fig. S1B). We tested for presence of an AP-2α/Kctd15 complex in embryos at the stage of NC formation. Such a complex could be visualized (Fig. 1C) after increasing sensitivity by using a high level of embryo extract.

Kctd15 Is a Highly Effective Inhibitor of AP-2 Activity.

To test for functional interactions between Kctd15 and AP-2 we used a reporter construct, AP2-Luc, that contains three copies of the AP-2 consensus binding site controlling the expression of luciferase (24, 34, 35) (Fig. 2A). Zebrafish AP-2α strongly activated the reporter in HEK293T cells, whereas cotransfection with zKctd15 dramatically reduced activity (Fig. 2B). Strong inhibition was also seen after coinjection into zebrafish embryos (Fig. 2C). Titration of Kctd15 levels highlights the fact that its inhibition of AP-2α activity is highly efficient (Fig. 2B), and human KCTD15 is even more efficient (Fig. S2 A and B). To compare protein expression levels we used one antibody to detect both proteins by their epitope tags. Kctd15 was highly expressed, with the zebrafish version approximately eightfold and the human version approximately 50-fold higher than AP-2α (Fig. 2D and Fig. S2C). Fig. S2D shows that zKctd15-FOS is as effective as the untagged version (Fig. 2B) in inhibiting reporter activity, validating the use of tagged proteins to judge expression levels. On this basis we estimate that at equimolar levels of Kctd15 and AP-2α protein the reporter is inhibited by approximately 50%.

Fig. 2.

KCTD15 represses AP-2 function. (A) AP2-Luc reporter, containing three AP-2 consensus sites driving expression of luciferase (Luc) (24, 34). (B) The reporter was strongly stimulated by zebrafish AP-2α (200 ng) and dramatically inhibited by zKctd15 (levels in ng indicated). RLU, relative light units. (C) Reporter activity in zebrafish embryos was inhibited by Kctd15. (D) Cells were transfected with FOS-tagged zebrafish AP-2α and Kctd15, and lysates were blotted. Kctd15 is expressed more efficiently than AP-2α.

AP-2α occurs in at least three isoforms varying at the N terminus; our standard construct is based on isoform 1b (35). To test whether Kctd15 inhibition is isoform specific, we tested zebrafish isoform 1a in the reporter assay; isoform 1a was active and inhibited by Kctd15 to a similar extent as isoform 1b (Fig. S2D). There are five AP-2 family members encoded by separate genes in vertebrate organisms (22). In Xenopus NC formation AP-2α is the dominant factor, whereas in zebrafish both AP-2α and AP-2γ are involved (20, 23, 28–30). We found that reporter activation by AP-2α, -2β, and -2γ is inhibited effectively by Kctd15 (Fig. S3A). We further used animal cap explants to test Kctd15 function in an additional context. AP-2α can cooperate with other factors to induce NC markers in this system (20, 29), and Kctd15 was able to inhibit this induction (Fig. S3 B and C).

AP-2 Binding to DNA Is Not Inhibited by KCTD15.

Multiple steps in AP-2 expression and function might be inhibited by KCTD15. Using HEK293T cells, we asked first whether KCTD15 reduces expression of AP-2α, the rationale being that some KCTD proteins are cofactors of ubiquitin ligases, leading to degradation of their substrates (36, 37). Eliminating this possibility we found that the level of AP-2α protein in the cell was not affected by KCTD15 (Fig. S4A). Regulation of intracellular localization controls the activity of many transcription factors, but we found that AP-2α remains predominantly nuclear in the presence of KCTD15 (Fig. S4B). AP-2α functions as an obligatory dimer in DNA binding and transcriptional activation. Using differentially tagged constructs we showed that KCTD15 does not prevent AP-2α dimer formation (Fig. S4C).

