Colour genes, oncogenes and melanocyte differentiation

Pigmentation has long been a favourite and easily used marker in the genetics of diverse organisms. Mammalian integumental pigments are melanins, synthesized by melanocytes in the epidermis and hair bulbs. Our understanding of mammalian pigmentation genes has been advanced significantly in the last few years (Hearing and Jiménez, 1989), partly by the advent of methods for the culture and immortalization of melanocytes. In cultured melanocytes, homozygous recessive germline mutations can be visibly expressed, complemented or even revert spontaneously, as will be discussed. In the mouse there are over 130 mutations that affect coat colour, mapping to well over 50 loci (Silvers, 1979). For the sake of brevity, most of this commentary will be concerned with the recent molecular characterization of mutations at just four of these loci, two of which encode developmentally controlled, melanocyte-specific products while the other two play a part in melanocyte development.

The corresponding cancer, melanoma, makes an excellent system for the study of cell differentiation and its relationship to malignancy, partly again because of the ready visibility of melanin (Bennett, 1989). The commentary will conclude with discussion of some relationships between malignancy and cell differentiation in melanocytes transformed by oncogenes or other agents.

Although short-term cultures of normal human melanocytes were obtained by Hu et al. (1957) and others, longterm cultures were first reported by Eisinger and Marko (1982), the key components of their culture medium being cholera toxin and the phorbol ester tetradecanoyl phorbol acetate (TPA), also known as PMA. Human diploid melanocytes senesced like other human cell types. However, murine melanocytes grown in similar media with TPA proved capable of spontaneous immortalization, and three groups of workers soon established murine melanocyte lines (Sato et al. 1985; Bennett et al. 1987; Tamura et al. 1987). The first such lines were from black mice and were followed by lines from mice carrying albino (c) and other germline coat-colour mutations (Abe et al. 1986; Halaban et al. 1988; Bennett et al. 1989).

Immortal melanocytes usually retain a high level of differentiated function in culture, synthesizing pigment characteristic of the mice of origin. For example, melanocyte lines from black (.nonagouti, a/a) and brown (b/b) mice produce black and brown pigment, respectively (e.g. see Sato et al. 1985; Bennett et al. 1989). Even albino melanocytes produce abundant premelanosomes, the unpigmented precursors of melanosomes (pigment organelles). Immortal melanocytes also tend to retain other properties of their normal counterparts including a diploid or near-diploid chromosome number, dependence on TPA for growth in standard culture media, inability to grow in suspension and lack of tumorigenicity in syngeneic or nude mice (Bennett et al. 1987; Dotto et al. 1989; Denner and Knowles, personal communication). It is these properties that have made such cultures useful for studies of both pigmentary gene expression and oncogenesis.

It was long suspected but not proven that albino (pinkeyed white) mutations in a number of organisms affected the gene for tyrosinase, the principal enzyme in melanin biosynthesis (Silvers, 1979). Recently, both mouse and human tyrosinase genes were cloned and sequenced, the original cDNA clones having been obtained from cultured melanocytes (Yamamoto et al. 1987; Kwon et al. 1987). Active cDNAs were identified by the property of transferring tyrosinase activity by transfection to cells that lacked it (Müller et al. 1988; Yamamoto et al. 1989). The mapping of the sequence at or near the mouse albino (c) locus on chromosome 7 (Kwon et al. 1987), its absence in mice with albino deletions (Ruppert et al. 1988) and especially its restoration of pigmentation on transfection to albino melanocytes (Yamamoto et al. 1989) combined to identify the c locus as containing the tyrosinase structural gene. Some functional aspects of the protein sequence, to be discussed, are shown in Fig. 1A (top half).

