Cell-autonomous and cell non-autonomous signaling through endothelin receptor B during melanocyte development

The neural crest is a pluripotent population of cells that originates along the dorsal line of the developing neural tube and gives rise to a variety of cell types, including pigment cells of the skin, inner ear and choroid,peripheral nervous system cells, enteric ganglia, endocrine cells, most of the craniofacial skeletal and connective tissues, and parts of the heart outflow tracts (Le Douarin and Kalcheim,1999). The generation of each of these cell types from neural crest stem cells depends on a complex interplay between cell-extrinsic and cell-intrinsic factors (Dorsky et al.,2000). Genetic evidence indicates that pigment cell development requires endothelin 3 (EDN3), a peptide ligand that along with two related ligands, endothelin 1 (EDN1) and endothelin 2 (EDN2), acts through the G-protein-coupled receptor endothelin receptor B (EDNRB)(Pla and Larue, 2003). It also requires KIT-ligand (KITL) (formerly known as Steel factor, stem cell factor or mast cell growth factor) that acts through the tyrosine kinase receptor KIT(Copeland et al., 1990; Geissler et al., 1988). Pigment cell development further depends on a set of transcription factors that include the SRY-like high mobility group protein SOX10 and the basic helix-loop-helix-leucine zipper protein MITF(Hodgkinson et al., 1993; Southard-Smith et al., 1998). Although interactions between signaling pathways and transcription regulation have to some extent been investigated in vitro(Goding, 2000), much needs to be learned about the interactions between distinct signaling pathways. The fact that melanocytes develop in an environment that is shared with many other cell lineages requires us to focus not only on signaling crosstalks within one cell type but also on crosstalks between different cell types. We address the question of whether EDNRB acts exclusively in melanoblasts themselves or effects melanocyte development also indirectly, by acting in other cell types.

It is well known that Ednrb is expressed in mouse melanoblasts(Lee et al., 2003; Opdecamp et al., 1998) and hence thought to act cell-autonomously(Hosoda et al., 1994), in contrast to its ligand EDN3 which is secreted and clearly works in a paracrine fashion during melanocyte development. Nevertheless, a study in chimeric mice composed of wild-type and Ednrb null mutant cells(Ednrb^s-l, piebald-lethal) has suggested that Ednrb can act cell non-autonomously during enteric neuroblast development (Kapur et al.,1995). Because of a lack of suitable markers for mutant melanoblasts, however, this chimeric study did not allow conclusions as to the fate of the mutant pigment lineage in the wild type environment, and so the question of whether melanocyte development is likewise subject to cell non-autonomous actions of Ednrb remained unanswered.

In order to be able to track Ednrb-expressing pigment cells even when they are deficient in functional Ednrb, we here used a novel null allele, Ednrb^lacZ, that was obtained by targeted insertion of the bacterial lacZ marker gene into the first exon of mouse Ednrb (Lee et al.,2003). Although chimeric mice composed of wild-type and mutant cells, as mentioned above, usually provide a straightforward approach to test for cell non-autonomous gene actions, they may offer limited possibilities to further characterize the cell types and growth factors involved. This limitation prompted us to perform tissue recombination experiments in vitro,using neural crest cell cultures established from explanted mid-gestation neural tubes (NTs).

The experiments show that in the absence of functional Ednrb, the development of the melanocyte lineage is initiated but no tyrosinase-positive or mature melanocytes are being generated. The mutant cells can be rescued,however, by a two-step procedure whereby a first step, involving cell non-autonomous EDNRB signaling in cells other than melanoblasts, leads to tyrosinase expression but not pigmentation, and a second step, mimicking cell autonomous EDNRB signaling in melanoblasts, leads to pigmentation. The results imply that Ednrb plays a significant role in melanocyte differentiation at time points beyond the narrow developmental period between E10.5 and E12.5 through which it has been shown to be required in vivo(Shin et al., 1999), and they suggest that this later role is mediated by both cell-autonomous and cell non-autonomous signaling pathways.

