Secrets to a healthy Sox life: lessons for melanocytes


Address correspondence to Dr Michael Wegner,


Sox proteins are transcriptional regulators with a high-mobility-group domain as sequence-specific DNA-binding domain. For function, they generally require other transcription factors as partner proteins. Sox proteins furthermore affect DNA topology and may shape the conformation of enhancer-bound multiprotein complexes as architectural proteins. Recent studies suggest that Sox proteins are tightly regulated in their expression by many signalling pathways, and that their transcriptional activity is subject to post-translational modification and sequestration mechanisms. Sox proteins are thus ideally suited to perform their many different functions as transcriptional regulators throughout mammalian development. Their unique properties also cause Sox proteins to escape detection in many standard transcription assays. In melanocytes, studies have so far focused on the Sox10 protein which functions both during melanocyte specification and at later times in the melanocyte lineage. During specification, Sox10 activates the Mitf gene as the key regulator of melanocyte development. At later stages, it ensures cell-type specific expression of melanocyte genes such as Dopachrome tautomerase. Both activities require cooperation with transcriptional partner proteins such as Pax-3, CREB and eventually Mitf. If predictions can be made from other cell lineages, further functions of Sox proteins in melanocytes may still lie ahead.

The chosen few

Approximately 5–10% of all genes in the mammalian genome code for transcriptional regulators. These are grouped into transcription factor superfamilies by their type of DNA-binding domain. In addition to the hundreds of zinc finger and leucine zipper proteins, helix–turn–helix and helix–loop–helix proteins, there are also a few dozen transcription factors with a high-mobility-group domain (Laudet et al., 1993). Twenty belong to the Sox (Sry-box) family of transcription factors (Schepers et al., 2002). Their high-mobility-group domain closely resembles the high-mobility-group domain of the Sry protein, the Sry-box (for review, see Bowles et al., 2000; Wegner, 1999). With the exception of Sry, which is an essential transcription factor for male sex determination encoded on the mammalian Y chromosome, all Sox proteins carry the term ‘Sox’ in their name followed by a number (Figure 1). According to sequence homologies within and outside the high-mobility-group domain, they can be classified into groups A–H (Figure 1). Two Sox proteins from the same group are more closely related in their primary sequence and in their biochemical characteristics than two Sox proteins from different groups (Bowles et al., 2000; Wegner, 1999). A compilation of the current and final nomenclature has recently been published (Schepers et al., 2002). Although Sox proteins represent only a tiny portion of all transcription factors, they are widely distributed and have many different biological functions which have been summarized in recent reviews (Bowles et al., 2000; Wegner, 1999). Here, the focus is on the unique biochemical properties of Sox proteins and a number of special features that have emerged from recent studies and allow a better understanding of their role in melanocyte biology.

Figure 1.

Classification of the 20 mammalian Sox proteins into groups A–H (Schepers et al., 2002). Note that group B is further divided into subgroups B1 and B2. Sox12 and Sox22 denominate the same Sox protein in mouse and human, respectively. The same holds true for Sox15 and Sox20.

The ‘omnipresent’ Sox protein

Although there are only 20 Sox proteins in mammals, they are so widely expressed during embryonic development and in the adult organism that all tissues and cell types that have been analyzed for their presence have turned out to express at least one Sox protein at some time of their development (Bowles et al., 2000; Wegner, 1999). Indeed, many developmental processes depend on the presence of Sox proteins, ranging from blastocyste formation, gastrulation and germ layer formation to formation of haematopoetic and nervous systems, skeleton, gonad, spleen, heart, blood vessels, and melanocytes (Bowles et al., 2000; Sock et al., 2004; Wegner, 1999). Thus, Sox proteins are everywhere.

Single Sox proteins nevertheless exhibit precisely defined expression patterns that correlate with temporal or spatial coordinates, particular processes or cell types, proliferative state or differentiation status. Strong expression of the group E protein Sox10, for instance, is restricted to neuroectodermal and neural crest derivatives (Kuhlbrodt et al., 1998a,b). Sox2, in contrast, occurs in stem cells (Avilion et al., 2003), whereas Sox11 can be found at sites of epithelial-mesenchymal interactions (Hargrave et al., 1997; Sock et al., 2004). As a consequence, organs, tissues and cell lineages often exhibit complex patterns of consecutive or overlapping expression of Sox proteins.

