Dual origin of melanocytes defined by Sox1 expression and their region-specific distribution in mammalian skin


  • Naoko Yoshimura,

    1. Department of Tissue and Organ Development, Regeneration and Advanced Medical Science, Gifu University Graduate School of Medicine, Gifu, Japan
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Tsutomu Motohashi,

    1. Department of Tissue and Organ Development, Regeneration and Advanced Medical Science, Gifu University Graduate School of Medicine, Gifu, Japan
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Hitomi Aoki,

    1. Department of Tissue and Organ Development, Regeneration and Advanced Medical Science, Gifu University Graduate School of Medicine, Gifu, Japan
    Search for more papers by this author
  • Ken-ichi Tezuka,

    1. Department of Tissue and Organ Development, Regeneration and Advanced Medical Science, Gifu University Graduate School of Medicine, Gifu, Japan
    Search for more papers by this author
  • Natsuki Watanabe,

    1. Department of Tissue and Organ Development, Regeneration and Advanced Medical Science, Gifu University Graduate School of Medicine, Gifu, Japan
    Search for more papers by this author
  • Takanori Wakaoka,

    1. Department of Otolaryngology, Gifu University Graduate School of Medicine, Gifu, Japan
    Search for more papers by this author
  • Takumi Era,

    1. Department of Cell Modulation, Institute of Molecular Embryology and Genetics (IMEG), Kumamoto University, Kumamoto, Japan
    Search for more papers by this author
  • Takahiro Kunisada

    Corresponding author
    • Department of Tissue and Organ Development, Regeneration and Advanced Medical Science, Gifu University Graduate School of Medicine, Gifu, Japan
    Search for more papers by this author

Author to whom all correspondence should be addressed.

Email: tkunisad@gifu-u.ac.jp


Melanocytes are pigment-producing cells generated from neural crest cells (NCCs) that delaminate from the dorsal neural tube. The widely accepted premise that NCCs migrating along the dorsolateral pathway are the main source of melanocytes in the skin was recently challenged by the finding that Schwann cell precursors are the major cellular source of melanocytes in the skin. Still, in a wide variety of vertebrate embryos, melanocytes are exclusively derived from NCCs. In this study, we show that a NCC population that is not derived from Sox1+ dorsal neuroepithelial cells but are derived from Sox1 cells differentiate into a significant population of melanocytes in the skin of mice. Later, these Sox1 cells clearly segregate from cells that originated from Sox1+ dorsal neuroepithelial cell-derived NCCs. The possible derivation of Sox1 cells from epidermal cells also strengthens their non-neuroepithelial origin.


Melanocytes themselves and/or the pigment they synthesize and sequester in melanosomes and deliver to keratinocytes provide photoprotection and thermoregulation for the skin. Melanocytes are originally derived from neural crest cells (NCCs) (Liem et al. 1995; Nakagawa & Takeichi 1998; Le Douarin & Kalcheim 1999; Wilson et al. 2004; Morales et al. 2005; Hall 2009) and undergo extensive migration to reach their various destinations in hair follicles, in the basal layer and the dermis of the skin, and also to various internal parts of the body, such as the uvea, Harderian gland, inner ear and heart (Erickson 1993; Nordlund et al. 2006; Uehara et al. 2009). NCCs delaminate from the dorsal neural epithelium of the neural tube by undergoing an epithelium to mesenchyme transition. NCCs then migrate through stereotypical pathways and differentiate into a variety of cell types in vertebrate embryos, including neuronal, glial, endocrine, skeletal and pigment cells. NCCs migrating along the dorsolateral pathway between the dermamyotome and the skin were thought to become melanocytes; however, recent work has uncovered that a significant population of follicular melanocytes in mice is produced by the re-differentiation of Schwann cell precursors that migrated from the dorsal root ganglia toward the dermal surface, which in turn had originated from NCCs that had taken the ventral migration pathway (Adameyko et al. 2012).

In mice, tissue or cell-type specific genetic tracing are frequently used to address the cellular and molecular identity of NCC-derived cells including melanocytes. When we used transgenic mice expressing Cre recombinase under control of the Sox1 promoter crossed to the Rosa26-YFP reporter strain that contains a floxed stop cassette preventing yellow fluorescent protein (YFP) expression until the Cre-induced deletion (Sox1-Cre/+; Rosa26R-YFP/+ mice: Srinivas et al. 2001; Takashima et al. 2007), we noticed that melanocytes were derived differently from Sox1-Cre+ and from Sox1-Cre populations in Sox1-Cre/YFP embryos. We show here that a significant population of melanocytes in the skin of mice are derived from trunk NCCs that originated from the non-neuroepithelium. We also observed a clear segregation of these two types of melanocytes in adult mouse skin suggesting a competitive acquisition of their stem cell niche.