The line of Ednrb^lacZ mice, in which the lacZreporter gene is inserted into the Ednrb locus thereby creating a null allele, has been described (Lee et al., 2003). It was kept on a mixed C57BL/6//C3H/HeJ background. As Ednrb^lacZ/Ednrb^lacZ homozygous mice do not usually survive beyond weaning because of megacolon, matings were set up between Ednrb^lacZ/+ heterozygotes. To generate Ednrb^lacZ homozygotes that were also heterozygous for Kit^lacZ, we first crossed Ednrb^lacZ/+mice with Kit^lacZ/+ mice(Bernex et al., 1996) and then intercrossed the double heterozygotes. Mice carrying the Kitl^Sl allele on a WC/ReJ background were obtained from Jackson Laboratories and bred with C57BL/6xC3H/HeJ F1 mice. Mice carrying the allele Mitf ^mi–ew on a Naw background were originally obtained form N. Jenkins/N. Copeland.

Noon on the day a vaginal plug was found was defined as embryonic day 0.5(E0.5). Embryos were harvested at E9.5. The polymerase chain reaction (PCR)was used to genotype embryos, using the head portion as source of genomic DNA. Three oligonucleotides were used: Ednrb1,5′-CCAGACTGAAAACAGCAGAGCGGC-3′ (forward); Ednrb2,5′-GGTCTCCCAGAGCCAGACTGGCGATA-3′ (reverse); and Ednrb-lacZ, 5′-CTGTTGGGAAGGGCGATCGGTGC-3′ (reverse). Because of a deletion associated with the targeted insertion, Ednrb^lacZ/Ednrb^lacZ embryos could be detected by the presence of a 270 bp band and the absence of a 495 bp band, wild-type embryos by the presence of the 495 bp band and the absence of the 270 bp band,and heterozygotes by the presence of both bands. To genotype embryos heterozygous or homozygous for Kitl^Sl, the oligonucleotides used were those described previously(Ono et al., 1998). For genotyping, genomic DNA was denatured (94°C for 3 minutes) in the presence of primers, and subjected to PCR in a 50 μl reaction for 30 cycles(94°C for 40 seconds, 55°C for 1 minute and 72°C for 1 minute)using Taq polymerase (Promega) and the manufacturer's buffer in the presence of 1.5 mM MgCl₂.

Embryos at E9.5 were obtained from the matings described above. Pure NT explants containing neural crest cells were isolated and cultured as described previously (Ito and Takeuchi,1984). For most experiments, culture medium consisted of 90% DMEM,1 mM L-glutamine, 1 mM penicillin-streptomycin, 10% FBS and EDN3 (Sigma) at 10 nM. When mentioned, KITL (R&D Systems) was added at 5 nM,12-O-Tetradecanoylphorbol 13-acetate (TPA) at 40 nM and endothelin 1 (EDN1,Sigma) at 40 nM. For in vitro reconstitution experiments, embryonic NT explants were isolated at E9.5 from the following matings: +/+ × +/+; Ednrb^lacZ/+ × Ednrb^lacZ/+; Mitf ^mi–ew/Mitf ^mi–ew × Mitf ^mi–ew/Mitf ^mi–ew; or Kitl^Sl/+ × Kitl^Sl/+. Neural crest cell cultures were established separately for each embryo and the embryos were genotyped. Twenty-four hours later, NTs from Ednrb^lacZ/Ednrb^lacZ cultures were removed and replaced with NTs from either wild type, Ednrb^lacZ/Ednrb^lacZ, Mitf ^mi–ew/Mitf ^mi–ew or Kitl^Sl/Kitl^Sl cultures. For each reconstitution experiment, the medium contained 10 nM EDN3.

At various days in culture, neural crest cells were fixed in 4%formaldehyde in PBS (pH 7.5) for 25 minutes at room temperature, and then permeabilized with 0.1% Triton-X-100 for 5 minutes. For double indirect immunolabeling, the cells were incubated with rabbit anti-Mitf(Opdecamp et al., 1997) or anti-tyrosinase antisera (Vincent Hearing) for 60 minutes, respectively, and then with anti-β-galactosidase (β-gal) monoclonal antibody (Promega)for 30 minutes. These antibodies were revealed with RITC-coupled goat anti-rabbit (Fab)₂ and FITC-coupled goat anti-mouse(Fab)₂.

