Modeling malignant melanoma in mice: pathogenesis and maintenanceLynda Chin*,11Department of Dermatology, Harvard Medical School, Department of Adult Oncology, Dana-Farber Cancer Institute, 44 BinneyStreet, M413 Boston, Massachusetts, MA 02115, USAKeywords: melanoma; RAS; tumor maintenance;mouse modelMalignant melanoma, current challengesThe signi®cance and impact of melanoma as a diseaseentity cannot be understated. As a cancer type,incidence of melanoma is rising at a rate second onlyto lung cancer in women and, on its present course, thelifetime risk will reach 1 in 75 among Caucasians in theUS by the year 2000 (Rigel et al., 1996). Furtherm ore,in contrast to most malignancies, melanoma aects amuch younger population, metastasizes early in thecourse of the disease, and fails to respond to currenttherapeutic regimens (Herlyn, 1993). Presently, clin-icians rely solely upon histopathologic criteria to stagemelanomas, guide therapeutic decisions, and predictdisease outcomes. One system, described by Breslow,measures tumor thickness from the uppermostnucleated layer of the epidermis (i.e., granular celllayer) to the greatest depth of tumor invasion. Anothersystem, known as Clark's classi®cation, correlatesanatomic level of invasion, coupled with mitoticindex, to prognosis. Such systems serve as usefulprognosticators but fall short as de®nitive pred ictorsof future clinical behavior of mela noma, particularlyearly melanocytic lesions ± an ongoing problem fuele din part by the lack of diagnostic reproducibility amongdermatopathologists evaluat ing such lesi ons (for re-view, see Barnhill et al., 1993). The inherent short-comings of pathology-based parameters havestimulated the search for molecular markers that canpredict more accurately the impending tumor pheno-type, forecast disease prognosis , and ultimately guidetherapeutic decisions.Despite the long history of clinical and moleculareorts directed towards this disease, surprisingly littleis known about the precise genetic lesions leading tomelanoma and even more vestigial is our under-standing of how the few recognized genetic lesionsrelate to disease classi®cation or progression. Theidenti®cation and validation of molecular markers hasbeen hampered by practical issues such as human tissueacquisition and would undoubtedly be accelerated bythe availability of a mouse model that faithfullyreproduces melanoma progression on the pathologicaland molecular levels. Furthermore, many of the geneticlesions identi®ed to date, with few exceptions, havecome from the cytogenetic analysis of far-advancedhuman melanomas, particularly metastatic disease.This fact, coupled with the inherent genomic instabil-ity of malignant melanoma, emphasizes the need toutilize primary, early stage disease samples. Moreover,a better genetic roadmap of tumor progression fromnormal melanocyte to metastatic melanoma will benecessary in order to correlate molecular events withtumor stages. On another level, the melanoma ®eld hasa need for in vivo systems to aid in the discovery,veri®cation and functional analysis of the growingcollection of melanoma susceptibility genes.In human studies, cytogenetic, linkage, and mole-cular analyses of familial and sporadic melanomashave implicated many genes and genomic loci in thegenesis and progression of melanoma, although veryfew of these candidate genes have been veri®ed as trueetiologic lesions in a formal genetic sense. In surveyingthe putative melanoma genes/loci for which reasonableexperimental support exists, two consistent themesemerge ± loss of 9p 21 locus, and activation of areceptor tyrosine kinase (RTK), such as EGFR, ormutations in its signaling surrogates such as RAS orPTEN. In this review, we will discuss our ongoingeorts to establish an accurate mouse model based onthese known genetic pro®les of human mela nomas.Genetics of human melanomasThe familial melanoma locus at 9p21: p16INK4a, p19ARFand p15INK4bThe 9p21 locus in man, and the syntenic region onmouse chromosome 4, contains the closely linked andhighly homologous INK4a and INK4b genes, encodingfor p16INK4aand p15INK4b, respectively. Both products ofthese genes inhibit CDK4/6-directed phosphorylationof pRB. Loss of INK4 function can directly driveCDK4/6 towards a more highly activated stateresulting in pRB hyperphosphorylation with asso-ciated de-repression of pRB-regulated genes (reviewedin DePinho, 1998), as well as enhanced activity ofCDK2/cyclin E by indirectly promoting sequestrationof the Cip/Kip proteins (p21 and p27) from CDK2/cyclin E (reviewed in Sherr and Roberts, 1999) leadingto S phase progression. The INK4a gene has thecapacity to encode a second unrelated non-CKIproduct by alternative exon usage (Quelle et al.,1995). This additional open reading frame initiates ina dierent ®rst exon (exon 1b) and continues in analternative reading frame in the shared exon 2 (Figure1). The novel protein was named `p19ARF' forAlternative Reading Frame protein. ARF functions asa poten t growth suppressor (Quelle et al., 1995, 1997),*Correspondence: L ChinOncogene (1999) 18, 5304 ± 5310ã1999 Stockton Press All rights reserved 0950 ± 9232/99 $15.00http://www.stockton-press.co.uk/oncblocks oncogenic transform ation and sustains p53-dependent apoptosis in RB null cells that have re-entered the cell cycle in vivo (Pomerantz et al., 1998) orin the setting of hyperproliferative oncogenic signals(de Stanchina et al., 1998; Zindy et al., 1998; Radfar etal., 1998). That p19ARFmay reside in a p53 pathwaywas ®rst proposed by genetic studies in the INK 4aD2/3(p16INK4aand p19ARF) null mouse showing an absence ofp53 mutat ions in melanomas (Chin et al ., 1997).Analogous genetic observations were demonstrated inp19ARF-de®cient MEFs that retained wildtype p53con®guration upon immortalization (Kamijo et al.,1997). These ®ndings, coupled with numerous assaysshowing that the anti-growth and anti-on cogenicactivities of ARF function in a p53-dependent manner(Kamijo et al., 1998; Pomerantz et al., 1998; Stott etal., 1998; Zhang et al., 1998), have forged a direct linkbetween the `INK4a/ARF' locus and the RB (p16INK4a)and p53 (p19ARF) (Figure 1b). Thus, tumor-associatedmutations at the INK4a/ARF locus have the potentialof crippling the two most important tumor suppressionpathways governing neoplasia, an arrangement thatlikely provides a basis for the prominent involvementof INK4a in human cancers.In melanoma, the 9p21 region commonly