Materials and methods


The Sox1-Cre knock-in strain of mice was kindly provided by Dr Y. Takashima and mice were maintained as previously reported (Takashima et al. 2007). Other strains of mice (Dct-lacZ, Dcttm1(Cre)Bee, Wnt1-Cre, ROSA26R-EYFP, ROSA26R-LacZ and K14-Cre) were kindly provided by Drs I. Jackson, F. Beermann, S. Iseki, F. Constantini and M. Osawa, respectively.


Skin specimens were fixed in 4% paraformaldehyde overnight and were then embedded in OCT compound for sectioning. The following primary antibodies were used: anti-DCT (αPEP8, a gift from Dr V. J. Hearing) at 1:200, anti-GFP (Nakarai) at 1:500 and Tuj1 (Covance) at 1:1000. Hoechst was used for nuclear staining. Samples were stained by appropriate fluorescence-tagged secondary antibodies and were examined using an Olympus IX-71 fluorescence microscope.

β-galactosidase staining

Samples were fixed for 30–60 min at 4°C in 1% formaldehyde, 0.2% glutaraldehyde, and 0.02% Nonidet P40 in phosphate-buffered saline (PBS) at pH 7.4. Fixed samples were incubated overnight at 30°C in the staining solution containing 5 mmol/L potassium ferricyanide, 5 mmol/L potassium ferrocyanide, 2 mmol/L MgCl2 and 0.5% Xgal (Gibco) in PBS. Samples were post-fixed in 4% paraformaldehyde.

Neural tube culture

Embryonic day 10.5 (E10.5) embryos were incubated in collagenase 0.75 mg/mL (Wako Jyunyaku) at room temperature for 20 min. After washing with PBS, embryos were dissected in the region corresponding to the end of branchial arches with fine scissors during observation through a binocular microscope (Carl Zeiss, Stereomicroscope DV4) and the cranial regions were discarded. The neural tube and the adjacent areas were removed from the trunk regions of embryos with fine forceps. The removed tissues were placed in culture as described (Morrison et al. 1999) for 24 h in 5% CO2 at 37°C.

Flow-cytometric analysis and cell sorting

Skin specimens were digested with 0.25% trypsin (Gibco) overnight at 4°C, then were washed once with medium containing 3% serum, and twice with serum-free medium. They were then digested with 0.35% collagenase (Gibco) for 1 h at 37°C, then were mechanically dissociated and filtered through a 70 μm cell strainer (BD Biosciences) and resuspended in PBS containing 3% fetal calf serum (FCS). The dissociated skin cells were blocked with rat anti-mouse Fc gamma receptor (2.4–G2, BD Bioscience) on ice for 30–40 min. After another wash with PBS containing 3% FCS, the cells were stained with fluorescein isothiocyanate (FITC)-conjugated rat anti-mouse CD45 (30–F11; BD Bioscience) on ice, washed and then incubated with allophycocyanin (APC)-conjugated rat anti-mouse c-Kit (2B8, BD Bioscience). Dead cells were removed by PI staining. All cell sorting and analyses were carried out with a fluorescence activated cell sorter (FACS) aria (Becton-Dickinson). The sorted cells were directly inoculated into six-well plates previously seeded with ST2 stromal cells.

Cell culture

For the induction of melanocytes, cells containing melanocyte precursors were inoculated into six-well plates previously seeded with ST2 stromal cells and containing α-minimal essential medium (Invitrogen) supplemented with 10% FCS, 10−7 mol/L dexamethasone (Sigma-Aldrich), 20 pmol/L fibroblast growth factor-2 (R&D Systems), 10 pmol/L cholera toxin (Sigma-Aldrich) and 100 ng/mL human recombinant endothelin-3 (Peptide Institute) in 5% CO2 at 37°C. The medium was changed every 3 days.


