The MADS box transcription factor MEF2C regulates melanocyte development and is a direct transcriptional target and partner of SOX10

Neural crest cells are multipotent progenitor cells that originate from the dorsal neural tube at the border of the neural plate and the lateral epidermis to give rise to various tissues, including the skeletal elements of the face and head, dorsal root and sympathetic ganglia, the glial cells of the peripheral and enteric nervous systems, and melanocytes (Baker and Bronner-Fraser, 1997; Knecht and Bronner-Fraser, 2002; Le Douarin et al., 2004; Selleck et al., 1993; Trainor, 2005). Neural crest-derived melanocytes originate as non-pigmented precursors called melanoblasts, which migrate along characteristic pathways to various destinations, such as the dermis and epidermis, the inner ear and the choroids of the eye (Crane and Trainor, 2006; Dupin and Le Douarin, 2003; Sauka-Spengler and Bronner-Fraser, 2006; Steel and Barkway, 1989). The development of mature pigment-producing melanocytes from their neural crest precursors is a complex process involving melanoblast specification, proliferation, survival, migration and differentiation. Several transcription factors are known to be involved in melanocyte development, and mutations in the genes encoding these factors have been linked to pigment disorders and other neurocristopathies (Baxter et al., 2004; Mollaaghababa and Pavan, 2003; Pingault et al., 2010; Spritz et al., 2003).

SOX10 is an HMG box transcription factor that is a well-established regulator of neural crest development (Hong and Saint-Jeannet, 2005; Mollaaghababa and Pavan, 2003). Loss-of-function mutations in human SOX10 result in haploinsufficiency and are associated with Waardenburg-Hirschsprung disease, a neurocristopathy that involves hypopigmentation, deafness and aganglionic colon (Baxter et al., 2004; Inoue et al., 2004; Parisi and Kapur, 2000; Pingault et al., 1998; Pingault et al., 2010; Spritz et al., 2003). Similarly, Sox10 heterozygous mice also exhibit hypopigmentation, as well as distal bowel aganglionosis, and, therefore, these mice serve as an excellent animal model of Waardenburg-Hirschsprung disease (Herbarth et al., 1998; Lane and Liu, 1984; Southard-Smith et al., 1999; Southard-Smith et al., 1998). Sox10-null mice completely lack pigmentation and fail to develop a functional peripheral nervous system, suggesting a role for SOX10 in the specification of neural crest cells into melanocytes and glia in the peripheral and enteric nervous systems (Britsch et al., 2001; Herbarth et al., 1998; Potterf et al., 2001; Southard-Smith et al., 1998).

The MADS domain transcription factor MEF2C is important in the development of several lineages, including cardiac and skeletal muscle, hematopoietic cells, the craniofacial and axial skeletons, and the central nervous system (Black and Cripps, 2010; Potthoff and Olson, 2007). Inactivation of Mef2c in the germline of mice results in embryonic lethality by embryonic day (E)10 due to severe defects in cardiac and vascular development (Bi et al., 1999; Lin et al., 1998; Lin et al., 1997). In the neural crest, Mef2c expression can be detected as early as E8.5 in the mouse in the region adjacent to the neural folds (Edmondson et al., 1994; Verzi et al., 2007). Inactivation of Mef2c specifically in the neural crest using a conditional knockout approach in mice results in lethality at birth due to airway obstruction and defective craniofacial development (Verzi et al., 2007).

In the present study, we have identified a novel role for MEF2C in neural crest-derived melanocyte development. Inactivation of Mef2c in the neural crest using Wnt1-Cre transgenic mice results in reduced expression of melanocyte genes during development and a significant reduction in the number of melanocytes at birth. We have also identified a highly conserved transcriptional enhancer from the Mef2c locus that directs expression to the neural crest and its derivatives, including the craniofacial mesenchyme, the peripheral and enteric nervous systems, and melanocytes. We demonstrate that Mef2c is a direct transcriptional target of SOX10 in developing melanocytes and peripheral nervous system via this evolutionarily conserved transcriptional enhancer. Finally, we show that MEF2C physically interacts with SOX10 and, together, these proteins cooperatively activate transcription. Thus, we propose that Mef2c has a role in melanocyte development as a transcriptional target and partner for SOX10.

