Neurofibromin as a regulator of melanocyte development and differentiation

Type I neurofibromatosis (NF1) is an autosomal dominant disorder caused by mutations in the gene NF1 (Ballester et al., 1990). NF1 encodes a 260 kDa protein called neurofibromin which possesses Ras–GTPase-activating protein (Ras-GAP) activity (DeClue et al., 1992). Patients with NF1 are prone to developing a set of different conditions involving benign and malignant tumors, including neural tumors, such as dermal neurofibromas, plexiform neurofibromas, astrocytomas and optic gliomas (Korf, 2000; Theos and Korf, 2006). Juvenile myeloid leukemia and other hematologic malignancies have also been associated with NF1 (Bader and Miller, 1978; Stiller et al., 1994). In addition, NF1 patients very frequently exhibit pigmentary abnormalities (Gutmann et al., 1997). These include localized, hyperpigmented, flat lesions of the skin termed café-au-lait macules (CALMs) and patches. Freckling of the intertriginous areas of the skin is also extremely common in NF1 (Esterly et al., 1998). Such findings of well-demarcated cutaneous lesions characterized by localized increases in pigmentation are pathognomonic of NF1 (Gutmann et al., 1997) and are likely to be associated with NF1 mutations. Apart from these specific lesions, a lesser known pigmentary phenomenon has been described in these patients, which is a mild general hyperpigmentation relative to their unaffected siblings (Riccardi, 1992). Hence, it is possible that the pigmentary dysregulation in NF1 patients occurs on two distinct levels, one involving all of the cutaneous melanocytes resulting in a generalized, constitutive mild hyperpigmentation and one restricted to localized fields of cutaneous melanocytes resulting in yet increased areas of hyperpigmentation.

Both malignant (Glover et al., 1991; Menon et al., 1990; Skuse et al., 1989; Xu et al., 1992) and benign tumors (Colman et al., 1995; Sawada et al., 1996; Serra et al., 1997) in NF1 have been associated with somatic loss of the second, wild-type NF1 allele. However, studies of CALMs have not found evidence of lesional loss of the second allele (Eisenbarth et al., 1997). Hence it is possible that, in the absence of additional evidence documenting loss of heterozygosity in CALMs, haploinsufficiency for NF1 underlies the development of pigmentary abnormalities in patients with NF1. However, haploinsufficiency in melanocytes alone renders it difficult to explain fully the localized nature of the hyperpigmented CALM lesions observed in these patients. Correspondingly, decreased neurofibromin was found in both the non-lesional and café-au-lait macule melanocytes from NF1 patients, with no difference between overall Ras-GTP levels and Ras-GAP activity in these cultures (Griesser et al., 1995). To support the notion that other cell types might participate in the formation of the hyperpigmented CALM, fibroblasts underlying CALMs were found to secrete basic fibroblast growth factor (bFGF) and hepatocyte growth factor (HGF) at higher levels than non-CALM-associated fibroblasts from NF1 patients (Okazaki et al., 2003). This finding provides a mechanism to contemplate a role for factors extrinsic to the NF1-deficient melanocyte to stimulate a hyperpigmentary response in NF1-haploinsufficient melanocytes.

To study experimentally how NF1 or Nf1 haploinsufficiency might cause either cell-autonomous or extrinsic cell-dependent changes in melanocyte gene expression underlying pigmentary changes, we have used mice deficient for the murine NF1 homolog Nf1 (Brannan et al., 1994) to explore melanocyte signaling and gene expression. Although pigmentary abnormalities have not been reported in these mice and in other Nf1 mouse models, previous work (Ingram et al., 2000) has shown that partial reversion of a Kit-dependent coat-color phenotype is possible when these mice are rendered Nf1-deficient. Here, we show that Nf1 deficiency also reverts a coat-color defect induced by a mutation in microphthalmia-associated transcription factor (Mitf), demonstrating that neurofibromin regulates signaling between Kit and Mitf (Hemesath et al., 1998; Wu et al., 2000) during melanocyte development; at a time when melanoblasts, the precursors of melanocytes, migrate during embryonic development from the neural tube through mesenchymal tissue to localize eventually in the hair follicle region of the skin. In differentiated melanocytes, we show that loss of a single Nf1 allele can both enhance the expression of melanogenic enzyme genes as well as potentiate Kit-dependent stimulation of the MAP-kinase pathway. Interestingly, our results point to a distinct difference in the regulation of melanogenic enzymes by MAP-kinase signaling in primary cells compared with immortalized cell types. Our results implicate neurofibromin as an important regulator both of melanocyte development and differentiation, and provide experimental support for the observation of both generalized and localized pigmentary abnormalities in NF1 patients carrying a defective NF1 allele.

Previous work had demonstrated that Nf1 haploinsufficiency partially restored the coat color defect in mice homozygous for a point mutation in the kinase domain of the Kit receptor tyrosine kinase Kit^W-41/W-41 (Nocka et al., 1990; Ingram et al., 2000), thereby establishing a role for neurofibromin in the modulation of Kit signaling during melanocyte development. Activation of Kit signaling results in phosphorylation of Mitf, enhancing its transcriptional activation capacity and targeting it for rapid degradation (Hemesath et al., 1998; Wu et al., 2000). The phenotypic overlap of mice harboring Kit and Mitf mutations suggests that these genes coordinately modulate the survival of the melanocyte lineage during development. The Kit^W-41/W-41 homozygote normally exhibits a ventral white belly spot, phenotypically indicative of a partial defect in the dorsal-to-ventral migration, survival and/or proliferation of melanoblasts during melanocyte development. Mitf^Mi-wh/+ heterozygotes exhibit a similar ventral spot in addition to a diluted, grayish-brown coat color, signifying effects of this mutation both upon melanocyte development and melanocyte differentiation, which are caused by a point mutation in the DNA-binding domain of Mitf (Hodgkinson et al., 1993).

