The making of a melanocyte: the specification of melanoblasts from the neural crest


Aaron Thomas, e-mail:


Melanocytes differentiate from the neural crest (NC), which is a transient population of cells that delaminates from the neural tube and migrates extensively throughout the embryo during vertebrate development. Melanoblast specification from NC precursors is a progressive process during which initially pluripotent cells become restricted to the melanogenic lineage and adopt the gene expression profile and morphology of melanocytes. This specification process is governed primarily by Wnt and BMP signaling molecules, although other signaling pathways, such as those activated by Kit and Endothelin 3, can also stimulate melanogenesis. The transcriptional repressor FoxD3 occupies a central role in melanocyte fate determination by repressing melanogenesis in premigratory NC cells and in other NC lineages.


With the exception of the retinal pigment epithelium (RPE) of the eye, all other pigment cells in vertebrates are derived from the neural crest (NC). The NC is a transient population of cells, unique to vertebrates, that arises from the dorsal neural tube or neural folds early during embryonic development. NC cells then migrate along defined pathways to specific sites in the embryo, where they complete the differentiation process to form bone, cartilage and adipose tissue, endocrine cells, several types of neurons and glia as well as pigment cells, depending on axial level (Le Douarin and Kalcheim, 1999).

NC cells from the axial level of the trunk differentiate primarily into neurons, glia, or melanocytes. When avian trunk neural tubes are explanted into culture, the first cells to emigrate from the neural tube differentiate almost exclusively into neurons and glia, whereas melanoblasts (melanocyte precursors) do not migrate until 6–12 h later (Henion and Weston, 1997; Reedy et al., 1998). These two populations of cells (neurogenic and melanogenic) have distinct migratory properties (Figure 1). Neurogenic NC cells migrate ventrally along the neural tube and then through the somites to sites where they will form ganglia. Melanoblasts migrate dorsolaterally (a few melanoblasts are occasionally seen in the ventral pathway), between the somites and ectoderm, and later invade the ectoderm, where they differentiate into melanocytes (Le Douarin and Kalcheim, 1999). The choice of pathway is determined by NC specification. In the chick, neurogenic NC cells are blocked from entering the dorsolateral pathway and cannot migrate in the dorsolateral pathway regardless of the developmental stage or axial level (Erickson and Goins, 1995). Conversely, melanoblasts isolated from NC cultures will immediately enter the dorsolateral migratory pathway when grafted into younger embryos in which dorsolateral migration has not yet begun. This difference in migratory behavior is determined by the expression of signaling molecules and receptors on the surface (reviewed in Harris and Erickson, 2007).

Figure 1.

 Migratory pathways of trunk NC cells and general scheme of gene expression during melanoblast specification. Neural crest cells migrate along the ventral (blue) and dorsolateral (red) migratory pathways. Pathway choice is determined by fate specification as discussed in the text. Premigratory NC cells in the dorsal neural tube express FoxD3, Sox10, and Pax3. Sox10 and Pax3 continue to be expressed during the delamination and migration of melanoblasts. [*In the mouse, Kit is reported to be upregulated prior to melanoblast migration or in the migration staging area (Manova and Bachvarova, 1991; Wilson et al., 2004)]. FoxD3 is downregulated and Mitf is upregulated in melanoblasts when they exit the neural tube. Mitf expression continues throughout melanoblast migration. As melanoblasts migrate, other genes are upregulated at different time points. In birds and mammals, tyrosinase is upregulated around the time of ectoderm invasion. Zebrafish melanophores upregulate tyrosinase much earlier. In birds and mammals, FoxD3 +  neuroblasts and glioblasts migrate in the ventral migratory pathway. Zebrafish and Xenopus melanophores are observed in both migratory pathways. nt, neural tube; no, notochord.

Thus, melanoblasts comprise a distinct subset of cells in the heterogeneous NC population. They are unique in their migratory behavior and in their gene expression profile, and these differences become apparent almost immediately after emigration from the neural tube. In this review, we will discuss the events that govern the specification of melanoblasts from the NC.

Specification is a process by which undifferentiated cells begin to express markers and exhibit behavior unique to one of its potential derivatives. These changes are accompanied by restrictions in developmental fate such that they generate only a subset of the potential progeny of the original cell. Specification is distinct from commitment in that specification is an operational term and refers to what a cell will differentiate into, whereas commitment refers to a cell’s potential when tested in alternate environments. For example, a specified melanoblast has begun expressing melanoblast-specific markers and will differentiate into a melanocyte under normal embryonic conditions; but if exposed to certain experimental conditions, it can be made to differentiate into another NC derivative. Such a melanoblast is specified, but not committed. This means that fate maps indicate fate restriction, but not necessarily fate potential. For example, a cell fated to become a melanoblast may retain the potential to differentiate into glia under other circumstances.