Kwon et al. (1989) pointed out a single G to C change, resulting in a Cys to Ser substitution, in the tyrosinase sequence derived from albino mice compared with the wild-type sequence (Fig. IB). Although only part of the albino sequence was described, so that further differences were possible, this Cys85 (numbering of Müller et al. 1988) falls in a region highly conserved between tyrosinases and tyrosine-related proteins (Müller et al. 1988; Hearing and Jiménez, 1989) (see next section and Fig. IB). The substitution is thus a candidate for the effective mouse albino mutation.

The spontaneous appearance of black, apparently revertant cells in a line of albino melanocytes (Fig. 2) provided a test for this (Jackson and Bennett, 1990). The black cells were cloned to give a black subline. White remutant cells were detected in this culture and were also subcloned. Amplification and sequencing of part of the tyrosinase gene from these sublines showed that the revertant line was heterozygous for an exact reversion to the wild-type codon for Cys85, while the white re-mutant sublines had lost this reversion, by allele loss. Thus this point substitution is sufficient to produce the mouse albino mutation (Jackson and Bennett, 1990). A restrictionfragment-length polymorphism revealed the same base change at this position in 15/15 other albino mouse strains, indicating a common genetic origin for all these strains (Jackson and Bennett, 1990).

Yamamoto et al. (1989) and Kwon et al. (1989) have published about 2500 and 300 bases, respectively, of the sequence 5’ to mouse tyrosinase, which are in agreement. This region contains various notable features including TATA- and CCAAT-like elements, palindromes, a long region of GA repeats, and a sequence with 8/10 homology to the consensus thyroid/retinoid-responsive element. Yamamoto et al. (1989) have constructed a tyrosinase ‘minigene’ capable of inducing pigmentation on expression in albino melanocytes, by linking their 5’ sequence to a full-length cDNA. They have reported tissue-specific expression of this construct in transgenic mice (Tanaka et al. 1990). The requirements for such expression can now be analysed, as well as those for the stimulation of tyrosinase gene transcription by melanocyte-stimulating hormone (MSH) or cyclic AMP (Kwon et al. 1988; Hoganson et al. 1989).

The first tyrosinase-related gene was identified after a sequence designated pMT4 was cloned from mouse melanoma cDNA using monoclonal anti-tyrosinase antibodies; it was thought originally to be a tyrosinase cDNA (Shibahara et al. 1986). However, pMT4 could not confer tyrosinase activity on transfected cells (Müller et al. 1988), and was mapped to the murine brown (b) locus (chromosome 4) rather than the c locus (Jackson, 1988). The sequence nonetheless showed high homology to authenticated tyrosinases (Müller et al. 1988), and was designated tyrosinase-related protein-1 or TRP-1 (Jackson, 1988). (A second related sequence, TRP-2 (Jackson, 1988) will not be discussed here.)

The wild-type TRP-1 sequence was inserted into a retroviral vector, which was used to infect cultures of immortal melanocytes homozygous for the brown (b) mutation (Bennett et al. 1990). These cells produce mutant brown pigment in culture. However, many cells with black to dark brown (wild-type) pigment were observed in cultures expressing the exogenous gene. Thus TRP-1 complements the brown mutation and is the product of this locus (Bennett et al. 1990). A human cDNA sequence for TRP-1 has also been reported (Cohen et al. 1990).

The function of TRP-1 is not known, although Hearing and Jiménez (1989) do not dismiss the possibility that it is another tyrosinase, suggesting that the failure to confer this activity by transfection could be for trivial reasons. The protein sequence is about 40 % homologous with that of tyrosinase and shares its known functional characteristics (Fig. 1A), including a leading signal sequence, a trans-membrane region, nearly all cysteine residues (indicating similar secondary structure) and two potential copper-binding sites (Müller et al. 1988; Jackson, 1988; Hearing and Jiménez, 1989). TRP-1 may thus be another pigmentary enzyme, located with tyrosinase on the inner surface of the membrane of the melanosome. It has also been reported to have catalase activity, which could prevent bleaching of melanin by the hydrogen peroxide that is a by-product of melanin synthesis (Halaban and Moellmann, 1990). As with tyrosinase, expression of TRP-1 appears to be specific to pigment-synthesizing cells (Jackson, 1988). Its genomic and flanking sequences remain to be described.