Derivation of melanocyte lineage cells both from Sox1-Cre+ and from Sox1-Cre cells

It is known that expression of Sox1 is first detected at E7.5 in the anterior half of the late-streak egg cylinder, which is destined to the neural plate, and that Sox1 expression is maintained until the neural tube is formed (Pevny et al. 1998). In fact, using mice with the Cre-recombinase gene inserted in the Sox1 allele (Sox1-Cre: Takashima et al. 2007) crossed with ROSA26R-LacZ strain (Sox1-Cre/LacZ) mice whose LacZ-encoded β-galactosidase expression is switched on by Cre-recombinase under constitutive expression of the ROSA promoter, we confirmed LacZ staining in the entire neural tube of E11.5 mice (Fig. 1a).

Figure 1.

Melanocytes induced from Sox1-Cre+ and Sox1-Cre cell lineages. (a) Transverse section of an embryonic day 11.5 (E11.5) Sox1-Cre/LacZ embryo; LacZ labeling is clearly detected in the whole neural tube and dorsal root ganglia. Dispersed LacZ-labeled cells in the dermis or somite are considered to be migrating neural crest cells (NCCs). (b) Flow cytometric analysis of neonatal Sox1-Cre/YFP skin for yellow fluorescent protein (YFP) and Kit expression. Cells were harvested from trunk skins from a total of 5 P0 mice; representative data from three independent analyses are presented. To exclude the hematopoietic cell lineage, samples were stained with CD45 and CD45 cells were collected. (c) In vitro differentiation of sorted YFPKit+ cells and YFP+Kit+ cells into melanocytes. YFPKit+ cells and YFP+Kit+ cells from Sox1-Cre/YFP neonatal skins were cultured for 20 days. Among the pigmented melanocytes induced (left, bright field microscopy), no YFP expression was observed in melanocytes induced from YFPKit+ cells (right, fluorescence microscopy). On the other hand, all melanocytes induced from YFP+/Kit+ cells express YFP. (d) Melanocytes induced from YFPKit+ or YFP+Kit+ cells in (c) were dissociated and cultured for another 20 days. (e) Expression of the melanocyte lineage marker Dct detected in YFPKit+ and YFP+Kit+ cells from the trunk skin region of P0 Sox1-Cre/YFP; Dct-LacZ mice. About 1000 sorted YFPKit+ and YFP+Kit+ cells were placed on slide glasses and were stained by X-gal. (f) Representative image of emigrating NCCs in vitro from a E10.5 Sox1-Cre/LacZ neural tube explant. A YFP+ cell is indicated by an arrow and YFP cells are indicated arrowheads. Scale bars, 100 μm (a); 25 μm (c, d, e, f).

We then crossed Sox1-Cre mice with ROSA26R-YFP (Sox1-Cre/YFP) mice. Sox1-Cre+ YFP-labeled (YFP+) neural tube and neural crest derivatives were detected in the trunk region. All nucleated cells forming the neural tube expressed YFP in E9.5 embryos, indicating that the Cre-recombinase mediated labeling system efficiently marks neuroepithelial cells destined to form the neural tube structure in E9.5 embryos (Fig. S1).

The Sox1-Cre lineage tracking system has been reported to identify the melanocyte cell lineage (Takashima et al. 2007). To confirm the developmental potential of YFP+ cells, we sorted out the Kit+ population, supposed to be enriched with melanocyte lineage cells (Kunisada et al. 1996), from Sox1-Cre/YFP mice using a FACS, and cultured them on a monolayer of ST2 stromal cells under conditions allowing for melanocyte development (Yamane et al. 1999; Motohashi et al. 2009). Skin-derived Kit+ cells were separated into YFP+ and YFP populations (Fig. 1b) and were then cultured for 20 days. Melanocytes were differentiated from the YFP+ neuroepithelial lineage cells; however, numerous colonies containing mature melanocytes appeared from the YFP population that originated from Sox1 non-neuroepithelial cells (Fig. 1c).

When the established neural crest marker gene Sox10 was used to isolate NCCs from E11.5 embryos of Sox10-IRES-GFP mice (Motohashi et al. 2011), melanocyte colonies were rarely formed from the Sox10 population, in contrast to the numerous melanocyte colonies from Sox10+ population (Fig. S2). Therefore, it was reasonably concluded that the Sox10+ neural crest population contains both Sox1-Cre+ and Sox1-Cre populations. These observations led us to revisit the issue of the non-neuroepithelial YFP origin of the melanocyte cell lineage.