To test the hypothesis that neurofibromin regulates Kit signaling through Mitf during melanocyte development, we performed a genetic experiment in which we intercrossed Nf1^+/– mice with both another Kit-receptor mutant exhibiting ventral white spotting, Kit^W/+ and Mitf^Mi-wh/+ mice to generate Nf1^+/–; Kit^W/+ and Nf1^+/–; Mitf^Mi-wh/+ compound heterozygotes whose spotting pattern could be compared directly with littermates that were wild-type at the Nf1 locus. The result of this genetic experiment (Fig. 1A) demonstrated that Nf1 haploinsufficiency caused a decrease in ventral belly spot size in both Kit^W/+ and Mitf^Mi-wh/+ mice.

Observing partial correction of a ventral spotting defect with both a Kit and a Mitf mutant allele with neurofibromin haploinsufficiency demonstrates that Nf1 interacts genetically with both Kit and Mitf. Although these findings suggest that neurofibromin regulates the Kit-Mitf signal transduction pathway operative during melanocyte development, it nonetheless remains formally possible that Nf1 exerts its interaction with these genes independently during melanocyte development. In other words, the possibility exists that the Nf1-Kit interaction occurs through a non-Mitf effector downstream of Kit, while at the same time the Nf1-Mitf interaction is regulated by a non-Kit effector upstream of Kit, and that these two regulatory axes are affected independently by neurofibromin (Fig. 1B). To investigate these possibilities, we developed Nf1^+/–; Kit^W-41/W-41 mice according to the scheme described previously (Ingram et al., 2000) and intercrossed them with Mitf^Mi-wh/Mi-wh homozygotes to generate Kit^W-41/+; Mitf^Mi-wh/+ compound heterozygotes. In addition to the uniform ventral belly spot present on Mitf^Mi-wh/+ heterozygotes, these mice also exhibit a large dorsal spot (Fig. 1C, right), demonstrating that these alleles interact during melanocyte development and confirming previous findings of this interaction (Hou et al., 2000). Nf1 haploinsufficiency partially reverted the large dorsal spot in Kit^W-41/+; Mitf^Mi-wh/+ mice (Fig. 1C, left). Since Nf1 haploinsufficiency reduced the Kit^W-41-dependent dorsal spot on the compound heterozygotes, we have further evidence that neurofibromin regulates the activity of the signaling pathway linking Kit and Mitf during melanocyte development.

The previous experiments addressed the role of neurofibromin during melanocyte development. However, the pigmentary manifestations of individuals with NF1 are more likely to reflect dysregulation of melanocyte differentiation than development, because they are generally not yet present at birth and do not include white spots. Because of the localized hyperpigmentation that is represented in café-au-lait macules and patches in NF1 patients, as well as the more subtle but also more generalized hyperpigmentation that has been noted in affected NF1 siblings (Riccardi, 1992), we hypothesized that Nf1^+/– melanocytes exhibit evidence of enhanced melanogenesis, whether observable grossly as darker hairs or as increased expression of key components of the melanogenesis pathway compared with their wild-type counterparts. Although Nf1 haploinsufficiency partially reverted the spotting phenotype in Kit and Mitf mutants, no effect upon coat color was observed. However, it is possible that the non-agouti genetic background of the mice used in this study rendered it difficult to observe any subtle differences in coat pigmentation. Additionally, murine melanocytes are follicular and not epidermal as in humans, rendering it unlikely that lesions resembling CALMs would be observed on the surface of murine skin. To gain deeper insight into the potential effects of neurofibromin on pigmentation gene expression in melanocytes, we purified and cultured melanocytes from mixed cell cultures obtained from dermal suspensions of C57BL6/J and C57BL6/J-Nf1^+/– neonatal littermates. We adapted a procedure originally developed for the purification of melanoblasts from murine embryos (Kunisada et al., 1996) for obtaining a pure population of melanocytes from these mixed primary cultures containing fibroblasts, melanocytes and unspecified other cell types. Because we were interested in determining how the status of the Nf1 gene affected levels of melanogenic gene expression, we used this technique to obtain pure populations of primary murine melanocytes within the first two passages after plating, in order to minimize potential complications resulting from either extended growth in culture or from treatment with geneticin (Halaban and Alfano, 1984), which is commonly used to eliminate fibroblasts from primary melanocyte cultures.

To purify primary melanocytes, cells cultured from dermal suspensions were subjected to limited trypsinization and labeled with antibodies against Kit and CD45, and the Kit⁺CD45^– population was collected by cell sorting. Typically this cell population comprised 1.0-1.5% of the total mixed cell population (Fig. 2A,B), when isolated within two weeks after plating of primary cultures. 98% of these cells expressed the melanocyte differentiation marker tyrosinase-related protein 1 (Tyrp1), confirming their identity as melanocytes (Fig. 2C,D).

To determine whether there was a difference in melanogenic gene expression between wild-type and Nf1^+/– melanocytes, we used real-time PCR to compare the RNA levels of the melanogenic enzymes Tyr, Tyrp1 and Dct, and of the transcription factor Mitf in primary melanocytes cultured from neonatal littermates that had been purified by FACS. Real-time PCR analysis revealed that the expression of Tyr, Tyrp1 and Dct was 2.5- to 3-fold higher in Nf1^+/– melanocytes compared with wild-type melanocytes (Fig. 3A-C). Additionally, the level of expression of Mitf was approximately fourfold higher in sorted Nf1^+/– melanocytes compared with those obtained from wild-type littermates (Fig. 3D). Owing to limiting numbers of cells, it was not possible to detect and compare the expression levels of the corresponding proteins from these FACS-purified cells using western blotting. We attempted to use immunofluorescence to assess differences between melanogenic protein expression in wild-type and Nf1^+/– melanocytes. Although the results, at least with Tyrp1, suggested higher expression levels in Nf1^+/– cells, we were unable to conclude this definitively because of a substantial amount of heterogeneity of expression between cells of a given genotype (supplementary material Fig. S2). However, the coordinate enhancement of melanogenic gene expression in these melanocytes is consistent with the property of enhanced melanogenic activity.