The 7039-bp Mef2c-F1 enhancer fragment was amplified by PCR from the mouse Mef2c locus using the primers F1-F: 5′-AGTGGGAAGCATAAGGCCCGGGAACTCTGAT-3′ and F1-R: 5′-ATGGTACCGTGTATGGTGGTCCCGGGAATGT-3′. The resulting PCR product was digested and cloned into the XmaI sites in plasmid hsp68-lacZ (Kothary et al., 1989) to create plasmid Mef2c-F1-lacZ. The 300-bp conserved region of Mef2c-F1 was generated as an XmaI-HindIII subfragment and was cloned into plasmid hsp68-lacZ to create plasmid Mef2c-F1[3-3.3]-lacZ for transgenic analyses or into plasmid pTK-β-gal to create plasmid Mef2c[3-3.3]-F1-TK-lacZ for transfection analyses. Mutations of cis-regulatory elements within Mef2c-F1 were generated using the gene SOEing method (Horton, 1997) and were confirmed by sequencing on both strands.

Transgenic mice were generated by oocyte microinjection as described previously (Dodou et al., 2003; Hogan et al., 1994). Sox10^Dom mice were purchased from the Jackson Laboratories. Wnt1-Cre, Mef2c^+/− and Mef2c^flox/flox mice have been described previously (Danielian et al., 1998; Lin et al., 1997; Vong et al., 2005). All experiments with animals complied with federal and institutional guidelines and were reviewed and approved by the University of California, San Francisco Institutional Animal Care and Use Committee.

β-galactosidase (β-gal) expression in lacZ transgenic embryos or tissues was detected by X-gal staining, as described previously (Dodou et al., 2003). Whole mount in situ hybridization was performed according to standard methods using digoxigenin-labeled antisense probes as described previously (Rojas et al., 2005). Dct (Image Clone ID 30539879), Mitf (Image Clone ID. 40047440), Pmel17 (Image Clone ID 30541702) and Sox10 probes were all designed by cloning the full-length cDNA into pBlueScript. Immunofluorescence labeling of cryosections was performed as described previously (Rojas et al., 2009) using the following primary antibodies at 1:100 dilution in PBS with 3% BSA and 0.1% Triton X-100: anti-SOX10 (R&D AF2864); anti-MEF2C (C-17) (Santa Cruz, sc-13268); anti-DCT (alpha-PEP8).

EMSAs were performed as described previously (Dodou et al., 2003). The sequences of the control SOX10 and MEF2 binding sites have been described previously (Dodou et al., 2003; Peirano and Wegner, 2000). The sense-strand sequences of the Mef2c-F1 oligonucleotides used for EMSA were: Mef2c SOX-1, 5′-GAATGCACTGACTACAAAGTGCATCCTGAAG-3′; Mef2c mutSOX-1, 5′-GAATGCACTGACCCGCGGTTGCATCCTGAAG-3′; Mef2c SOX-2, 5′-GAATAGCTCTATAACAAAGTAACTACAGAGT-3′; Mef2c mutSOX-2, 5′-GAATAGCTCTATCCGCCGTTAACTACAGAGT-3′; Mef2c SOX-3, 5′-GGCCATTTAGCTCACAATGAAGGTCTGTGTT-3′; Mef2c mutSOX-3, 5′-GGCCATTTAGCTCGAGGGTAAGGTCTGTGTT-3′; Mef2c MEF2, 5′-GGGCTCCAACTATTTATAGAACTGAGTA-3′; and Mef2c mutMEF2, 5′-GGGCTCCAACTATCTGTAGAACTGAGTA-3′.

B16F10 mouse melanoma cells were transfected in nearly confluent 10-cm plates with 6 μg of pRK5-Sox10[Flag] using Lipofectamine LTX reagent (Invitrogen), according to the manufacturer's protocol. Twenty-four hours post-transfection, ChIP was performed using the ChIP Assay Kit (Upstate/Millipore) according to the manufacturer's instructions, using anti-FLAG antibody (clone M2, Sigma) and Protein G-PLUS agarose beads (Santa Cruz). Immunoprecipitated fragments and input samples were subjected to PCR using the following primers: Mef2c-F1-A, 5′-GTCCTGGAGTCTTGCACAG-3′ and Mef2c-F1-B, 5′-CAGCCCAAGCTTCCGTATGG-3′ to amplify the region of Mef2c-F1 containing the SOX binding sites; or primers VEGFR2-F1, 5′-ATCATGTGACAGCAAGACCG-3′ and VEGFR2-R1, 5′-TCTTGGTATGTTGGGTCACTC-3′, designed to amplify a region of the Vegfr2 promoter not expected to be bound by SOX10.