Stability of the inserted Sox1-Cre recombinase gene in the melanocyte cell lineage and the existence of the Sox1-Cre melanocyte cell lineage

Although previous work showed the persistent labeling of neuroepithelial cells and their descendants by the Sox1-Cre and Rosa26R-YFP system (Takashima et al. 2007), we further tested the stability of this system in melanocyte lineage cells. We cultured YFP+Kit+ cells from Sox1-Cre/YFP mice for 20 days and found that most of the pigmented cells as well as the non-pigmented cells expressed YFP (Fig. 1c, right). Conversely, virtually no YFP+ cells were observed in colonies derived from YFP cells (Fig. 1c, left). To further test the stability of Sox1-Cre/YFP mice, we re-plated the initial cultures containing melanocyte colonies on the ST2 stromal cell monolayer and continued to culture them for an additional 20 days. Our previous work indicated that melanocyte colonies formed from melanocyte precursor cells present in the first culture during this second cycle of culture (Motohashi et al. 2007, 2009). In YFP+ and in YFP cell-derived secondary cultures, all cells maintained their expected expression patterns (Fig. 1d), confirming the genetic stability of the inserted Sox1-Cre and loxP genes during melanocyte development. In other words, the YFP melanocyte precursors observed could not be transient cells destined to eventually express Sox1.

To confirm the identity of YFP+ or YFP cells in the Kit+ population, we used the melanocyte lineage marker Dct which has been shown to be expressed in melanocytes and their progenitors (Tsukamoto et al. 1992; Pavan & Tilghman 1994; Wehrle-Haller & Weston 1995). We dissected skins from Sox1-Cre/YFP mice crossed with Dct-LacZ transgenic mice, in which the β-galactosidase gene is controlled by the melanocyte lineage marker Dct promoter (Mackenzie et al. 1997). Kit+ cells sorted from postnatal day 0 (P0) Sox1-Cre/YFP; Dct-lacZ skins were further separated into YFP+ and YFP populations. Hematopoietic lineage cells were gated out using the pan-hematopoietic CD45 marker. As shown in Figure 1e, both cell populations clearly showed LacZ expression (75.9 ± 10.69% YFPKit+ cells and 97.6 ± 4.12% YFP+Kit+ cells expressed Dct, an average of three experiments in which 1000 sorted and stained cells were measured), indicating that not only YFP+ cells but also non-neural epithelium-derived YFP cells have the potential to generate the melanocyte cell lineage.

To further investigate the nature of YFP melanocyte lineage cells, we prepared NCCs in vitro from dissected neural tubes prepared from E 10.5 Sox1-Cre/YFP embryos. Cells that migrated out from the neural tubes in vitro were considered as NCCs as most of these cells expressed the neural crest specific marker HNK1 in chick embryos (Catala et al. 2000). This was also suggested in mouse embryos as most of the cells that migrated out from the neural tube explants expressed the Sox10 marker (Fig. S1 i–k). As explained in Figure 1f, and Supplementary Figure S1 a–h, 65.0% of NCCs that migrated out from the neural tube explants expressed YFP (2124 YFP+ cells in total 3268 emigrated cells from three neural tubes). As most of the neural tube cells of Sox1-Cre/YFP embryos expressed YFP as early as E9.5 (Fig. S3) and since the dissected neural tube contains surrounding tissues, it is likely that YFP cells that migrated out from the explants are not of neural tube origin. This notion is also supported by the fact that the Sox10 population contains substantially no melanocyte precursors (Fig. S2) while the Sox1 population isolated and tested by the same method contains a number of melanocyte precursors (Fig. 1c). In other words, although a significant part of the Sox1 population is not derived from the developing neural tube, this population is likely to contain Sox10+ melanocyte precursors of NCC lineage.