The ‘omnifunctional’ Sox protein

Sox proteins have been associated with a multitude of functions. There is strong evidence that Sox proteins are required to preserve stem cell characteristics and to maintain a pluripotent state under certain conditions (Avilion et al., 2003; Bylund et al., 2003; Graham et al., 2003; Kim et al., 2003). Sox proteins also influence cell death, survival and proliferation (Akiyama et al., 2002; Honore et al., 2003; Hur et al., 2004; Kapur, 1999; Panda et al., 2001; Sonnenberg-Riethmacher et al., 2001) as well as cell fate decisions and consecutive lineage progression (Akiyama et al., 2002; Stolt et al., 2003). In other cases, Sox proteins are essential for terminal differentiation (Peirano et al., 2000; Stolt et al., 2002).

These many different functions offer an explanantion for the sometimes complex pattern of expression of multiple Sox proteins in a single cell lineage. One Sox protein may be essential for maintaining stem cell characteristics of the early multipotent progenitor, while the second may define the already specified progenitor and yet another Sox protein may be responsible for its terminal differentiation. Such a situation is, for instance, observed in the central nervous system, where pluripotent, proliferating neural stem cells express Sox2 (Bylund et al., 2003; Collignon et al., 1996; Graham et al., 2003). Once the cells loose their capacity to proliferate and become determined neuronal precursor cells, they switch off Sox2 and instead begin to express Sox11 (Uwanogho et al., 1995).

In some cases, one and the same Sox protein can be required for several consecutive steps of lineage development. Thus, Sox10 maintains pluripotency in early migrating neural crest progenitors (Kim et al., 2003), but also influences their fate decisions at later stages by preventing neuronal differentiation and forcing them instead to develop into peripheral glia or melanocytes in a context-dependent manner (Aoki et al., 2003; Britsch et al., 2001; Dutton et al., 2001; Honore et al., 2003; Paratore et al., 2001). Different amounts of Sox10 are required for these functions (Kim et al., 2003). At least in some peripheral glia such as myelinating Schwann cells, Sox10 has yet another essential function during terminal differentiation (Peirano et al., 2000). Similarly, Sox9 is essential at multiple stages of chondrogenesis during skeletal development including the early formation of mesenchymal condensations, the specification of chondrocyte progenitors and their development to collagen-expressing pre-hypertrophic chondrocytes (Akiyama et al., 2002; Bi et al., 1999, 2001).

Not all functions of a Sox protein are equally visible in mouse models that carry a deletion or inactivation of the corresponding Sox gene as evident, for example, from knockout mice for Sox11, Sox18 and Sox8. Neural development in Sox11-deficient mice appears undisturbed despite the strong expression of Sox11 in maturing neuronal precursor cells (Sock et al., 2004). Although Sox18 is strongly expressed in developing vascular epithelium and hair, its loss only leads to a mild coat defect in affected mice with a reduced proportion of zigzag hairs, pheomelanin reduction in hair shafts and follicles and a resulting darker appearance (Pennisi et al., 2000a,b). In Sox8-deficient mice, neural crest cells, the central nervous system and the male gonad all develop normally, although these tissues normally express Sox8 (Chaboissier et al., 2004; Sock et al., 2001; Stolt et al., 2004). Instead of a tissue-specific defect, Sox8-knockout mice are smaller and have a reduced body weight, but are viable and fertile (Sock et al., 2001).

The easiest explanation for an absent or mild phenotype or an incomplete phenotypic penetrance is the continued presence of Sox proteins with similar and therefore redundant function. The ability to compensate the loss of a given Sox protein is especially pronounced for closely related Sox proteins from the same group (Figure 1). Accordingly, tissues which are not or mildly affected in a knockout mouse often express two or more Sox proteins of the same group. Neuronal precursors in the central nervous system, for instance, express Sox4 in addition to Sox11, both of which belong to group C (Cheung et al., 2000). Developing vascular epithelium in forming blood vessels is not only positive for Sox18, but also for the other group F proteins Sox17 and Sox7 (Downes and Koopman, 2001). In the case of Sox8-deficient mice, cells of the neural crest or the central nervous system continue to express the related group E proteins Sox9 or Sox10 (Stolt et al., 2004; Maka et al., 2005), while strong Sox9 expression persists in the Sox8-deficient gonad (Chaboissier et al., 2004).