Transfections were performed in C3H10T1/2 cells in 6-cm plates using either Superfect (Qiagen) or Fugene 6 (Roche) according to the manufacturer's instructions. Each transfection contained 1 μg of reporter and 1 μg of expression plasmid. In transfections with two expression plasmids, the amount of each expression plasmid was adjusted to 0.5 μg each. In transfections without an expression construct, the parent expression plasmid was added to keep the total amount of DNA in each transfection constant at 2 μg. Cells were cultured for 48 hours after transfection, then harvested and assayed by using the Luminescent β-gal Detection System (Clontech). SOX10 and MEF2C-VP16 were each expressed from the pRK5 mammalian expression vector (BD Pharmingen).

Staining of 3,4-dihydroxy-l-phenylalanine (DOPA) and electron microscopy (EM) on skin samples isolated from neonatal pups from a region of the back located between the shoulders were performed as previously described (Nguyen and Wei, 2007). For EM analyses, DOPA-stained back skin was fixed in modified Karnovsky's fixative (2% paraformaldehyde, 2% glutaraldehyde and 0.06% CaCl₂ in 100 mM cacodylate buffer, pH 7.3). The tissue was then washed twice in 100 mM cacodylate buffer, fixed in 1.5% potassium ferrocyanide and 2% osmium tetraoxide for 2 hours, rinsed in water, dehydrated through a series of ethanol washes, and embedded in Epon resin overnight. Epon blocks were sectioned and mounted onto formvar-coated grids. Grids were stained in 10% uranyl acetate in 50% methanol and viewed on a Zeiss 10A electron microscope at 60 kV. Melanosome density per unit area of melanocyte cytoplasm was determined by measuring the cytoplasmic area with NIH ImageJ software and counting DOPA-stained melanosomes within the cytoplasmic area.

GST pull-down assays were performed as described previously (Chen et al., 2000). The GST-MEF2C expression plasmid was generated by inserting the entire coding sequence of MEF2C and an in-frame C-terminal FLAG epitope (Khiem et al., 2008) in-frame into the PGEX-2T expression vector (Pharmacia). SOX10 full length and SOX10 DNA binding domain truncation proteins were generated from pCITE-Sox10[FL] and pCITE-Sox10[DBD], respectively. pCITE-Sox10[DBD] contains a 235-amino acid N-terminal fragment of SOX10 that includes the DNA binding domain cloned in-frame into pCITE. The SOX10 full length and DBD proteins were synthesized and radiolabeled with [³⁵S]-methionine in vitro using the TNT Quick Coupled Transcription/Translation System (Promega), according to the manufacturer's directions.

Co-immunoprecipitation was performed in lysates from co-transfected B16F10 mouse melanoma cells. Nearly confluent 10-cm plates were co-transfected with 3 μg each of pRK5-MEF2C-FLAG (Khiem et al., 2008) and pRK5-Sox10[HA] using Lipofectamine LTX reagent according to the manufacturer's protocol. Cells were grown for 30 hours after transfection and then harvested in NP-40 lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 1% NP-40, pH 7.4) containing protease inhibitors for 30 minutes at 4°C with gentle agitation followed by centrifugation to remove insoluble material. Lysates were then incubated overnight at 4°C with 3 μg of mouse anti-MEF-2 antibody (sc-17785, Santa Cruz) or mouse IgG1 antibody (ab18447, Abcam) as an isotype control. Samples were then incubated for 2.5 hours at 4°C with 100 μl of 50% Protein G Sepharose bead in NP-40 lysis buffer. Bound beads were washed 4 times with NP-40 lysis buffer and eluted in SDS sample buffer containing 100 mM DTT and boiled for 4 minutes before analysis by SDS-PAGE. Co-immunoprecipitation was determined by western blot with rabbit anti-HA antibody (sc-805, Santa Cruz) at a 1:200 dilution.