Spatiotemporal distribution of Sox1 and Sox1+ melanocytes during development

In order to obtain spatial information about the YFP and the YFP+ melanocyte cell lineages, we performed immunohistochemical detection of YFP and Dct expressing cells in trunk skin sections from various developmental stages of Sox1-Cre/YFP mice. In E11.5 embryos, YFP+ cells were detected in the dorsal root ganglion and in the dorsal dermal trunk region in addition to the neural tube (Fig. 2a,b). However, the number of YFP+Dct+ melanocyte precursors in the dorsal dermal region was 30 times lower than the number of YFPDct+ melanocyte precursors in this early stage (Fig. 2g). In E13.5 embryos, YFP+Dct+ cells showed a significant increase in the dorsal dermal region (Fig. 2g), then at around E15.5, the frequency of YFPDct+ cells was reversed by the continuous increase of YFP+Dct+ precursors (Fig. 2c,g). During the first hair cycle at P0, hair follicles including YFPDct+ cells could still be detected (Fig. 2d); however, around P30, some hair follicles became occupied mostly with YFP+Dct+ cells (Fig. 2e,g). Even in P170 skin, hair follicles containing YFPDct+ cells were rarely detected (Fig. 2f,g). As observed in Figure 2c–e, not all YFP+ cells expressed the Dct marker. The recent observations that the re-differentiation of melanocytes from Schwann cell precursors occurs not earlier than E13 and that these Schwann cell precursor-derived melanocytes comprise nearly 70% of P11 hair follicles (Adameyko et al. 2012) coincided with our quantitative analysis of a sudden increase in YFP+Dct+ cells around E13.5 and the final dominance of YFP+Dct+ cells in the trunk skin (Fig. 2g), suggesting that YFP+Dct+ melanocytes are Schwann cell derived.

Figure 2.

Dynamic changes in the distribution of yellow fluorescent protein negative (YFP) and YFP+ melanocyte lineage cells. (a) Cross-section of an embryonic day 11.5 (E11.5) trunk stained with antibodies against TuJ1 (red) and YFP (green); nuclei were stained by Hoechst. (b) E11.5 trunk section stained with the melanocyte lineage marker Dct (red) and YFP (green); most Dct+ cells were YFP (arrowhead) and YFP+ Dct+ cells were rarely detected (shown by the arrow). (c–f) Dct (red) and YFP (green) staining of trunk skin containing hair follicles. After birth, both YFPDct+ cells and YFP+Dct+ cells were detected in the hair follicles. YFPDct+ cells (arrowhead) and YFP+Dct+ cells (arrow) are indicated. (g) Fractions of YFP+ and YFP in Dct+ cells were calculated for each developmental stage. About five sections each from three to five embryos or skins were observed and a total 200–500 cells were counted in various stages. (h) Hair follicles were classified into three types as indicated in P0 to P3 mice. (i) The fractions of these three types among the follicles containing YFPDct+ cells. DRG, dorsal root ganglion; NT, neural tube. Scale bars, 30 μm (a); 100 μm (b); 40 μm (c); 100 μm (d, e, f).

Since we frequently observed an uneven distribution of YFP+Dct+ and YFPDct+ cells in individual hair follicles, we classified hair follicles containing YFPDct+ cells into three patterns, as indicated in Figure 2h. According to this analysis, only 25% of these follicles contained YFPDct+ cells (TypeA or TypeC in Fig. 2h) in the bulge region where follicular melanocyte stem cells have been reported to reside (Nishimura et al. 2002) in P3 skin. As most of the hair follicles not containing YFPDct+ in the bulge region did contain YFP+Dct+ cells (TypeB in Fig. 2h), hair follicles containing only YFP+Dct+ cells might eventually dominate in trunk hair follicles.

Sox1 melanocytes are preferentially distributed and maintained in the facial and tail regions

It should be noted that Sox1+ neuroepithelium-derived cells comprise a significant part of the mesenchymal cells in the facial and tail regions (Fig. 3a) and numerous YFP+Dct cells were observed in these mesenchymal regions (Fig. 3b–g). These kinds of neural crest-derived cells have also been reported (Osumi-Yamashita et al. 1994; Sieber-Blum et al. 2004). We noticed that a major part of Dct+ melanocyte lineage cells in the facial and tail regions of E12.5 embryos were YFP (Fig. 3b,e). These YFPDct+ cells were maintained constantly in E15.5 embryos (Fig. 3c,f) and even in P3 newborn mice (Fig. 3d,g). The fraction of YFPDct+ melanoblasts in the P3 facial and tail skin regions was significantly higher than in the trunk region (Fig. 3i). Even in P170 mice, YFPDct+ melanocytes were maintained in the hair follicles of these regions (Fig. 3h). To confirm that these YFPDct+ melanoblasts are of NCC lineage, we used mice with the Cre-recombinase gene inserted in the Wnt1 allele (Wnt1-Cre: Danielian et al. 1998) crossed with ROSA26R-YFP (Wnt1-Cre/YFP) mice. Dct+ cells in the hair follicles were mostly Wnt1-Cre positive, even in the head and tail regions (Fig. S4b), which clearly indicates that Dct+ cells are categorized as a neural crest lineage (Echelard et al. 1994). It should be noted that most Dct+ cells did not express YFP in E12.5 Wnt1-Cre/YFP embryos (Fig. S4a). These data suggest that melanocytes derived from the non-neuroepithelium also have the potential to establish a melanocyte stem cell system throughout life, at least in specific environments such as the facial and tail regions.

