Two distinct types of mouse melanocyte: differential signaling requirement for the maintenance of non-cutaneous and dermal versus epidermal melanocytes

Melanocytes develop from the pluripotent neural crest, which also gives rise to a number of other cell types, including neurons and glial cells of the peripheral nervous system as well as bone and cartilage cells of the head skeleton. Their immature form, called the melanoblast, migrates along characteristic pathways to various destinations, such as the dermis and epidermis, the inner ear and the choroids of the eye (for reviews, see Hall, 1999; Le Douarin and Kalcheim,1999).

The proliferation and differentiation of melanocytes in hair follicles are closely coupled with the hair regeneration cycle. The follicular melanocytes comprise a stem cell system, and melanocyte stem cells reside in the upper permanent portion of the hair follicles throughout the hair cycle(Nishimura et al., 2002). The eye, harderian gland and ear are also abundant in melanocytes; however, unlike in the hair follicles, no obvious spatially restricted niche of melanocyte stem cells has been described in these organs(Boissy, 1999).

Numerous signaling systems and transcription factors regulate all aspects of melanocyte development. These include the Wnt signaling pathway(Ikeya et al., 1997; Dorsky et al., 2000); the G-coupled endothelin B receptor (EDNRB) and its ligand, endothelin 3 (ET3;EDN3) (Baynash et al., 1994; Hosoda et al., 1994; Pavan and Tilghman, 1994); the tyrosine kinase receptor KIT and its ligand, KITL [also known as stem cell factor (SCF) or mast cell growth factor (MGF)](Geissler et al., 1988; Cable et al., 1995; Wehrle-Haller and Weston,1995); the hepatocyte growth factor (HGF) and its ligand, c-MET(McGill et al., 2006; Kos et al., 1999); and the transcription factors PAX3 (Watanabe et al., 1998; Potterf et al.,2000), SOX10 (Pingault et al.,1998; Southard-Smith et al.,1998; Potterf et al.,2000) and MITF, which is a basic helix-loop-helix leucine-zipper protein (Hodgkinson et al.,1993; Opdecamp et al.,1997; Nakayama et al.,1998; Lister et al.,1999). More recently, it was reported that PTEN(Inoue-Narita et al., 2008; Sarin and Artandi, 2007),Notch (Moriyama et al., 2006; Aubin-Houzelstein et al., 2008; Osawa and Fisher, 2008), FGF2(Weiner at al., 2007; Barsh and Cotsarelis, 2007; Yonetani et al., 2008) and BCL2 (Mak et al., 2006; Nishimura et al., 2005; McGill et al., 2002) are also involved in melanocyte development. Among the genes that encode these factors,the extensively studied signaling ligand and receptor pairs, Kitl-Kitand Et3-Ednrb, provide a particularly intriguing set, as mutations in any one of them leads to a strikingly overlapping phenotype. Thus, mice with such mutations are viable but show the early loss of cells of the melanocyte lineage, which leads to white spotting phenotypes and, sometimes, to a predominately white coat color (Baynash et al., 1994; Hosoda et al.,1994; Yoshida et al.,2001; Botchkareva et al.,2001).

Using cultures of embryonic stem (ES) cells for the induction of melanocyte differentiation in vitro, we previously investigated the requirement for EDNRB signaling throughout the entire process of melanocyte development, in association with that for KIT signaling(Aoki et al., 2005). Kit homozygous knockout(Kit^W-lacZ/Kit^W-lacZ) ES cells, which were completely defective for melanocyte development, were significantly and dose dependently compensated by the addition of ET3 in vitro, and forced expression of ET3 in the skin reduced the white spot area of Kit^W57/Kit^W57 mice in vivo(Aoki et al., 2005). Although KIT and EDNRB signaling are individually indispensable for the survival and maintenance of melanocyte lineage cells, at least during a certain period of development (Yoshida et al.,1996a; Shin et al.,1999), these two signals are thought to influence each other.