Experimental evidence for functional redundancy among Sox proteins of the same group comes from different approaches. Overexpression of ectopic Sox8, Sox9 or Sox10 in the early chick neural tube similarly induces neural crest formation showing that all three group E Sox proteins are equivalent in this function (Cheung and Briscoe, 2003). Equally informative is the combination of multiple gene deletions in compound mutant mice which may reveal tissue-specific functions invisible in the single Sox gene knockouts. Combining the Sox8 deletion with loss of either Sox9 or Sox10, for instance, uncovers previously hidden roles for Sox8 in the development of the central nervous system, the enteric nervous system and the male gonad (Chaboissier et al., 2004; Maka et al., 2005; Stolt et al., 2004). Spontaneous mouse mutants which exist for some Sox proteins can also be revealing. The ragged mouse mutant, for instance, carries a single-base deletion in the Sox18 gene that introduces a frameshift into the coding sequences (Pennisi et al., 2000a,b). The resulting truncated protein consists of the first 313 amino acids of Sox18 fused to 122 unrelated carboxyterminal amino acids, and shows all the characteristics of a dominant negative in vitro. In contrast to Sox18-deficient mice, ragged mutant mice have a much stronger coat phenotype and also come down with severe vascular defects and edema that result in prenatal death. This argues that the ragged Sox18 protein also functions as a dominant negative in vivo and interferes with the function of all group F proteins including Sox7 and Sox17 at sites of co-expression (Downes and Koopman, 2001). Other Sox18 mutations lead to intermediate phenotypes between the ragged mutant and the Sox18-deletion (Fitch et al., 2003), and in humans lead to hypotrichosis-lymphedema-telangiectasia (Irrthum et al., 2003).

The hidden Sox protein

With all these different functions, many genes must be under the transcriptional control of Sox proteins. However, few gene promoters have so far been described as bound and regulated by Sox proteins indicating that Sox proteins often escape detection in standard transcription studies. It is likely, that some of the features of Sox proteins are responsible.

As already mentioned, most cells express at least one Sox protein. Thus, if a promoter is analyzed in reporter gene assays by transfection in a heterologous cell line, the requirement for a Sox protein may go unnoticed because endogenous Sox proteins may be able to substitute for the Sox protein normally working on the promoter.

Furthermore, mapping of regulatory elements within the promoter is usually performed in electrophoretic mobility shift or DnaseI footprint assays. Reactions not only contain fragments of the promoter and nuclear extracts, but also the synthetic double-stranded desoxy-ribonucleic acid dIdC as competitor to suppress unspecific binding of extract proteins to the promoter fragment. This DNA is chosen because its simple repetitive sequence with hypoxanthine as sole purine base is unlikely to resemble many specific binding sites for transcription factors. In fact, dIdC only resembles AT-rich DNA in the minor groove as well as GC-rich DNA in the major groove (Solomon et al., 1986). Resemblance to the minor groove of AT-rich DNA is usually irrelevant for binding studies, as most transcription factors establish their key contacts in the major groove. Unfortunately, proteins with a high-mobility-group domain including Sox proteins present one of the few exceptions as they primarily interact with the minor groove (Werner et al., 1995). The consensus recognition site for Sox proteins (5′-A/TA/TCAAA/TG-3′) is furthermore AT-rich (Harley et al., 1994) so that dIdC is a very effective competitor for Sox proteins. As a consequence, standard binding assays systematically suppress the detection of Sox proteins. To avoid this, dIdC must be replaced by dGdC.

Sox proteins are also no easy targets for assays designed to identify transcriptional regulators by their ability to interact with a promoter in vivo such as the yeast one-hybrid screen. A promoter-specific transcription factor is best identified in a yeast one-hybrid screen, if the transcription factor elicits a strong transcriptional response by itself or in cooperation with the ubiquitous transcriptional regulators of the yeast cell. This screening assay is, therefore, biased towards transcription factors that function autonomously on promoters of ubiquitously expressed genes. Sox proteins do just the opposite, as they usually require other transcription factors as partner proteins to perform cell-specific functions (Kamachi et al., 2000).

Sox2 is a good example. In the inner cell mass of the mammalian blastocyst, Sox2 teams up with the POU-domain transcription factor Oct-3/4 to regulate such genes as the transcriptional co-activator UTF (Nishimoto et al., 1999) or FGF-4 (Yuan et al., 1995) whose products are an important downstream regulator of inner cell mass development (UTF) and an essential paracrine factor for extraembryonic trophoblasts (FGF-4). The requirement for transcriptional partners furthermore appears to vary between tissues. Thus, Sox2 functionally interacts with Pax-6 at later stages of development in the mammalian eye to ensure lens induction and lens-specific expression of differentiation genes such as δ-Crystallin (Kamachi et al., 1998, 2001) and with POU-domain proteins such as Brn-2 in the neural primoridum for Nestin expression (Tanaka et al., 2004). To achieve this cooperativity, Sox2 and its partner bind to adjacent sites in composite elements of tissue-specific intronic (for UTF, δ-Crystallin and Nestin) or exonic (for FGF-4) enhancers and interact with glue-like surface patches that lock the proteins in distinct conformational arrangements onto DNA (Figure 2). The interactions strongly depend on the exact spacing and orientation of the binding sites within the composite element (Remenyi et al., 2003; Williams et al., 2004).

