Over 10 years have passed since the first Sox gene was implicated in melanocyte development. Since then, we have discovered that SOX5, SOX9, SOX10 and SOX18 all participate as transcription factors that affect key melanocytic genes in both regulatory and modulatory fashions. Both SOX9 and SOX10 play major roles in the establishment and normal function of the melanocyte; SOX10 has been shown to heavily influence melanocyte development and SOX9 has been implicated in melanogenesis in the adult. Despite these advances, the precise cellular and molecular details of how these SOX proteins are regulated and interact during all stages of the melanocyte life cycle remain unknown. Improper regulation of SOX9 or SOX10 is also associated with cancerous transformation, and thus understanding the normal function of SOX proteins in the melanocyte will be key to revealing how these proteins contribute to melanoma.
Recently deemed ‘omnifunctional’, the SOX family of transcription factors has a reputation for their involvement in the development and normal physiology of numerous tissues through their regulation of a wide range of seemingly disparate biological events. From maintaining stem cell properties, to lineage restriction, and terminal differentiation, SOX proteins appear to coordinate these functions through precise temporal and spatial expression patterns that are unique to individual cell types and tissues (Kiefer, 2007; Wegner, 2009). The melanocyte lineage is no exception. The participation of SOX proteins in melanocyte development was first uncovered when the mouse Dominant megacolon phenotype (Dom) was genetically linked to a truncating mutation in Sox10 (Sox10Dom) (Figure 1, Table 1; Herbarth et al., 1998; Southard-Smith et al., 1998). The Sox10Dom mouse serves as a model for Waardenburg–Shah syndrome type 4 (WS4), which is actually a combination of two disorders – WS4 patients often experience sensorineural deafness and hypopigmentation associated with Waardenburg syndrome and enteric aganglionosis associated with Hirschprung’s disease. Further studies of Sox10 mouse mutants have revealed that both aspects of WS4 can be attributed to early defects in neural crest development, specifically the inability of neural crest cells (NCCs) to be specified as melanocytes and glial cells (Paratore et al., 2001; Potterf et al., 2001; Southard-Smith et al., 1998). Since this initial discovery, numerous Sox10 mutations have been uncovered in humans, mouse, and zebrafish (Figure 1, Table 1), as well as evidence demonstrating that SOX10 plays a key role in the transcriptional control of the master regulatory gene for melanogenesis, Mitf (microphthalmia-associated transcription factor) (Bondurand et al., 2000; Elworthy et al., 2003; Lang and Epstein, 2003; Lee et al., 2000; Potterf et al., 2000; Verastegui et al., 2000). In vitro analysis additionally demonstrates that SOX10 has the ability to transcriptionally regulate a number of genes that are required for melanin synthesis – these include dopachrome tautomerase (Dct, aka tyrosinase-related protein 2 or Trp2) (Jiao et al., 2004; Ludwig et al., 2004; Potterf et al., 2001), tyrosinase (Tyr) (Hou et al., 2006; Murisier et al., 2007), and tyrosinase related protein 1 (Tyrp1) (Murisier et al., 2006).
Table 1. Coding region mutations and polymorphisms of Sox10
Despite our advances in understanding the role of SOX10 in melanocyte development, there is still debate regarding the extent of influence that SOX10 has over various melanogenic targets in vivo. Whether this transcription factor participates similarly in adult melanocytes, particularly those that arise from melanocyte stem cells (MSCs) during pigment cell regeneration, is also unknown. Furthermore, additional SOX family members, including SOX9, SOX18 and SOX5, are also implicated in regulating aspects of the melanocyte life cycle. In this review we will focus on summarizing the current state of our knowledge on the function of SOX proteins in defining and maintaining the melanocyte lineage. We will use melanocyte development as our primary example while also extending our analysis to recent discoveries of SOX proteins in melanocytes of the adult organism. Lastly, based on their critical regulatory functions, we will consider how and/or if these SOX proteins might be misregulated in the development and progression of melanoma. Note, in regards to Sox gene and SOX protein nomenclature, species-specific rules are used when appropriate and mouse-specific rules are used for general references.
The SOX protein family
The SOX family is comprised of approximately 20 transcription factors named after the original member, Sry (Sry HMG-box), because they all share a similar high-mobility-group (HMG) domain (Schepers et al., 2002). Through their HMG domain, SOX proteins mediate sequence-specific DNA binding initiated by the consensus motif, 5′-(A/T)(A/T)CAA(A/T)G-3′ (Harley et al., 1994). SOX protein interactions can be monomeric or dimeric, and once bound to DNA, they regulate transcription through numerous methods – they interact either directly or through coactivators with general transcription machinery to mediate both gene activation and repression; they can indirectly influence transcription by affecting the availability of other factors by sequestering them into complexes (reviewed in Wegner, 2009); and through their L-shaped HMG domain, they can bind the minor groove of their DNA target and impose conformational changes that result in widening of the minor groove, unwinding of the DNA helix and DNA bending (Werner et al., 1995). The chromatin remodeling ability of SOX proteins is thought play a major role in providing the architectural framework for larger transcriptional complexes that regulate gene expression (Werner and Burley, 1997). Flexibility within the HMG domain allows for a range of DNA-bending angles and it has been postulated that the degree of bending is influenced by flanking DNA target sequences and neighboring protein/DNA complexes (Weiss, 2001). The combination of these interactions is thought to drive SOX protein specificity and, depending on the cellular context, may elicit varying levels of promoter activation.