It is very typical of transgenic mice in which KITL expression in the skin is induced by the cytokeratin 14 promoter that melanocyte development in specific areas of the skin is extremely progressive, and in such mice, the nose and foot-pad regions are filled with fully pigmented melanocytes as early as the newborn stage (Kunisada et al.,1998). The molecular basis for these observations is still not clear, but it appears that the environment is not uniformly favorable for melanocyte development even under the controlled expression of a single influential factor. Instead, melanocytes may react differentially to the signal in a spatially distinctive manner and region-specific melanocyte cell types might exist.

In the present study, we compared the dependencies of non-cutaneous and cutaneous (epidermal and dermal) melanocytes on KIT, EDNRB, HGF and other signals, using various genetic and cytological approaches. The results suggested that the non-cutaneous and dermal melanocytes were not as sensitive to KIT signaling as the epidermal melanocytes and instead depended more on ET3 and HGF signaling than KIT signaling. As for the cutaneous melanocytes,experimentally induced dermal melanocytes behaved more similarly to those of non-cutaneous tissues. In the light of these findings, we propose the existence of two major melanocyte populations: KIT-sensitive cutaneous melanocytes in the epidermis and weakly KIT-sensitive non-cutaneous and dermal melanocytes. Our classification, based on the growth requirements of the melanocytes, might provide new insight into the characterization of melanocytes in various animals and into the origins of human melanocyte-related diseases.

All animal experiments were approved by the Animal Research Committee of the Graduate School of Medicine, Gifu University. ICR mice, C57BL/6 mice and W/W^v mice were obtained from Japan SLC (Shizuoka, Japan). The following transgenic mice were maintained in our animal facility: those generated with the human cytokeratin 14 promoter (hk14) driving cytokine/growth factor cDNAs [hk14-ET3(Yamazaki et al., 2005);hk14-KITL (Kunisada et al.,1998); hk14-HGF (Kunisada et al., 2000)]; Kit Val620Ala transgenic mice(Tosaki et al., 2006); W^v/W^v and W/W mice derived from W^v/+ and W/+ mice; DCT-lacZ transgenic mice(Mackenzie et al., 1997); and C57Bl/6 background CAG-EGFP mice (RIKEN, BRC). CAG-CAT-EGFP mice (a gift from J. Miyazaki, Osaka University, Osaka, Japan)(Kawamoto et al., 2000) were bred with Dct^tm1(Cre)Bee mice (a gift from F. Beermann,Swiss Institute for Experimental Cancer Research, Epalinges, Switzerland) to generate compound heterozygotes (Yonetani et al., 2008; Osawa et al.,2005). Genotyping was performed as described previously(Guyonneau et al., 2004).

Mice were housed in standard animal rooms with food and water ad libitum under controlled humidity and temperature (22±2°C) conditions. The room was illuminated by fluorescent lights that were on from 8:00 to 20:00 hours.

Postnatal day (P) 0 or P3 mice were sacrificed by decapitation and ear capsules, nose, eyes, harderian glands and skin were quickly dissected on ice. The ear capsules were dissected into very small pieces. The nasal vibrissae were collected from the opposite side of the epidermis (vibrissae hair follicle) one by one, and dissected into very small fragments. The eyes were separated into cornea, lens, and neural retina; each was dissected into very small pieces (Aoki et al.,2008a). The harderian glands were also dissected into very small fragments. These small pieces of all these tissues were treated for 30 minutes at 37°C or overnight at 4°C with 0.25% trypsin/1 mM EDTA (Invitrogen,USA), 0.1% collagenase 1 (Sigma-Aldrich, USA) and 1× dispase (Roche,Switzerland). The skin was treated with 0.25% trypsin/1 mM EDTA for 30 minutes at 37°C or overnight at 4°C. Then, the epidermis was peeled off the underlying dermis and each layer collected separately and dissected into very small pieces. The cells of these small pieces were dissociated by gentle pipetting, and the cell suspensions strained through 400-mesh nylon (Sansho,Japan).