The timing of melanoblast migration has been studied in detail in several species. To assist the reader’s understanding of the events discussed below and how that relates to melanoblast migration, we summarize these events for mouse, chick, and zebrafish embryos. Mouse trunk NC cells delaminate from the neural tube and begin migration at E9–E9.5. Cells enter both migratory pathways simultaneously, although melanoblast markers are not upregulated until E10.5–E11 in the dorsolateral migratory pathway (Baxter and Pavan, 2002, 2003; Nakayama et al., 1998; Opdecamp et al., 1998). Whether those cells that migrate dorsolaterally before expression of melanoblast markers differentiate into melanocytes is not known. Mouse melanoblasts begin populating the dermis and epidermis and producing pigment at E16.5 (Serbedzija et al., 1990). Chick NC cells begin migration in the trunk at stage 12–13 (Loring and Erickson, 1987; Tosney, 1978) and immediately enter the ventral migratory pathway. Dorsolateral migration is first observed at the level of the forelimb in stage-20 embryos, and cells first begin invading the ectoderm at about stage 22 (Erickson et al., 1992). Melanin production is reported to begin as early as day 5 (about stage 27) or as late as day 13 in chick embryos (Hulley et al., 1991). Before entering the dorsolateral pathway, melanoblasts appear to accumulate in the Migration Staging Area (MSA), an acellular extracellular matrix-rich wedge between the somite and the dorsal neural tube (Weston, 1991), as though waiting for some cue to enter the dorsolateral pathway. In zebrafish embryos, NC cells enter the ventral migratory pathway at 16–20 h post-fertilization (hpf), depending on axial level. Dorsolateral migration begins approximately 4 h later (Raible et al., 1992) and pigmentation is first seen at about 24 hpf. Zebrafish melanoblasts can migrate in both migratory pathways, although neural and glial cells are never observed migrating dorsolaterally (Camp and Lardelli, 2001).

Clonal NC studies

The question of when a NC cell is specified as a melanoblast is not easy to answer. The question has been addressed experimentally with clonal studies (using quail or chick NC cells) in which individual nascent NC cells are followed as they proliferate and differentiate. These studies give us perhaps the best understanding of when a NC cell has begun down the pathway to the melanocyte phenotype. In the first published report of limited-dilution clonal analysis of the NC, NC cells from Japanese quail neural tubes were seeded at clonal density 48 h after the neural tubes were dissected and cultured (Cohen and Konigsberg, 1975). After differentiation, approximately 5% of colonies were of mixed cell type, and the rest were either entirely pigmented or non-pigmented, indicating that by 48 h after emigration from the neural tube, most melanoblasts were specified (Cohen and Konigsberg, 1975). In a later study (Sieber-Blum and Cohen, 1980), nascent NC cells were dissociated and plated at clonal density after only 18 h in culture. More than 50% of the resulting clones were composed of melanocytes only, demonstrating that more than 50% of melanoblasts are specified within 18 h of emigrating from the neural tube. When quail cranial NC cells were cultured on a feeder layer of 3T3 fibroblasts by microscope-controlled single-cell seeding, no colonies of pure melanocytes developed (Baroffio et al., 1988). However, in this study, the NC was dissected from the neural tube before delamination and dissociated into single cells before plating.

These early in vitro clonal studies yielded conflicting results concerning the question of melanoblast specification, which can be explained depending on when the cells were cloned relative to the timing of their migration from the neural epithelium. More recently, Henion and Weston (1997) further refined the timing of melanoblast specification. The authors dissected quail neural tubes, placed the tubes in culture, then labeled a single cell with lysinated rhodamine dextran (LRD) at defined time points after plating. They found that during the first 6 h, most NC cells are either bipotent neural/glial progenitors or specified neuroblasts or glioblasts. Specified melanoblasts do not migrate during the first 6 h. Only about 10% are capable of generating melanocytes and most of those are bipotent glial/melanogenic precursors (Henion and Weston, 1997). It was subsequently determined that those neural/glial precursors can be identified within the first 6 h by the expression of TrkC (Luo et al., 2003). Henion and Weston (1997) also discovered that the cells that migrate at later time points consist of many more melanogenic cells. Some of those cells are glial/melanogenic bipotent precursors, but most are specified melanoblasts. In addition, over half of the early migrating NC cells are pluripotent progenitors capable of giving rise to multiple lineages, whereas 87% of cells labeled 30–36 h after delamination are unipotent (Henion and Weston, 1997). This is consistent with work from our lab in which we discovered that in avian embryos, melanocyte precursors emigrate later than neural/glial precursors, and that only melanoblasts can exploit the dorsolateral migratory pathway (Erickson and Goins, 1995; Reedy et al., 1998). With regard to melanoblast specification, in vitro clonal analysis of NC specification tells us that early migrating NC cells in avian embryos do not have melanogenic potential. Furthermore, almost all NC cells are lineage restricted when they delaminate from the neural tube; some are bipotent and many are unipotent.

In vivo clonal studies of the NC have yielded similar results to in vitro studies. When individual chick neural tube cells were injected with LRD, labeled cells of every NC derivative were detected in at least some of the resulting clones (Bronner-Fraser and Fraser, 1988, 1989). Some clones were comprised of pigment cells only, whereas other labeled cells gave rise to multiple derivatives. When premigratory NC cells were labeled in the zebrafish, few pluripotent cells were identified (Raible and Eisen, 1994; Schilling and Kimmel, 1994). Instead, most clones were composed of a single cell type, and the few pluripotent NC cells that were observed migrated early and divided into specified progeny early during migration (Raible and Eisen, 1994).