Although Mef2c function is required in the neural crest (Verzi et al., 2007), its expression in that lineage is not well defined. This is likely to be due to the strong expression of Mef2c in developing cardiac, skeletal muscle and vascular lineages, making the relatively weaker, but essential, expression of Mef2c in the neural crest difficult to visualize by in situ hybridization or immunohistochemistry (Black and Cripps, 2010; Edmondson et al., 1994). It has been observed previously that the expression of Mef2c during mouse embryogenesis is controlled by several discrete modular enhancers that each direct expression to a subset of the complete endogenous expression domain of Mef2c (Black and Cripps, 2010). Therefore, we reasoned that isolation of Mef2c enhancers with activity restricted to the neural crest might be used to determine the detailed expression of Mef2c within the neural crest without obfuscation from stronger expression in other lineages proximal to or overlying neural crest-derived tissues. Towards that goal, we identified a highly conserved enhancer that exhibited transcriptional activity in the developing neural crest and its derivatives (Fig. 1).

Mef2c-F1 enhancer activity in the neural crest was detected as early as E8.5 in the neural fold mesenchyme (data not shown). Between E9.0 and E9.5, β-gal was strongly expressed in craniofacial mesenchyme in a pattern similar to endogenous Mef2c at this stage (Verzi et al., 2007) (data not shown). Mef2c-F1 enhancer activity in the craniofacial mesenchyme began to decline by E10.5, but activity became more pronounced in the developing cranial nerves (Fig. 1B). β-Gal activity was also evident in cranial and sympathetic ganglia at E10.5 (Fig. 1B). In addition, Mef2c-F1-lacZ expression was detected in the developing heart (Fig. 1B), but the bona fide regulation of Mef2c in the heart via this enhancer is not a focus of the present study. By E12, β-gal expression in the craniofacial mesenchyme was no longer observed. However, activity in the cranial nerves, sympathetic chain ganglia, peripheral nerves and nerve roots, and enteric ganglia was evident (Fig. 1C). At E14, Mef2c enhancer activity continued throughout the developing peripheral and enteric nervous systems (Fig. 1H). Sections through the developing gut at E16.5 revealed that β-gal expression directed by the Mef2c-F1-lacZ enhancer was within the myenteric plexus (Fig. 1I).

Mef2c-F1-lacZ expression was also evident in the developing melanocyte lineage, beginning at ~E10.5 (Fig. 1B) and continuing throughout embryonic development (Fig. 1J). At E12, robust β-gal activity was apparent in single cells in the peri-ocular region of the embryo (Fig. 1D). This pattern of enhancer activity was consistent with melanocyte expression of the transgene. In perinatal skin, transgene expression was observed in follicular melanocytes and in the melanocytes that occupy the dermal-epidermal border (Fig. 1J). To determine whether the endogenous Mef2c gene was expressed in melanocytes, we examined Mef2c expression by in situ hybridization. As noted above, Mef2c is expressed broadly in many tissues, including the vasculature, somites and nerves (Black and Cripps, 2010; Dodou et al., 2004; Edmondson et al., 1994), making endogenous expression in melanocytes difficult to observe. However, Mef2c expression was evident at E12.5 in the dorsal region of the embryo between the hind limbs in a pattern highly reminiscent of melanocytes (Fig. 1E). To confirm that expression of MEF2C was in melanocytes, we compared the expression of MEF2C protein and the melanocyte marker dopachrome tautomerase (DCT) in the dermis at E12.5 by immunofluorescence (Fig. 1F). Numerous cells co-expressed MEF2C, which was confined to the nucleus, and DCT, which was present in the cytoplasm, confirming MEF2C expression in melanocytes. We also compared expression of MEF2C with SOX10 expression in the dermis at E12.5 and found that all SOX10-expressing cells in the nascent skin also expressed MEF2C (Fig. 1G), although, as expected, many MEF2C-expressing cells did not express SOX10. These results indicate that SOX10 and MEF2C are co-expressed in vivo in a subset of cells in the dermis.

Because of the strong expression of Mef2c and robust activation of the Mef2c-F1 enhancer in melanocytes and in the peripheral and enteric nervous systems, we examined those lineages in detail in Mef2c neural crest knockout (NC KO) neonates to determine whether MEF2C might have requirements in those neural crest-derived lineages. Although Mef2c NC KO mice die at birth of craniofacial defects (Verzi et al., 2007), they exhibited no obvious defects in peripheral or enteric innervation at birth (data not shown). As an initial analysis of melanocyte development, we stained epidermis and dermis taken from the anterior region of the back from Mef2c NC KO and control neonates by incubation with DOPA, which detects pigmented cells, and examined the stained tissue by light microscopy (Fig. 2). We observed a significant reduction in the number of melanocytes in epidermis (Fig. 2, compare panels A and B) and dermis (Fig. 2, compare panels C and D) in Mef2c NC KO mice compared with littermate controls. Within the dermis, we observed a strong reduction in the number of both follicular and interfollicular melanocytes (Fig. 2C,D). Indeed, the number of DOPA-stained cells per field showed that the number of melanocytes in the epidermis was reduced by 87% in Mef2c NC KO neonates compared with littermate controls (Fig. 2E). These observations demonstrate that MEF2C is required for establishing or maintaining the proper number of melanocytes in vivo.