The activity of neurofibromin as a negative regulator of Ras (DeClue et al., 1992) led us to predict that Nf1^+/– melanocytes, like Nf1^+/– mast cells (Ingram et al., 2000), would exhibit enhanced activation of signaling downstream from Ras. Because of the previously described signaling cascade in melanocytes, linking activation of the Kit receptor tyrosine kinase to direct phosphorylation of Mitf by the activated form of the MAP kinase Erk (Hemesath et al., 1998; Wu et al., 2000), we thought that stimulation of Nf1^+/– melanocytes with stem cell factor (SCF)/Kit ligand might lead to greater phosphorylation of Erk compared with stimulation of wild-type melanocytes. Since we had difficulty visualizing western blot protein signals from the limiting number of cells used after FACS separation, we used mixed cell primary cultures pooled by genotype for these experiments. (A table summarizing the different categories of cells used for the experiments in this paper, including the immortalized melan-a melanocyte line used later, is included in the supplementary material Figs S1 and S2.)

Prior to Kit stimulation, Nf1^+/– cells showed increased phosphorylation of Erk relative to wild-type cells (Fig. 4A). They also showed increased expression of the melanogenic proteins TRP-1 and Dct/TRP-2, though comparable levels of tyrosinase (Fig. 4A). Levels and mobility of Mitf remained unchanged. In mixed primary cultures, expression of Tyrp1 and Dct is higher in Nf1^+/– cells compared with wild-type cells (Fig. 4B), similar to observations of relative gene expression levels that were made in purified melanocyte cultures. However, both the protein and mRNA levels of tyrosinase and Tyr remained unchanged in these mixed primary cultures (Fig. 4A,B). This result differed from those obtained from purified primary melanocytes, where Tyr gene expression as well as Tyrp1 and Dct expression was found to be increased in Nf1^+/– cells.

Mixed primary cultures were also studied to determine the effect of Kit stimulation upon Erk phosphorylation in Nf1^+/– cells relative to wild-type cells. The results of this experiment showed that, in addition to the enhanced expression of phosphorylated Erk that is observed prior to stimulation of cells by SCF/Kit ligand (Fig. 4C), addition of 20 ng/ml SCF/Kit ligand to the cultures resulted in higher levels of phosphorylated Erk at 5 and 15 minutes in Nf1^+/– cells, relative to its induction observed in wild-type cells. These findings demonstrate that mixed cultures containing primary melanocytes derived from Nf1^+/– haploinsufficient mice exhibit an increased responsiveness of MAP kinase signaling to Kit stimulation, consistent with an elevated activity of Ras in these cells.

We performed a number of control experiments to address our concerns that the results demonstrated in Fig. 4A-C might represent differences between the number of Kit⁺ melanocytes in mixed cultures or differences in the responses of non-melanocytic cell types in the cultures to Kit signaling, rather than reflecting hyperresponsiveness of the MAP kinase signaling pathway in Nf1^+/– melanocytes. In flow cytometry analysis of mixed primary cultures of Nf1^+/– and wild-type cells, the Kit⁺CD45^– populations representing melanocytes were present in comparable numbers (supplementary material Fig. S1A). This result confirms that the expression of the melanogenic proteins observed in Fig. 4A represents intrinsic differences in levels, rather than differences due to increased numbers of melanocytes in the Nf1^+/– cultures. To support this contention, neonatal mice expressing the Dct-lacZ transgene as a marker of melanoblasts and melanocytes (Hornyak et al., 2001) exhibited similar numbers of these cells in histologic sections from wild-type and Nf1^+/– littermates (supplementary material Fig. S1B). To establish that the expression of activated Erk in these cultures is restricted to melanocytes and not present in other cell types, we used identical methods to prepare cultures from the skin of Mitf^Mi-wh/Mi-wh neonatal mice, which lack melanocytes (Opdecamp et al., 1997). In comparison with wild-type and Nf1^+/– mice, cultures prepared from neonatal Mitf^Mi-wh/Mi-wh dermal cell suspensions showed no Erk phosphorylation with or without specific stimulation by Kit (supplementary material Fig. S1C), indicating that the phosphorylated Erk we observed (Fig. 4A and C) originated solely from melanocytes. Mitf^Mi-wh/Mi-wh neonates have normal numbers of mast cells, so it is unlikely that any Kit-expressing mast cells (Kim et al., 1999) in these cultures are contributing to the effects observed. Western blotting (supplementary material Fig. S1D) confirmed that Nf1^+/– cultures exhibit lower levels of neurofibromin protein that their wild-type counterparts. Finally, a previous report has demonstrated elevated levels of SCF/Kit ligand secretion in murine Nf1^–/– Schwann-cell-conditioned medium (Yang et al., 2003). To test whether enhanced SCF/Kit ligand secretion in the culture is responsible for enhanced Erk activation in Nf1^+/– melanocytes, conditioned medium (CM) from a day-10 mixed primary culture from WT and Nf1^+/– melanocytes was tested for SCF/Kit ligand concentration by ELISA. The results showed no significant difference between wild-type and Nf1^+/– SCF/Kit ligand concentrations (supplementary material Fig. S1E). Together, these observations suggest that Nf1-haploinsufficient melanocytes in mixed primary cultures express higher levels of the accessory melanogenic proteins Tyrp1 and Dct, exhibit a greater degree of Erk activation without specific growth factor stimulation, and demonstrate hyperresponsiveness to stimulation with the melanocyte survival factor and mitogen SCF/Kit ligand.

