REVIEW ARTICLE

Pathways to Melanoma Development: Lessons from theMouse

Graeme J. Walker and Nicholas K. HaywardQueensland Cancer Fund Research Unit, Joint Experimental Oncology Program, Queensland Institute of Medical Research, Post Of®ce Royal

Brisbane Hospital, Brisbane, 4029, QLD, Australia

Because of subtle differences between mouse andhuman skin, mice have traditionally not been anideal model to study melanoma development.Understanding of the molecular mechanisms ofmelanoma predisposition, however, has been greatlyimproved by modeling various pathway defects inthe mouse. This review analyzes the latest develop-ments in mouse models of melanoma, and sum-marizes what these may indicate about thedevelopment of this neoplasm in humans. Mutationsof genes involved in human melanoma have beenrecapitulated with some unexpected results, particu-larly with respect to the role of the two transcripts(Ink4a and Arf) encoded by the Cdkn2a locus. Boththe Ink4a/pRb and Arf/p53 pathways are involved inmelanoma development in mice, and possible mech-anisms of cross-talk between the two pathways arediscussed. We also know from mouse models that

Ras/mitogen-activated protein kinase pathway acti-vation is very important in melanoma development,either through direct activation of Ras (e.g., HrasG12V), or via activation of Ras-effector pathways byother oncogenes (e.g., Ret, Hgf/Sf). Ras can co-operate with the Arf/p53 pathway, and probably theInk4a/Rb pathway, to induce melanoma. Thesethree growth regulation pathways (Ink4a/pRb, Arf/p53, and Ras/mitogen-activated protein kinase)seem to represent three major ``axes'' of melanomadevelopment in mice. Finally, we summarize experi-ments using genetically modi®ed mice that havegiven indications of the intensity and timing of ultra-violet radiation exposure that may be most respon-sible for melanoma development. Key words: knockoutmice/melanoma/transgenic mice. J Invest Dermatol119:783±792, 2002

Historically, the mouse has not been a good model forstudying melanoma. Advances in methodologies togenerate transgenic and knockout animals, how-ever, have enabled researchers to assess the in vivoeffects of pathways thought to be dysregulated in

human melanoma development. Furthermore, new experimentalultraviolet radiation (UVR) treatment regimens have shed somelight on the environmental events thought to contribute to thegenesis of this neoplasm.

Melanoma is a tumor of pigment-producing cells (melanocytes)that differentiate from neural crest progenitor cells during embry-onic development. Melanocytes are located in the skin, hair

follicles, stria vascularis of the inner ear, and uveal tract of the eye.Ocular (uveal) melanomas arise from neural crest derivedmelanocytes within the eye. The only pigment cells not arisingfrom neural crest precursors are those of the retinal pigmentepithelium (RPE), which are derived from epithelial cells in theoptic cup. During embryonic development neural crest cellsdifferentiate within the embryo through two major pathways: (i)the ``ventral'' pathway, which gives rise to neurons and glial cells ofthe peripheral nervous system, and (ii) the ``dorsolateral'' pathwayfrom which pigment cells arise. Pigment cell precursors (melano-blasts) migrate ®rst to the dermis and differentiate (possibly underthe in¯uence of a-melanocyte-stimulating hormone and otherfactors), then to the epidermis (Quevedo and Fleischmann, 1980).

In contrast to humans, in which there is a strong associationbetween UV exposure and melanoma development (Whiteman etal, 2001), normal adult mice do not develop melanomas, even afterchronic exposure to UVR. Traditionally, shaved or hairless micehave been exposed to various acute (intense, short-term) andchronic (low level, long-term) UV treatments to simulate humansun exposure. These treatments promote various types of skincancers, including squamous cell carcinoma, papilloma, and®brosarcoma, but not melanoma (Gallagher et al, 1984). Evencotreatment with carcinogens such as 7,12-dimethylbenz[a]anthra-cene (DMBA) promote little or no melanoma development,although there is a signi®cant increase in the frequency of the otherskin cancers. The architecture of mouse and human skin showssome subtle differences, and it is thought that differences inmelanocyte structure and location within the skin may be

Manuscript received May 31, 2002; accepted for publication June 7,2002

Reprint requests to: Dr N.K Hayward, Cancer Unit, QueenslandInstitute of Medical Research, Post Of®ce Royal Brisbane Hospital,Herston, Qld 4029 Australia. Email: NickH@qimr.edu.au Explanationof CDK-inhibitor nomenclature/abbreviations: The cyclin-dependentkinase inhibitor 2A (CDKN2A) gene encodes the p16INK4A (INK4A)and p19ARF (ARF) proteins. Gene names are italicized, whereas proteinsare not. Upper case signi®es genes or proteins human and title case (e.g.,Cdkn2a) indicates orthologues in mouse. Similarly, other CDK inhibitorsinclude CDKN2B, which encodes p15INK4B; CDKN2C, p18INK4C;CDKN2D, p19INK4D; CDKN1A, p21CIP1; CDKN1B, p27KIP1 andCDKN1C, p57KIP2.