The cultured neural crest cells were fixed at 14 days in 4% formaldehyde in PBS. X-gal staining was performed in X-gal buffer (2 mM MgCl₂,0.02% NP-40, 2 mM K₃Fe(CN)₆, 2 mM K₄Fe(CN)₆, 0.05% X-gal in PBS, pH 7.5), and postfixed in 4% formaldehyde. The appearance of melanocytes was examined at various time points in live cultures. Their differentiation from neural crest cells was judged by the presence of melanin granules (one of the markers of mature melanocytes) and the characteristic dendritic morphology under phase-contrast and bright-field microscopy.

To be able to experimentally manipulate the development of Ednrbmutant melanocytes, we first established an in vitro culture system using murine trunk neural tubes (NTs) harvested from embryos at E9.5, a stage at which melanoblasts have not yet emigrated from the neural crest. As shown in Fig. 1A, EDN3 promoted the generation of pigmented cells from Ednrb^lacZ/+ NTs, as found previously with wild-type NTs (Hou et al., 2000; Reid et al.,1996). When stained for the expression of β-gal, 14-day-old Ednrb^lacZ/+ cultures showed β-gal-positive cells that were pigmented (Fig. 1A,arrow), along with some that lacked pigmentation and some that were pigmented but lacked clear β-gal staining (Fig. 1A, arrowhead). These latter cells may reflect a progressive downregulation of Ednrb in mature melanocytes. By contrast, in cultures established from Ednrb^lacZ/Ednrb^lacZ NTs, no pigmented cells were seen, although β-gal-positive cells survived(Fig. 1B). Theseβ-gal-positive cells lacked characteristic melanoblast markers, including MITF (see below), and probably represent cells of other lineages. The results indicate that in vitro, Ednrb is required for the generation of melanocytes and suggest that Ednrb is expressed in the melanocyte lineage beyond E12.5, i.e. beyond the developmental period through which it has been shown to be important in melanocyte development in vivo(Shin et al., 1999).

To determine up to which stage Ednrb^lacZ/Ednrb^lacZ cells progress during in vitro development, we stained NT explant cultures for markers that are sequentially expressed in the melanocyte lineage(Hou et al., 2000). The transcription factor MITF, for example, normally begins to be expressed in melanoblasts at E10.5 and is required for the development of the melanocyte lineage (Hodgkinson et al.,1993; Hou et al.,2000; Nakayama et al.,1998; Opdecamp et al.,1997). As shown in Fig. 2A,B, after 2 days of in vitro culture in the absence of EDN3,MITF protein is expressed in the nuclei of β-gal-positive cells,regardless of whether they were derived from Ednrb^lacZ/+or Ednrb^lacZ/Ednrb^lacZ embryos. Not surprisingly, the transcription factor SOX10, which is needed for Mitf expression and plays crucial roles in the development of neural crest derivatives (Potterf et al.,2000; Potterf et al.,2001; Southard-Smith et al.,1998; Verastegui et al.,2000), is also normally expressed in both β-gal-positive Ednrb^lacZ/+ and Ednrb^lacZ/Ednrb^lacZ cells (not shown). These results are consistent with the previous observation by whole-mount in situ hybridization that the early melanoblast marker dopachrome tautomerase(Dct) is expressed in Ednrb^lacZ/Ednrb^lacZ embryos(Lee et al., 2003) and suggest that Ednrb is not required for the initiation of melanoblast development.

We then tested for the expression of the tyrosine kinase receptor KIT, an additional marker characteristic of early melanoblasts. As direct antibody labeling of KIT may not give high level signals and requires fixation criteria that are not easily compatible with double-labeling for other antigens(Opdecamp et al., 1997), we decided to test for KIT expression in an indirect way, using Kit^lacZ mice in which sequences encoding a nuclearβ-gal have been inserted into the Kit gene(Bernex et al., 1996). By crossing Ednrb^lacZ and Kit^lacZ mice,we obtained doubly marked mice in which the two signals are distinct in that the Ednrb-derived β-gal is cytoplasmic and the Kit-derived β-gal is nuclear. As shown in Fig. 2C,D, double nuclear/cytoplasmic β-gal is visible independent of whether the cells were derived from NTs that were heterozygous or homozygous at Ednrb^lacZ. Thus, as for the onset of MITF or SOX10 expression, Ednrb is also not required for the onset of KIT expression.