The enhanced phosphorylation of Erk that was observed in Nf1^+/– cultures both prior to (Fig. 4A,C) and after Kit stimulation (Fig. 4C) implied activation of a kinase upstream of Erk responsible for maintaining Erk phosphorylation in these cells. The role of neurofibromin as an inhibitor of Ras-dependent signaling (DeClue et al., 1992) suggested that this kinase could be MEK, the kinase immediately upstream from Erk in the canonical MAP kinase pathway. To determine the role of MEK in both Kit-independent and Kit-dependent Erk activation in Nf1^+/– cells, we used a specific inhibitor of MEK, PD98509 (Davies et al., 2000), to inhibit its activity in primary murine melanocytes. We found that PD98509 inhibited not only the enhanced phosphorylation of Erk observed in Nf1^+/– cells prior to stimulation by SCF/Kit ligand (Fig. 5A, lane 3) but also the hyperresponsiveness of Erk to SCF/Kit ligand in these cells (Fig. 5A, lane 4). These levels of Erk phosphorylation were similar to those of unstimulated, wild-type cells (Fig. 5A, lane 1) in this experiment.

In a previous report (Englaro et al., 1998), murine B16 melanoma cells treated with PD98059 to inhibit MEK signaling showed an increase of expression of the melanogenic enzyme tyrosinase. This result reveals a different relationship between MAP kinase activation and tyrosinase expression than our previous results (Fig. 3A), showing elevated expression of Tyr in FACS-sorted primary Nf1^+/– melanocytes with elevated MEK signaling. However, these paradoxical results could be explained by other cellular differences between primary melanocytes and B16 melanoma cells, driving similar responses of Tyr expression to neurofibromin reduction in the former case and MEK inhibition in the latter. To explore an apparent difference between the response of primary murine melanocytes and B16 murine melanoma cells to changes in the activity of MEK, we hypothesized that events resulting in melanocyte immortalization might switch the response to MEK inhibition of melanocytic cells from anti-melanogenic to pro-melanogenic. We used melan-a cells, an immortalized mouse melanocyte cell line derived from C57BL6 mice (Bennett et al., 1987), for these experiments.

In contrast to the result previously reported with B16 cells, Tyr expression decreased in FACS-sorted wild-type and Nf1^+/– primary mouse melanocytes (Fig. 5E) upon MEK inhibition with PD98059. In fact, the inhibition of MEK-dependent phosphorylation of Erk by PD98059 decreased the relative expression of Tyr, Tyrp1, Dct and Mitf gene expression in these sorted primary cells (Fig. 5E-H). Similar to results from B16 melanoma cells, PD 98059 increased the expression of Tyr, Tyrp1 and Dct in melan-a cells (Fig. 5B-D). These results demonstrate a clear difference between the response of primary melanocytes and immortalized melanocytes to MEK inhibition. The changes we observe in primary cells with MEK inhibition are also consistent with the increase in melanogenic gene expression observed in Nf1^+/– melanocytes where Erk is activated (Fig. 3, Fig. 4A).

Since we observed a difference between the responses of primary and immortalized melanocytes to the inhibition of MEK, we were interested in whether these two cell types also exhibited similar or paradoxical responses to loss of neurofibromin. We used small interference RNAs (siRNAs) to reduce the levels of neurofibromin in melan-a cells. Three independent siRNAs designed to mediate knockdown of the neurofibromin each resulted in efficient reduction of neurofibromin protein, relative to that observed with a scrambled control, following transfection (Fig. 6A). Subsequent real-time PCR analysis of total RNA samples isolated from melan-a cells 48 hours after transfection showed that these cells exhibited increased expression of Tyr, Tyrp1 and Dct relative to cells transfected with a scrambled siRNA (Fig. 6B).

Hence, loss of neurofibromin, due to haploinsufficiency in Nf1^+/– primary cells or to siRNA-mediated deletion in melan-a cells, results in an increase in melanogenic gene expression. In contrast to the paradoxical effects of MEK inhibition, primary murine melanocytes and their immortalized counterparts, melan-a cells, exhibit a similar, enhanced melanogenic gene response to a reduction of neurofibromin, implying that the level of cellular neurofibromin may be dominant to activity of the canonical MAP kinase pathway in determining melanogenic gene activity.

Studies of Nf1 knockout mice have demonstrated that germline loss of Nf1 leads to embryonic lethality due to cardiac defects (Brannan et al., 1994), while heterozygous loss of Nf1 predisposed mice to tumor development (Jacks et al., 1994). Pigmentary abnormalities were not noticed in these mice, probably because pigmented melanocytes in mice are predominantly localized in the hair follicle, instead of both the epidermis and hair follicle as in humans. Our finding that haploinsufficiency for neurofibromin partially restores the coat-color defect in Kit^W/+ and Mitf^Mi-wh/+ heterozygotes as well as Kit^W/+; Mitf^Mi-wh/+ compound heterozygotes demonstrates that neurofibromin negatively regulates the signal transduction pathway linking Kit to Mitf during melanocyte development. Mitf is activated by SCF/Kit ligand stimulation of the MAP kinase pathway (Hemesath et al., 1998; Wu et al., 2000) and has been shown to be a melanocyte survival factor during development (Hornyak et al., 2001; Opdecamp et al., 1997) as well as during post-natal life (McGill et al., 2002; Nishimura et al., 2005). These results suggest that the similar effect observed in Kit^W-41/Kit^W-41 homozygotes (Ingram et al., 2000) and the Kit-ligand-independent survival of Nf1^–/– melanoblasts in vivo and in vitro (Wehrle-Haller et al., 2001) is mediated by the enhanced activity of Mitf as an effector. An interesting corollary results from considering the regulation of Kit by Mitf that has been described both in mast cells (Tsujimura et al., 1996) and melanoblasts (Opdecamp et al., 1997). Loss of Nf1, thereby potentiating Kit-dependent signaling to Mitf, could conceivably strengthen a positive feedback loop between these factors, thereby contributing to enhanced survival of both mast cells and melanoblasts. Furthermore, studies of Nf1^+/– melanoblast cultures could be used to explore this dependence, and also how Nf1 haploinsufficiency modifies the temporal expression of melanocytic gene expression during differentiation.