Even after specification and overt pigmentation, the phenotype of melanocytes is not completely stable. A recent study found that only 1–2% of cultured mouse NC cells that express melanoblast markers on day 4 of culture differentiate into melanocytes after 2 weeks (Dunn et al., 2005). When melanocytes are isolated from quail epidermis and placed in culture, they will, over several days, downregulate expression of melanocyte-specific genes and some will differentiate into other NC derivatives (Dupin et al., 2000; Real et al., 2006; Richardson and Sieber-Blum, 1993). Endothelin 3 (Edn3) added to the culture medium dramatically enhances and hastens this phenotypic switch (Dupin et al., 2000; Real et al., 2006). Schwann cells isolated from avian embryos are likewise induced by Edn3 to dedifferentiate to a bipotent precursor and differentiate into melanocytes (Dupin et al., 2003; Real et al., 2005). After several days in culture with Edn3, melanocytes begin expressing genes indicative of early, pluripotent NC cells (Real et al., 2006). Based on these data, it appears that Edn3 causes the reversion of both melanocytes and Schwann cells to a biopotent precursor of both cell types. Although Edn3 causes the dedifferentiation of Schwann cells and melanocytes, it is unclear what drives their redifferentiation into the other cell type in the continued presence of Edn3. There is yet no ready explanation of this phenomenon. It is clear, however, that otherwise fully differentiated Schwann cells and melanocytes retain some ability, under certain experimental conditions, to revert to a more immature phenotype.

There is some disagreement regarding how long this plasticity lasts during embryogenesis. In some studies, epidermal melanocytes from E7.5 quail embryos were routinely used and they displayed this ability to revert to a glial-melanogenic bipotent precursor and differentiate along other lines (Dupin et al., 2000; Real et al., 2006). However, another study found that by E6, epidermal quail melanocytes were fully committed to the melanogenic lineage and would not regress to a more primitive cell type (Richardson and Sieber-Blum, 1993). This disparity is not fully understood, but may be explained by different culture conditions. In the studies performed in the Dupin lab, cells were cultured in medium containing 2% embryo extract and numerous growth hormones were added to the medium (Dupin et al., 2000; Lahav et al., 1998; Ziller et al., 1983). In contrast, cells in the Richardson study were cultured with 10% embryo extract and no added growth hormones (Richardson and Sieber-Blum, 1993). The role of the undefined embryo extract and added hormones is not known. In spite of the disagreement about how long melanocytes remain phenotypically plastic, we can conclude that melanocytes retain some ability to revert to an uncommitted precursor and that reversion is enhanced by Edn3. However, whether those phenotypically plastic melanocytes in the ectoderm would actually differentiate into anything other than melanocytes in vivo is unknown.

It is not fully understood how NC cells are recruited to the various lineages, given that they originate from the same location in the embryo. In the zebrafish, NC cells appear to be specified according to their position in the neural tube, with the most medial cells developing into neural cells and pigment cell precursors being located lateral to neural precursors (Schilling and Kimmel, 1994). In the chick, the timing of signaling events is likely the deciding factor (see below). However, this does not adequately explain how individual cells make cell fate choices. At times, differently fate-specified cells are emigrating from the neural tube simultaneously. Stochastic events or localized environmental differences may also play important roles in the specification of individual NC cells. Much work remains to be done before a clear picture of NC specification will emerge.

Progressive specification of the NC

These clonal studies tell us that, other than the occasional pluripotent cell, the early migrating NC do not have melanogenic potential. Furthermore, the specification of melanoblasts does not happen at once, but rather in steps. This was best demonstrated in zebrafish NC cells by Raible and Eisen (1994). They labeled individual NC cells in zebrafish embryos, and then followed those cells at frequent time intervals to determine when they made cell fate decisions. They were able to determine when individual cells made fate decisions and discovered that initially pluripotent cells generated fate-restricted progeny within the first two cell divisions (Raible and Eisen, 1994). Fate restriction occurs in a hierarchical fashion (Figure 2) in which initially pluripotent cells divide to generate two daughter cells with different developmental potential (Le Douarin and Dupin, 2003; Weston, 1991). In the case of melanoblasts, the immediate precursor is a glial/melanogenic bipotent cell that divides to generate a specified melanoblast and a specified glial precursor. Pluripotent cells capable of generating both neurons and melanocytes also generate glia (Baroffio et al., 1988; Bronner-Fraser and Fraser, 1988, 1989; Henion and Weston, 1997; Raible and Eisen, 1994). In the zebrafish, approximately 20% of the NC cells that migrate in the first 2 h are pluripotent and the rest are neural progenitors. Subsequently, almost all are restricted to a single fate at the time they exit the neural tube (Raible et al., 1992; Schilling and Kimmel, 1994), indicating that specification usually occurs before delamination. Similar work done in the chick indicates that many more are pluripotent (Bronner-Fraser and Fraser, 1988, 1989), but those experiments labeled neural epithelial cells, not NC cells. Therefore, we can only conclude that specification does not occur, in most cases, before a NC cell is generated by the division of a neural epithelial cell. Furthermore, from in vivo and in vitro clonal studies, it is clear that most melanoblasts are specified when they exit the neural tube or shortly thereafter. It is unclear if all melanoblasts pass through the pluripotent and bipotent stages of lineage restriction or if some are already specified at the time they are born as NC cells.