Abbreviations: RPE, retinal pigment epithelium; DMBA, 7,12-dimethylbenz[a]anthracene; TSG, tumor suppressor gene; PTK, proteintyrosine kinase; CDK, cyclin-dependent kinase; OMM, ocular melanomas;CMM, cutaneous melanomas; RTK, receptor tyrosine kinase.

0022-202X/02/$15.00 ´ Copyright # 2002 by The Society for Investigative Dermatology, Inc.

783

responsible for the inability to induce melanomas in mice withUVR. Epidermal melanocyte numbers increase substantially afterbirth for about 2 wk; however, in mouse skin, but not human skin,nonfollicular melanocyte numbers decline as they follow thebasement membrane in the invagination process involved in theformation of hair follicles (Hirobe, 1995). Thus in adult mice mostmelanocytes are located in hair follicles (Hirobe, 1995), althoughsome are still located in the epidermis, particularly in nonhairy skin(e.g., ears, footpads). It is thought that due to their location,follicular melanocytes may be more protected from UVR, althoughthis has not been proven.

A number of genetic abnormalities involved in human mela-noma susceptibility and tumor progression have been recapitulatedin mice. The purpose of this review is to outline the latestdevelopments in mouse models of melanoma, and to summarizewhat these may indicate about the development of this neoplasm inhumans. To begin, genes known to be involved in the develop-ment of human melanoma will be discussed, followed by asummary of each of the mouse models of melanoma generated todate. Lastly, the role of UVR in melanoma genesis in these animalswill be critically assessed.

GENES AND PATHWAYS INVOLVED IN HUMANMELANOMA DEVELOPMENT

The hallmark of solid tumor development is the acquisition ofmultiple genetic defects involving the inactivation of tumorsuppressor genes (TSG) and the activation of oncogenes. Incutaneous melanoma, nonrandom deletions and rearrangements areseen in several chromosomal regions, including 1p, 7q, 9p, 10q, and11q (Dracopoli and Fountain, 1996). At present, the two mostimportant TSG involved in human melanoma are CDKN2A andPTEN, mapping to 9p and 10q, respectively. Potential oncogenesinclude CDK4, NRAS, and epidermal growth factor receptor(EGFR), and various protein tyrosine kinases (PTK), such as EPH-A2 and EPH-B3, that are overexpressed in up to 90% of melanomacell lines (reviewed in Easty and Bennett, 2000). Interestingly,other PTKs, such as KIT and FES, are consistently downregulatedin melanoma cell lines (Easty and Bennett, 2000).

CDKN2A, mapping to a frequently deleted region of 9p21,encodes p16INK4A (hereafter termed INK4A), a cyclin-dependentkinase (CDK) inhibitor that binds to and inhibits CDK4 andCDK6. When complexed with D-type cyclins these kinases driveentry into the cell cycle by phosphorylating the retinoblastomafamily of proteins (pRb), which causes the release of E2Ftranscription factors and expression of E2F-regulated genes, therebyallowing progression from G1 to S phase. CDKN2A can beinactivated by several mechanisms, homozygous deletion, mutation

or promoter methylation. Such alterations are frequently detectedsomatically in sporadic melanomas (e.g., Dracopoli and Fountain,1996; Pollock and Trent, 2000), and constitutionally in familialmelanoma patients (e.g., Hussussian et al, 1994; Kamb et al, 1994).Loss of INK4A function partially inactivates the G1 block, and alsoleads to escape from cell senescence in culture.