Analysis of neonatal skin by electron microscopy revealed that the few remaining individual melanocytes in the dermis of Mef2c NC KO mice had fewer melanosomes than did the melanocytes present in littermate controls (Fig. 2, compare panels F and G). Quantification indicated a 65% reduction in the number of electron-dense melanosomes in dermal melanocytes from Mef2c NC KO neonates compared with their littermate controls (Fig. 2H). These data suggest that, in addition to being required for the proper number of melanocytes in the skin, MEF2C function is also required in the neural crest or in neural crest-derived cells for melanocyte function or differentiation in vivo.

We also examined the expression of well-established markers of melanocyte differentiation and function in Mef2c NC KO and control embryos (Fig. 3). Interestingly, Mitf, Dct and Pmel17 all had reduced expression in Mef2c NC KO embryos compared with their littermate controls (Fig. 3). At E12.5, Pmel17 expression was reduced, including in the trunk region around the hind limbs (Fig. 3, compare panels A and B). Similarly, Mitf expression was reduced in the same region at E12.5 in Mef2c NC KO embryos (Fig. 3, compare panels C and D). Reduced expression of melanocyte markers was apparent in multiple regions of the embryo and over several days of development in Mef2c NC KO embryos. For example, the expression of the melanocyte differentiation marker Dct was clearly reduced in the region around the eyes in Mef2c NC KO embryos at E14.5 (Fig. 3, compare panels E and F). Quantification of Dct-labeled melanocytes in the interlimb region at E12.5 showed that Mef2c NC KO embryos had on average only 69% as many labeled cells as control embryos (356±98 labeled cells/embryo, n=4 embryos for Mef2c NC KO versus 514±133 labeled cells/embryo, n=4 embryos for control, P=0.0027 by Student's t-test).

To place Mef2c into a pathway for neural crest and melanocyte development, we next examined the Mef2c-F1 neural crest enhancer for essential cis-regulatory elements by generating a series of deletions within the 7-kb conserved fragment (Fig. 4A). These analyses identified a 300-bp region (3-3.3 kb) that was sufficient for Mef2c-F1 enhancer function in vivo and directed β-gal expression in the same pattern as the full length Mef2c-F1 fragment at E11.5 (Fig. 4, compare panels B and C). This 300-bp region was also required for enhancer function in the context of the full-length 7-kb Mef2c-F1 fragment, as deletion of this region completely abolished enhancer activity in transgenic embryos at E11.5 (Fig. 4D).

We examined the highly conserved, crucial 300-bp region from Mef2c-F1 for candidate cis-regulatory sequences that might be involved in neural crest expression. These analyses revealed three putative SOX binding sites and a single candidate MEF2 binding site (Fig. 4E). The presence of the conserved SOX elements in the Mef2c-F1 enhancer further suggested the possibility that a SOX transcription factor might be a direct upstream activator of Mef2c.

Previous studies have shown that Sox10-null embryos have a nearly complete lack of melanocyte marker expression as early as E11 (Kapur, 1999; Potterf et al., 2001; Southard-Smith et al., 1998). The observations that Mef2c and Sox10 are each required for proper melanocyte differentiation and share common downstream targets in melanocytes, including Dct, Mitf and Pmel17, and that the two factors are co-expressed in melanocytes in the dermis (Fig. 1G) is consistent with the idea that the two transcription factors might function in a common pathway in the neural crest and its derivatives. Based on the observation that the Mef2c NC KO pigmentation phenotype was similar, but substantially less severe, than the pigmentation phenotype in Sox10-null embryos, we hypothesized that Mef2c might function downstream of Sox10 during trunk neural crest development. Moreover, we found that the expression of Sox10 was unaffected in the peripheral nervous system in Mef2c NC KO embryos when compared with control embryos (data not shown), which is consistent with the notion that Sox10 might function upstream of Mef2c in the trunk neural crest and its derivatives.