Figure 3.

Follicular melanocytes in the facial and tail regions consist largely of non-neuroepithelial cells. (a) Embryonic day 14.5 (E14.5) Sox1-Cre/LacZ embryo stained with X-gal indicates that Sox1+ cells are detected in the facial and tail regions (arrowheads). (b–g) Skin sections prepared from facial (b–d) and tail (e–g) regions of various developmental stages stained with antibodies against Dct (red) and yellow fluorescent protein (YFP) (green); nuclei were stained with Hoechst. In the facial region, the majority of Dct+ cells are YFP. Also, in the tail region, about half of the Dct+ cells are YFPeven in P3 mice. (h) Vibrissa follicles of P170 mice contain Dct+YFP cells. (i) Fractions of YFPDct+ (image) and YFP+Dct+ (■) cells in the indicated regions of P3 mice. About 200 cells of five sections each from three embryos were counted in each region. Arrowheads, YFPDct+ (d and g); arrows, YFP+Dct+ (d and g). Scale bars, 1 mm (a); 40 μm (b, c, e, f); 100 μm (d, g, h).

Embryonic epidermal skin as a possible cell lineage origin corresponding to the Sox1-Cre population

Although we demonstrated that Sox1-Cre cells represent a cell lineage that is developmentally distinguished from Sox1-Cre+ cells and those do not belong to Sox1-Cre derivatives that somehow have been led not to express YFP, a positive discrimination of the cell lineage is preferable for the Sox1-Cre population. We suggested in the previous section that most of the Sox1-Cre melanocytes are derived from Sox10+ cells; however, Sox10 is considered as a pan-neural crest lineage marker. We therefore introduced another genetic cell lineage marker K14-Cre, since epidermal skin is spatially the most closely allied tissue next to the forming neural tube in the embryo. Uniform labeling of epidermal skin by the K14-Cre transgene was previously reported (Vasioukhin et al. 1999) and is also confirmed in our K14-Cre/YFP line (Fig. 4d,g,j). Dissociated K14-Cre/YFP cells including head and trunk regions were stained with anti-Kit and anti-CD45 antibodies and were analyzed by flow cytometry (Fig. 4a). Kit+CD45 cells representing melanocyte lineage cells (the blue dotted cell population in Figure 4a) were tested for the expression of YFP in K14-Cre/YFP skin, and as expected, a significant amount of Kit+CD45 skin-derived cells expressed the K14-Cre driven YFP marker (Fig. 4b). In contrast, Kit+CD45+ hematopoietic lineage cells (the green dotted cell population in Fig. 4a) was mostly K14-Cre (Fig. 4c). While we could not still discriminate K14-Cre+ cells simultaneously as a Sox1-Cre population, it is highly likely that Sox1-Cre cells correspond to the K14-Cre+ epidermal cells. In the trunk skin of K14-Cre/YFP mice, YFP+Kit+ melanocytes were detected in the hair follicle, as were those with YFPKit+ population (Fig. 4d–i). In the facial region, the majority of Kit+ follicular cells are YFP+ (Fig. 4j–m), and this is predicted from Figure 3i in the Sox1Cre (YFPDct+) cell populations that correspond to K14-Cre+ melanocyte lineage cells.

Figure 4.