Further evidence implicating the INK4A/cyclin D/CDK4/pRbpathway comes from the ®nding of two rare germline CDK4mutations in melanoma kindreds (Zuo et al, 1996; Sou®r et al,1998). Both mutations (R24C and R24H) prevent CDK4 frombeing inhibited by INK4A, underlining the importance of CDK4as a second melanoma susceptibility gene. CDK4 can act as anoncogene in some tumors (mainly melanomas and gliomas) whereits somatic overexpression usually results from gene ampli®cation(He et al, 1994; Schmidt et al, 1994). To date, activating mutationsof CDK4 (e.g., R24C) have only been detected somatically inmelanoma (e.g., Tsao et al, 1998). Mutations that affect the activityof any component of the INK4A/CDK4/cyclin D/pRb pathwayhave important rami®cations for melanocyte transformation.Nearly all melanomas have been found to have a defect in thispathway (Castellano et al, 1997; Walker et al, 1998).

In a situation so far unique in the human and mouse genomes,CDKN2A also encodes a distinct and otherwise unrelated tumorsuppressor protein, p14ARF (p19Arf in mouse; hereafter termedARF), which acts through a different pathway involving stabiliza-tion of p53 through abrogation of HDM2 (Mdm2 in mouse)-induced p53 degradation (Zhang et al, 1998). The alternativelyspliced INK4A and ARF mRNA are transcribed off different ®rstexons, and utilize the same second exon, but in a different readingframe (Fig 1). CDKN2A germline mutations in exon 1a affectonly the INK4A transcript, whereas some of those occurring inexon 2 can affect both INK4A and ARF. Evidence is mountingthat germline deletion (Bahuau et al, 1998; Petronzelli et al, 2001;Randerson-Moor et al, 2001) or mutation (Rizos et al, 2001) ofexon 1b may also predispose to melanoma as well as tumors of theneural system. Somatic exon 1b mutations have not been detectedin uncultured melanomas, although melanoma cell lines have beenreported with speci®c deletions of exon 1b, leaving INK4A intact(Kumar et al, 1998).

Regulation of the G1±S phase cell cycle transition is made morecomplex by the existence of other CDK4 inhibitors, including theINK4-speci®c inhibitors p15INK4B, p18INK4C, and p19INK4D, andthe less speci®c CIP1/KIP1 inhibitors p21CIP1, p27KIP1, andp57KIP2, which bind many CDKs (Fig 2). The latter can bind to,but do not inhibit CDK4. To proceed fully through G1, completephosphorylation and inactivation of pRb may require the kinaseactivity of both CDK4 and CDK2, made possible in part byreassortment of inhibitors, particularly p27KIP1 shuf¯ing betweencyclin D/CDK4 and cyclin E/CDK2 complexes, leading toactivation of the latter (reviewed in Sherr and Roberts, 1999). Ineffect, CDK4 can also help promote G1±S phase progression bytitrating p27KIP1 away from CDK2/cyclin E complexes. CDKN2Ais the only CDK-inhibitor gene mutated in human melanoma, andnone of the knockout mouse models of the other inhibitorsdevelop skin cancers, although some are susceptible to thedevelopment of other tumor types, particularly of neuroendocrineorigin (reviewed in Chin et al, 1998).

Cytogenetic deletions and loss of heterozygosity on chromosome10q is a common feature of human melanoma. A candidate tumorsuppressor, PTEN, was isolated from 10q23, and subsequentlyshown to be mutated or deleted in many tumor types (reviewed inCantley and Neel, 1999). Germline PTEN mutations have beendetected in several familial hamartoma syndromes, includingCowden disease and Bannayan±Riley±Ruvulcaba syndrome (re-viewed in DiLiberti, 1998). In the largest study of PTEN inmelanoma carried out to date (Pollock et al, 2002), deletion ormutation of the gene was detected in 23% of melanoma cell lines,which, in terms of mutation frequency, makes it possibly the mostimportant ``classical'' TSG gene in melanoma after CDKN2A(although downregulation by mechanisms other than mutation or

Figure 1. Genomic organization, alternative splicing andproducts of the Cdkn2a locus conserved between mouse andhuman. E1b is exon 1b, E1a, exon 1a, E2 is exon 2, and E3 is exon3. Both transcripts share the second and third exons but not the ®rst.The Ink4a transcript uses exon 1a, whereas the Arf transcript uses exon1b. The unshaded portions of each exon correspond to the untranslatedregions of the transcripts.