To determine whether the putative SOX elements in the Mef2c-F1 enhancer might represent bona fide cis-acting elements, each was tested for SOX10 binding in vitro by EMSA (Fig. 5A). Indeed, SOX10 bound robustly and specifically to each of the putative SOX sites from the Mef2c-F1 neural crest enhancer (Fig. 5A, lanes 2, 8 and 14). We also examined the ability of SOX10 to bind to the Mef2c-F1 SOX binding sites in the endogenous Mef2c locus by ChIP in vivo in B16F10 mouse melanoma cells transfected with FLAG epitope-tagged SOX10 (Fig. 5B). Importantly, the DNA fragment encompassing the SOX sites in the endogenous Mef2c-F1 enhancer was precipitated and amplified by the anti-FLAG antibody (Fig. 5B, lane 3) but not by the isotype control IgG antibody (Fig. 5B, lane 2). Furthermore, precipitation with anti-FLAG antibody did not result in PCR amplification of a bona fide enhancer from the Vegfr2 (Kdr – Mouse Genome Informatics) gene (De Val et al., 2008), which was used as a non-specific enhancer target (Fig. 5B, lane 7). These results indicate that SOX10 binds to the SOX sites in the Mef2c-F1 enhancer in vitro and in vivo in a relevant cell line.

To determine whether the binding of SOX10 to the Mef2c-F1 enhancer was functionally relevant, we tested the ability of SOX10 to trans-activate the Mef2c-F1 enhancer (Fig. 5C). Co-transfection of a SOX10 expression plasmid with the 300-bp Mef2c reporter plasmid Mef2c[3-3.3]-F1-TK-lacZ resulted in robust activation of the reporter gene (Fig. 5C, lane 4). This activation was dependent upon the presence of the SOX binding sites, as mutation of these elements in the Mef2c enhancer eliminated SOX10-dependent activation (Fig. 5C, lane 6). Taken together with the EMSA results shown in Fig. 5A and the ChIP results shown in Fig. 5B, these trans-activation studies strongly support the notion that the SOX binding sites in the Mef2c-F1 enhancer represent bona fide SOX response elements and that Mef2c is a direct transcriptional target of SOX10.

We also examined the MEF2 site in the Mef2c-F1 neural crest enhancer for binding by MEF2C in EMSA (Fig. 6A). MEF2C bound robustly to the bona fide MEF2 element from the myogenin promoter (Dodou et al., 2003) and to the MEF2 site from Mef2c-F1 (Fig. 6A, lanes 2 and 8). The MEF2 element in the Mef2c-F1 enhancer was also responsive to MEF2-dependent activation in C3H10T1/2 cells (Fig. 6B). Co-transfection of a MEF2C-VP16 expression plasmid with the 300-bp Mef2c-F1 enhancer plasmid Mef2c[3-3.3]-F1-TK-lacZ resulted in significant trans-activation (Fig. 6B, lane 4). The activation of the Mef2c-F1 enhancer by MEF2C was dependent on the MEF2 site in the enhancer as mutation of the site completely abolished activation (Fig. 6B, lane 6). Together, the results of Fig. 6A,B supports the notion that the MEF2 element in the Mef2c-F1 enhancer represents a bona fide binding site, and suggest that Mef2c is subject to auto-regulation by itself or cross-regulation by other members of the MEF2 family in developing neural crest lineages.

Based on the observation that MEF2C and SOX10 each trans-activated the Mef2c-F1 enhancer (Fig. 5C and Fig. 6B), we tested whether the two could cooperatively activate the enhancer (Fig. 7A). Indeed, when SOX10 and MEF2C-VP16 expression plasmids were co-transfected with the 300-bp Mef2c-F1 reporter plasmid, the two transcription factors synergistically activated the enhancer (Fig. 7A, lane 7). The strong transcriptional synergy by MEF2C and SOX10 suggested that the two transcription factors might physically associate. We tested this possibility in GST pull-down assays in vitro and co-immunoprecipitation assays in vivo (Fig. 7B,C). In control reactions, neither SOX10 nor an N-terminal fragment of SOX10, which includes the HMG DNA-binding domain, interacted with GST alone (Fig. 7B, lanes 2 and 4). However, full length SOX10 and the HMG domain of SOX10 each specifically interacted with GST-MEF2C fusion protein (Fig. 7B, lanes 1 and 3). Interaction of GST-MEF2C with the N-terminal portion of SOX10 was more robust than the interaction with full length SOX10 protein (Fig. 7B, compare lanes 1 and 3), which is consistent with the interaction that has been observed previously between another SOX protein, SOX18, and MEF2C (Hosking et al., 2001). SOX10 and MEF2C also interacted in vivo (Fig. 7C). B16F10 mouse melanoma cells were transfected with MEF2C and HA-tagged SOX10 expression constructs and were harvested and subjected to immunoprecipitation with anti-MEF2C antibody. Following immunoprecipitation, cells were subjected to western blot analyses with anti-HA to detect co-precipitated SOX10. HA-SOX10 was specifically detected only in the presence of anti-MEF2C antibody (Fig. 7C, lane 4) and not by isotype-matched IgG (Fig. 7C, lane 3) or by beads alone (Fig. 7C, lane 2), indicating that SOX10 was specifically precipitated by MEF2C co-expression in B16F10 cells. Taken together, these results show that MEF2C and SOX10 cooperatively activate the Mef2c-F1 neural crest enhancer, and suggest that this transcriptional synergy might be dependent on a physical association between the two transcription factors.