Despite the fact that lineage development is initiated in the absence of functional Ednrb, the cultures do not generate pigmented cells. As the rate-limiting step in pigmentation is tyrosinase activity, we reasoned that cells devoid of functional Ednrb might not reach the tyrosinase-positive stage. Indeed, double-labeling for β-gal and tyrosinase showed that even after 14 days in culture, i.e. 8 days after the usual onset of tyrosinase expression (Hou et al., 2000), Ednrb^lacZ/Ednrb^lacZ NTs did not generate tyrosinase-positive cells, while Ednrb^lacZ/+ NTs did (Fig. 2E,F). This result suggested that in vitro, Ednrb signaling is required for the development of cells to the differentiated, tyrosinase-positive stage and hence explains why Ednrb^lacZ/Ednrb^lacZcells never become pigmented.

To test whether Ednrb-deficient melanoblasts may be rescued in the presence of wild-type NT or neural crest cells, we performed in vitro recombination experiments in which the Ednrb^lacZ/Ednrb^lacZ NTs, after a 24 hour culture period during which the neural crest cells emigrated, were removed and replaced with Ednrb^+/+ NTs. For these experiments, we used Ednrb^+/+ NTs from embryos homozygous for the Mitf allele, Mitf ^mi–ew, which cannot generate melanoblasts that survive for more than 24-48 hours and hence neither contain tyrosinase-positive nor pigmented cells(Opdecamp et al., 1997). In this way, we could avoid the potential complication that a transfer of melanosomes from pigmented Ednrb^+/+ cells to the non-pigmented Ednrb^lacZ/Ednrb^lacZcells might have created. As a control, Ednrb^lacZ/Ednrb^lacZ NTs were exchanged for Ednrb^lacZ/Ednrb^lacZ NTs obtained from parallel cultures, and, as expected under these conditions, no tyrosinase-positive cells were seen (Fig. 3A). The reconstitution with Mitf ^mi–ew/Mitf ^mi–ew NTs, however, led to the generation ofβ-gal/tyrosinase double-positive cells in Ednrb^lacZ/Ednrb^lacZ cultures(Fig. 3B). This result suggested that the Ednrb-positive neural crest cells provided an exogenous signal capable of rescuing tyrosinase expression in the co-cultured Ednrb-negative cells, and that melanocytes were not required for this rescue. Nevertheless, the possibility, however remote, that at least some of the double-positive cells might have been the product of cell fusions cannot formally be excluded with this experiment.

It has been shown previously that KIT signaling, much as EDNRB signaling,is a crucial regulator of melanocyte development(Copeland et al., 1990; Geissler et al., 1988). In vitro, Kitl-deficient (Kitl^Sl/Kitl^Sl)NT explants do not support the generation of melanocytes(Ono et al., 1998), and Kit-deficient(Kit^lacZ/Kit^lacZ) explants do not give rise to tyrosinase-expressing melanoblasts and do not generate pigmented cells(Hou et al., 2000). These observations prompted us to test whether Kitl^Sl/Kitl^Sl NTs, which carry a wild-type Ednrb locus but lack KITL, would rescue Ednrb^lacZ/Ednrb^lacZ melanocyte precursors. Consistent with previous results(Ono et al., 1998), NT explants from E9.5 Kitl^Sl/Kitl^Sl embryos neither generated tyrosinase-positive nor pigmented cells, even after prolonged periods in culture and in the presence of EDN3 (not shown). The addition of Kitl^Sl/Kitl^Sl NTs to Ednrb^lacZ/Ednrb^lacZ cultures indeed showed that in contrast to Kitl-positive Mitf ^mi–ew/Mitf ^mi–ew NTs, Kitl-deficient NTs do not allow for Ednrb-deficient cells to become tyrosinase positive (Fig. 3C). As there is no reason to assume that the conditions for potential cell fusion in these co-cultures was any different from those in co-cultures between Mitf ^mi–ew/Mitf ^mi–ew +Ednrb^lacZ/Ednrb^lacZ cells, cell fusion can safely be excluded as an explanation for the rescue in the latter cultures. Furthermore, we can conclude that even though the Ednrb-deficient cultures carried a wild-type Kitl locus,their production of KITL was apparently neither sufficient to rescue their own melanocyte precursors nor, conversely, those derived from the Kitl-deficient cultures. Hence, Ednrb signaling is required for appropriate KITL expression in these cultures, and KITL is required for tyrosinase expression in Ednrb-deficient melanoblasts.