Figure 2.

 Model of the progressive specification of melanoblasts in the trunk NC. Initially pluripotent progenitors are progressively fate-restricted to generate specified melanoblasts. Pluripotent cells divide to generate bipotent neural/glial and glial/melanogenic precursor cells. This glial/melanogenic precursor then divides to generate a specified melanoblast. However, it is not clear whether all melanoblasts go through all stages, or are already beyond the pluripotent precursor stage when they segregate from the neural epithelium.

Although most NC cells are specified before migration, a few remain pluripotent. Clonal studies, whether in vitro or in vivo, have occasionally identified pluripotent cells long after delamination from the neural tube. These cells are likely NC stem cells, and may persist into adulthood (reviewed in Delfino-Machin et al., 2007).

The key role of Mitf in the lineage switch

Based on the clonal analysis described above, it appears that at some point in the life of a NC cell, it makes the switch from a bipotent glial-melanogenic cell to a specified melanoblast (see Figure 3 for a summary of the factors known to regulate this fate switch). As this switch is thrown, the cell’s gene expression profile must necessarily change to determine the eventual fate. So, what are the changes in gene expression that occur during this switch and how are these changes controlled and coordinated? Several transcription factors play key roles in upregulating many of the genes responsible for melanogenesis. These include the Microphthalmia-associated transcription factor (Mitf), the paired domain- and homeodomain-containing transcription factor Pax3, and the Sry-related transcription factor Sox10. Loss of any of the three results in failure of melanoblast development. Sox10 and Pax3 are expressed in the dorsal neural tube before the first NC cells delaminate and are important for other NC lineages as well as melanocytes. For example, Pax3 mutant mice exhibit defects in the neural tube and other NC derivatives (Moase and Trasler, 1992), and Sox10 mutants often have megacolon because of a deficiency of enteric NC cells (Tachibana et al., 2003).

Figure 3.

 Regulatory interactions controlling specification of the NC. Numerous interactions among various transcription factors and signaling molecules control the specification of melanoblasts in the NC, as discussed in the text. The lineage switch from a neurogenic fate to melanogenic fate is governed by numerous signals interacting at the level of FoxD3 and Mitf.

Although Pax3 and Sox10 are clearly necessary for melanoblast specification, it is Mitf that is often referred to as the master regulator of melanogenesis (not to suggest that Mitf is solely responsible for the melanogenic phenotype, of course). Mitf is a basic-helix-loop-helix-leucine-zipper transcription factor expressed in numerous tissues. In avian and mammalian species, the Mitf gene consists of common exons 2–9 and numerous alternative first exons, each with a unique promoter for expression in different tissues (reviewed in Shibahara et al., 2001; Steingrimsson et al., 2004). For the remainder of this review, we will only discuss Mitf-M, produced from the 3′-most first exon, 1M, which is expressed exclusively in and required for development of NC-derived melanocytes. Mitf is expressed in melanoblasts soon after emigration from the neural tube and well before they enter the dorsolateral migratory pathway, coincident with lineage specification (Kumasaka et al., 2004; Lister et al., 1999; Nakayama et al., 1998; Opdecamp et al., 1997; A. J. Thomas and C. A. Erickson, unpublished data). In the trunk of chick embryos (Figure 4), a few Mitf+ cells are seen as early as stage 15, well before the onset of dorsolateral migration (Erickson et al., 1992). Mitf+ cells accumulate in the MSA until stage 19 at the forelimb axial level when they begin to migrate over the somite. By stage 21, a stream of individual Mitf+ cells extend from the dorsal midline of the neural tube into the dorsolateral migratory pathway (A. J. Thomas and C. A. Erickson, unpublished data). Mitf is first observed in mouse melanoblasts at E10.5–E11 in cells dorsal to the neural tube and in the MSA (Nakayama et al., 1998; Opdecamp et al., 1997).

Figure 4.

 Mitf expression pattern in chick trunk NC cells. Antibodies to Mitf (green, arrowheads) and the NC marker HNK-1 (red) label NC cells of the chick trunk. White cells are autofluorescing red blood cells. Mitf expression is first seen in a few cells along the length of the neural tube as early as stage 15 (A). Such cells are always found in the migration staging area and have not entered the dorsolateral migratory pathway. By stage 19 (B), many Mitf+ cells have accumulated in the migration staging area at the axial level of the forelimb. Melanoblasts are occasionally seen in the dorsolateral pathway. By stage 21 (C), few cells are seen in the migration staging area and melanoblasts are organized as a stream of Mitf+ cells from above the dorsal neural tube into the dorsolateral pathway. Scale bar = 10 μm.