To test the possible influence of KCTD15 on binding of AP-2 to its cognate sites in the DNA, we explored NC genes that are also expressed in HEK293T cells, such as MSX1, PAX3, FOXD3, and SNAI2, components of the NC gene regulatory network (4). We checked the sequence of 1 kb upstream of the transcription start site and chose one region for each gene that contains a canonical AP-2 binding site. To carry out ChIP experiments we transfected expression constructs for AP-2α alone or together with KCTD15 into the cells, because the endogenous content of these two proteins is low. The ChIP results show that the chromatin regions containing cognate sites did bind AP-2α, both in the absence and presence of KCTD15 (Fig. 3A). In some cases the presence of KCTD15 increased AP-2α binding, but such an effect was not seen in all regions tested. An additional ChIP experiment was carried out on cells transfected with AP2-Luc reporter (Fig. 2A) to provide abundant optimized binding sites. The reporter DNA showed binding to AP-2α that was unaffected by the presence of KCTD15 (Fig. 3B). Thus, it seems that KCTD15 does not prevent AP-2α binding to chromatin.

Fig. 3.

KCTD15 does not affect specific binding of AP-2 to DNA. (A) ChIP assay showing binding of AP-2α to promoter regions of PAX3, MSX1, SNAI2, and FOXD3 in HEK293T cells. A region of approximately 200 bp containing an AP-2 binding site was amplified. Cells were transfected with AP-2α alone or together with KCTD15, as shown. PAX3 exon 5 was used as negative control. (B) ChIP assay of AP2-Luc. The AP2-Luc reporter construct was overexpressed in HEK293T cells alone (L), together with AP-2α (L-A), KCTD15 (L-K), or all (L-A-K). KCTD15 did not inhibit AP-2α binding to its cognate sites. (C) Oligonucleotide binding assay. Two oligos, each containing an AP-2 binding site, were used. Extracts of cells overexpressing AP-2α (A), KCTD15 (K), or both (A-K) were incubated with WT 3′-biotinylated (WT*) double-stranded oligos. The oligo–protein complexes were isolated, proteins eluted, separated by gel electrophoresis, and blotted with anti-AP-2α. Competition was carried out with WT or mutant (mut) unlabeled oligos (absence of asterisk indicates no biotinylation).

Specificity of AP-2/DNA interactions was further tested in an oligonucleotide pull down assay. Biotinylated double-stranded oligonucleotides containing an AP-2 binding site were incubated with lysates of AP-2α–expressing cells, the complexes isolated, and the eluted proteins visualized by gel electrophoresis and immunoblotting. Using two different oligonucleotides we find that AP-2α is bound and is effectively competed by WT but not by binding-site mutant nonbiotinylated oligonucleotides (Fig. 3C). Again, binding of AP-2α was unaffected by the presence of KCTD15 in one case, whereas being slightly strengthened in the second case.

KCTD15 Inhibition Requires Proline 59-Dependent Binding to the AP-2α Activation Domain.

AP-2α contains an activation domain in its N-terminal region and a DNA binding and dimerization domain in its C-terminal region (22, 24, 38) (Fig. S5 A and B). Using constructs that express the two regions of zebrafish AP-2α we found that the N-terminal domain binds KCTD15 about as well as the full-length molecule, whereas no binding could be detected with the C-terminal domain (Fig. S5B). Previous studies have shown that the activation function of AP-2α is concentrated in an ∼60–70 amino acid region that is proline-rich, although it does not correspond to a typical proline-rich activation sequence (24, 38) (Fig. S5A). A fusion of the proline-rich domain (PRD) to the GAL4 DBD interacts with Kctd15 and further strongly activates an upstream activation sequence (UAS)-Luc reporter, as shown previously (24, 38) (Fig. 4 A–C). Most importantly in the present context, the reporter activity elicited by the PRD-Gal4 fusion is inhibited by Kctd15 as effectively as AP-2α–dependent activity (Fig. 4C; compare with Fig. 2 B and C). Thus, Kctd15 affects the activation function of AP-2α.

Fig. 4.