In mammals there are nine groups of SOX proteins: A, B1, B2 and C–H (Schepers et al., 2002). Those that participate as developmental regulators of melanogenesis belong to the SOXE group and include SOX9 and SOX10. SOXE group genes have a very similar structure and are characterized by the presence of the HMG box, a DNA-dependent dimerization domain, and two separate transactivation domains (one central and one at the carboxyterminus) (Figure 1; Bowles et al., 2000; Peirano and Wegner, 2000; Schreiner et al., 2007). Despite these similarities, and the fact that they often share overlapping expression patterns, the necessity of each during embryogenesis is quite varied; for example, SOX9 plays a dominant role in both gonad and chondrocyte development as well as neural crest generation, while SOX10 is essential for the maintenance and differentiation of NCCs once they become migratory (reviewed in Guth and Wegner, 2008). Although each of the SOXE proteins can provide functional redundancy in the neural crest if expressed at the appropriate time as shown in xenopus and chick (Cheung and Briscoe, 2003; Taylor and LaBonne, 2005), development of the melanocyte lineage in mammalian organisms is particularly dependent on unique functions attributed to SOX10 (Kellerer et al., 2006). Interestingly, evidence from human cell culture suggests that after the establishment of the melanocyte precursor, or melanoblast, SOX9 plays a similar role in melanocyte differentiation in the adult organism to that of SOX10 during development (Cook et al., 2005; Passeron et al., 2007).
Two other SOX proteins that appear to play more modulatory roles in melanogenesis include SOX5, of the D group (discussed below), and SOX18 of the F group. Mice with mutations in Sox18 (ragged), when bred on an agouti background, are missing the subapical pheomelanin-pigmented band in the hair shaft, giving these mice an overall darker coat color. This effect appears to be not autonomous to melanocytes, but rather may be due to an effect of SOX18 on agouti signaling protein, both of which are expressed in the dermal papilla of the hair follicle (Pennisi et al., 2000).
Consistent with the neuroectodermal origin of melanocytes, the first time SOX proteins influence the neural crest appears to be during its generation. In most species, SOX8, 9 and 10 are expressed in the area of the dorsal neural tube preceding or coincident with the emergence of the neural crest (Figure 2; reviewed in Hong and Saint-Jeannet, 2005). Of these SOX proteins, SOX9 plays an early role in neural crest development, however its requirement by NCCs varies depending on the axial level in which the NCC originates, the particular neural crest subpopulation in which it is expressed, and the species. In mouse between E8.5 and E9.5, SOX9 is expressed in the premigratory NCCs as they arise along the length of the embryo. Due to essential roles for Sox9 in chondrocyte development, Cheung et al. (2005) employed Wnt1-Cre to specifically knockout Sox9 in NCCs, and at E10.5 observed a drastic reduction in trunk neural crest derivatives due to apoptosis. At this timepoint, the specification and migration of cranial neural crest appears unaffected, however later assessment reveals that SOX9 is required to maintain the chondrogenic potential of the cranial neural crest (Mori-Akiyama et al., 2003). While an essential role for Sox9 in the development of the melanocyte lineage can be inferred from the complete loss of trunk NCCs when Sox9 is missing, whether loss of Sox9 affects pigmentation in mouse requires further assessment. In xenopus, sox9 is also expressed in the developing neural crest, however loss of sox9 by treatment of embryos with morpholinos or antisense oligos results in the specific reduction of cranial NCCs that contribute to the craniofacial skeleton, but does not affect the development of pigment cells (Spokony et al., 2002). In zebrafish, there are two orthologs of sox9, sox9a and sox9b, the latter of which is expressed by both cranial and trunk NCCs. Similar to xenopus, sox9b mutants show severe defects in the formation of the craniofacial cartilage, but only moderate effects on the development of pigment cells. In sox9b−/− embryos iridophores are reduced in numbers, and melanocytes, although present in the correct number and distribution, have more dispersed melanosomes (Yan et al., 2005).
On the other hand, gain-of-function experiments indicate that SOX9 is competent to influence NCCs in the acquisition of melanocyte-like properties in chick (Cheung and Briscoe, 2003). Similarly, in xenopus, SOX9 has the ability to direct the formation of melanoblasts in a sox10-deficient background (Taylor and LaBonne, 2005). Despite its sufficiency, endogenous SOX9 is downregulated in trunk NCCs in mouse, chick, xenopus and zebrafish as they become migratory (Cheung et al., 2005; McKeown et al., 2005; Spokony et al., 2002; Yan et al., 2005). At least in mouse, prolonged Sox9 expression may adversely affect normal melanoblast development, as Sox9 ectopically expressed under the control of the melanocyte-specific promoter Dct results in light coat pigmentation, dorsal white spotting, and microphthalmia (Qin et al., 2004).