As the above reconstitution experiments suggested that KITL is required to allow tyrosinase expression in Ednrb-negative melanocyte precursors,we then asked whether it was sufficient. In fact, we found that the treatment of Ednrb^lacZ/Ednrb^lacZ cultures with soluble KITL from the day of NT explantation led to the generation of tyrosinase-expressing cells (Fig. 4). Thus, KITL was both required and sufficient to promote the differentiation of Ednrb-deficient melanocyte precursors.

Although the above results demonstrated a cell non-autonomous rescue of Ednrb-negative melanocyte precursors to the tyrosinase-positive stage, differentiation to mature, pigmented melanocytes was not observed,either in co-cultures or after the addition of KITL. This is probably not due to the possibility that KITL or other rescue factors were not present at high enough concentrations. Rather, it suggests that EDNRB signaling has an additional, obligatory cell-autonomous role in terminal cell differentiation. EDNRB signaling is known to activate PKC, which in turn activates tyrosinase(Park et al., 1999). As TPA activates PKC (Oka et al.,1996), as does EDNRB signaling, we tested whether TPA might rescue pigmentation in KITL-treated, Ednrb-deficient cultures. As shown in Fig. 5A, after 6 days in the presence of KITL and another 8 days in the presence of both KITL and TPA, Ednrb^lacZ/+ control cultures generated pigmented cells,consistent with previous results with wild-type cells(Murphy et al., 1992). In Ednrb-deficient cultures, despite the presence of KITL, no pigmented cells appeared if TPA was omitted (Fig. 5B). By contrast, when TPA was added at 6 days, mature melanocytes were generated during the following 8 days(Fig. 5C). Thus, TPA can rescue pigmentation in KITL-treated, Ednrb-deficient cells.

It has been suggested earlier that signaling through the paralogous G-coupled receptor EDNRA, which responds preferentially to endothelin 1 (EDN1)and endothelin 2, but has no evident role in early melanoblast development(Opdecamp et al., 1998), could possibly account for pigmentation at later stages(Reid et al., 1996). We,therefore, also tested whether EDN1 might induce pigmented cells in KITL-treated, Ednrb^lacZ/Ednrb^lacZcultures. Whereas 40 nM of EDN1 (which can interact equally well with EDNRA and EDNRB) was perfectly able to induce pigmented cells in Ednrb^lacZ/+ cultures, it was unable to do so in Ednrb^lacZ/Ednrb^lacZ cultures(Fig. 5D,E). This suggests that in these cells, signaling through EDNRA is either quantitatively insufficient or qualitatively distinct from signaling through EDNRB. Taken together, the above experiments suggest that while tyrosinase-positive melanoblasts can be generated through cell non-autonomous signaling pathways, terminal cell differentiation still seems to require Ednrb signaling or a drug like TPA that mimics EDNRB signaling.

The analysis of the in vitro development of lacZ-marked, Ednrb-deficient melanocyte precursors allowed us to reach four main conclusions. First, regardless of the presence or absence of functional Ednrb, cells expressing early lineage markers such as SOX10, MITF and KIT are generated normally, indicating that EDNRB signaling has no instructive role in melanocyte development. Second, the Ednrb mutant cells do not express the melanogenic enzyme tyrosinase and hence are blocked in cell differentiation. Third, the mutant cells can be differentiated to express tyrosinase, if a source of KITL is provided. KITL can be added directly to the medium or can come from co-cultured neural tubes or neural crest cells that are wild type at Ednrb. This suggests that Ednrb signaling stimulates the synthesis and/or secretion of KITL and hence is capable of inducing melanocyte differentiation indirectly. Fourth, the presence of Ednrb wild-type cells, or of KITL, does not allow the tyrosinase-expressing, Ednrb-deficient cells to become pigmented. Pigmentation can be induced, however, by stimulating PKC, a downstream signal of Ednrb, suggesting that the activation of tyrosinase is a cell-autonomous function of Ednrb.