To test the requirement for the MEF2 and SOX cis-regulatory elements in the Mef2c-F1 enhancer in vivo, the MEF2 site or all three SOX sites were mutated in the context of the full-length 7-kb Mef2c-F1 enhancer, which was then tested for activity in transgenic embryos (Fig. 8). Mutation of the MEF2 site in the enhancer had only a partial effect on the activity of the enhancer in the nascent neural crest at E8.5 (Fig. 8, compare panels A and B, arrowheads) but resulted in a nearly complete loss of transgene expression by E11.5 (Fig. 8, compare panels C and D). These observations suggest that the initial activation of Mef2c-F1 does not require the MEF2 site but that maintenance of Mef2c expression in the neural crest via the Mef2c-F1 enhancer requires auto- and cross-regulation by MEF2C and other MEF2 factors.

Mutation of the three SOX sites in the Mef2c-F1 enhancer also abolished expression in all neural crest-derived lineages at all stages of embryonic development, including at E11.5 (Fig. 8, compare panels C and E). Interestingly, the SOX sites were not required for Mef2c-F1 enhancer activity in the heart at E11.5 (Fig. 8E), demonstrating the specificity of those sites for neural crest-specific activity of the enhancer in vivo. To test specifically the role of SOX10 as a possible upstream activator of Mef2c via the Mef2c-F1 enhancer, we crossed mice harboring the wild-type Mef2c-F1-lacZ transgene into a heterozygous Sox10-mutant background (Sox10^Dom/+). Mef2c-F1 enhancer activity was significantly reduced in all neural crest lineages when one functional copy of Sox10 was missing (Fig. 8, compare panels C and F), further supporting a possible role for SOX10 specifically as an upstream regulator of Mef2c in the neural crest.

MEF2C is a well-known regulator of gene expression in skeletal and cardiac muscle, the central nervous and immune systems, and the axial and craniofacial skeletons (Black and Cripps, 2010; Potthoff and Olson, 2007). Here, we show that loss of Mef2c in the neural crest results in a profound deficit in pigmentation due to a reduction in the number of melanocytes in the skin and impairment of melanocyte differentiation and function. Interestingly, in other lineages in which MEF2C function is required, such as in cardiomyocytes and endothelial cells, initial lineage specification occurs, but the differentiated phenotype fails to stabilize and eventually development of the heart and vasculature collapses in the absence of MEF2C (Bi et al., 1999; Lin et al., 1998; Lin et al., 1997). The phenotype that we observed here is similar in that markers of the melanocyte lineage, including Mitf, Dct and Pmel17 are reduced but still clearly present in Mef2c NC KO fetuses (Fig. 3). However, by birth, pigmentation was much more severely impaired (Fig. 2). We considered the possibility that the observed dramatic loss of pigmentation might be caused by loss of melanocytes due to apoptosis or impaired proliferation between E13.5 and birth, but we observed no differences in TUNEL staining or in BrdU incorporation from E11.5 to E18.5 (data not shown). It is possible that proliferation or apoptosis were affected but below the level of detection of our assays. However, we favor the hypothesis that MEF2C function is required to maintain the differentiation phenotype in melanocytes. The observation that MEF2C function was required for appropriate melanosome number (Fig. 2F-H) supports this notion.