Studies on the effect of neurofibromin on melanogenesis have been limited. Cultured human melanocytes from the café-au-lait macules of NF1 patients have been shown to contain a higher melanin content than non-lesional melanocytes, consistent with increased melanogenesis in these cells (Kaufmann et al., 1991). Our results show both in purified and in mixed primary cultures a higher level of melanogenic gene expression in Nf1^+/– melanocytes than wild-type cells, providing nominal experimental support to the observation of an increased constitutive level of pigmentation in affected NF1 siblings (Riccardi, 1992). Increased melanogenic gene expression was also demonstrated in the immortal murine melanocyte line melan-a upon siRNA-mediated knockdown of neurofibromin, demonstrating that this effect was consistent between primary and immortal murine melanocytes.

Our data nonetheless shows a paradoxical difference between the effect of MEK inhibition on melanogenic gene expression between primary and immortal melanocytes. Consistent with the negative regulation of Ras-dependent MAP kinase signaling, resulting from a complete cellular complement of neurofibromin, inhibition of MEK in primary cells inhibited melanogenic gene expression. However, in immortalized melan-a cells, where neurofibromin reduction also increased melanogenic gene expression, MEK inhibition increased melanogenic gene expression. This suggests that neurofibromin and MEK signaling are uncoupled in immortalized cells, with neurofibromin playing a greater role in the regulation of pigmentation compared with MEK signaling. Gain-of-function studies in human melanoma cell lines have demonstrated activation of human TYR (Suzuki et al., 1994) and TYRP2/DCT (Suzuki et al., 1998) expression by full-length rat neurofibromin and the human neurofibromin GAP-related domain. These observations are difficult to reconcile with our results demonstrating activation of the murine form of these genes with reduced neurofibromin. However, it is possible that cellular changes leading to malignant transformation in human cells alter the relationship between neurofibromin dosage and regulation of pigmentation gene expression that is observed in the primary murine melanocyte cultures we used. Additionally, a limited study of cultured melanocytes from a small number of NF1 patients and healthy humans showed reduced expression of MITF, DCT and TYRP1 in cultured NF1^+/– melanocytes (Boucneau et al., 2005). Although genetic or species differences might be responsible for this difference in results, our results nonetheless show enhanced melanogenic gene expression in genetically-defined murine Nf1^+/– melanocytes, results most consistent with the observation of increased pigmentation in NF1 patients.

Dysregulated Ras signaling has been described in NF1- and Nf1-deficient hematopoietic cells (Bollag et al., 1996; Zhang et al., 1998; Ingram et al., 2000). Our results demonstrate hyperactivation of MAP kinase signaling via Erk phosphorylation (Fig. 4C) and extend the finding of dysregulated Ras signaling with Nf1 haploinsufficiency to melanocytes, a neural crest-derived cell type. The observation that, in contrast to mast cells (Ingram et al., 2000), there is hyperactivation of MAP kinase without specific growth factor stimulation (Fig. 4A, Fig. 5A) suggests that control of MAP kinase signaling in melanocytes is even more sensitive to neurofibromin dosage than in mast cells.

A model for the development of CALMs can be proposed on the basis that increased secretion of SCF occurs from dermal fibroblasts underneath CALMs in NF1 patients (Okazaki et al., 2003). We propose that increased secretion of SCF and, perhaps, related factors from regional clones of fibroblasts, combined with the hyperresponsiveness to SCF we have described in Nf1^+/– melanocytes, underlies the development of these localized pigmentary lesions. The interaction of soluble, fibroblast-derived SCF with Nf1^+/– melanocytes, sensitized by dysregulated Ras signaling to hyperactivate MAP kinase and, thus, melanogenesis, is a plausible explanation for the development of these further hyperpigmented lesions occuring within a generalized, mildly hyperpigmented background.

Kit receptor activation and treatment with TPA induce phosphorylation of Mitf through the Ras-Raf-MEK-Erk pathway, potentiating Mitf activity and inducing proteasomal degradation (Hemesath et al., 1998; Wu et al., 2000). However, in Nf1^+/– melanocytes, we noted equivalent amounts and mobilities of Mitf protein despite increased phosphorylated Erk in these cells (Fig. 4A) and increased Mitf expression (Fig. 3D). One possibility accounting for these observations is that the sustained activation of Erk, in the absence of specific, growth factor-induced stimulation, in Nf1^+/– cells may lead to different outcomes than transient activation. For example, activation of feedback pathways may result in negligible levels of Mitf phosphorylation under conditions of sustained Erk activity and tight post-transcriptional regulation of Mitf levels even with fluctuations in the level of Mitf transcripts. This phenomenon may be akin to the distinct responses of PC12 cells to transient versus sustained activation of Erk caused by different growth factors (Marshall, 1995). Furthermore, we propose that environmental factors account for the discrepancy we noticed between the increase in tyrosinase gene expression in purified Nf1^+/– primary melanocytes and the similar levels of expression in mixed primary cultures. Factors expressed upon or secreted by non-melanocytic cells in those cultures might be responsible for blunting the effect of Nf1 haploinsufficiency upon tyrosinase gene expression selectively.