Ten thousand cells prepared from each tissue as described above were inoculated into 6-well plates previously seeded with ST2 cells or into gelatin-coated plates and cultured in Alpha Minimum Essential Medium(Invitrogen) supplemented with 10% fetal calf serum (FCS) (Equitech-Bio, USA or Nichirei Bioscience, Japan). They were maintained under 5% CO₂at 37°C. Then, 40 pM basic fibroblast growth factor (bFGF; FGF2) (R&D Systems, USA), 10 nM dexamethasone (Dex) (Sigma) and 10 pM cholera toxin (CT)(Sigma) were added and remained present during the culture period. The medium was changed every 3 days (Aoki et al.,2008b).

The specific inhibitors and corresponding concentrations used for cell culture were: 10 μg/ml ACK2, which is a neutralizing antibody for KIT (see final section of Materials and methods); 100 ng/ml BQ788, which inhibits EDNRB(Peptides International, USA); 5 μM γ-secretase inhibitor, which is used as a Notch inhibitor; 25 μM PD98059, a MEK kinase inhibitor; 5 μM MET kinase inhibitor, used as an HGF inhibitor; 10 μM SU1498, a VEGFR inhibitor; 10 μM SU5402, an FGFR inhibitor; 10 μM AG1433, an EGFR inhibitor; and 5 μM SB203580, a p38 inhibitor (all from Merck,Germany).

Cell suspensions of ear capsules, vibrissae, choroid, harderian glands,epidermis and dermis prepared from P0 DCT-lacZ mice or P3 Dct^tm1(Cre)Bee/CAG-CAT-EGFP mice as described above were centrifuged and resuspended in staining medium [SM, PBS containing 3% fetal calf serum (FCS)]. The dissociated cells were then blocked with rat anti-mouse Fc gamma receptor (2.4-G2, BD Bioscience) on ice for 30-40 minutes. After another wash with SM, the cells were stained with phycoerythrin(PE)-conjugated rat anti-mouse CD45 (PTPRC) (30-F11, BD Bioscience) on ice for 30-40 minutes, washed, and incubated with allophycocyanin (APC)-conjugated rat anti-mouse c-KIT (2B8, BD Bioscience) on ice for 30-40 minutes. The cells were washed and resuspended in SM containing 3 mg/ml propidium iodide (PI)(Calbiochem, USA). All cell sorting and analyses were performed with a FACS Aria dual-laser flow cytometer (BD Bioscience). The sorted cells were directly inoculated into 6-well plates previously seeded with ST2 stromal cells using the ACDU System (BD Bioscience).

Mice were killed with an overdose of sodium pentobarbital (200 mg/kg). The eyes were enucleated and fixed by immersion overnight in 10% formalin in phosphate buffer (pH 7.2). The methods used for histological analysis have been described in detail previously (Aoki et al., 2008c). Briefly, eyes were dehydrated with ethanol, soaked in xylene and embedded in paraffin. Horizontal serial sections were prepared at 3 μm and stained with Hematoxylin and Eosin (HE).

lacZ staining was performed as reported in detail previously(Yoshida et al., 1996b). In brief, cells were fixed for 10 minutes in 2% paraformaldehyde supplemented with 0.2% glutaraldehyde and 0.02% Tween 20. After three washes in PBS, the cells were stained overnight at 37°C in 10 mM phosphate buffer (pH 7.2)containing 1.0 mM MgCl₂, 3.1 mM K₄[Fe(CN)₆],3.1 mM K₃[Fe(CN)₆] and 2 mg/ml X-Gal. The staining reaction was stopped by washing in PBS. The specimens were post-fixed overnight in 10% formalin in phosphate buffer (pH 7.2).

The rat monoclonal antibody ACK2 against the c-KIT receptor protein, which can block c-KIT function, and its class-matched control monoclonal antibody were described previously (Nishikawa et al., 1991).