Expression of epidermal lineage marker K14-Cre among the Kit+ melanocyte cell population. (a) Flow cytometric analysis of neonatal K14-Cre/YFP skin for Kit and CD45 expression. Cells were harvested from trunk skins from a total of 5 P0 mice; representative data from two independent analyses are presented. (b, c) Kit+CD45 melanocyte lineage cells marked with blue dots in (a) and Kit+CD45+ hematopoietic lineage cells marked with green dots were analyzed, respectively, for the expression of K14-Cre/YFP and are shown by the histogram. (d–i) P0 K14-Cre/YFP trunk skin sections stained with anti-YFP antibody (d and g, indicated by the green color), anti-YFP and anti-Kit (e and h, anti-Kit stain indicated by the red color). Boxed regions in e and h are enlarged in f and i, respectively. (j–m) Facial epidermal region of K14-Cre/YFP stained as in d to i. Kit+K14(YFP)+ cells in the hair follicles were enlarged in (l, m). Nuclear staining by Hoechst dye is indicated by the blue color. Scale bars, 40 μm (d, e, g, h, j, k); 15 μm (f, i, l, m).


The dorsal neuroepithelium origin of neural crest lineage cells including melanocytes is largely accepted (Le Douarin & Kalcheim 1999; Wilson et al. 2004; Hall 2009). In mice, direct labeling studies clearly indicate the neural tube origin of migratory NCCs (Chan & Tam 1988; Serbedzija et al. 1990, 1992). Labeled individual cell clones were mostly divided and generated two or more cells within the dorsal neural tube and those were induced to either become ventrally migrating cells destined to the dorsal root ganglia, sympathetic ganglia and adrenal medulla, or to become dorsolaterally migrating cells that possibly differentiate into melanocytes, or stay as neural tube cells (Serbedzija et al. 1994). These observations are also consistent with the multipotentiality of NCCs, as well as with melanocyte fate decision within the neural tube (Bronner-Frazer 1995). In this study, we traced melanocytes identified by stable lineage markers and demonstrated two distinct pathways for the melanocyte lineage, one from the neuroepithelium and the other from the non-neuroepithelium. Sox1 non-neuroepithelium-derived melanocytes that migrated along the dorsolateral pathway mostly disappeared with age in the trunk region; however, they were permanently maintained in the facial and tail regions.

Two waves of melanocyte differentiation

According to our immunohistochemical analysis, the majority of Dct+ melanocyte precursors were YFP in E11.5 Sox1-Cre/YFP embryos, but then a sudden increase in the YFP+Dct+ population was detected by E15.5. Previous studies demonstrated the existence of two waves of melanocyte differentiation. In zebrafish, the second wave of melanocytes occupies the lateral region (Budi et al. 2008; Hultman et al. 2009).

It has also been reported that the cranial non-neuroepithelium in close proximity to the neuroepithelium gives rise to a specific population of cranial NCCs marked by Wnt1-Cre and platelet derived growth factor (PDGFRα) (Breau et al. 2008). Although those cells are different from the YFPDct+ cells observed in the trunk region of Sox1-Cre/YFP mice, neural crest-like cells might not necessarily be derived only from the neuroepithelial region. Kit+ melanoblast-like cells that appear within the dorsal midline of the early neural tube (Wilson et al. 2004) might also be a candidate for the origin of YFP cells. Our data suggest that YFPDct+ cells express Kit (Fig. 1); however, our close immunohistochemical investigations failed to detect YFPDct+ cells in the midline of the early neural tube. YFPDct+ cells might not be a typical NCC lineage; however, the cells were marked by Wntl-Cre (Fig. S4), a well established NCC marker. Also, we indicated that most of the Sox1-Cre melanocytes were derived from NCC marker Sox10 expressing cells. Although YFPDct+ cells are regarded as the NCC lineage based on the expression of these NCC markers, it would be better to indicate the spatial information about their origin. Additional experiments using K14-Cre, which allows tracing of epidermal skin cell lineage revealed a K14-Cre+ population in Kit+ melanocytes, which strongly suggests the epidermal origin of Sox1-Cre melanocytes. Our study shows that the melanocyte lineage has two distinct origins; one from the Sox1+ neuroepithelium and the other from the Sox1 non-neuroepithelium.

Dynamic changes in the distribution of YFP and YFP+ melanocyte lineage cells

In Sox1-Cre/YFP embryos around E11.5, most Dct+ melanoblasts were YFP, but from E13.5 and later, the number of YFP+ melanocytes increased (Fig. 2g). While both lineages of melanocytes are distributed throughout the body skin, YFP melanocytes were more frequently found in the head and tail regions. The uneven distribution of early melanoblasts biased to the head (Wilkie et al. 2002) may correspond to this observation.