In both Sox10Dom/Dom mice and cls−/− zebrafish, loss of Sox10 is associated with an almost complete absence of NCCs on pigment cell migratory pathways (Dutton et al., 2001; Potterf et al., 2001). In cls−/− zebrafish embryos, non-ectomesenchymal NCCs, or those that would give rise to neurons, glia and pigment cells, fail to migrate away from the neural tube, do not differentiate, and eventually undergo apoptosis. The occurrence of this phenotype is preceded by the absence of normal mitf and kit expression, both of which are required for melanocyte development. This suggests that the role of sox10 in this aspect of zebrafish pigmentation is to specify melanocyte fate (Dutton et al., 2001). Forced expression of mitf in cls−/− or mitf/nacre−/− embryos rescues melanophores to the same extent in both mutant lines and indicates that, at least for this pigment cell subtype, the transcriptional activation of mitf can account for the sole function of Sox10 (Elworthy et al., 2003). Similar phenotypic observations have been made in Sox10 mutant mice, namely, that there is an absence of Mitf expression in primary NCC cultures derived from Sox10LacZ/LacZ null mice and that Sox10Dom/Dom mice exhibit higher numbers of apoptotic cells in neural crest migratory pathways early in development (Hou et al., 2006; Southard-Smith et al. 1998). However, by lineage-directed gene transfer in primary NCC cultures it was also shown that although SOX10 acts through MITF to induce the expression of the melanocyte-specific proteins DCT, PMEL17 and TYRP1, MITF is not able to induce the expression of TYR or to rescue pigmentation in the absence of SOX10 (Hou et al., 2006). While the extent of SOX10’s influence on melanocyte-specific gene expression appears to be slightly different in mouse and zebrafish, it appears that in both species the quintessential role of SOX10 in pigmentation is to define the melanocyte lineage through the upregulation of MITF which then goes on to initiate the differentiation steps required for melanocyte migration and survival.
Due to early embryonic lethality in Sox10Dom/Dom or Sox10LacZ/LacZ mice, the requirement for SOX10 throughout melanoblast development in this species has not been fully ascertained. In mouse, SOX10 is detected in melanoblasts that have colonized the matrix of the developing hair follicle at E16.5 and 18.5, and suggests that SOX10 expression may be maintained throughout melanoblast development or at least transiently expressed at key developmental timepoints (Osawa et al., 2005). In contrast, in zebrafish Sox10 is downregulated in all NC cells (except glial cells) after their initial specification (Dutton et al., 2001). This differential SOX10 expression pattern observed between species may simply reflect the fact that mouse melanoblasts require both SOX10 and MITF to generate TYR-expressing, mature melanocytes, whereas this is not the case in zebrafish.
SOX proteins in the adult organism
SOX proteins appear to also participate in the melanocyte life cycle after birth (Figure 2). Human neonatal foreskin contains both non-pigmented melanoblasts and pigmented melanocytes and can be grown in media that supports one or the other population. Melanoblasts derived from these cultures express SOX10 at high levels which decrease as these cells differentiate into melanocytes (Cook et al., 2003). Similarly, in histological sections of normal (non-palmoplantar) skin, SOX10 was only detectable in approximately one of six melanocytes (Hasegawa et al., 2008). Conversely, SOX9, previously unrecognized for its role in melanocyte differentiation, is upregulated as melanoblasts transition to a more melanocytic state (Cook et al., 2005). The production of SOX9 by melanocytes was previously speculated as antibodies against this protein are detected in the sera of vitiligo patients, and recently its in vivo expression in human melanocytes was confirmed (Hedstrand et al., 2001; Krahl and Sellheyer, 2009; Passeron et al., 2007). The role of SOX9 in adult skin appears to mimic that of SOX10 during embryogenesis. Upon stimulation with UVB, SOX9 is upregulated and goes on to directly activate the MITF promoter, increase DCT and TYR expression, and consequently, the production of melanin (Figure 3). In contrast to the embryonic development of melanocytes, SOX9 expression in adult melanocytes is dependent on the activation of the cAMP pathway by way of α-MSH. This pathway can be perturbed by treatment with the α-MSH antagonist, Agouti signaling protein, and underscores the role of SOX9 in acquired pigmentation or tanning (Passeron et al., 2007). Interestingly, SOX10 expression in adult melanocytes is never fully extinguished, and a further report suggests that SOX10 binding sites contribute significantly to the cAMP responsiveness of the MITF promoter (Huber et al., 2003). Whether or not SOX9 mediates its control of melanogenesis independently of or in parallel with the actions of SOX10 remains to be discovered. Although evidence thus far does not support any major contribution of SOX9 to the development of the melanocyte lineage in the embryo, the nature of SOX protein evolution suggests that SOX group family members like SOX9 and SOX10 may retain some of the same functionality. One current model also speculates that the different requirements for SoxE genes during neural crest development may chiefly reflect their different temporal and positional expression domains, an idea that can be easily extended to the adult organism (Guth and Wegner, 2008).