The ACK2 antibody was microinjected into each fetal amniotic cavity at embryonic day (E) 11-12.5 as previously described(Yoshida et al., 1993). In brief, 1 μl of antibody solution (0.2-40 mg/ml) was injected into each amniotic cavity of pregnant females with a glass micropipette of ∼50-70μm diameter. The pipette was manually inserted anywhere in the antimesometrial third of the embryo beneath the placenta. Some of the fetuses that received a high concentration of ACK2 (e.g. over 15-20 mg/ml) died as embryos or around birth. Some surviving littermates treated with this procedure were maintained for more than 1.5 years to observe the long-term effect of the antibody treatment on coat color spotting patterns, the stability of the pattern with age and fertility, and to monitor for any behavioral abnormalities.

A mouse model of human piebaldism comprising the Val620Ala mutation in the Kit gene shows various coat-pigmentation patterns among the transgenic lines generated (Kit^V620ATg -1 and Kit^V620ATg-4) (Tosaki et al., 2006). Occasionally, phenotypes varied within the same transgenic line. For example,in Kit^V620ATg -4, which is a relatively less pigmented line, we sometimes found almost completely white individuals(Fig. 1B). Spontaneous Kit mutants, such as W/+ and W^v/+, also had white spots, and homozygous W/W,W^v/W^v or W/W^vmice showed a completely white coat color(Fig. 1C-E).

For these mice, which exhibit different KIT signal transduction levels, we investigated the localization of melanocytes, especially those in the non-cutaneous tissues such as the uvea (i.e. the choroid, ciliary body and iris), the cochlea and the harderian gland. C57BL/6 and Kit^V620ATg, W^v/W^v, W/W^v and W/W mutant mice are arranged in Fig. 1 from left to right in accordance with the expected KIT signal transduction level, under which are shown histological representations of melanocytes in their choroid, ciliary body, iris, cochlea and harderian gland. Melanocytes were present in all these tissues of C57BL/6 control mice (Fig. 1Aa-e), whereas no melanocytes were found in the same tissues from W/W^v or W/W mice(Fig. 1Da-e,Ea-e). In Kit^V620ATg mice, although they had no melanocytes in their hair follicles, melanocytes were detected in the ear capsules, and the melanocytes in their eyes (choroid, ciliary body, iris and harderian gland), as well as those in the cochlea in the inner ear, were found to be distributed normally(Fig. 1Ba-e). In W^v/W^v mice, which also had no melanocytes in their hair follicles, there were no melanocytes in their eyes(choroid, ciliary body and iris) or in their cochlea(Fig. 1Ca-d); however,melanocytes were distributed normally in their ear capsules and harderian gland (Fig. 1Ce). Based on these observations, we hypothesize that these non-cutaneous melanocytes might be less dependent on KIT signaling than the follicular melanocytes, although extensive loss of KIT signals does lead to a complete lack of these non-cutaneous melanocytes.

To estimate the requirements for KIT signaling in the non-cutaneous mealnocytes, we utilized a neutralizing antibody for the KIT receptor, ACK2,to temporarily inhibit the KIT signaling(Nishikawa et al., 1991). C57BL/6 mice injected with ACK2 (0.2-40 μg) at E11.5-12 lost melanocyte stem cells either partially or completely, a loss that was proportional to the amount of ACK2 injected, as judged by the area of the white spots(Fig. 2; see also Materials and methods). The pigment patterns of these mice remained unchanged throughtout their life (Fig. 2A-L). Among these, completely white mice injected with more than 15 μg of ACK2 had no melanocytes in their ear capsule (Fig. 2J-L), inner ear (Fig. 2U), choroid (Fig. 2R), ciliary body (Fig. 2S), iris (Fig. 2T)or harderian gland (Fig. 2V),whereas the partially pigmented mice injected with 0.5-10 μg of ACK2 had melanocytes in theses areas (Fig. 2G-I,M-Q, arrowheads), in almost the abundance seen in the control C57BL/6 mice (Fig. 1A). These data further support the notion that non-cutaneous melanocytes are less dependent than epidermal melanocytes on KIT signaling during development.