We noticed that MEK inhibition decreased melanogenic gene expression in primary melanocytes but increased expression in immortalized melanocytes (Fig. 5) and murine melanoma cells (Englaro et al., 1998). Surprisingly, these findings, in combination with the similar response of primary and immortalized melanocytes to reduction of neurofibromin (Fig. 4B, Fig. 6B), might point to a role for neurofibromin described previously (Johnson et al., 1994; Johnson et al., 1993) that is distinct from its Ras-GTPase activity. In these studies, overexpression of neurofibromin altered the growth and differentiation characteristics of certain cell types, including human melanoma cells, without altering levels of Ras-GTP. Here, we observed that similar effects upon melanogenic gene expression occurred with a reduction of neurofibromin in primary and immortalized melanocytes each derived from the C57BL6 mouse strain (Bennett et al., 1987) despite diverging effects of direct MEK inhibition. The strongest explanation for our findings is that neurofibromin has a dominant role to Ras-dependent MAP-kinase signaling in melanocytic cells in the regulation of melanogenic gene expression. Other differences, such as the loss of p16 expression, which has been reported both in melan-a cells (Sviderskaya et al., 2002) and in melanoma cells (Walker et al., 1998), may explain the divergent effects upon pigmentation of MAP-kinase inhibition between primary and immortalized melanocytes. Activation of RAF signaling in melanocytes induces a p16-dependent senescence (Gray-Schopfer et al., 2006; Michaloglou et al., 2005), and the absence of this response from p16-deficient melan-a cells might trigger other intracellular events leading to a paradoxical response on melanogenic gene expression compared with primary cells. It is interesting to speculate whether, in a different experimental setting, a deficiency of Nf1 might promote either cellular senescence or transformation due to its effects on RAS-RAF pathway activation. Use of primary melanocytes, rather than immortalized cell lines, might be important for evaluating accurately the effects of cell signaling modulators upon melanogenesis. We recommend to confirm the results of cell differentiation experiments obtained with immortalized and transformed melanocytes as well as other cell types in primary cells to rule out potentially confounding factors that modify the differentiation response.

C57BL/6-Nf1^+/– mice (Brannan et al., 1994) were obtained both from The Jackson Laboratories and the NCI-Frederick Mouse Repository. Heterozygotes were intercrossed with C57BL/10 mice to generate mice for primary cultures but maintained on a C57BL/6 background. Nf1^+/– mice were genotyped from tail DNA using previously described procedures (Brannan et al., 1994). C57BL/6-Kit^W/+ mice (Jackson) were intercrossed with C57BL/6-Nf1^+/– mice to generate C57BL/6-Kit^W/+; Nf1^+/– and C57BL/6-Kit^W/+; Nf1^+/+ (wild-type) mice for comparison. The genotype of Kit^W/+ mice was inferred from the characteristic presence of a ventral white spot on the heterozygote. C57BL/6-Mitf^Mi-wh/Mitf^Mi-wh homozygotes (gift of Lynn Lamoreux, Texas A & M College of Veterinary Medicine) were intercrossed with C57BL/6-Nf1^+/– mice to generate C57BL/6-Mitf^Mi-wh/+; Nf1^+/– compound heterozygotes and C57BL/6-Mitf^Mi-wh/+; Nf1^+/+ mice for comparison. Mitf^Mi-wh/+ heterozygotes were genotyped by their lighter grey-brown coat color and the presence of ventral white spotting. C57BL/6-Nf1^+/–; Kit^W41/W41 mice generated according to a previously described scheme (Ingram et al., 2000) were intercrossed with C57BL/6-Mitf^Mi-wh/Mi-wh homozygotes to generate Kit^W41/+;Mitf^Mi-wh/+ compound heterozygotes that were either wild-type at the Nf1 locus or Nf1^+/–. Housing and care of all mice were performed under animal care protocols under the auspices of the Institutional Animal Care and Use Committee of the Henry Ford Health System and the Animal Care and Use Committee of NCI-Bethesda.

Nf1^+/– and wild-type primary melanocytes were cultured from dermal cell suspensions, as described previously (Abdel-Malek et al., 2001), derived from 0.5-day-old to 1.5-day-old pups obtained from a cross between C57BL/6-Nf1^+/– and C57BL/10 with some modifications. Neonatal mice were sacrificed by decapitation and the entire dorsal skin between the limbs removed. Dorsal skins from each pup were individually trypsinized in 0.25% trypsin in calcium and magnesium-free PBS containing 1% penicillin/streptomycin overnight at 4°C. The following day, the epidermis was removed and the dermis was suspended in 8 ml of melanocyte growth medium (MGM). MGM in this study was MCDB-153 medium (Sigma), prepared according to the manufacturer's protocol, containing 4% heat-inactivated fetal bovine serum (FBS), 1% penicillin/streptomycin, 5 mg/l insulin, and 1 mg/l transferrin. The dermis was shaken by vortexing for 5 minutes at room temperature and the dermal suspensions from each skin were plated individually into a 25 cm² flask and supplemented with 32 nM tetradecanoyl phorbol 12-myristate-acetate (TPA), 5 μM dibutyryl cAMP (dbcAMP) and 0.6 ng/ml human recombinant bFGF. The medium was changed daily for the first 3 days followed by further three changes per week. The cells generally reached confluency 7-10 days after plating.

Mixed primary melanocyte cultures were trypsinized with 0.05% trypsin-EDTA and pelleted. Pelleted cells were washed twice by resuspension in 2 ml of wash buffer (2% FBS in PBS) followed by centrifugation at 800 g for 10 minutes. Pelleted cells were then resuspended in 400 μl of wash buffer, followed by addition of 50 μl normal mouse serum and incubated on ice for 1 hour. After a 10-minute incubation with rat anti-mouse CD16-CD32 (Fc γIII-Fc γII receptor; 1 μg per 1 ×10⁶ cells, BD Pharmingen), cells were double-labeled with R-phycoerythrin (R-PE)-conjugated rat anti-mouse CD117 (Kit; 0.2 μg per 1 ×10⁶ cells, BD Pharmingen) and allophycocyanin (APC)-conjugated rat anti-mouse CD45 (0.2 μg per 1 ×10⁶ cells; BD Pharmingen) and incubated on ice for 1 hour. Some cells were labeled with mouse PE-anti-IgG (0.2 μg per 1 ×10⁵ cells) alone and mouse APC anti-IgG (0.2 μg per 1 ×10⁵ cells) alone as isotype controls and incubated on ice for 1 hour. Antibody-containing solutions were removed by washing cells twice in 400 μl wash buffer at 800 g for 5 minutes. Pelleted cells were resuspended in 400 μl of wash buffer and filtered through a 45-μm filter prior to cell sorting. Cells were sorted using FACS Vantage SE with DiVa option (BD Biosciences, San Jose, CA). The Kit⁺CD45^– population which typically constitutes 1-2% of the total population was collected.