Mitf is crucial for the specification and survival of melanoblasts. Loss of Mitf results in the absence of melanocytes (Lister et al., 1999; Mochii et al., 1998; Steingrimsson et al., 2004). More than 20 Mitf mutants have been identified in the mouse, all of which exhibit pigmentation defects (Steingrimsson et al., 2004). When Mitf is experimentally reduced in nascent avian melanoblasts in cell culture by short hairpin RNA (shRNA), they transdifferentiate into glia (A. J. Thomas and C. A. Erickson, unpublished data). This may happen because Mitf upregulates Tbx2, part of the family of T-box transcription factors known to function in maintenance of cell identity (Carreira et al., 2000). Misexpression of Mitf in NIH 3T3 cells causes them to adopt a melanocyte-like morphology and express melanocyte-specific genes (Tachibana et al., 1996). Misexpression of Mitf in quail neuroretinal cells (Planque et al., 1999, 2004) or medaka embryonic stem-like cells (Bejar et al., 2003) causes them to differentiate into melanocytes. In zebrafish embryos, ectopic expression of mitfa (the zebrafish melanocyte-specific Mitf homolog) induces ectopic melanin-containing cells, although these cells often differ in both morphology and location from normal melanophores (Lister et al., 1999). Mitf also affects melanocyte growth and survival by regulating Bcl2 (Mcgill et al., 2002), p21Cip1 (Carreira et al., 2005), Dia1 (Carreira et al., 2006), and INK4A (Loercher et al., 2005). It is clear that Mitf is essential at the earliest stages of melanoblast specification, and Mitf may, in fact, be the earliest gene upregulated in the specification of melanoblasts (Figure 1).

Mitf regulates the expression of many of the genes required for melanogenesis, including tyrosinase and the related genes, tyrosinase-related protein 1 (Tyrp-1), and dopachrome tautomerase (Dct, also known as tyrosinase-related protein-2). It does so by binding to E-box (CATGTG, binding site of basic helix-loop-helix transcription factors) and M-box (an 11-bp motif with an E-box at its core) (Aksan and Goding, 1998) elements conserved in their regulatory regions (reviewed in Murisier and Beermann, 2006). Of these three, Dct is the first gene expressed. It is first detectable in mouse melanoblasts at E11 (Wehrle-Haller and Weston, 1995), shortly after Mitf is upregulated. Tyrp-1 and tyrosinase expression begins later: Tyrp-1 is upregulated not long after Dct, and tyrosinase expression is delayed until shortly before pigmentation becomes evident. These three genes have different regulatory mechanisms, all involving Mitf. Other factors involved are Sox10, Pax3, Tbx2, and signaling events through the Kit and Ednrb receptors (Murisier and Beermann, 2006). In addition, Mitf regulates the expression of Pmel17/Silver (Baxter and Pavan, 2003; Du et al., 2003), MATP (Baxter and Pavan, 2002), MLANA/MART1 (Du et al., 2003), and Qnr-71 (Turque et al., 1996), and has been implicated in the expression of Kit (Tsujimura et al., 1996).

The regulation of Mitf itself has been studied in detail (see Levy et al., 2006 for a recent review). Transcription of Mitf is driven (primarily) by the transcription factors Sox10 and Pax3 (Bondurand et al., 2000; Potterf et al., 2000; Watanabe et al., 1998), both of which are expressed in the dorsal neural tube long before NC migration commences and well before Mitf is expressed (Bondurand et al., 1998; Cheng et al., 2000; Cheung and Briscoe, 2003; Kuhlbrodt et al., 1998; McKeown et al., 2005; Otto et al., 2006). Mitf is also upregulated by Wnt3a in melanocytes (Takeda et al., 2000) and growth in Wnt3a-conditioned medium induces melanoblast differentiation in cultured quail NC cells (Jin et al., 2001). Wnt signaling is important for NC induction and specification at multiple times during NC development (Raible, 2006). Wnt3a is expressed in the dorsal neural tube during NC migration, and that expression coincides approximately with the initiation of neuroblasts and glioblasts (Jin et al., 2001). Frzb-1, a secreted frizzled-related protein that competes with frizzled receptors for Wnt binding, is strongly expressed in neuroblasts and glioblasts, but at lower levels in melanoblasts (Jin et al., 2001). The increase in Wnt signaling associated with the downregulation of Frzb-1 may be important for the upregulation of Mitf. Signaling through cAMP can also upregulate Mitf expression and induce melanogenesis in NC cells (Ji and Andrisani, 2005). However, there is little evidence that this pathway is important in the initial expression of Mitf in melanoblasts, but is important for the expression of Mitf in mature melanocytes (Levy et al., 2006).