Our previous study using transgenic mice generated with Kitl, Et3or Hgf cDNAs driven by the human cytokeratin 14 (KRT14)promoter (denoted hk14) indicated that the expression of KITL in the skin induced epidermal skin melanocytosis (Fig. 3C,G,K) and that the expression of ET3 or HGF induced dermal skin melanocytosis (Kunisada et al.,1998; Yamazaki et al.,2005; Kunisada et al.,2000) (Fig. 3A,B,E,F,I,J). Therefore, we determined the effect of the overexpression of these factors on the non-cutaneous melanocytes in the harderian glands of these transgenic mice.

Overall, the melanocytes in the harderian glands were increased in number in these transgenic mice (Fig. 4). In hk14-ET3 mice, the number of melanocytes in the harderian gland was increased (Fig. 4A,E,I), and a significant increase was also observed in hk14-HGF mice (Fig. 4B,F,J), whereas only a slight increase was observed in hk14-KITL mice(Fig. 4C,G,K,Q). In vitro colony formation from the dissociated cells from the dissected harderian gland tissues also revealed that the number of colonies formed was dramatically increased in hk14-ET3 and hk14-HGF mice(Fig. 4M,N), with only a slight increase in hk14-KITL mice (Fig. 4O,R), as compared with the control(Fig. 4P).

In the ocular tissues, the cornea is usually clear, and we observed no melanocytes in the control C57BL6 mice (n=20/20). However, in hk14-ET3 mice we found that dendritic melanocytes had migrated into the cornea(n=20/20; Fig. 5A,D,G,J). Also, pigmented melanocytes, which were not dendritic in shape, were detected in hk14-HGF mice (n=18/20; Fig. 5B,E,H,K). In hk14-KITL mice, we never observed melanocytes in the cornea (n=20/20; Fig. 5C,F,I,L). These observations indicate that transgenically supplied KITL is less effective for the development of non-cutaneous melanocytes of the ocular tissues or harderian gland than ET3 or HGF supplied in the same manner.

To investigate the molecular basis of the relative ineffectiveness of KITL in the development of non-cutaneous melanocytes, we analyzed KIT expression in the cutaneous and non-cutaneous melanocytes. We previously reported that KIT⁺/CD45^- cells in the neonatal skin comprised the developing melanoblast population(Kunisada et al., 1996; Motohashi et al., 2007). In the cutaneous tissues, such as the epidermis, dermis, ear capsule and vibrissa, KIT⁺/CD45^- populations were clearly detected by flow cytometry (Fig. 6A-E). However, fewer KIT⁺/CD45^- cells were observed in non-cutaneous tissues, such as the uvea in the eye and harderian gland, and the average fluorescence intensity, which indicates the amount of KIT expressed on the cell surface, was clearly reduced in these tissues(Fig. 6F,G). This finding suggests that the lower expression of KIT in the melanocytes of these non-cutaneous tissues is responsible for their lower dependence on KIT signaling.

To further test the differentiation of non-cutaneous melanocytes with respect to KIT expression, we purified 1000 KIT⁺/CD45^-cells from P0 DCT-lacZ mice (Mackenzie et al., 1997) and cultured them on ST2 cells(Yamane et al., 1999). After 14 days, pigmented melanocytes and DCT-lacZ-positive melanoblasts were detected among the KIT⁺/CD45^- precursors isolated from cutaneous tissues such as the epidermis, dermis, vibrissa and ear capsule,whereas only a few pigmented melanocytes and DCT-lacZ-positive melanoblasts were generated from non-cutaneous tissues such as the uvea and harderian gland(Fig. 7A,B). Most of the colonies generated contained lacZ-positive cells, thus confirming the melanocyte cell lineage (Fig. 7B). Thus, very few melanocyte precursors were present in KIT⁺ populations in the non-cutaneous tissues.