Figure 2.

Cooperativity between Sox proteins and their partner proteins is often mediated by composite elements within promoters or enhancers that contain adjacent binding sites for both transcription factors. With only the Sox protein (Sox) or its partner (P) bound, the composite element remains inactive. In vivo activation requires joint binding of both proteins and protein–protein interactions between them. Genes reported in the literature to be cooperatively activated by a Sox protein and its partner are listed.

Sox2 is not the only Sox protein known to require partners for promoter activation (for review, see Kamachi et al., 2000). Sox9, for example, interacts with a long isoform of c-Maf and the β-catenin co-activator in chondrocytes (Akiyama et al., 2004; Huang et al., 2002), as well as with Steroidogenic factor 1 (SF-1) in Sertoli cells (De Santa Barbara et al., 1998). Sox10, on the other hand, co-operates with Pax3, Krox-20 and Mitf in particular neural crest-derived cell types to ensure cell type specific target gene activation (Bondurand et al., 2000, 2001; Ghislain et al., 2003; Jiao et al., 2004; Lang and Epstein, 2003; Ludwig et al., 2004; Potterf et al., 2000). Adjacent binding sites for these group E Sox proteins and their respective partners have again been identified in some of the target promoters (Figure 2). However, it is too early to tell whether such adjacent binding is an essential requirement or whether functional interaction on a given promoter might also take place over greater distances (see below).

The requirement for transcriptional partners and the ability to team up with different partner proteins may explain why the same Sox protein can be used for different purposes in various cell types or at consecutive stages of development in the same cell type. At the same time, it generates a specificity problem. There is evidence for a code in which a particular Sox protein matches best with a certain subset of transcription factors (Kamachi et al., 2000; Kuhlbrodt et al., 1998a,b; Tanaka et al., 2004). Additionally, interactions in solution are probably often too weak to be functionally significant and may only become relevant once the transcription factors are jointly bound to DNA (Remenyi et al., 2003; Williams et al., 2004).

The architectural Sox protein

Sox proteins regulate target gene transcription not only through physical interaction with other transcription factors, but may additionally influence the three-dimensional structure of a promoter or an enhancer (Figure 3A). Upon binding to DNA, the minor groove is widened so that Sox proteins introduce a strong bend of 70–85° (Connor et al., 1994; Ferrari et al., 1992; Werner et al., 1995). Multiple binding sites are present in several target gene promoters and enhancers, including the Collagen 2a1 and Collagen 11a2 enhancers in case of Sox9 and the Mitf, Dct and Myelin protein zero promoters for Sox10 (Bridgewater et al., 1998; Lefebvre et al., 1997; Ludwig et al., 2004; Ng et al., 1997; Peirano et al., 2000). Sox protein-dependent DNA-bending at all these sites strongly alters the overall three-dimensional arrangement of the enhanceosome, defined as the promoter or enhancer with all their bound transcription factors (Figure 3B).

Figure 3.

Architectural properties of Sox proteins. (a) Binding of Sox proteins introduces a strong bend into DNA. A DNA region with multiple binding sites for Sox proteins is shaped into a defined conformation. (b) This architectural activity of Sox proteins may enable contacts between transcription factors bound to the same region and support the formation of an enhanceosome. (c) Binding of Sox protein dimers alters DNA conformation differently from monomers.

Thus, Sox proteins may function as architectural proteins (Werner and Burley, 1997). They may actually determine how other transcription factors interact with each other in the enhanceosome (Figure 3B) and which transcriptional co-factors are recruited. Both Sox2 and Sox9, have also been reported to directly interact with transcriptional co-factors such as CBP/p300 (Nowling et al., 2003; Tsuda et al., 2003). Sox9 additionally binds to the TRAP230 subunit of the thyroid hormone receptor-associated protein complex (Zhou et al., 2002), whereas Sox6 has been shown to interact with the transcriptional repressor CtBP2 (Murakami et al., 2001).