An attractive model for the regulation of melanoblast specification is that the expression of Mitf is the key event. According to his model, Mitf expression is induced in NC cells at the time melanoblasts are to be specified and Mitf, in turn, upregulates the other genes or gene cascades that confer melanocyte characteristics on the cell. However, although Mitf regulates the expression of many of the genes required for melanogenesis and appears capable of causing some cells to differentiate into melanocytes, it may not directly control the pathfinding ability of melanoblasts. When we knocked down Mitf activity in migrating NC cells by electroporating an α-Mitf shRNA construct into chick neural tubes (A. J. Thomas and C. A. Erickson, unpublished data), we observed numerous electroporated cells migrating in the dorsolateral pathway, suggesting that Mitf does not directly regulate the unique pathfinding ability of melanoblasts. It would be interesting to see if the melanocytes generated by transdifferentiation of neuroretinal cells by misexpression of Mitf (Planque et al., 2004) are capable of migrating in the dorsolateral pathway, and if so, when they achieve the ability.

FoxD3 represses melanogenesis

Because Sox10 and Pax3 are expressed in the dorsal neural tube and in migrating neuroblasts and glioblasts, and because Mitf may not directly regulate the pathfinding ability of melanoblasts, we suggest a modification to the model that Mitf regulates all aspects of specification. Another transcription factor, the forkhead box transcription factor FoxD3, lies upstream of Mitf and regulates all aspects of melanoblast character–melanogenesis and migratory capability.

FoxD3 is expressed in the neural tube before NC migration and in migrating neural and glial precursors (Dottori et al., 2001; Kos et al., 2001; Lister et al., 2006; Yamagata and Noda, 1998). In the mouse, FoxD3 is also expressed in some NC cells in the dorsolateral migratory pathway, but those cells are Mitf-negative (Dottori et al., 2001). Loss-of-function studies of FoxD3 have proven difficult in mammalian species because, in addition to its role in the NC, FoxD3 is absolutely required during early mammalian development, as FoxD3-null mouse embryos fail to form a primitive streak, undergo gastrulation, or develop mesoderm. FoxD3 is also required to maintain pluripotency in mouse embryonic stem cells (Hanna et al., 2002). This difficulty was recently overcome by Teng et al., who eliminated FoxD3 specifically in the NC by crossing FoxD3flox/− and Wnt1-Cre mice. The resulting embryos died perinatally and had a massive loss of NC derivatives (Teng et al., 2008). Three published reports detail the effects resulting from loss of FoxD3 in the zebrafish (Lister et al., 2006; Montero-Balaguer et al., 2006; Stewart et al., 2006). All three report that the NC is normally specified without FoxD3, but NC-derived structures are dramatically reduced or missing. Melanophores are not affected when FoxD3 is knocked down by injection of morpholino oligonucleotides (Lister et al., 2006). In contrast, colgate zebrafish embryos, deficient in the histone deacetylase hdac1, exhibit prolonged expression of foxd3 in the NC and have profoundly reduced number of melanophores, but other NC derivatives are normal (Ignatius et al., 2008). In chick embryos, morpholino oligonucleotide-mediated knockdown of FoxD3 in the neural tube causes NC cells to enter the dorsolateral migratory pathway prematurely and misexpression of FoxD3 prevents cells from exploiting that pathway. Likewise, loss of FoxD3 in cultured quail NC cells causes a dramatic increase in the number of pigment cells (Kos et al., 2001). Taken together, these studies suggest that FoxD3 represses melanogenesis.

FoxD3 is a transcriptional repressor (Freyaldenhoven et al., 1997; Ignatius et al., 2008; Pohl and Knochel, 2001; Sasai et al., 2001; Sutton et al., 1996; A. J. Thomas and C. A. Erickson, unpublished data; Yaklichkin et al., 2007) that directly represses Mitf (Ignatius et al., 2008; A. J. Thomas and C. A. Erickson, unpublished data) in melanoma cells and avian melanoblasts. In the zebrafish, foxd3 represses melanogenesis by repressing mitfa (Ignatius et al., 2008). Based on our work in quail and chick NC cells, FoxD3 is capable of repressing all aspects of melanoblast character, including melanogenesis and migratory behavior (Kos et al., 2001).