Critical role of P59 for Kctd15 interaction and inhibition of the PRD. (A) Mutants in the conserved PPxY motif within the AP-2α PRD, and two additional mutants in the adjoining region are shown. PRDs were fused to the Gal4-DBD and the FOS tag. (B) Kctd15 interaction depends on P59. WT and mutant PRD fusion constructs were expressed in HEK293T cells alone or with Kctd15, and complexes were pulled down using Strep-Tactin matrix. PRDs could bind Kctd15 except in the constructs containing the P59A mutation (P59A, 4A, and 6A). (C) PPxY-mutants activate the UAS-Luc reporter two- to fivefold higher than WT PRD-Gal4. PRD-Gal4 WT and P60A, Y62A, and 2A were dramatically inhibited by zebrafish Kctd15, whereas P59A and mutants containing this change were inhibited twofold or less. The constructs were expressed at similar levels (B and Fig. S6 C and D).

The AP-2α PRD contains a PPxY sequence (here PPPY), known as a protein motif interacting with the WW domain (39), preceded by a similar four-residue motif (PPYF) (Fig. 4A and Fig. S5A). Because this motif is highly conserved among vertebrate AP-2 proteins except AP-2δ (22), we systematically mutated proline and tyrosine residues in the PRD-Gal4 fusion construct and tested Kctd15 interaction and transcriptional activity (Fig. 4 A–C and Fig. S6A). The activation function of the AP-2α PRD was not abolished by mutating individual or all proline residues within the eight-residue sequence (Fig. 4C and Fig. S6A). Rather, the activity was increased several fold above WT. In contrast, Kctd15 binding and inhibition were affected differentially. The P59A mutant in which the first proline of the PPxY motif was mutated could not bind Kctd15 (Fig. 4B), and reporter activation by this mutant was largely unaffected by Kctd15; although up to a twofold inhibition was seen, we consider it insignificant compared with the 80- to 200-fold inhibition of other constructs (Fig. 4C). The other single-proline mutants tested were bound and inhibited by Kctd15 as well as the WT (Fig. 4 B and C and Fig. S6A). Likewise, mutation of the conserved tyrosine in the PPxY motif (Y62A) retained activity and Kctd15 sensitivity. Further, the 2A mutant in which both prolines of the degenerate PPYF motif were mutated (P55/56A) maintained activity and Kctd15 sensitivity. In contrast, the 4A and 6A mutants, which involved multiple residues including P59, lost their affinity for and sensitivity to Kctd15 (Fig. 4 B and C and Fig. S6A). To test whether these results depend on the artificial Gal4 fusion and use of UAS reporter we introduced the P59A mutation, alone or as part of the 4A mutant, into full-length AP-2α and tested these mutant proteins against the AP2-Luc reporter. As shown in Fig. S6B, the experiment confirmed the results based on the PRD-Gal4 fusion. Although WT AP-2α was Kctd15 sensitive, both AP-2α (P59A) and (4A) were active but Kctd15 insensitive. Fig. S6 C and D shows that the various mutant constructs were expressed at closely similar levels.

These results lead to the following conclusions. In agreement with and extension of the data of Wankhade et al. (38), who mutated two of the residues we tested, we conclude that the prolines and the tyrosine in the PPxY motif in AP-2α are not required for its activation function. However, P59 alone among the residues tested is critical for binding of the activation domain to Kctd15 and its sensitivity to inhibition. The tight correlation of binding with functional impairment implies that Kctd15 inhibits AP-2α by direct interaction with a region in the activation domain. Because P59 is not required for activity, we suggest that Kctd15 binding occludes a functionally critical region or leads to conformational changes that preclude transcriptional activation.

Discussion

Kctd15 Inhibits AP-2 Activity.

We originally focused our attention on Kctd15 because of its highly effective inhibition of NC formation in zebrafish and Xenopus embryos (31). The observed interference with Wnt activity provides a partial explanation of the effect, which is complemented by blockage of AP-2 function in the embryo, as described in the present report. Canonical Wnt activity is not only essential for NC induction, it also acts very high within the signaling hierarchy that initiates this process, and expression of tfap2a itself depends on a Wnt signal (29). Because Kctd15 inhibits tfap2a expression in the NC domain one might argue that Kctd15 affects an upstream event, except for the fact that the NC regulatory network is not a linear pathway due to multiple feedback loops. Thus, disruption of any one component can have both direct and indirect effects. This is particularly true of AP-2, which acts at different levels of the regulatory hierarchy (20). Thus we suggest that Kctd15 affects NC formation through both the Wnt and AP-2 modules.