During the catagen stage of hair cycling, melanocytes in the hair matrix undergo apoptosis and are replaced during anagen by premature melanocytes that originate from dividing MSCs. How MSCs initiate a lineage-specific gene program while still maintaining an undifferentiated state is not completely understood. Lang and colleagues provide evidence for a model where the transcription factor PAX3, in concert with SOX10, plays a dual role (Lang et al., 2005). First, PAX3 and SOX10 initiate the melanogenic cascade through the activation of MITF. Then, PAX3 and LEF/TCF factors prevent terminal differentiation by blocking MITF binding sites on the Dct promoter. Accordingly, MITF would accumulate in MSCs and act as a ‘biological capacitor,’ ready to initiate rapid melanocyte-specific gene transcription once PAX3 repression is relieved. However, independent analysis of the molecular expression profile of MSCs in comparison to melanoblasts and melanocytes suggests that MSCs downregulate a number of key factors, including SOX10, MITF and LEF1, that are required for this model to be applicable to MSCs. Specifically, over half of the melanoblasts that initially colonize the lower permanent portion of the hair follicle, where MSCs reside, lose SOX10 expression in the mouse by P2 (Osawa et al., 2005). Furthermore, in the adult mouse during the catagen stage of the hair cycle, when the existing hair regresses and only MSCs survive, no expression of SOX10 is detected in the remaining follicular melanocytes (Sharov et al., 2005). Perhaps one explanation for this discrepancy is that these melanocyte-critical factors are maintained in MSCs at relatively low levels or fluctuate dependent on the stage of the hair cycle. Considering a recent study demonstrating how two different Sox10 gene dosage levels are used to maintain neuronal lineage potential while inhibiting terminal differentiation (Kim et al., 2003), it will be interesting to see whether and how SOX proteins might play similar roles in the MSC. In light of SOX9’s upregulation in human melanocytes upon differentiation it will also be necessary to assess changes in SOX expression during hair cycling. Currently there is little histological expression data for SOX9 in hair follicles of humans and mice, and the expression data for SOX10 in the human hair matrix appears contradictory (Choi et al., 2008; Commo et al., 2004; Krahl and Sellheyer, 2009).
SOX proteins and the regulation of melanogenesis
Our mechanistic understanding of how SOX proteins function as transcriptional regulators of melanogenesis primarily comes from studies of SOX10. Despite only a decade of research we have amassed a considerable amount of information on the targets of SOX10, how SOX10 interacts with other transcription factors, and the regulation of SOX10 itself. From here we may be able to gain insight into how other SOX proteins may participate in melanocyte development and maintenance.
Additional direct targets of SOX10, as determined by in-vitro transcriptional activation assays, are Dct, Tyr, and Tyrp1– all melanogenic enzymes that are essential for the production of melanin (Figure 3). For Dct, multiple studies demonstrate that SOX10 directly activates its transcription in synergy with MITF (Jiao et al., 2004; Ludwig et al., 2004). Six potential SOX10 binding sites reside within the Dct promoter, both monomeric and dimeric, however, conflicting data exist as to which might be required in melanocytes and melanoma cells (Jiao et al., 2004; Ludwig et al., 2004). While these different results may simply reflect the cell types used for in vitro assays, it may also hint at regulatory plasticity enabled by SOX10’s ability to interact with a variety of binding partners. In vivo evidence to support the idea that Dct is a direct target of SOX10 also remains somewhat unclear. Analysis of Sox10Dom/+ mice demonstrates that haploinsufficiency for SOX10 results in transiently low Dct transcript levels from E11.5 to E12.5. This is not observed in Mitfmi/+ stage-matched embryos suggesting that the reduction in Dct expression seen at this stage is due to the direct control of Dct by SOX10 rather than indirectly through SOX10-dependent changes in Mitf expression (Potterf et al., 2001). However, primary neural crest cultures derived from Mitfmi-ew/mi-ew mice that are also overexpressing Sox10 do not exhibit detectable levels of DCT, indicating that SOX10 alone is not sufficient for Dct transcriptional activation (Hou et al., 2006). Nevertheless, if SOX10 does regulate Dct during early melanoblast development, its role is likely non-essential, as Dct expression is quickly recovered and the ventral white spotting observed in Sox10Dom/+ adult mice can be attributed to an earlier reduction in overall melanoblast numbers (Potterf et al., 2001). Thus, the biological purpose for SOX10’s early regulation of Dct in the mouse melanocyte lineage requires further investigation.