FoxD3 is part of a collection of transcription factors, including Msx1, Pax3, Zic1, Snail2, AP-2, and Sox10, among others, that are expressed early in the induction of the NC (Steventon et al., 2005). Because the downregulation of FoxD3 is a critical step in the specification of melanoblasts, understanding how FoxD3 is regulated should give us considerable insight into the lineage switch that occurs as NC cells are specified as melanoblasts from a glia/melanoblast bipotent precursor. Unfortunately, we know more about what regulates the initial expression of FoxD3 among this suite of transcription factors during NC induction, rather than the later event, when FoxD3 is downregulated at the time of melanoblast specification. FoxD3 expression is induced in the NC by both BMP and Wnt signaling molecules (Pohl and Knochel, 2001; Taneyhill and Bronner-Fraser, 2005), which are essential in the initial generation and later lineage specification of the NC (Raible, 2006; Yanfeng et al., 2003). BMP and Wnt signaling pathways interact at the level of three transcription factors, Msx1, Pax3, and Zic1, to induce NC formation, and therefore, FoxD3 expression. Intermediate levels of BMP signaling induce expression of Msx1, which in turn is genetically upstream of FoxD3 (Tribulo et al., 2003). Msx1 also initiates the expression of Pax3 (Monsoro-Burq et al., 2005), which is also genetically upstream of FoxD3 (Dottori et al., 2001). BMP signaling can also induce Pax3 (Taneyhill and Bronner-Fraser, 2005), although high levels of BMP repress Pax3 (Sato et al., 2005). Wnt signaling upregulates the expression of Pax3 (Bang et al., 1999) and Wnt is required in combination with Pax3 for NC generation and FoxD3 expression (Sato et al., 2005). In Xenopus embryos, the expression domains of Pax3 and Zic1 in the ectoderm partially overlap prior to neurulation and it is in this region of overlap that FoxD3 is expressed and the NC is specified (Sato et al., 2005). A model has emerged in which Wnt, Fibroblast Growth Factor (FGF), and retinoic acid (RA) signaling activities are needed in concert with intermediate levels of BMP signaling to induce the expression of Msx1, Pax3, and Zic1 transcription factors, which in turn upregulate NC-inducing transcription factors, including FoxD3 (Monsoro-Burq et al., 2005; Steventon et al., 2005).

Although these studies have given us insight into some of the mechanisms that regulate FoxD3 expression, they do not provide evidence for what factors directly regulate FoxD3 as opposed to being merely somewhere upstream and capable of altering expression. Nor do they explain how FoxD3 is downregulated in melanoblasts. Two recently published studies offer some insight into the downregulation of FoxD3 with specific regard to melanoblast specification. First, a family with an unusual vitiligo phenotype was discovered to have a single-nucleotide polymorphism in the FoxD3 promoter that results in increased expression of FoxD3 in melanoblasts and melanocytes (Alkhateeb et al., 2005). The SNP is in an evolutionary conserved region (ECR), but the transcription factor whose binding is altered by the mutation is unknown. Second, a recent study has identified the histone deacetylase hdac1 as a repressor of FoxD3 in zebrafish embryos (Ignatius et al., 2008). They report that hdac1−/− mutant embryos have pigmentation defects because of prolonged expression of foxd3 and a concomitant delay and reduction in the expression of mitfa. Histone deacetylases regulate transcription by modifying the acetylation of histones and other proteins as part of larger protein complexes and their activity on specific genes is directed by other proteins (Jenuwein and Allis, 2001; Narlikar et al., 2002; Strahl and Allis, 2000). Thus, it is not known whether hdac1 acts directly at the level of foxd3 transcription or if it is further upstream.

When we performed an in silico analysis (Ovcharenko et al., 2004) of the FoxD3 locus (Figure 5), extending upstream approximately 200 kb and downstream approximately 100 kb from FoxD3’s single intron, we discovered several ECRs containing many conserved and aligned transcription factor binding sites. Among them were binding sites for Pax3, Msx1, Smad family transcription factors (downstream of BMP signaling), and Tcf/LEF1 (downstream of Wnt signaling). While these pathways and transcription factors have been implicated in the regulation of FoxD3, it is not yet known if any of them directly regulates FoxD3. No studies examining the function of any of these regions or binding sites have been published to date.

Figure 5.

 Evolutionary conserved regions at the FoxD3 locus. The FoxD3 locus Xenopus, zebrafish, chick, mouse, dog, and Rhesus macaque genomes are compared using the ECR browser at (Ovcharenko et al., 2004). Peaks indicate areas of conservation. The colors are defined as: red—intergenic sequence, green—repetitive elements, blue—coding region, yellow—untranslated region. The transcription factors Msx1 and Pax3, and Wnt and BMP signals (which affect gene transcription at Tcf/LEF and Smad transcription factor binding sites, respectively) have been implicated in regulation of FoxD3 expression. Regions of high conservation with conserved binding sites for these factors are marked with gray boxes.

Signaling pathways that affect melanoblast specification

In addition to regulating NC induction, the Wnt signaling pathway promotes melanoblast specification in the NC. Wnt1 and Wnt3a double-knockout mice exhibit defects in several NC derivatives, including melanocytes (Ikeya et al., 1997). When the Wnt signaling pathway is targeted by knocking out β-catenin directly in NC cells, melanocytes and sensory neurons are lost (Hari et al., 2002). The loss of melanocytes is because of a change in cell fate specification, rather than a proliferation defect (Lee et al., 2004). This is consistent with data from our lab, in which quail NC cells cultured in the presence of Wnt3a results in an increase in the number of pigment cells at the expense of neural and glial cells (Jin et al., 2001). Upregulation of Wnt-controlled genes by injection of β-catenin mRNA into NC cells in zebrafish embryos directs them to a pigment cell fate (Dorsky et al., 1998; Lewis et al., 2004). This effect is thought to be mediated by the binding of LEF1 to the Mitf promoter in response to nuclear translocation of β-catenin (Dorsky et al., 2000; Takeda et al., 2000) and by Mitf directly interacting with β-catenin to alter the suite of genes regulated by β-catenin (Schepsky et al., 2006).