For flow cytometry, mixed primary melanocyte cultures were double-labeled with R-PE-anti-mouse CD117 and APC-anti-mouse CD45 antibody as described above and flow cytometric analysis was performed using a FACS Calibur (BD Biosciences). For immunofluorescence analysis, sorted Kit⁺CD45^– cells were analyzed using anti-mouse MEL-5 (anti-Tyrp1, Signet Labs) after fixation in 4% paraformaldehyde (PFA) in PBS for 20 minutes at room temperature. Cells were rinsed in PBS three times for 5 minutes and incubated in anti-mouse MEL-5 antibody (1:200) for 1 hour at room temperature followed by three more washes for 5 minutes with PBS. Cells were then incubated in goat anti-mouse Alexa-Fluor-546 (1:200) secondary antibody for 1 hour at room temperature. Following three more washes in PBS for 5 minutes, slides were mounted with Vectashield mounting medium with DAPI (Vector Laboratories) and visualized under a fluorescence microscope.

Mixed primary cultures of wild-type and Nf1^+/– melanocytes were grown in MGM in a 75-cm² flask for 10 days to generate conditioned medium. From these cultures, 10 ml of conditioned medium were used to determine concentrations of Kit ligand (R&D Systems Inc., Minneapolis, MN) using the ELISA Quantikine M kit to measure Kit ligand, according to the manufacturer's protocol.

FACS-sorted primary melanocytes of each genotype were plated at a density of 1 ×10⁴ cells per well in 24-well plates and grown in MGM supplemented with 32 nM TPA, 0.6 ng/ml bFGF and 5 μM dbcAMP for 7 days. Medium was replaced with medium without TPA, bFGF and dbcAMP 72 hours prior to RNA isolation. Real-time PCR of wild-type and Nf1^+/– melanocytes was performed using TaqMan primers and probes for Tyr, Tyrp1, Dct, and Mitf. Primers and probes for Tyr(NM_011661), Tyrp1 (NM_031202), Dct (NM_010024), Mitf (NM_008601), β-actin (NM_001101) were designed using Applied Biosystems Assay-by-Design primer design software. RNA was isolated and quantitative PCR performed as previously described (Urzua et al., 2006) with β-actin as a reference transcript assayed under identical conditions respective to the gene of interest.

1 μg of total RNA isolated from 10-day old mixed wild-type and Nf1^+/– cultures was reverse transcribed to cDNA using the Superscript First Strand Synthesis System for reverse transcriptase (RT)-PCR (Invitrogen). RNA was treated with DNase prior to cDNA synthesis according to the Superscript protocol. A PCR mastermix consisting (per 20 μl volume unit) of 1.5 × RT buffer (Invitrogen), 1 × SYBR Green 1 (Molecular Probes, Invitrogen Detection Technologies), 5% DMSO, 0.2 μl of Hot MasterTaq DNA polymerase (Eppendorf, Hamburg, Germany), and cDNA reverse transcribed from 50 ng of total RNA was prepared for each sample. Primer sequences based on GenBank accession numbers NM_011661 (Tyr), NM_031202(Tyrp1), NM_010024(Dct/Tyrp2) and NM_001101 (β-actin) were designed using Primer Express (Applied Biosystems, Foster City, California) as previously described (Lanning et al., 2005). The primers for Tyr, Tyrp1, Dct, Mitf and β-actin were tested for primer-dimer formation prior to their use in real-time PCR analysis. The final primer pair concentration was 0.3 μM. Reactions in a total volume of 30 μl per tube were run for 40 cycles as follows: 96°C, 15 seconds; 60°C, 1 minute. The relative expression levels of Tyr, Tyrp1 and Dct were determined for RNA isolated from Nf1^+/– cultures by normalizing them to the expression levels of Tyr, Tyrp1 and Dct from wild-type cultures. β-actin was used as internal control for Nf1^+/– and wild-type samples. The fold change in Tyr, Tyrp1 and Dct gene expression was calculated using the 2^-ΔΔCt method (Lanning et al., 2005; Livak and Schmittgen, 2001).

FACS-separated wild-type and Nf1^+/– melanocytes were plated in a 24-well plate at a density of 1.3 ×10⁴ cells per well or 0.8 ×10⁴ cells per well. TPA-, bFGF- and dbcAMP-containing MGM was replaced with medium without TPA, bFGF and dbcAMP 72 hours prior to RNA isolation and the MEK inhibitor PD 98059 (50 μM, Calbiochem) was added 12 hours prior to RNA isolation with Trizol. Quantitative real-time PCR analysis was performed using TaqMan primers and probes for Tyr, Tyrp1, Dct, Mitf and β-actin (Applied Biosystems) as previously described. Wild-type and Nf1^+/– cells treated with an equal volume of DMSO were used as controls.

Wild-type and Nf1^+/– mixed primary cultures were replaced with MGM lacking TPA, bFGF and dbcAMP, and were treated with the MEK inhibitor PD98059 (50 μM final concentration) (Wang et al., 2001) 12 hours prior to cell lysate preparation. Murine SCF (Peprotech) at 20 ng/ml was added to selected cultures for 15 minutes. Cell lysates obtained were used to evaluate MAP kinase activation by western blot analysis using mouse monoclonal antibody against phosphorylated Erk (1:100; Santa Cruz Biotechnology) and rabbit polyclonal anti-Erk-2 (1:200; Santa Cruz Biotechnology).

Melan-a cells (gift of Dorothy Bennett, St. George's Hospital Medical School, London, UK) (Bennett et al., 1987) were cultured in medium consisting of 10% FBS and 40nM TPA in RPMI-1640 medium. Prior to cell lysis, cells were treated for 12 hours or 24 hours with 50 μM PD98059. Trizol-isolated RNA from cells either incubated with or not incubated with PD98059 was amplified using primers specific for Tyr, Tyrp1 and Dct expression using the SYBR green quantitative real time PCR protocol previously described for mixed primary cultures.