Feed-forward transcriptional regulation is a common theme in developmental gene regulation, in which this positive-acting mechanism helps to reinforce developmental decisions to differentiate (Amin et al., 2010; Penn et al., 2004; Sandmann et al., 2007; Tapscott, 2005; Verzi et al., 2007). MEF2 transcription factors are known to participate in this type of self-reinforcing transcriptional circuitry in other systems, in which they often function to potentiate the activity of other transcription factors in a feed-forward fashion (Black and Cripps, 2010). Here, we identified MEF2C as a direct transcriptional target of SOX proteins and demonstrate that it can function as a partner of SOX10. The function of SOX10 in different neural crest-derived lineages is spatially and temporally modulated by its interaction with other transcriptional factors, including PAX3, MITF, KROX-20 (EGR2 – Mouse Genome Informatics) and SP1, and the functional interaction of these transcriptional partners with SOX10 is often facilitated by direct physical association (Jiao et al., 2004; Kuhlbrodt et al., 1998; Lang and Epstein, 2003; Ludwig et al., 2004; Melnikova et al., 2000). Consistent with this cofactor-based role for SOX10 in neural crest development, we observed a direct physical association between SOX10 and MEF2C (Fig. 7). We also observed co-expression of MEF2C and SOX10 in the same cells during development (Fig. 1G), further supporting the notion that SOX10 might function as a bona fide regulator and cofactor of MEF2C. However, we cannot rule out the possibility that other SOX factors, such as SOX9, might also be involved in Mef2c regulation during neural crest development. Based on the direct regulation of Mef2c by the novel Mef2c-F1 enhancer identified here, the synergistic activation of transcription by MEF2C and SOX10, and the direct physical interaction and co-expression of the two factors, we propose a model in which SOX10 cooperates with MEF2 to activate Mef2c expression in melanocytes in a feed-forward transcriptional pathway.

Mutations in human MEF2C are strongly associated with neuronal disorders, including seizures, mental retardation and epilepsy (Le Meur et al., 2010; Novara et al., 2010; Nowakowska et al., 2010), demonstrating the important role for MEF2C in the development of neuronal lineages in humans. Mutations in human SOX10 often cause a severe form of Waardenburg syndrome (Type IV, also known as Waardenburg-Hirschsprung disease), which involves aganglionic colon and severe pigmentation defects (Parisi and Kapur, 2000; Pingault et al., 1998; Pingault et al., 2010). In addition to SOX10, mutations in genes encoding other transcription factors known to function as SOX10 cofactors or to function in a common pathway with SOX10, also cause various forms of Waardenburg syndromes, including PAX3, MITF and SNAI2 (Pingault et al., 2010). Interestingly, only 50% of human patients with classic symptoms of Waardenburg-Hirschsprung disease have mutations in known disease loci, and even in cases with known genetic causes, the severity or the symptoms are often highly variable (Parisi and Kapur, 2000; Pingault et al., 1998; Pingault et al., 2010; Southard-Smith et al., 1999). Distinct mutations in human SOX10 are associated with different severity of disease, depending on the effect of the various mutations on SOX10 protein function (Inoue et al., 2004). It has also been postulated that some of the variability observed in Waardenburg-Hirschsprung patients with SOX10 mutations is due to modifier genes, which are largely hypothetical, because few modifiers of SOX10 have been defined in humans (Parisi and Kapur, 2000; Pingault et al., 2010). In mice, however, several modifiers of Sox10 function have been identified, including genes encoding endothelin signaling components and the related SOX protein SOX8 (Maka et al., 2005; Stanchina et al., 2006). It is important to note that neither heterozygous Mef2c mice nor Mef2c neural crest conditional knockout mice exhibit aganglionosis of the bowel (Lin et al., 1997; Verzi et al., 2007), a hallmark of Waardenburg-Hirschsprung disease, and we did not observe any exaggeration of the Sox10^Dom/+ phenotype when crossed into a Mef2c heterozygous background (data not shown). However, we do show here that loss of Mef2c function in the neural crest results in hypopigmentation at birth, which is characteristic of many forms of Waardenburg syndromes; therefore, the MEF2C gene should be considered for a direct or modifier role in congenital neural crest disorders in humans.

The authors thank Ivy Hsieh and Jillian Jarrett for excellent technical assistance. This work was supported by Grant #6-FY08-311 from the March of Dimes and R01 DE019118 from the NIDCR to B.L.B. and by grants from the ACS and the Melanoma Research Alliance to M.L.W. Deposited in PMC for release after 12 months.