In addition to Wnt, other signaling pathways also influence melanoblast specification and differentiation. For example, BMP signaling acts antagonistically to Wnt signaling in the specification of melanoblasts. Culture of quail NC cells in the presence of BMP4 directs them to a neural fate at the expense of melanogenesis (Jin et al., 2001). Considering that the concentration of BMP4 molecules in the neural tube is critical for the initial expression of FoxD3 and induction of the NC, it is possible that downregulation of BMP4 is the event that leads to downregulation of FoxD3 and melanoblast specification. Consistent with this, BMP4 expression is reduced in the neural tube approximately at the time of melanoblast specification (Jin et al., 2001).

Much attention has been given to Kit in the development of melanocytes from the NC as it is clearly important for their development. Mutation of Kit or its ligand, Kitl, results in pigment deficiencies in humans and mice. In humans, a heterozygous mutation of Kit results piebaldism, characterized by white spotting in the forehead and belly (Giebel and Spritz, 1991). Melanoblasts are specified normally in KitW mutant mouse embryos, but they migrate no farther than the MSA, where they disappear soon after upregulating Dct (Wehrle-Haller and Weston, 1995). The KitW-v mutant, which encodes a dominant negative Kit allele, has reduced numbers of melanoblasts as early as E10.5 and exhibit reduced melanoblast proliferation (Mackenzie et al., 1997). Moreover, stimulation of the Kit signaling pathway increases the number of melanocytes in NC cultures (Lahav et al., 1994; Price et al., 1998).

Studies have identified periods during melanoblast specification and development when Kit signaling is critical (Hou et al., 2000, 2004; Mackenzie et al., 1997; Nishikawa et al., 1991; Yoshida et al., 1996) as opposed to other times when it is dispensable. Activation of Kit initiates a signaling cascade that results in the phosphorylation of Mitf and a concomitant increase in Mitf transcriptional activation activity (reviewed in Goding, 2000; Levy et al., 2006). Although Kit is essential for proper melanoblast differentiation, it is not required for the initial specification of melanoblasts (Hou et al., 2000).

Edn3 dramatically increases the number of melanocytes that differentiate from NC cultures (Lahav et al., 1996), and, as mentioned above, Edn3 can induce the reversion of differentiated melanocytes and Schwann cells to a bipotent precursor (Dupin et al., 2000, 2003; Real et al., 2005, 2006). Edn3 is a small (21-aa) vasoactive peptide synthesized by cleavage from a larger precursor and which binds to G-protein-coupled heptahelical receptors, endothelin receptor A or B (Ednra or Ednrb). Edn3-null mice are melanocyte deficient and Edn3 signaling is required early in the migration of melanoblasts (Baynash et al., 1994; Hosoda et al., 1994). Although Edn3 increases the number of melanocytes differentiating in NC cultures, this is because Edn3 is highly mitogenic for melanoblasts and, to a lesser extent, glial precursors (Lahav et al., 1996, 1998). Whether Edn3 affects early specification of melanoblasts is unlikely, however, because melanoblasts can be specified in the absence of signaling through Ednrb (Hou et al., 2004).

Some researchers have explored melanoblast specification and melanocyte differentiation by inducing melanocyte differentiation in embryonic stem (ES) cells. These experiments offer further insight into the signaling pathways required for the differentiation of melanocytes. Mouse ES cells can be induced to differentiate into melanocytes through a NC-like intermediate (Motohashi et al., 2007). In the process, Edn3 is absolutely required for melanoblast specification, and Edn3 can compensate for lack of Kit signaling (Aoki et al., 2005). Human ES cells can also be made to differentiate into melanocytes in medium containing Wnt3a, Edn3, Kitl, cholera toxin, the phorbol ester 12-O-tetradecanoylphorbol-13-acetate (TPA), and bFGF (Fang et al., 2006). However, as these studies begin with ES cells rather than NC cells, they may not accurately represent the requirements of NC cells in vivo.

Concluding thoughts

At this point in our discussion, there is a gap in our knowledge of how melanoblasts are specified. We know the downstream effects and how they are regulated. Mitf is upregulated almost immediately upon melanoblast specification and is responsible for expression of many of the gene products that participate in melanin biosynthesis. It is indispensable for melanoblast specification and can, in some cases, cause other cells to transdifferentiate into melanocytes. FoxD3 represses Mitf during the early phases of NC migration and generally represses melanoblast specification. We also know that Wnt and BMP signaling regulate FoxD3 expression. However, the gap between these signaling pathways and melanoblast specification remains to be elucidated. Whether changes in Wnt and/or BMP signaling cause the downregulation of FoxD3 or other signaling events are responsible is yet unknown. To understand how melanoblast specification is regulated at a molecular level, we will need to understand what signaling events are responsible for FoxD3 downregulation and then connect those events, through FoxD3, to further downstream effectors. Mitf is one of those effectors, but there are certainly others. The process is likely to be governed, at least in part, by Wnt and/or BMP signaling, and the mechanism may or may not be related to the upregulation of FoxD3 during NC induction.