SOX10 transcriptional activation of Dct as well as that of Mitf can be modulated by the competitive binding of SOX5 to the Dct and Mitf promoters (Stolt et al., 2008). This was discovered when the melanoblast loss caused by Sox10 haploinsufficiency was partially rescued by the concomitant loss of Sox5. It was subsequently shown that the long isoform of SOX5 recruits the transcriptional co-repressors CtBP2 and HDAC1 to the promoters of Mitf and Dct, thus both occupying potential SOX10 transcriptional activation sites and bringing about direct transcriptional repression. It is interesting to note that SOX5 is expressed in mouse melanoblasts at the precise timepoint, E10.5–E12.5, when DCT is transiently reduced in Sox10Dom/+ mice. These data point to a sensitive modulatory role for SOX5 in regulating SOX10 transcriptional activity.
For both Tyr and Tyrp1, SOX10 can activate their transcription via direct binding to evolutionally conserved distal enhancers located similar distances – approximately 15 kb – from the transcription start sites of each gene (Murisier et al., 2006, 2007). In the Tyr enhancer, two conserved SOX10 binding motifs, along with two MITF binding motifs, are required for Tyr enhancer activity (Murisier et al., 2007). In the Tyrp1 enhancer, three conserved SOX10 binding motifs are present, one of which is essential for SOX10-mediated transcriptional activation (Murisier et al., 2006). Interestingly, lineage-directed gene transfer of Mitf into NCCs lacking functional SOX10 reveals that while SOX10 is required for Tyr expression in mammalian cells, it is not for Tyrp1 (Hou et al., 2006). Thus, although SOX10 can regulate both Tyr and Tyrp1 expression via structurally similar enhancers, as shown in vitro, it appears that their absolute requirement for SOX10 differs.
Taken together, the current data on SOX10 transcriptional regulation indicates that SOX10 target activation may be temporally specified, as in Dct regulation, and may utilize a number of coactivators, such as PAX3. In addition, the dependency of SOX10-MITF synergy, observed at the Dct promoter, on the presence of the SOX10 C-terminal transactivation domain, but not on a direct interaction of these two proteins, points to a higher order of transcriptional organization (Ludwig et al., 2004). These examples correlate with the model proposed for SOX protein family members, that they appear to function as a part of a complex, in which the SOX protein interacts with another transcription factor whose binding site is nearby, and unique SOX protein-binding partner combinations confer specificity of transcriptional target activation (Kamachi et al., 2000; Kondoh and Kamachi, 2009).
Further examples of the complexity of SOX10 protein–protein interactions are seen in non-melanocyte expressed genes that are regulated by SOX10. Although PAX3 and SOX10 physically interact and synergistically activate both the MITF and cRET promoters (important in enteric nervous system development), SOX10 and PAX3 each independently bind the MITF promoter, while at the cRET promoter, initial protein–protein interactions between SOX10 and PAX3 facilitate promoter binding. This provides a molecular explanation for a specific class of human SOX10 DNA binding domain mutations in Yemenite Deaf Blind Syndrome that affect melanocyte development but not enteric ganglia development; in these cases, PAX3 successfully recruits these mutant SOX10 proteins to the cRET promoter (Lang and Epstein, 2003). SOX10 has also been shown to synergistically interact with Schwann cell-expressed transcription factors POU3F1 (OCT6), POU3F2 (BRN2) and EGR2 (KROX-20) (Kuhlbrodt et al., 1998). Specifically, SOX10 synergizes with the two POU factors to activate EGR2, and then SOX10–EGR2 together activates further downstream genes (Bondurand et al., 2001; Ghislain and Charnay, 2006; LeBlanc et al., 2007). This SOX10-dependent transcriptional cascade in Schwann cells is comparable to the SOX10-dependent cascade in melanocytes, where SOX10 activates Mitf and then partners with MITF protein to activate a downstream gene, Dct (Kondoh and Kamachi, 2009). As an aside, human melanoma cells lacking POU3F2 expression also lack SOX10 expression, suggesting melanocytes themselves may require interactions of POU and SOX10 regulatory pathways (Cook et al., 2005; Smit et al., 2000).
Although much is known regarding these SOX10 interactions and targets, there is clearly more to be discovered. For example, abundant evidence indicates overlap between the SOX10 and EDNRB genetic pathways in melanocytes, but to date a direct interaction of SOX10 with the EDNRB promoter has been observed only in human melanoma cell lines, where SOX10 can bind to the EDNRB promoter region and interact with the transcription factor SP1 (Cook et al., 2005; Southard-Smith et al., 1998; Stanchina et al., 2006; Yokoyama et al., 2006). Ednrb regulation by SOX10 may be modulatory in nature, as Sox10 overexpression in developing mouse melanoblasts was unable to rescue hypopigmentation in Ednrb mutants, and at E10.5 Ednrb expression persists in migrating NCCs that lack Sox10 (Hakami et al., 2006).