At first sight, a role for Ednrb in melanocyte differentiation would seem to contradict previous in vivo findings that showed, based on conditional downregulation, that Ednrb is no longer needed for melanocyte development after E12.5 (Shin et al., 1999); at this developmental time point, tyrosinase is not yet expressed, at least not in the trunk area, and pigmentation also is not seen until several days later. We have to consider, though, that our in vitro results are based on analyzing isolated NT and neural crest cells, not intact embryos. Conceivably, the role of Ednrb in melanocyte differentiation may not be revealed in vivo because of compensatory mechanisms that operate efficiently only in the embryo. Thus, parallel signaling pathways, whether acting cell-autonomously or not, might substitute for the lack of Ednrb beyond E12.5 in vivo. In fact, Ednrb-deficient mice can have pigmented spots on the head and at the base of the tail, indicating that in vivo, melanocytes can be generated, albeit rarely, in the complete absence of Ednrb. The fact that, in vitro, conditions can be found that rescue Ednrb-negative cells to the fully mature, pigmented stage is entirely consistent with the appearance of these pigmented spots in vivo,and suggests that they are derived from a few somatically variant precursor cells that have crossed a threshold of sensitivity to rescuing pathways that are similar, or identical, to those we have identified in vitro. Although it was believed that signaling through EDNRA might be responsible for these pigment spots (Reid et al.,1996), this is unlikely, because our in vitro studies with EDN1 provide no evidence for a role for EDNRA in pigmentation.

There are several alternative explanations for the seeming discrepancy between a role for EDNRB beyond the equivalent of E12.5 in vitro and not in vivo. It is possible, for example, that the conditional downregulation of the Ednrb mRNA in vivo was not followed by an equally rapid downregulation of EDNRB protein or of its activated downstream targets. In addition, the conditional downregulation of Ednrb, even if complete in melanoblasts, may not have been complete in other cell types, and incomplete downregulation in other cell types may then have added indirectly to the rescue of melanocytes as suggested by our in vitro analysis.

The cell non-autonomous action of Ednrb during melanocyte development lends support to previous observations in chimeric mice in which a wild-type environment helped Ednrb mutant enteric neuroblasts to colonize the large intestine (Kapur et al., 1995). The details of the underlying mechanisms may differ,though, as it has been suggested that EDNRB signaling may normally delay neuroblast differentiation (Wu et al.,1999), allowing for prolonged cell proliferation and migration,and not promote cell differentiation, as described here for melanoblasts. In any event, cell non-autonomous actions of growth factor receptors are not without precedents. A telling example are embryos deficient in a tyrosine kinase receptor, fibroblast growth factor receptor 1 (Fgfr1), which is expressed in several cell types, including presomitic mesoderm precursors. In such embryos, somites fail to form, suggesting Fgfr1 plays a cell-autonomous role in somite specification, but a chimeric analysis showed that upon development in a wild-type environment, the mutant cells can contribute to somites (Rossant and Spence,1998).

The existence of cell non-autonomous mechanisms would predict that genetic modifiers causing phenotypic variabilities in receptor mutants, including Ednrb mutants of mice (Pavan et al., 1995) and humans, need not necessarily influence the expression or activity of the corresponding signaling pathways in the very cell types that cause disease. It is often unknown, however, what cell types might be responsible for such indirect rescue. In the case of Ednrb,we have only demonstrated that the melanocyte lineage, at least its later tyrosinase-positive cell population, is not involved, and we can but speculate as to the cell type that is involved (see below). In addition, we do not know whether in the Ednrb-deficient NTs and their derivatives, the respective `helper' cell type(s) are missing altogether, or are present but simply unable to respond to EDN3. Regardless of the nature of the responsible cell type, our results clearly indicate that the helper cells need to provide at least one factor – KITL.