784 WALKER AND HAYWARD THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

deletion may be more common for some PTK, e.g., KIT and FES).PTEN functions by dephosphorylating the lipid second messengerphosphotidylinositol (PI) 3,4,5 triphosphate and other proteins inthe cascade that controls aspects of cell growth and survival,including apoptosis (reviewed in Maehama and Dixon, 1999). Oneof the major PTEN pathways involves Ras, PI3K, and AKT (PKB).Uncontrolled activity of any of these proteins is oncogenic. BecausePTEN and NRAS mutations seem to be mutually exclusive inhuman melanoma, it is thought that they may act through the samepathway (Tsao et al, 2000). This is supported by functional evidenceshowing that PTEN is also capable of suppressing activated Ras-mediated transformation of NIH3T3 cells (Tolkacheva and Chan,2000; Tsao et al, 2000), associated with suppression of the PI3Ksignaling cascade stimulated by Ras. Thus the activation of Ras andloss of PTEN may be substantially equivalent, at least in terms oftheir actions through the PI3K pathway. PTEN has also beenshown to induce cell cycle arrest mediated by PI3K in a pRb-dependent manner (Paramio et al, 1999), possibly by down-regulating cyclin D and upregulating p27Kip1 (Atkas et al, 1997;Weng et al, 2001).

Activating mutations of the Ras family (in general Nras, but to alesser extent Hras and Kras) are detected in about 15% ofmelanocytic lesions (van Elsas et al, 1996), but appear to be a lateevent in melanoma progression. Although the genetic evidenceoutlined above points to a role for the Ras/PI3K/Akt pathway inmelanoma, Ras has multiple effectors (reviewed in Hunter, 1997;Sears and Nevins, 2002) that may also stimulate other pathwaysinvolved in melanoma development. These may include the Ras/Raf/mitogen-activated protein kinase (MAPK) pathway thatpromotes transcription of CCND1 (the gene encoding cyclinD1), and also assists in post-translational assembly of cyclin D/CDK4 complexes. Stimulation of CCND1 transcription is alsomediated by activation of another Ras effector, RalGDS.Furthermore, Ras stimulation through the RhoA GTPase pathwaycan induce degradation of p27KIP1 (Hu et al, 1999), thus releasing

inhibition of cyclin E/CDK2 activity needed for cells to proceedinto S-phase. Ras effectors are capable of inducing senescence,apoptosis, or activating cell proliferation, with the various effectorpathways collaborating to achieve speci®city of signaling dependenton the setting (Sears and Nevins, 2002).

GENETICALLY MODIFIED MOUSE MODELS TO STUDYMELANOMA DEVELOPMENT

Attempts to induce transformation of melanocytes in mice have, ingeneral, utilized transgenic methods to overexpress oncogenes(Table I), and homologous recombination techniques to ``knockout'' portions of TSG or ``knock in'' mutations of either of TSG oroncogenes (Table II).

TRANSGENIC MODELS

Tyr-SV40E Among the ®rst genetically modi®ed mice reportedto develop melanoma were those generated by Mintz and co-workers. These transgenic animals, termed Tyr-SV40E,overexpressed SV40 T antigen (Tag) under the control of themelanocyte-speci®c tyrosinase (Tyr) gene promoter, and weresusceptible to melanoma (Table I), developing mainly ocularmelanomas (OMM) at an early age (Bradl et al, 1991; Klein-Szantoet al, 1991). The tumors grew rapidly and metastasized to otherparts of the body. Melanoma development was greatly dependenton copy number of the transgene. Low copy number animals wereused to study cutaneous melanomas (CMM), as the high expressermice died very early from OMM, which developed mainly in theRPE. The CMM arose more frequently from hair folliclemelanocytes than those in the epidermis (Silvers and Mintz,1998). SV40 Tag disrupts various cellular pathways, butsigni®cantly, one of the ways it causes cellular transformation isby inactivating both pRb and p53 (reviewed in Ali and DeCaprio,2001).

Figure 2. Cyclin-dependent kinase inhibitors capable of binding to Cdk4. To completely proceed through G1, successive phosphorylation ofpRb by both Cdk2 and Cdk4 may be required.