As we observed a number of non-pigmented but lacZ-positive melanoblasts in the neonatal harderian gland of DCT-lacZ mice (see Fig. S1A in the supplementary material), we dissected small pieces of harderian gland from P0 mice and cultured them without stromal cells. Successive observations every other day revealed an increase in the number of pigmented melanocytes by this culture method (see Fig. S1B in the supplementary material). Observation every 12 hours revealed that black pigmented melanocytes present at the start of the culture did not divide actively (see Fig. S2A in the supplementary material). Pigmented cells marked in 36-hour cultures only divided once or twice in 264 hours (see Fig. S2B,C in the supplementary material). Thus, it is likely that the highly proliferative melanocyte lineage cells were not mature melanocytes and that the harderian gland or eye contained unpigmented melanoblasts. To further investigate the identity of these non-cutaneous melanoblasts, we tested whether the DCT-positive melanocyte lineage cells that were negative for the KIT receptor were really present as melanocyte precursors. For this purpose, transgenic mice were generated by crossing Dct-Cre knock-in mice [Dct^tm1(Cre)Bee](Guyonneau et al., 2004) with CAG-CAT-EGFP reporter mice (Kawamoto et al., 2000) that express green fluorescent protein (GFP) under the control of a strong ubiquitous promoter (CAG) after Cre-mediated removal of a floxed chloramphenicol acetyl transferase (CAT) gene cassette so as to specifically label melanoblasts with GFP. Melanocyte lineage-specific GFP expression was confirmed in the transgenic mice, as described previously (Osawa et al.,2005). From these mice, we sorted 1000 GFP⁺ cells from the CD45^-/KIT^- gated population by FACS(Fig. 7C,D) and cultured them on ST2 monolayers for 14 days. From both harderian gland and ocular tissues,pigmented melanocytes were induced from the GFP⁺ population;however, no melanocyte colony was detected from GFP^- cells in the CD45^-/KIT^- population(Fig. 7E). Thus, precursors for the non-cutaneous melanocytes were mostly present as GFP⁺ cells in the CD45^-/KIT^- melanoblast population.

To address the additional molecular signal(s) required for the growth and/or differentiation of melanocytes in non-cutaneous tissues, we tested the effect of signal inhibitors in vitro. First, we confirmed the effect of ET3 by using its receptor (EDNRB) antagonist, BQ788. When non-cutaneous melanocytes(i.e. those in harderian gland and eye) were cultured and treated with BQ788 or the ACK2 antibody, the growth and/or differentiation of the melanocytes was blocked effectively by BQ788 but not by ACK2(Fig. 8A). Cutaneous melanocytes (i.e. epidermis) treated with the same concentration of these inhibitors totally disappeared (Fig. 8B). The numbers of melanocytes derived from the harderian gland and uvea were partially reduced by ACK2 treatment, but were stably maintained over a 1-50 μg/ml concentration range(Fig. 8C,D); by contrast, the number of melanocytes derived from the epidermis was sharply reduced at higher ACK2 concentrations (Fig. 8E). These results are consistent with those shown in Fig. 2, which indicated that non-cutaneous melanocytes can survive when treated with an amount of ACK2 that is effective in blocking the cutaneous melanocytes.

To investigate whether there are factors crucial for the growth and differentiation of non-cutaneous melanocytes in addition to KITL and ET3, we added various inhibitors to melanocyte cultures prepared from P0 DCT-lacZ mice. When melanocytes from the harderian gland were cultured and exposed to an effective amount of PD98059, which is a MEK kinase inhibitor, or to a MET kinase inhibitor that inhibits HGF signaling, their growth and/or differentiation was severely inhibited(Fig. 9). PD98059 was effective toward all melanocyte populations tested(Fig. 9). However, the MET kinase inhibitor was more suppressive toward melanocytes from non-cutaneous tissues (Fig. 9E,F) than toward those from cutaneous tissues (Fig. 9A-D), indicating a role for HGF in the development of non-cutaneous melanocytes. Other inhibitors tested, includingγ-secretase inhibitor, VEGFR (KDR) inhibitor (SU1498), FGFR inhibitor(SU5402), EGFR inhibitor (AG1433), p38 (MAPK14) inhibitor (SB203580) and Rho/Rock inhibitor (Y27632), did not show a statistically significant inhibitory effect on the melanocyte populations tested (see Fig. S3 in the supplementary material).