Around the first hair cycle, YFP+ and YFP melanocyte lineage cells were randomly distributed in the hair follicles of trunk skin (Fig. 2d,e), but after the second hair cycle most trunk follicles were occupied mostly by YFP+ melanocytes (Fig. 2f) and this continued to later ages. On the other hand, the facial and tail skin regions contained preferentially YFP melanoblasts from as early as E12.5, after which hair follicles of the facial and tail regions were mostly occupied by YFP melanocytes. The notion that two different melanocyte lineages follow specific spatial and time-dependent changes may indicate the further biological significance of these two melanocyte lineages.

A possible mechanism for the specific distribution pattern of YFP+ and YFP melanocyte lineages

In the facial and tail regions of Sox1-Cre/YFP mice, YFP melanocytes were concentrated in the region where Sox1+ neural crest-derived mesenchymal cells were densely distributed as indicated by the intensively lacZ stained area in Fig. 3a. It is therefore possible that these neural crest-derived mesenchymal cells selectively support the proliferation and/or the survival of YFP melanoblasts. In many vertebrates, the tail bud is of mixed origin because germ-layer decisions are delayed (Griffith et al. 1992; Kanki & Ho 1997; Davis & Kirschner 2000). Thus, different stromal conditions, such as the trunk dermis derived from the dermamyotome and the facial/tail dermis derived from the cranial neural crest and mesenchymal tail bud, might induce different developmental conditions for YFP and YFP+ melanocyte lineages.

To explain the dynamic change of YFP and YFP+ melanocytes in hair follicles of the trunk skin of Sox1-Cre/YFP mice, we found the preferential localization of YFP+ melanocytes in the bulge region where the niche for follicular melanocyte stem cells is thought to reside (Nishimura et al. 2002). Although many of the YFP melanocytes reside only within the hair matrix region (Fig. 2i), there are still some YFP cells in the bulge region reminiscent of YFP+ and YFP cells competing in the niche for their self-renewal. While molecular requirements for melanocyte stem cells in the hair follicle are obscure, it is likely that the stromal conditions as a possible niche for melanocyte stem cells are favorable for YFP+ melanocytes in the trunk hair follicles. Keratinocytes are good candidates for the niche of melanocyte stem cells; however evidence supporting different properties between keratinocytes in the trunk/head and facial regions are scarce. In addition to these environmental cues, differences between YFP and YFP+ melanocyte lineage cells have to be further investigated to elucidate the nature of the dynamic distribution pattern of these melanocyte lineages.

Evolutionary aspect of the two distinct waves of melanocytes

Although migratory NCCs have not been described in non-vertebrate urochordates, cells at the border of the neural plate or within the neural tube are often proposed as evolutionary precursors of the neural crest (Baker & Bronner-Fraser 1997; Corbo et al. 1997; Langeland et al. 1998; Holland & Holland 2001; Stone & Hall 2004). One of those migratory cell populations resembling NCCs in an ascidian urochordate was reported to finally differentiate into pigment cells expressing HNK-1 antigens or the Zic gene (Jeffery et al. 2004). A proposed hypothesis that neural crest evolution, beginning with the release of migratory cells from the central nervous sysytem to produce body pigmentation in a common ancestor of urochordates and vertebrates, which then gained an additional multipotential cell fate (Jeffery et al. 2004; Abitua et al. 2012), could be reconsidered because YFP neural crest lineage cells of Sox1-Cre/YFP mice are very likely to exclusively differentiate into melanocytes. We propose that the initial migrating cells that appear in non-vertebrate urochordates are conserved in vertebrates as Sox1 non-neuroepithelium-derived NCCs and that the multipotentially differentiating Sox1+ neuroepithelium-derived NCCs developed subsequently during early vertebrate evolution, in accordance with the evolution of the neural tube structure.


We thank Dr Shin-Ichi Nishikawa for Sox1-Cre mice, Dr V. J. Hearing for αPEP8 antibody, Dr I. J. Jackson for Dct-lacZ mice, Dr F. Beermann for Dcttm1(Cre)Bee mice, Dr S. Iseki for Wnt1-Cre mice, Dr F. Constantini for ROSA26R-EYFP and ROSA26R-LacZ mice and Dr M. Osawa for K14-Cre mice. This work was supported by grants from the Japan Society for the Promotion of Science and Research, a Fellowship from the Japan Society for the Promotion of Science for Young Scientists (HA), and a grant from the Ministry of Education, Science, Sports, and Culture of Japan.