Experimental evidence for the architectural function is still scarce. Nevertheless, mutations within the high-mobility-group domain of SRY and Sox2 exist that leave DNA-binding specificity and affinity unaltered, but selectively interfere with DNA-bending (Pontiggia et al., 1994; Scaffidi and Bianchi, 2001). In the case of Sox2, loss of DNA-bending completely abolished the ability of the mutant protein to activate transcription. In the case of SRY, loss of DNA-bending caused phenotypic sex reversal.

Interestingly, the exact structural alteration that is introduced into DNA may vary among different Sox proteins because some Sox proteins bind DNA as dimers, whereas the majority prefers monomeric binding (Figure 4). Sox proteins can be constitutive dimers in solution as observed for the group D Sox proteins (Sox5, Sox6 and Sox13) which contain a zipper-like coiled-coil dimerization domain (Lefebvre et al., 1998). Alternatively, dimers may form only upon cooperative binding to adjacent binding sites on DNA (Bondurand et al., 2001; Bridgewater et al., 2003; Peirano et al., 2000; Sock et al., 2003). In case of group E Sox proteins (Sox8, Sox9 and Sox10), this cooperative binding is mediated by a conserved region situated in front of the high-mobility-group domain (Peirano and Wegner, 2000; Schlierf et al., 2002). It leaves open the possibility that group E Sox proteins also bind as monomers to single binding sites on DNA (Figure 4). With two high-mobility-group domains in a dimer each binding to DNA, the resulting bend is a function of the exact spacing of adjacent sites. Thus, it is attractive to assume that via differential spacing of adjacent binding sites, Sox protein dimers could introduce bending angles into promoters that are most suited for the overall conformation of the enhanceosome (compare Figure 3A with Figure 3C). The importance of the DNA-dependent dimerization of group E Sox proteins is underlined by human SOX9 mutations that selectively disturb this ability and lead to campomelic dysplasia, the skeletal malformation syndrome associated with SOX9 mutations (Bernard et al., 2003; Sock et al., 2003).

Figure 4.

Different modes of DNA binding among Sox proteins. Most Sox proteins such as those belonging to group B (SoxB) exist as monomers in solution and bind independently to DNA as monomers. Group E Sox proteins (SoxE) also exist as monomers in solution and may bind to DNA as monomers. Alternatively, two group E Sox proteins bind cooperatively to adjacent DNA sites and form functionally relevant dimers. In contrast, group D Sox proteins (SoxD) already exist as constitutive dimers in solution and thus invariably bind DNA as dimers.

The regulated Sox protein

The complex expression pattern and many functions of Sox proteins make it necessary to tightly regulate their expression as well as their activity. As developmental regulators, Sox genes are controlled in their expression by the essential developmental signalling pathways. Sox9 expression, for instance, has been shown to be under the positive control of BMP, FGF, Hedgehog and Wnt signalling in chondrocytes and embryonic intestinal epithelial cells (Blache et al., 2004; Murakami et al., 2000a,b; Uusitalo et al., 2001; Zeng et al., 2002). Similarly, Sox10 expression in the Xenopus neural crest is induced both by FGF and Wnt signals (Aoki et al., 2003; Honore et al., 2003). Signalling molecules have also been identified that counteract this activation and repress expression. For Sox9, these repressive signalling molecules in chondrocytes include the cytokines interleukin-1α and TNF-α (Murakami et al., 2000a,b) and the lipophilic retinoic acid and thyroid hormone (Okubo and Reddi, 2003; Sekiya et al., 2001a,b; Weston et al., 2002). Steroid hormones also influence Sox gene expression as evident from a strong activation of Sox9 expression by glucocorticoids in chondrocytes (Sekiya et al., 2001a,b) and progestin-dependent Sox4 expression in breast cells (Graham et al., 1999). Thus, Sox genes are not preferential targets of a particular signalling pathway, but rather are influenced in their expression by many different signalling pathways. As a consequence, different signals converge on Sox genes and become integrated into a single readout (Kolettas et al., 2001; Schaefer et al., 2003).

The regulatory regions that mediate these effects on Sox gene expression have not been thoroughly characterized. However, it is known that expression of many Sox genes is controlled by numerous regulatory regions that are spread over significant distances (Bagheri-Fam et al., 2001; Brunelli et al., 2003; Qin et al., 2004; Uchikawa et al., 2003). In the case of the mammalian Sox9 gene, they extend over an interval of over 1 Mb (Wunderle et al., 1998). Thus, it is likely that the effects of the various signals will also be mediated by several of these regulatory regions.