Additional factors in SOX10 pathways that regulate neural crest development are being identified using screens with N-ethyl-N-nitrosourea (ENU) mutagenesis or yeast two-hybrid analyses. ENU screens provide the means to identify both direct and indirect targets or regulators of SOX10. An ENU screen, in which Sox10tm1Weg or Sox10LacZ mice were used to sensitize offspring to alterations in neural crest-derived tissues that are dependent on SOX10 function, demonstrated the importance of GLI3 in melanoblast function and ERBB3 in the development of cranial and sympathetic ganglia (Buac et al., 2008; Matera et al., 2008). Yeast two-hybrid analyses revealed that the N-terminal portion, including the HMG domain of SOX10, directly interacts with many proteins that may modulate SOX10 transcriptional activity during development, including the following: REB, OLIG2, JUN, C/EBPalpha, KROX-20, SP1, and PAX3 (confirmed by co-immunoprecipitation); PAX6, MEOX1, HIVEP1, DLX5, HHEX, ALX4, HOXA3, BRN1, UTF1 and the mitochondrial outer membrane protein ARMCX3 (Mou et al., 2009; Wissmuller et al., 2006). These ongoing studies will greatly expand our understanding of the downstream pathways of SOX10, and in the case of ENU mutagenesis, also have the potential to identify upstream regulatory proteins that direct expression of SOX10 itself.
A theory has been proposed for MITF that it acts as a molecular rheostat, whereby varying MITF levels provide a precise biochemical mechanism to regulate opposing cellular functions depending on the level of MITF expression (Carreira et al., 2006). Potentially, SOX10 functions in a similar manner. SOX10 expression in the NC correlates with maintenance of undifferentiated cell properties, suggesting SOX10 levels regulate NC pluripotency (Kim et al., 2003). SOX10 clearly has a role in the specification of melanoblasts, as described above, yet is downregulated (as demonstrated in human cell culture and zebrafish) as melanocytes begin to express markers of differentiation (Cook et al., 2005; Dutton et al., 2001). During hair cycling, SOX10 expression is also diminished in the MSC population (Osawa et al., 2005). Because SOX10 expression is most often found in melanocytes that are migrating from the hair bulb to the hair matrix and are positive for markers of proliferation, it is tempting to postulate that high levels of SOX10 are required by melanocytes to remain proliferative and migratory (Belmadani et al., 2009; Osawa et al., 2005). Conversely, lower SOX10 expression levels may promote stem cell maintenance as well as allow differentiation to progress. The modulation of SOX10 expression levels that has been observed at different stages of the melanocyte life cycle suggests that close regulation of SOX10 levels is crucial for SOX10 function.
The details of Sox10 regulation are just beginning to emerge, and are of great interest given SOX10’s central role in melanocyte development and function. Using the tools of mutant/transgenic mice and zebrafish along with comparative sequence analyses, the complex upstream regulatory regions that control Sox10 gene expression are being discovered (Figure 4). Four studies revealed 14 multiple-species conserved sequences (MCS) that display high levels of evolutionary conservation and variable control of Sox10 expression (Antonellis et al., 2006, 2008; Deal et al., 2006). Five MCSs target expression to developing melanocytes (MCS4, -5, -7, -9, and -1c). In both zebrafish transgenics and in transgenic mice harboring LacZ under the control of each of these regions, MCS4 and MCS7 direct expression that coincides with most of the endogenous Sox10 expression patterns (Figure 4; Antonellis et al., 2008). MCS4 and MCS7 contain dimeric, head-to-head SoxE family binding sites that are functionally significant to the enhancer capabilities of these regions. Identification and analysis of functional dimeric SoxE binding sites upstream of zebrafish sox10 support the relevance of dimeric SoxE binding sites for enhancer activity (Antonellis et al., 2008). Analysis of zebrafish sox10 also identified regions in intron 1 that contain binding sites for Lef1, Sox9b, Notch, and β-Catenin, and are conserved among zebrafish and mammalian species. These four proteins are able to activate transcription of reporter constructs under the control of this putative sox10 intron 1 regulatory region (Dutton et al., 2008).
Data regarding additional factors that bind these SOX10 enhancer regions is emerging. For example, gel shift assays showed that mouse Sox10 enhancers can be bound by SOX9, SOX10, PAX3, AP2, LEF, however, these have not yet been functionally tested for transcriptional activation. Autoregulation has been shown for SOX10 in Schwannoma cells (Lee et al., 2008), so it is intriguing to consider this possibility for melanocytes as well. SOX9 does not appear to be one of the SOX10 regulatory proteins in mature melanocytes, as neither silencing nor overexpression of SOX9 in normal human epidermal melanocytes has any affect on SOX10 protein levels (Passeron et al., 2007). Recently, Sox10 expression was shown to be directly activated in immortalized mammary gland epithelial cells by the TRAP/Drip/Mediator complex, which includes Mediator complex subunit 1 (MED1) and functions to activate gene transcription. MED1 is recruited to the Sox10 promoter at MCS4 and MCS7, and knockdown of MED1 expression completely ablates Sox10 expression in this cell line (Zhu et al., 2009b). Further evidence, such as chromatin immunoprecipitation mapping, will be needed to identify all of the proteins that regulate Sox10 expression via these many enhancer regions, and to determine how these conserved enhancers in the Sox10 promoter orchestrate the temporal and cell-specific regulation of Sox10 expression.