Although our interest in Kctd15 and AP-2 derives from their role in NC formation, AP-2 has many biological roles, including the activation of genes such as metallothionein (24), p21WAF1/CIP1 (40), several epidermal genes in Xenopus (18), and others (22, 27). A large number of target genes have been identified by genome-wide approaches, including genes activated and suppressed by AP-2 (26, 41, 42). It remains to be studied whether Kctd15 inhibits gene regulation by AP-2 in general or only in a certain context. Although the expression patterns of AP-2 and Kctd15 overlap in some regions, they show considerable differences (23, 29, 31, 43, 44), and thus not all AP-2 functions are likely to be affected by Kctd15.

Physical and Functional Interactions Between Kctd15 and AP-2.

We find that AP-2α and Kctd15 can bind to each other when coexpressed in cultured cells. Although we have not excluded the possibility that the interaction is mediated by other proteins in the extract, binding in extracts of mammalian cells and zebrafish embryos is consistent with direct binding. Physical interaction in cell extracts between AP-2 and the related protein KCTD1 has been reported previously, as was the inhibition of reporter activation (32), but no biological role is known for KCTD1. The extent of inhibition seen with KCTD15 was substantially higher, but it is not clear whether this reflects inherent differences between the proteins or experimental conditions. The inhibitory effect was similar with all three AP-2 family members tested and was seen both in cultured human cells and in zebrafish embryos.

Kctd15 inhibition of AP-2 activity is not based on a reduction in nuclear abundance of the transcription factor or its ability to bind to cognate sites in the DNA. Instead, the effect depends on a specific binding of Kctd15 to the proline-rich activation domain (PRD) within the N-terminal half of AP-2α. This binding involves a specific proline residue that is part of a PPxY motif. Mutations in any or all prolines in this region resulted in active AP-2α molecules, as indicated previously (32, 38), but just a single proline mutant abolished the sensitivity to Kctd15 (Fig. 4 and Fig. S6). This striking effect emphasizes the specific structural requirements for Kctd15 binding and inhibition of AP-2. It is possible that Kctd15 binding affects the interaction of AP-2 with coactivators or corepressors. Although such factors have been described (45–47), it is likely that additional cofactors exist that interact with AP-2 and have a role in Kctd15 regulation.

The PPxY motif is not known to bind to the BTB domain, the characteristic feature of the Kctd family. Rather, PPxY is a widely distributed motif that binds to the WW domain and in this manner mediates multiple biologically important interactions (39). Because there is no WW domain in Kctd15 and no similarity between the BTB and WW domain structures (39, 48), the structural basis for the AP-2/Kctd15 interaction remains to be elucidated.

Multiple Functions in the Kctd Protein Family.

The Kctd family, with more than 20 members, is characterized by a BTB domain with similarity to potassium channels (36, 48), but Kctd proteins do not function as channels. BTB domain-containing proteins were classified into four subgroups with variable properties, yet the basic folding structure of BTB domains remains similar (48). The Kctd factors belong to the T1 subfamily of BTB domain-containing proteins that are believed to assemble as tetramers (48), and a tetrameric form has been suggested for KCTD11 (37). However, the only structure solved for a family member, KCTD5, shows a pentameric assembly, albeit with maintenance of the common BTB fold (49). Because AP-2 is known to dimerize, it is likely that the complex between AP-2 and Kctd15 represents a higher-order structure.