Sox proteins in turn influence the activity of signalling pathways by controlling the expression of signalling molecules or their receptors. Sox9, for instance, regulates expression of the TGF-β family member anti-Müllerian-hormone in Sertoli cells (De Santa Barbara et al., 1998), while Sox10 is responsible for the upregulation of the erbB3 component of the neuregulin receptor in the developing peripheral nervous system (Britsch et al., 2001) and the endothelin receptor B in enteric neural crest cells (Zhu et al., 2004). As a consequence, Sox proteins can be part of regulatory feedback loops. In chondrocytes, for instance, BMP-2 activates Sox9 expression which in turn leads to upregulation of the BMP antagonist noggin and, as a consequence, downregulation of BMP-2 activity (Zehentner et al., 2002).

Sox proteins also regulate each others’ expression as evident from induction of Sox5 and Sox6 expression by Sox9 in chondrocytes (Akiyama et al., 2002). Once induced, Sox5 and Sox6 cooperate with Sox9 in the activation of common target genes such as Collagen 2a1 (Lefebvre et al., 1998). Sox5 and Sox6 thus increase the transcriptional activity of their inducer Sox9.

Other mechanisms by which the activity of Sox proteins may be modulated are post-translational modifications. Sequence inspection of Sox proteins reveals that they frequently contain consensus recognition sites for phosphorylation, acetylation or sumoylation. So far, little is known about the actual occurrence of such modifications in Sox proteins and their role in vivo. Studies on Sox9 prove, however, that post-translational modifications indeed occur in Sox proteins and cause functional changes. Sox9 has been found to interact with the catalytic subunit of protein kinase A (Huang et al., 2000). It furthermore contains two sites that are phosphorylated by this kinase. This phosphorylation in turn increases the DNA-binding affinity and the transcriptional activity of Sox9 (Huang et al., 2000). One of the factors, which activates protein kinase A under physiological conditions in chondrocytes through elevated intracellular cAMP levels, is the parathyroid hormone related peptide PTHrP. As a consequence, PTHrP increases the transcriptional activity of Sox9 specifically in prehypertrophic chondrocytes through protein kinase A-dependent phosphorylation (Huang et al., 2001).

Sequestration may be equally important for regulating the transcriptional activity of some Sox proteins. In addition to the two nuclear localization signals that are conserved among all Sox proteins (Poulat et al., 1995; Südbeck and Scherer, 1997), a nuclear export signal is present within the high-mobility-group domain of the group E Sox proteins Sox8, Sox9 and Sox10 (Gasca et al., 2002; Rehberg et al., 2002). Proteins with both nuclear import and export sequences will continuously shuttle between the cytoplasm and the nucleus, and can be modified or retained in the cytoplasm through interactions with cytosolic proteins. Post-translational modification can furthermore influence the activity of both import and export sequences and shift the balance between both processes so that at the extreme ends, Sox proteins could be completely nuclear in one situation and completely cytoplasmic under different circumstances. Thus, it can be imagined that even before Sox gene expression is turned off in a cell and before all remaining Sox protein molecules are polyubiquitinated and degraded by the proteasome (Akiyama et al., 2005), the protein is excluded from the nucleus and sequestered away from its target genes. Alternatively, cytoplasmic storage may allow the production of transcriptionally inert Sox protein that can be rapidly mobilized on demand. Whether nuclear export or import is indeed used for such rapid regulation of Sox protein activity in vivo, is not clear at the moment. However, Sox proteins have been found under certain circumstances in a cytoplasmic localization. There have been reports, for instance, that Sox9 which has a role in male sex determination, is first present in the sexually indifferent gonad of genotypically male and female embryos, but exclusively localized in the cytoplasm (da Silva et al., 1996; de Santa Barbara et al., 2000). Once gender-specific differentiation of the gonad sets in, Sox9 is translocated to the nucleus of the male gonad where its expression persists. In the female gonad, Sox9 remains in the cytoplasm, and its expression is shut off a bit later.

Sox proteins and melanocyte biology

Similar to many other cell types throughout the developing embryo, melanocytes depend on Sox proteins. Melanocytes are derived from the neural crest which compared with adjacent neuroectoderm down-regulates Sox2 and instead turns on Sox9 expression (Cheung and Briscoe, 2003; Spokony et al., 2002; Wakamatsu et al., 2004). Loss of Sox9 or expansion of Sox2 thus translates into an early defect in neural crest cell generation and a resulting reduction or loss of melanocytes. Whereas Sox9 is primarily expressed in premigratory neural crest cells and shut off soon after epithelial-mesenchymal transition, Sox10 just starts to be expressed in the premigratory neural crest immediately before the onset of migration and remains on in migrating neural crest cells (Aoki et al., 2003; Britsch et al., 2001; Dutton et al., 2001; Honore et al., 2003; Kapur, 1999; Kuhlbrodt et al., 1998a,b; Southard-Smith et al., 1998). This includes those neural crest cells that migrate along a dorsolateral pathway and give rise to the melanocyte lineage (Figure 5).