Further modes of Sox10 regulation
In addition to the transcriptional modulation described above, interacting proteins modify SOX10 function or stability. Sumoylation, the attachment of a SUMO polypeptide to specific lysine residues that results in varied protein activity, stability, or localization, has been shown to occur for SOX10 in mammalian cells (Girard and Goossens, 2006) and in xenopus (Taylor and LaBonne, 2005). In vitro assays showed that mutation of any of the three potential SUMOylation sites of SOX10 (Figure 1, Table 1) results in increased transcription of MITF and increased SOX10-PAX3 synergy at MITF, suggesting that SUMOylation serves to inhibit SOX10 transcriptional activity (Girard and Goossens, 2006).
The regulation of SOX10 subcellular localization appears critical, and the presence of both nuclear import and nuclear export signals (NES) in many SOX proteins, including SRY and all three SOXE proteins, suggests this may be a conserved means of SOX protein regulation (Smith and Koopman, 2004). Because these import and export signals occur in the HMG domain, any SOX10 mutations within these signals, such as the human Ser135Thr mutation or the zebrafish m618 mutation (Figure 1, Table 1), also result in reduced DNA binding, so defects of sequestering/shuttling cannot be discerned apart from DNA binding defects in a whole organism (Dutton et al., 2001). However, in vitro experiments clearly demonstrate that the NES signal within the HMG domain of SOX10 acts to promote active shuttling of SOX10 between the nucleus and the cytoplasm, but non-intuitively, mutation of this NES actually reduces SOX10 transcriptional activation (Rehberg et al., 2002). This suggests that nuclear-cytoplasmic shuttling serves a more complex function than simply downregulating SOX10 activity by removing it from the nucleus. This hypothesis is supported by the recent description of the mitochondrial-associated protein ARMCX3 interacting with SOX10, and by transient transfection data suggesting that this interaction results in increased SOX10 transcriptional activity (Mou et al., 2009).
Sub-cellular localization affects the function of other SOX proteins, including SOX2, SRY and SOX9 (Gontan et al., 2009; Kiefer, 2007; Smith and Koopman, 2004). In the case of SOX9, temporally specific nuclear-cytoplasmic shuttling of SOX9 is observed during sexual differentiation, and also during chondrocyte differentiation (de Santa Barbara et al., 2000; Gasca et al., 2002; Morais da Silva et al., 1996; Topol et al., 2009). Interestingly, a novel mechanism for regulation of Wnt signaling by SOX9 in chondrocytes was recently discovered, in which SOX9 bound to and translocated a ‘destruction complex’ to the cytoplasm that bound to and degraded beta-catenin (Topol et al., 2009). Transfection experiments suggest SOX10 can also repress the canonical Wnt signaling pathway (Sinner et al., 2007), although it is not known whether SOX9 or SOX10 use this mechanism to affect neural crest and/or melanocyte development in vivo. In light of the above evidence suggesting that the subcellular localization of SOX10 may play a role in its activity, and that both SOX10 and beta-catenin participate in Mitf activation, it would be intriguing to determine whether SOX10-dependent degradation of beta-catenin provides a higher order of Mitf regulation.
SOX proteins in melanoma
As described above, much of what we currently understand of SOX gene function in the melanocyte cell lineage comes from developmental studies. Given that tumor cells are often found to express genes that are characteristic of their tissue of origin, it is not surprising that both SOX9 and SOX10 are expressed during various stages of melanoma progression and in established melanoma cell lines. Recent work has focused on assessing potential roles for SOX9 and SOX10 in melanomagenesis. For example, transcriptional profiling of metastasis-derived cell lines using unbiased hierarchical clustering correlated high levels of expression for SOX10, MITF and other SOX10 transcriptional target genes with a proliferative cell gene expression signature phenotype; inversely, decreased expression of these genes correlated with a more invasive melanoma gene signature profile (Hoek et al., 2008, 2006). Consistent with this pathway being important to melanoma progression, the MITF locus is amplified in ∼10% of primary cutaneous and ∼20% metastatic melanomas (Garraway et al. 2005) and suggests that MITF acts as an oncogene. Furthermore, somatic mutations of either SOX10 or MITF have been found in 7% of primary and 20% of metastatic tumors (Figure 1 and Table 1; Cronin et al., 2009). The extent to which the mutations identified in SOX10 may directly impact target gene expression and tumor progression in vivo remains to be assessed. It is interesting to note that the six SOX10 somatic, primary lesion mutations identified by Cronin et al. were derived from mucosal, lentigo malignant, and superficial spreading melanomas rather than cutaneous melanomas. Follow-up analysis will be required to assess if SOX10 somatic mutations are more likely to be identified within these histologic melanoma subtypes.