A biological role has been determined for only a few of Kctd family members. Kctd5 and Kctd11 are adaptors for Cullin E3 ubiquitin ligases, and Kctd11 affects the hedgehog pathway (36, 50). Copy number variation of a chromosomal region that includes KCTD13 is associated with neurological defects, and overexpression of KCTD13 in zebrafish induced microcephaly resembling the human condition, suggesting an involvement of KCTD13 in brain development (51). The related factors Kctd12.1 (also named Leftover), Kctd12.2 (Right on), and Kctd8 (Dexter) show asymmetric expression in the left and right habenula in zebrafish and have been valuable markers for studying brain asymmetry (52). As mentioned above, Kctd1 interacts with AP-2 (32, 33). Kctd15, which we have shown to inhibit NC development and affect Wnt signaling and AP-2 function, is regulated by FGF in Xenopus, and an enhancer trap line that mimics kctd15 expression in zebrafish has been described (43, 44). Further, the KCTD15 gene is associated with obesity in whole-genome association studies (53, 54), but no molecular basis for such a link has been reported. Factors important in adipogenesis are regulated by AP-2 (55-56), suggesting a possible molecular basis for the observed linkage of KCTD15 to obesity.

Materials and Methods

cDNAs and Plasmids.

ORFs from ESTs of human TFAP2a isoform 1b (National Center for Biotechnology Information reference sequence: NM_001032280) was obtained from Open Biosystems; zebrafish Tfap2a isoform 1b (Mammalian Gene Collection: 76915; IMAGE: 6524319) and Tfap2a isoform 1a (NM_176859.2, AF457191) amplified from cDNA of 24-h embryos were subcloned into pCS2⁺ vector. Human Kctd15 ORF (1–283 aa) was obtained from HEK293T cells by RT-PCR. Zebrafish Kctd15a has been previously described (31). Human KCTD15, zebrafish Kctd15, and Tfap2a (isoforms a and b and domains) were subcloned into pcDNA3.1/myc-His containing the C-terminal FOS tag, generously provided by J. Magadan [Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), National Institutes of Health (NIH), Bethesda, MD] (57). zKctd15-FOS was transferred into pCS2⁺ vector for producing RNA for embryo injection. Xenopus Tfap2a and -b were generously provided by T. Sargent (NICHD, NIH, Bethesda, MD) and AP2-Luc reporter by H. Hurst (Queen Mary University, London, UK). GFP-TFAP2a was generated in the pEGFP-C1 vector and TFAP2a-Flag in the pFLAG-CMV-6c vector (Sigma-Aldrich). Wild-type and mutant zebrafish AP-2α PRD fused to the Gal4-DBD, and full-length Tfap2 mutants were subcloned into pcDNA3.1/myc-His. 6×-Gal4-UAS-Luc reporter (58) was from Addgene (plasmid 33020).

Antibodies.

Anti-AP-2α 3B5 and anti AP-2α 5E4 were purchased from DSHB, and anti-zebrafish AP-2α antibody from LifeSpan Biosciences (LS-C87212/25863). Anti-FLAG clone M2 antibody was from Sigma, and anti-GFP-HRP from Miltenyi Biotec. Human KCTD15 was detected with anti-KCTD15 MaxPab mouse polyclonal antibody (H00079047-B01P; Abnova) or rabbit anti-human KCTD15 polyclonal antibody (LS C110024/25441; LifeSpan Biosciences). The latter was also used to detect zebrafish Kctd15. Anti-α-tubulin antibody (Calbiochem) was used as control. HRP-conjugated secondary antibodies were from Jackson Laboratories.

Cell Culture and Transfection.

HEK293T and HeLa cells were grown in Dulbecco’s modified Eagle’s medium supplemented 10% FBS. HeLa cells (American Type Culture Collection) were transiently transfected by using Lipofectamine 2000 (Invitrogen). HEK293T cells were transfected using Xtreme-Gene HP (Roche).

Pull Down, Immunoprecipitation, and Immunoblotting.

Cells or embryos were lysed in cold lysis buffer [0.5% Triton X-100, 50 mM Tris·HCl (pH 7.5), 150 mM NaCl, 5 mM EDTA] supplemented with complete Mini protease inhibitor mixture (Roche). General procedures were followed as previously described (57). NuPAGE Bis-Tris Gel system 4–12% (Invitrogen) was used for protein separation. SuperSignal West Pico or Dura Chemiluminescent Substrates from Thermo Scientific were used for detection of HRP-conjugated antibodies.