The sequences of cDNAs for the murine brown mutant TRP-1 gene and an induced revertant to wild-type (partial sequence) have been reported (Zdarsky et al. 1990). There are several differences between the brown and (original) wild-type sequences, but only one of these is restored in the revertant, showing that it is the critical mutation. This is in the codon for a conserved cysteine residue, Cys86 which, interestingly, falls very close to the Cys85 of the albino mutation of tyrosinase when the two protein sequences are aligned by homology (Zdarsky et al. 1990; Fig. IB). Both cysteine residues lie within a region particularly highly conserved between the two proteins, and presumably vital for protein function, although without any identified external homology (Fig. IB). There are several other such regions (not shown; Zdarsky et al. 1990).

An interesting link between malignancy and development was found in the identification of the murine dominant white spotting (W) locus with the proto-oncogene c-kit, the oncogenic homologue of which is carried by a feline sarcoma virus. Mutant alleles of W, even when heterozygous, reduce the numbers of melanocytes developing in the skin, in some cases severely. Germ cells and some haemopoietic stem cells are similarly affected, from early embryonic to adult stages (Silvers, 1979). After c-kit was mapped close to the W locus (mouse chromosome 5), c-kit genomic or cDNA sequences from a number of W mutant strains were analysed by Southern blotting or sequencing. Some restriction fragment patterns and all the sequences were found to differ from those of normal c-kit (Geissler et al. 1988; Nocka et al. 1990). This constituted convincing evidence for the identity of the two genes, c-kit transcripts are detectable in cell types including melanocytes, mast cells and the other known targets of W mutations (Nocka et al. 1989). Now, the c-kit product has tyrosine kinase activity, and from its sequence is a trans-membrane receptor, with homology to a family of growth factor receptors (Nocka et al. 1990; Stenman et al. 1989). Interestingly a different receptor in the same family, the platelet-derived growth factor A (PDGF-A) receptor, is encoded close to the human c-kit gene (Stenman et al. 1989).’

Several W mutations are associated with impaired tyrosine kinase activity in immunoprecipitable c-kit (Nocka et al. 1990), and impaired or absent responses of mast cells and pluripotent haemopoietic stem cells to a growth- and differentiation-inducing factor produced by cultured cells of several types (Witte, 1990). These responses have now been used by several groups to purify mouse and rat ligands, and sequence the mouse, rat and human genes, as reported in a striking set of eight articles (Martin et al. 1990, and adjacent papers, reviewed by Witte, 1990). The ligands appear to be soluble or membrane-bound products of the same novel gene, despite receiving various names including MGF (mast cell growth factor), KL (kit ligand), SLF (Steel factor) and SCF (stem cell factor) (Witte, 1990). This author prefers SCF, as the other terms each describe only one aspect of the factor - and MGF already denotes a melanocyte growth factor from brain. It is satisfying that the SCF gene maps to the Steel (SI) locus (mouse chromosome 10) and is deleted in mutant alleles of SI, while its product corrects haemopoietic deficiencies in SI mutant mice (Witte, 1990). This is because the SI locus was widely speculated to encode a W ligand as its mutations have similar phenotypes to W mutations, yet SI melanocytes and other target cells are normal if transplanted into normal mice (Silvers, 1979). In summary, SCF is encoded at the SI locus, its receptor at the W/c-kit locus, and both are required for the normal development of germ cells, various haemopoietic stem cells and melanocytes.