In the melanocytosis model hk14-ET3 and hk14-HGF transgenic mice, the melanocytes were maintained mostly in the dermis, whereas those in hk14-KITL transgenic mice were restricted to the epidermis, even in the aged individuals(Fig. 3E-L). Interestingly,when hk14-ET3 or hk14-HGF mice were crossed with hk14-KITL mice, the melanocytes of the double-transgenic mice were distributed to both dermis and epidermis, as a simple superimposition of both phenotypes(Fig. 10A,B). It could be that if dermal and epidermal melanocytes and their precursors maintained in the skin were mutually compatible, i.e. dermal melanocyte precursors could be recruited to become epidermal melanocytes by the exogenous ET3, an imbalance of dermal and epidermal melanocyte distribution might occur. Thus, the phenotype of the double-transgenic mice suggests the fairly discriminate and independent maintenance of these two melanocyte populations.

When a hk14-ET3 or hk14-HGF mouse was crossed with a Kit^V620ATg piebaldism model mouse, the coat color pattern of the double-transgenic hk14-ET3/+;Kit^V620A/+ Tg or hk14-HGF/+;Kit^V620A/+ Tg mice was indistinguishable from that of Kit^V620A/+ Tg or Kit^V620A/Kit^V620A Tg mice(Fig. 11D,I). However, the skin of the double-transgenic mice was heavily pigmented, even in the area covered with white hair (Fig. 11E,J). The characteristic, heavily pigmented skin pattern of hk14-ET3 and hk14-HGF mice, and the characteristic white spot of the central head region and coat color of Kit^V620ATg mice, did not change during the postnatal stage (Fig. 11A-D,F-I) (Tosaki et al.,2006). Histological analysis revealed that the dermal skin of hk14-ET3/+;Kit^V620A/+ Tg or hk14-HGF/+;Kit^V620A/+ Tg mice was pigmented (Fig. 11K,L), whereas the epidermis was not. By contrast,hk14-KITL/+;Kit^V620A/+ Tg mice with similar white spots showed pigmented skin under the black hair and totally unpigmented skin under the white hair (Fig. 11M-O). These results indicate that the ectopic localization of melanocytes in the dermis is induced and maintained by the expression of ET3 or HGF in the skin and that these dermal melanocytes do not contribute to hair pigmentation. These observations allow us to conclude that cutaneous melanocytes can be classified into two types, one being epidermal melanocytes, which are highly dependent on KIT, and the other being dermal melanocytes, which are more responsive to ET3 and HGF. As judged by their response to these factors, dermal melanocytes share the characteristics of non-cutaneous melanocytes found in the harderian glands and uvea of the eye.

Melanocytes are generally classified into cutaneous and non-cutaneous(extra-cutaneous) types. Cutaneous melanocytes are pathophysiologically distinguished as epidermal and dermal melanocytes. In this report, we described non-cutaneous melanocytes that developed in the eye, ear and harderian gland that were highly dependent on ET3 and HGF signaling and less dependent (although not totally independent) on KIT signaling, and, by contrast, epidermal melanocytes that were more dependent on KIT signaling. Cutaneous melanocytes in the dermis were revealed to be more selectively responsive to ET3 and HGF, like non-cutaneous melanocytes, and could be distinguished from epidermal melanocytes by not contributing to hair pigmentation.

In the context of the structure of skin tissue, the epidermis and dermis are composed of completely different cell populations, and therefore the cells distributed to these two regions are supposedly different in their molecular components, such as adhesion molecules, and in their growth signal dependency. Skin dendritic cells are classified into epidermal Langerhans cells and dermal dendritic cells based on differences in their cell surface marker expression patterns, gene expression profiles, growth signal requirements and immunological functions (Steinman et al.,1997). In the case of melanocytes, the expression of transgenes such as Et3 or Hgf (Fig. 3) could have induced a permanent maintenance of dermal melanocytes that normally disappear after birth, indicating the specific growth requirements of dermal melanocytes. In other words, if dermal and epidermal melanocytes are of an identical cell lineage and are arbitrarily distributed among the skin tissues, then melanocyte distribution in the Kitl transgenic skin should have been similar to that in the Et3 and Hgf transgenic mice.