Figure 5.

Summary of known Sox10 functions during melanocyte development. Sox10 is required both in the neural crest cell (NCC) and the melanoblast (MB) where it performs several functions in cooperation with different partner proteins, including activation of Mitf and Dct gene expression. It is not clear whether Sox10 is still expressed in fully differentiated melanocytes (MC).

Mice that are deficient for Sox10 have lost most or all of their melanocytes (Britsch et al., 2001; Herbarth et al., 1998; Kapur, 1999; Southard-Smith et al., 1998). Heterozygous mouse mutants and human patients with one inactivated Sox10 allele exhibit a partial melanocyte defect which in humans usually leads to Waardenburg syndrome characterized by partial depigmentation of hair, skin or iris and additional sensorineural deafness due to melanocyte loss from the stria vascularis of the inner ear (Pingault et al., 1998). Additional symptoms observed in human patients with heterozygous SOX10 mutations include a partial loss of the enteric nervous system known as Hirschsprung disease (Pingault et al., 1998) and – depending on the type of SOX10 mutation – a combination of peripheral demyelinating neuropathy and central dysmyelinating leukodystrophy (Inoue et al., 2004).

The melanocyte defect is partially a consequence of the survival function of Sox10 in migrating neural crest cells (Paratore et al., 2001). In this respect, Sox10 behaves similar to the Pax-3 transcription factor which is also broadly expressed in the neural crest and whose mutations are the most common cause for Waardenburg syndrome in human patients (for a review, see Read and Newton, 1997). Supporting this similarity in expression patterns, melanocyte development can actually be rescued by ectopic expression of a Sox10 transgene under the control of regulatory regions from the Pax-3 gene in virally transformed Sox10-deficient neural crest cells (Hou et al., 2004).

Not all derivatives of migrating neural crest cells are as severely affected by Sox10 mutations as the melanocyte lineage, hinting at an additional melanocyte-specific function for Sox10 (Britsch et al., 2001; Herbarth et al., 1998; Kapur, 1999; Southard-Smith et al., 1998). Sox10 remains expressed during specification of neural crest precursor cells to melanoblasts (Figure 5). The role of Sox10 in migrating neural crest cells and later in melanocyte specification is a good example for the repeated use of a particular Sox protein during consecutive phases of lineage development. Whereas Sox10 is down-regulated early in pigment cell precursors from zebrafish (Dutton et al., 2001), it remains detectable for quite some time during embryonic melanocyte development in mammals (Potterf et al., 2001; Southard-Smith et al., 1998). Whether Sox10 is still present in fully differentiated melanocytes of the adult, has not yet been reported. Melanocyte cultures and melanomas at least express significant amounts of Sox10 (Kamaraju et al., 2002; Khong and Rosenberg, 2002).

The first and probably foremost role of Sox10 during melanocyte development is in the initial specification event (Figure 5). A number of studies have shown that Sox10 is directly responsible for activating the expression of the melanocyte-specific isoform of Mitf (Mitf-M), the key transcriptional regulator of melanocyte development (Bondurand et al., 2000; Lee et al., 2000; Potterf et al., 2000; Verastegui et al., 2000). Lending support to the genetic relationship between Sox10 and Mitf, heterozygous mutations in the Mitf gene have a strong phenotypic overlap with heterozygous Sox10 mutations and similarly lead to Waardenburg syndrome in humans. Additionally, double heterozygous animals with mutations in both genes exhibit a more severe hypopigmentation than animals with single heterozygous mutations (Potterf et al., 2000).

Mitf is also under the transcriptional control of Sox10 in zebrafish arguing that this function is conserved in vertebrates (Elworthy et al., 2003). In zebrafish, it has furthermore been shown that expression of an Mitf transgene under the control of the Sox10 promoter rescues the pigment cell defect in Sox10-deficient colourless mutants supporting that Mitf activation is the most essential function for Sox10 in this cell lineage. Both in zebrafish and in mammals, Sox10 directly interacts with several binding sites in the melanocyte-specific promoter of the Mitf gene (Bondurand et al., 2000; Elworthy et al., 2003; Lee et al., 2000; Potterf et al., 2000; Verastegui et al., 2000) arguing that Sox10 may have a profound impact on the three-dimensional structure of the enhanceosome at the Mitf promoter. Binding of Sox10 to these sites is indeed necessary for promoter activation both in tissue culture and in transgenic zebrafish (Bondurand et al., 2000; Lee et al., 2000; Potterf et al., 2000; Verastegui et al., 2000; Elworthy et al., 2003). In mammals, additional binding sites for Sox10 are present in an enhancer situated approximately 14.5 kb in front of the Mitf promoter. They likely contribute to full Sox10-dependent gene activation (Watanabe et al., 2002).