For immunoprecipitation of endogenous complex, one-somite embryos were dechoryonated and lysed in RIPA buffer containing protease inhibitors. Anti-Kctd15 antibody (2.5 μg; LifeSpan Biosciences) was bound to the beads (Protein G Sepharose 4 Fast Flow; GE Healthcare). Precleared lysates corresponding to 100 embryos were used in each sample and incubated overnight at 4 °C. General procedures were followed as indicated above. The complex was eluted using 2× sample buffer with 50 mM DTT for 5 min at 95 °C.

Subcellular Fractionation.

This was performed using NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Scientific) according to the manufacturer’s instructions.

mRNA Microinjection.

mRNAs were synthesized from linearized constructs using the mMESSAGE mMACHINE kit (Ambion) according to the manufacturer’s instruction. RNA quality was checked on formaldehyde gels. RNA or plasmids were injected into the yolk or the cell of one-cell-stage embryos. Embryos were collected at bud or one- to two-somite stage for in situ hybridization (ISH) or luciferase assay.

Whole-Mount ISH.

Zebrafish tfap2a ISH probe was prepared by subcloning the 3′ UTR into pCS2⁺ after BamHI/EcoRI digestion. The foxD3 probe has been previously described (31). Antisense riboprobes were generated according to the manufacturers’ instructions (Roche). Whole-mount ISH was performed as previously described (59).

ChIP Assay.

Human genome browsers and PromoSer predictor (http://biowulf.bu.edu/zlab/PromoSer) were used to identify putative AP-2-binding sites in PAX3, MSX1, FOXD3, and SNAI2 genes within 1 kb upstream of the transcription start site. Primer sets were designed to amplify a 180- to 200-bp fragment containing the consensus sites (Table S1). HEK293T cells were transfected with TFAP2a alone or together with KCTD15. ChIP was performed as previously described (60) with small modifications (61). Cross-linked chromatin was sheered to ∼300 bp using the Bioruptor Next Gen (Diagenode). Monoclonal anti-AP-2α 3B5 (DSHB) antibody and control IgG were used for immunoprecipitation. PCR was performed using AccuPower PCR PreMix (Bioneer Inc.).

Oligonucleotide-Binding Assay.

Two 40mer oligonucleotides containing AP-2 consensus sites (Tables S1) were 3′-biotynilated (WT*). WT and mutant unlabeled forms were used in competition experiments. Hybridized oligos (10 nmol) were mixed with precleared whole-cell extracts overexpressing AP-2α, KCTD15, or both. Binding buffer containing 10 mM Tris·HCl (pH 7.5), 50 mM NaCl, 1 mM DTT, 0.3 mg/mL BSA, 10% (vol/vol) glycerol, and 10 μg polydI/dC per sample (Sigma) was used. WT oligos at 10×, 50×, and 200×, and mutant oligos at 50× and 200× were used for competition. NeutrAvidin resin (Sigma) was used to pull down the complexes. General procedures were as previously described (62).

Luciferase Assay.

pGL3-AP2-Luc and pGL2-Gal4-UAS-Luc reporters were used to study AP-2 or PRD and mutant activity. HEK293T cells were transfected with 250 ng of AP2-Luc or 6×-UAS-Luc reporter, 25 ng of pRL-CMV (renilla), 200 ng of Tfap2a (full length or PRD and mutants), and 100 ng of Kctd15 unless specified. AB*/TL embryos were injected with 50 pg AP2-Luc reporter and 5 pg pRL-CMV DNAs, 50 pg Tfap2a mRNA alone or together with 80 pg Kctd15 mRNA. Injected embryos (one-somite stage) or HEK293T cells (24 h after transfection) were lysed in 1× Passive Buffer (Promega). Luciferase assay was performed using the Dual Luciferase Reporter Assay System from Promega. Each luciferase activity was measured at least three times.

Animal Cap Assay.

These were carried out as previously described (31, 63).