Although there is no evidence to date for somatic mutations occurring in SOX9, increased SOX9 expression reduces the proliferation of multiple melanoma cell lines in both in vitro and in vivo assays (Passeron et al., 2009). As mentioned above, UV exposure, a known contributor to melanoma progression, upregulates SOX9 expression and increases nuclear localization of SOX9 via a cAMP/PKC-dependent pathway. This in turn leads to an increase in activation of MITF and DCT that is independent of SOX10 (Passeron et al., 2007). Protein kinase C (PKC)-dependent shuttling of SOX9 is consistent with observations made in Sertoli-like testicular carcinoma cells, where protein kinase A (PKA) phosphorylation of SOX9 at Ser64 and Ser181 mediates its nuclear translocation (Malki et al., 2005).
Although it is appreciated that both SOX9 and SOX10 are expressed in melanoma-derived cell lines, only recently have these proteins been evaluated directly in melanoma tissues (Table 2). Recent immunohistochemical analysis has focused on evaluating the temporal and positional expression patterns for SOX9 and SOX10 in melanoma samples and determining whether this correlates with other diagnostic markers and/or disease state. For example, analysis of SOX10 expression in sentinel lymph nodes shows that SOX10 expression is comparable to the commonly used metastatic melanoma diagnostic markers S100, melan-a and HMB-45 (Blochin and Nonaka, 2009). SOX10-positive cells are present in 31% of melanocytic nevi, 43% of primary melanoma and 50% of metastatic melanoma (Bakos et al., 2009; Passeron et al., 2009). Conversely, SOX9 expression is observed in a high percentage of samples (75–100% and 83.9% by individual studies), decreasing moderately as the severity of melanoma progresses (Passeron et al., 2009; Rao et al., 2010). Direct comparison of SOX10 and SOX9 expression showed statistically significant co-expression between SOX9 and SOX10 in primary metastatic lesions, along with their common downstream target gene Nestin (Bakos et al., 2009). This co-expression is notable as both SOX9 and SOX10 can synergistically regulate the expression of Nestin, which itself is considered a ‘stem cell’ marker and is implicated in proliferation and migration in cells (Dahlstrand et al., 1992a,b; Flammiger et al., 2009).
Table 2. Expression of SOX9 and SOX10 proteins in human tissue samples
aVariations in the intensity of SOX9 staining have been observed when using antibodies from different commercial sources. The specific antibody used to produce the results listed is noted in parentheses.
SOX9+ cells colocalize with MART1+ cells in adult dorsal skin, SOX9 increases upon UVB exposure
Closer evaluation of the cellular expression patterns of SOX9 and SOX10 in melanoma sections by immunohistochemistry reveals distinct subcellular localizations for each protein. In both primary lesions and metastatic melanoma cells, SOX9 is detected in the nucleus and perinuclear area, whereas SOX10 is perinuclear and cytoplasmic (Bakos et al., 2009; Passeron et al., 2009; Rao et al., 2010). Although the mechanism for SOX9 localization to the nucleus appears to be regulated through PKA signaling (Malki et al., 2005), it is unclear if SOX10 can also be regulated through PKA-induced phosphorylation. Sequence alignment of SOX10 and SOX9 demonstrates that these proteins exhibit only 59% amino acid sequence identity, and neither of the SOX9 PKA sites are conserved in SOX10 (Pusch et al., 1998). However, there is evidence that the subcellular localization of SOX10 is controlled in melanoma. The tyrosine kinase TYRO3 can mediate the translocation of SOX10 to the nucleus, resulting in a subsequent SOX10-dependent increase in the M isoform of MITF. High levels of TYRO3 correlate with the cancerous transformation of primary melanocytes, as well as melanoma proliferation, survival and tumorigenicity (Zhu et al., 2009a). Whether TYRO3 mediates this effect through direct phosphorylation of SOX10 or indirectly through the activation of other downstream signaling components is not known. In addition, it remains to be determined whether SOX9 nuclear localization can also be mediated through TYRO3 signaling.
The extent to which SOX9 and SOX10 contribute to melanoma progression is only beginning to be addressed. Whether they function independently or in combination during melanocyte transformation and metastasis is unclear. These initial studies suggest that higher SOX10 expression is correlated with a proliferative and transformative cell phenotype, while high levels of SOX9 result in a less proliferative, more differentiated cell phenotype. This suggests a possible antagonistic relationship between these two SOX proteins. Since both SOX9 and SOX10 are present in a range of melanoma tissues, it will be imperative to take a closer look at how altering the levels of these proteins, either at the transcriptional or post-translational level, affects various cellular processes. Additionally, our overall understanding of SOX proteins in melanoma will benefit from the further characterization of SOX9 and SOX10 expression and function during the melanocyte life cycle.
Correction added after online publication June 2010: Figure 2 has been replaced as the original figure had incorrect text labels for the three layers of the skin.
This review was supported by the Intramural Research Program of the National Human Genome Research Institute, National Institutes of Health.