There is a consistent proliferative difference between melanocytes and melanoma cells in culture, which is proving useful although it is not understood. Melanocytes (mouse or human) will not grow in a standard culture medium supplemented with serum, but they will grow if TPA is also added. However, many melanomas can grow in such a standard medium, and in general the growth of metastatic melanomas is inhibited by TPA. Several groups have ‘transformed’ immortal murine melanocytes with exogenous oncogenes or chemical carcinogens, and have described properties of the resulting cells including their responses to TPA. The results fell into four patterns: (1) there was no detectable effect on the melanocytes, reported only for an activated p53 gene (Dotto et al. 1989). (2) Cells expressing oncogenes v-Ha-ras, v-neu or adenovirus gene Ela, and cells transformed by two chemical carcinogens, all showed an alteration from TPA dependence to independence (or to inhibition of growth in the case of Ha-ras), combined with a loss of pigment and the acquisition of the ability to grow in suspension or (where tested) to form tumours in mice (Wilson et al. 1989; Dotto et al. 1989; Denner and Knowles, personal communication). (3) Clones expressing v-rnyc or polyoma middle T antigen initially showed a mitogenic response to TPA, and were pigmented. However, growth soon accelerated in the absence of TPA, while a loss of pigmentation of the cultures was observed (Dooley et al. 1988; Dotto et al. 1989). These changes were observed in each of four initially pigmented clones expressing middle T antigen (B. Nester and DCB, unpublished data). All these types of unpigmented cultures were tumorigenic (Dooley et al. 1988; Dotto et al. 1989). Pattern (4) was reported only for melanocytes ectopically expressing basic fibroblast growth factor (bFGF) rather than an oncogene. These cells became independent of TPA and unpigmented, but not tumorigenic (Dotto et al. 1989).

Two aspects of this work will be emphasized. The first is the loss of pigmentation, which in 6/7 treatments was associated with tumorigenicity. To put it another way, all cultures that became malignant also became unpigmented. One interesting question about this is that of whether the melanocytes dedifferentiated. While the cells described by Dotto et al. (1989) were uncloned, the other groups used the cloned line melan-a (Bennett et al. 1987), or a subline of it (mel-ab; Dooley et al. 1988), so all cells were progeny of one black melanocyte. Denner and Knowles (personal communication) tested for the expression of the melanocytic mRNAs for tyrosinase and TRP-1. They detected both gene products in melan-a cells, but neither was found in depigmented cells transformed by chemicals, v-Ha-ras or Ela. This indicates some dedifferentiation at the level of transcription, which would not be unduly surprising in view of the evidence for instability of differentiation in melanoma cells (Bennett, 1989).

The second point to consider is a comparison between patterns (1) and (2) above. In pattern (2), cells seem to undergo at least two steps of alteration, with unpigmented cells emerging from and overgrowing initially pigmented transfected clones. Thus, although melanocytes expressing these oncogenes become malignant, some extra spontaneous cellular change may play a part. However, in pattern (1), malignancy may have arisen in one step with integration of v-ras or v-neu genes, although this has not been proven. It should be remembered however that the melanocytes used were already immortal, i.e. not entirely normal. Thus it cannot be inferred that melanoma can be caused by any dominant oncogene, although up to 25 % of human melanomas are reported to have one or more activated ras genes (Shukla et al. 1989). Indeed, observations including the high frequencies of breakpoints in chromosomes 1 and 6 in melanoma, the existence of a familial form of the disease (Dracopoli and Bale, 1988; Trent et al. 1989) and the loss of malignancy of human melanoma cells following the réintroduction of a normal human chromosome 6 (Trent et al. 1990) provide evidence for involvement of the loss of a normal gene.

The genetics of pigmentation and of pigment-cell malignancy have both entered a phase of analysis at the molecular level. Moreover, a region of overlap has been indicated, in which exogenous oncogenes can affect the expression of two coat colour genes, while another coat colour gene actually is a proto-oncogene. This area of overlap has yielded some fascinating results recently and, as indicated by a surge of related research, promises to be a fruitful field in the near future.

The author is grateful to Joachim Denner, Ian Hart and Ian Jackson for communication of results prior to publication, and to Ian Jackson for valuable comments on the manuscript. Research in the author’s laboratory is supported by the Cancer Research Campaign and the Wellcome Trust.