Dark skin (DSK) mutant mice also support our notion of distinct types of melanocyte because these mice are classified into dermal and epidermal DSK types (Fitch et al., 2003; Van Raamsdonk et al., 2004). Importantly, no DSK mutant mice are known to contain melanocytes in both their epidermis and dermis, indicating factor-specific localization of skin melanocytes. In accordance with our results, Dsk3 and Dsk4(Rps19 and Rps20) mutants would contain epidermal melanocytes generated by the augmented expression of skin KITL(McGowan et al., 2008). Recent identification of the uveal melanoma and blue naevi-specific mutation of the human counterpart of DSK1 (GNAQ), the heterotrimeric G-proteinα-subunit, as opposed to the BRAF and NRAS mutations exclusively found in neoplasms originating from epidermal melanocytes(Van Raamsdonk et al., 2009),also indicate the distinct nature of non-cutaneous and epidermal melanocytes. Still, as suggested by Van Raamsdonk et al.(Van Raamsdonk et al., 2004),increased ET3 or HGF signaling might cause excessive proliferation of the early melanoblasts, followed by their restricted entry into the epidermis across the basal layer. In this scenario, melanocytes left behind in the dermis may develop into the dermal cell type in response to dermal-specific environmental cues.

The precise origins of the non-cutaneous and dermal melanocytes, as well as of epidermal melanocytes, are still open to further investigation. Epidermal melanocytes, including those of hair follicles, are maintained in a KITL-rich environment, whereas dermal melanocytes are supposed to be selectively maintained by ET3 or HGF expression. Since the number of epidermal melanocytes and their distribution pattern were not affected by ET3 or HGF expression(Kunisada et al., 1998; Kunisada et al., 2000) (H.A. and T.K., unpublished), common precursors for these two melanocyte populations would probably have separated from each other at a considerably early developmental stage. If not, common precursors would have been exhaustively used for the dermal melanocyte population. In addition,hk14-ET3/+;Kit^V620A/+ Tg mice are white in coat color and have dark pigmented dermal skin underneath, suggesting that dermal- and epidermal-restricted niches are required for each melanocyte population. This lifetime-stable localization further supports the distinct nature of these two populations. Recently, Notch signaling was reported to be selectively employed for the maintenance of epidermal melanocytes: in Notch1 and/or Notch2 conditional knockout mice, the ear and choroid layer were pigmented normally, in contrast to the gradual loss of coat color pigmentation(Schouwey et al., 2007). This finding also highlights the discrimination of epidermal melanocytes from dermal or non-cutaneous melanocytes.

In relation to skin diseases caused by aberrant melanocyte development, the characteristics of the melanocytes should be considered. In the case of dermal melanocytosis, a broad group of congenital and acquired melanocytic lesions(Zembowicz and Mihm, 2004), ET or HGF are more likely to be involved in the pathogenesis of this disease than is KIT signaling. However, it should also be emphasized that both melanocyte cell types are derived from common neural crest-derived precursors, as W/W mutation or extensive blockade of KIT signaling during embryogenesis completely eliminates the melanocytes from the whole body. In fact, we occasionally found W^v/W^v mice with white ear capsules accompanied by totally unpigmented harderian glands,in contrast to the majority of the W^v/W^v individuals with pigmentation in both their ear capsules and harderian glands (as shown in Fig. 1C). This finding might indicate a decisive effect of a small fluctuation in the KIT signal among W^v/W^v individuals, i.e. reflecting an exacting threshold of KIT signal intensity for the survival of developing melanocytes, at least in their common precursors. At the same time, we demonstrated the strict discrimination of non-cutaneous and dermal melanocytes from epidermal melanocytes based on their reactions to variously manipulated external ligands, including KITL, ET3 and HGF. These findings might help towards a clearer understanding of the pathogenesis of melanocyte-related diseases, including melanomas.