Melan-a cells cultured as above were transfected with three different siRNA duplexes (Dharmacon) using the following conditions. One day before transfection, cells were plated in 2.5 ml of growth medium without antibiotics at 30% confluency in six-well plates. For each transfection in a six-well plate using Oligofectamine (Invitrogen), the manufacturer's protocol was used to transfect cells with 10 μl of 20 μM siRNA duplex dissolved in a 190 μl final volume stock solution. This was added to the cells and they were incubated at 37°C in a CO₂ incubator for 4 hours. RNA was isolated 48 hours after transfection and relative levels of Tyr, Tyrp1 and Dct gene expression determined by real-time reverse transcriptase (RT)-PCR.

siRNA duplexes used were (1) sense 5′-UGAUGAUGCCAAACGACAAdTdT-3′, antisense 3′-dTdTCUACUACGGUUUGCUGUU-5′; (2) sense 5′-GCCAAAAUGGAAGAUGGCCdTdT-3′, antisense 3′-dTdTCGGUUUUACCUUCUACCGG-5′; (3) sense 5′-GUGGAUCCUACCAGGUUAGUUdTdT-3′, antisense 3′-dTdTCACCUAGGAUGGUCCAAUCAA-5′. siRNA duplexes 1, 2 and 3 targeted exons 40, 20 and 29, respectively, of the mouse Nf1 transcript (Genbank Acc. No. NM_010897). A scrambled siRNA sequence 5′-UGAUAGUCGAACGCAAACdTdT-3′ was used as a control.

Western blotting was used to determine the efficiency of siRNA-mediated knockdown of neurofibromin in melan-a cells. Cell lysates were collected 72 hours post-transfection with neurofibromin siRNAs and used for SDS-PAGE and western blotting. The transfection was carried out using the protocol described above. Western blot analysis was performed using rabbit polyclonal anti-neurofibromin antibody (1:100, Santa Cruz Biotechnology).

Real-time PCR primers for this experiment were: Tyr Fwd 5′-CGGCCAACGATCCATT-3′ and Rev 5′-TGCCTTCGAGCCATTG-3′; Tyrp1 Fwd 5′-ATTTCTCTCACGAAGGACCCGCTT-3′ and Rev 5′-ATTGGTCCACCCTCAGTGCTGTTA-3′; Dct/Tyrp2 Fwd 5′-CAACCGGACATGCAAATG-3′ and Rev 5′-TGTGGTGCCAGATCTTCTCCATGT-3′; β-actin Fwd 5′-TGTGATGGTGGGAATGGGTCAGAA-3′ and Rev 5′-TGTGGTGCCAGATCTTCTCCATGT-3′. The Tyr primer spans exons 1 and 2, Tyrp1 exons 2 and 3, Dct exons 1 and 2, β-actin exons 3 and 4. Experimental details for real-time PCR analysis have been described previously (Lanning et al., 2005).

Mixed primary cultures from Nf1^+/– and wild-type neonates were initiated as described and used within the second passage. Medium was replaced with medium lacking TPA, dbcAMP and bFGF, 72 hours prior to the preparation of cell lysates. Murine SCF/Kit ligand (Peprotech) at 20 ng/ml final concentration was added to Nf1^+/– and wild-type cells for 5 minutes and 15 minutes prior to cell lysis. Cell lysates were prepared by rapidly scraping the cells in PBS and centrifuging twice for 2 minutes at 1000 g. Cell pellets were suspended in RIPA buffer containing 10 μM Na₃VO_4, 200 μM PMSF and 10 μg/ml aprotinin and incubated in ice for 20 minutes and spun at 20,000 g for 2 minutes at 4°C. Protein concentrations were determined by Bio-Rad protein assay and 5 μg of total protein was solubilized in 2× SDS-sample buffer (Bio-Rad) and heated for 5 minutes. Following SDS-PAGE and transfer of the western blot to PVDF membrane, the blots were blocked in 5% nonfat dry milk (Carnation) in 0.05% Tween 20 in TBS and probed using specific antibodies. MAP kinase pathway activity was determined using mouse monoclonal antibody against phosphorylated Erk (1:100, Santa Cruz Biotechnology) and rabbit polyclonal anti-Erk2 (1:200; Santa Cruz Biotechnology). Proteins were detected with peroxidase-conjugated secondary antibody and chemiluminescence reagents (Pierce). Relative levels of melanogenic protein expression were determined using the following antibodies: rabbit polyclonal anti-Tyrp1 (a-PEP1, 1:200), gift of Vincent Hearing, NCI, Bethesda, MD; rabbit polyclonal anti-Dct/Tyrp2 (α-PEP8, 1:200) (Tsukamoto et al., 1992); goat polyclonal anti-tyrosinase (1:100, Santa Cruz Biotechnology, M-19); mouse monoclonal anti-Mitf (1:10, C5, ascites, gift of David Fisher, Dana-Farber Cancer Institute, Boston, MA) and rabbit polyclonal anti-neurofibromin (1:100, Santa Cruz Biotechnology)

C57BL6-Nf1^+/– mice were crossed with [Tg]Dct-LacZ mice (Hornyak et al., 2001), and wild-type and Nf1^+/– skins from neonatal litters at P1 were fixed in 0.4% PFA in PBS for 90 minutes at 4°C and rinsed with PBS (3 × 10 minutes), and incubated overnight with X-gal staining solution. The following day, the skins were washed with PBS (3 × 10 minutes) and post-fixed in 4% PFA in PBS. Wild-type and Nf1^+/–LacZ-positive stained skins were counterstained with eosin-hematoxylin prior to embedding in paraffin and sectioning in a paravertebral orientation. Sections were visualized and photographed using bright-field microscopy.

The technical assistance of Mathew Abraham is appreciated. We thank Karlyne Reilly (NCI-Frederick, Frederick, MD) for providing some Nf1^+/– mice. This work was supported by DAMD17-01-1-0709 (Department of Defense Neurofibromatosis Research Initiative) and by NIH R01 AR47951 (to T.J.H.). It was also supported in part by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research.