Among the Sox10 binding sites in the Mitf promoter, one stands out as the major contributor. This binding site is located approximately 250 bp in front of the transcriptional start site and is immediately adjacent to a binding site for Pax-3 (Bondurand et al., 2000). Several studies were able to show a synergistic activation of the Mitf promoter by Sox10 and Pax-3 (Bondurand et al., 2000; Potterf et al., 2000) thus arguing that these co-expressed transcription factors may represent partner proteins during melanocyte specification (Figure 2). In support of this partnership, down-regulation of Pax-3 in a melanoma line causes it to loose Mitf expression and instead acquire characteristics of peripheral glia in its expression pattern (Kamaraju et al., 2002). Pax-3 is, however, not the only transcription factor that cooperates with Sox10 in Mitf gene activation (Figure 5). Additionally, CREB only mediates the stimulatory effects of α-MSH on Mitf gene expression in the presence of Sox10. Sox10, thus confers cell-specificity to CREB function in melanocytes (Huber et al., 2003).

Although activation of Mitf expression is definitely the most important role of Sox10 during melanocyte development, it appears likely that Sox10 as a transcription factor has more than a single target. Several observations indeed argue that the gene for Dopachrome tautomerase (Dct, Tyrosinase-related protein-2) is also under Sox10 control (Figure 5). Dct expression is affected more strongly than other melanocyte markers in the Sox10 heterozygous mouse (Britsch et al., 2001; Potterf et al., 2001). A transient loss of Dct expression during early melanocyte development is furthermore only observed in Sox10 heterozygous mice, but not in Mitf heterozygotes (Potterf et al., 2001). This argues that Sox10-dependent activation of Dct expression cannot be completely mediated through Mitf despite the fact that Mitf is known to directly activate the promoters of Dct and other pigmentation genes (Bertolotto et al., 1998).

In agreement with a direct transcriptional regulation, the Dct promoter can be activated by Sox10 in transient transfection assays and contains several Sox binding sites arguing again that Sox10 may mediate part of its effect by shaping the overall conformation of the enhanceosome (Ludwig et al., 2004). As observed in other promoters, binding sites differ from each other by their affinity for Sox10 as well as by their ability to bind either Sox10 monomers or dimers (Figure 4). One of the more important binding sites is furthermore in immediate vicinity to an M-box (Figure 2), the major binding site for Mitf (Bertolotto et al., 1998). Although there is currently no evidence that M-box and adjacent Sox10 binding site constitute a composite element similar to the ones essential for Sox2 function, it is clear that both Sox10 and Mitf are able to synergistically activate the Dct promoter, particularly under conditions where both transcription factors are present in suboptimal amounts (Jiao et al., 2004; Ludwig et al., 2004). Thus, it appears that in the specified melanoblast, Sox10 may use Mitf as a partner protein to activate melanoblast-specific genes (Figure 5). This cooperativity might be especially important during early phases of melanocyte development when Mitf amounts are still limiting or when other signalling pathways, that positively influence expression of melanocyte genes, are not yet fully active.

Although Sox10 is firmly established as a regulator of melanocyte development, many questions are still unresolved. Thus, it is not yet clear whether Sox10 has any functions in fully differentiated melanocytes and whether these functions are distinct from the ones exerted during melanocyte development. It remains to be shown how exactly Sox10 cooperates with its various partner proteins during melanocyte development and how its activity may be modulated by changes in expression level, subcellular localization or post-translational modifications. Finally, it needs to be addressed whether the expression of Sox10 in melanoma is causal for tumorigenesis or simply a consequence of its expression in all melanocyte-derived cells. It should also be reminded that most cell types express more than a single Sox protein. Thus, it may not be too surprising if melanocyte development additionally depends on the presence of other Sox proteins yet to be identified.


Research on Sox proteins in my laboratory is supported by grants from the Deutsche Forschungsgemeinschaft, the Thyssen-Stiftung, the Schram-Stiftung and the Fonds der Chemischen Industrie.