The rational for testing KITL as the prime rescue factor for Ednrb-deficient cells was based on a number of previous studies that showed that Ednrb and Kit are both required for melanocyte development (Baynash et al.,1994; Geissler et al.,1988) and that EDN3 and KITL cooperate to effect melanogenesis(Hou et al., 2000; Reid et al., 1996). Moreover,we have previously found that although Kit-deficient neural crest cells respond to EDN3 with enhanced survival, they do not develop to the tyrosinase-positive state (Hou et al.,2000), and we report here that Ednrb deficiency does not interfere with Kit expression. Last, variant alleles of the Kitl locus on mouse chromosome 10 serve to phenotypically modify the hypomorphic, non-lethal Ednrb^s (piebald) allele(Pavan et al., 1995; Rhim et al., 2000). Taken together, these findings provided strong indications that it was the lack of sufficient amounts of KITL that was responsible for the absence of tyrosinase expression in Ednrb-deficient melanocyte precursors. Indeed, KITL was able to induce tyrosinase expression in Ednrb-deficient cells, as was the addition of Ednrb^+/+ NTs, even Ednrb^+/+ NTs that were incapable of generating melanocytes of their own as long as they could synthesize KITL. It appears, then, that EDNRB, by still unknown mechanisms, stimulates KITL synthesis and/or secretion. A likely source for this KITL is the neural tube or its derivatives where both Ednrb (Lee et al.,2003) and Kitl (Guo et al., 1997; Wehrle-Haller and Weston, 1999) are expressed.

Although the above results showed a role for KITL in rescuing the Ednrb-deficient cells to the tyrosinase-positive stage, KITL was not sufficient to render them pigmented. This was unlikely to be because of insufficient expression of tyrosinase; the intensity of its immunofluorescent signal in KITL-treated Ednrb^lacZ/Ednrb^lacZcells was similar to that observed in equally treated Ednrb^lacZ/+ cells. Rather, it appears that tyrosinase is not properly activated in the rescued cells. In fact, even in Ednrb-wild type cultures, KITL is insufficient to stimulate melanogenesis (Morrison-Graham and Weston,1993; Murphy et al.,1992; Reid et al.,1996), but the addition of TPA(Murphy et al., 1992) or EDN3(Reid et al., 1996) induces pigmentation. Similarly, the addition of TPA induces pigmentation in our rescued cells. A major site of action of both EDNRB signaling and TPA is the activation of protein kinase C (PKC), which has been shown to activate tyrosinase by serine phosphorylation (Park et al., 1999). If indeed melanogenesis in Ednrb-deficient cells requires the stimulation of PKC, then our results strongly suggest that for melanin synthesis to occur, the cells also depend on cell-autonomous EDNRB signaling pathways. In fact, during terminal differentiation, the KITL-mediated cell non-autonomous action of Ednrb may be replaced by its cell-autonomous signaling. This view is consistent with the observation,based on the in vivo use of neutralizing KIT antibodies, that KIT signaling is no longer needed for melanogenesis after ∼E14.0 when melanocytes begin their terminal differentiation in the epidermis(Nishikawa et al., 1991).

Although our observations do not assess the relative importance of cell non-autonomous Ednrb signaling during melanocyte development in wild-type embryos in vivo, they clearly show that such mechanisms exist. This fact suggests, therefore, that the well-known dual dependency of the melanocyte lineage on two major signaling pathways, G-protein coupled signaling through EDNRB and tyrosine kinase receptor signaling through KIT,may actually result from complex interplays between cell-autonomous and cell non-autonomous actions. As schematically shown in Fig. 6, these different modes of action probably operate at distinct stages in lineage development, allowing for an intricate mechanism to fine-tune the proper deposition of melanocytes in the different bodily compartments.

We thank Dr J.-J. Panthier for providing Kit^lacZ mice,Dr V. Hearing for providing tyrosinase antibody, Ms S. Skuntz for help in genotyping Ednrb^lacZ/+ mice, laboratory members for sharing reagents and helpful suggestions, and the staff of the NINDS animal facility for excellent animal care. We also thank Dr M. Dubois-Dalcq for critical reading of the manuscript.