VOL. 119, NO. 4 OCTOBER 2002 MOUSE MODELS OF MELANOMA 785

Some of the most interesting ®ndings to come from the Tyr-SV40E overexpresser animals were those from grafting experimentsto analyze the contribution of age and skin from different regions ofthe body to the development and latency of CMM (Mintz andSilvers, 1993). Donor skin excised from different body sites on highcopy number transgenic mice were grafted on to less melanomasusceptible, longer-living, low copy number transgenic animals ofvarious ages. Grafts on to neonates resulted in CMM developmentwith much shorter latency than those on to adult animals. Thisobservation probably stems from the large number of incompletelydifferentiated, growth factor-receptive melanocytes in neonatalskin, or extrinsic factors from surrounding keratinocytes that maystimulate melanocyte cell division. These hypotheses are supportedby the observation that CMM on grafts from body skin alwaysappeared near the edge of the grafts, presumably stimulated byin¯ammatory and wound-healing cytokines and other factors.Furthermore, grafts of snout skin on to low copy numbertransgenic hosts resulted in multiple lesions at the graft site,compared with single lesions at the sites of grafts taken from dorsalskin. Thus, the surrounding microenvironment, as well as the stateof initiation of the melanocytes themselves, contributes to CMMdevelopment.

Tyrosinase-related protein 1 (Tyrp1) SV40 Tag The Tyrp1gene promoter has also been used to overexpress SV40 Tag in mice(Penna et al, 1998). In this case overexpression was generallyrestricted to the RPE of the eye, whereas expression in melanocytesof the skin was very low. Pigmented tumors developed in the RPEof these animals and metastasized to lymph nodes and the spleen in aslittle as 3 mo (for a review of transgenic models of melanoma of theskin and RPE see Beermann et al, 1999).

Metallothionein (Mt1)-Ret and Tyrp1-Ret In transgenic miceoverexpressing the Ret proto-oncogene, driven by an inducible

Mt1 gene promoter (Iwamoto et al, 1991; Takahashi et al, 1992;Kato et al, 1998), melanocytes proliferated and underwentneoplastic transformation. The tumors arose from cutaneousmelanocytes, and also from the choroidal neural crest-derivedpigment cells within the eye. Unlike some of the Tyr-SV40Emelanomas, those in the Mt1-Ret lines did not metastasize. In asimilar study, when Ret was overexpressed in the RPE using theTyrp1 gene promoter (Schmidt et al, 1999), the animals developedmicrophthalmia (small eyes), with thickened RPE and benigntumors of the RPE observed in some animals. RET, CDK4, andMET are at present the only three cases of an oncogene beingresponsible for susceptibility to cancer in humans. RET mutationspredispose to multiple endocrine neoplasia type 2 (Mulligan et al,1993). RET is a cell surface receptor tyrosine kinase (RTK). It isinvolved in differentiation and proliferation of neural crest cells asthe receptor for glial cell line-derived neurotrophic factor. Retfunctions through a variety of signaling pathways, including theRas and PI3K cascades (reviewed in Van Weering and Bos, 1998).RTK are also implicated in melanoma development in theXiphophorus ®sh model, where the role of Xmrk (an EGFRhomolog) has been well established (reviewed in Wellbrock et al,1997). EGF also acts through the Ras/Raf/MAPK pathway, andhuman EGFR maps to chromosome 7p11±13, a region commonlyampli®ed in melanoma. The effect of Ret in eye development inthe transgenic animals may be partly mediated by the effect of theRet/RTK/MAPK pathway on increasing the activity of themicrophthalmia transcription factor (Hemesath et al, 1998).

Mt1-hepatocyte growth factor/scatter factor (Mt1-Hgf/Sf) The metallothionein gene promoter was also utilized inanother mouse model to drive overexpression of Hgf/Sf (Otsuka etal, 1998). These mice developed melanocyte dysplasia andmetastatic melanoma. HGF/SF is a multifunctional cytokine,which mediates proliferation partly by stimulating p38MAPK

Table I. Transgenic mouse models of melanoma

Geneticmodi®cation(promoter/gene) Strain

Spontaneousmelanoma Tumor type

UV-inducedmelanomasa UV wavelengths UV protocol Reference

Tyr-SV40 Tag(high expresser)

C57/BL6 Yes, up to 100%from 4 wk

PredominantlyOMM, but someCMM, metastatic

Mortality too highto be assessed

Bradl et al (1991)Klein-Szanto et al (1991)

Tyr-SV40 Tag(low expresser)

C57/BL6 Infrequent OMM,no CMM

26%CMM;latency37±98 wk

70% UVB280±320nm;29.7% UVA,320±380 nm

3.28 kJ per m2 to4 d old neonates,repeated for 5 d

Klein-Szanto et al (1994)Kelsall and Mintz (1998)

Tyrp1-SV40 Tag

NMRI/HAN Yes, 100%OMM by 3 mo

MetastaticRPE tumours

Not assessed Penna et al (1998)

Mt1-Ret BALB/C,C57BL6

Yes, up to 93%OMM by 5 mo

CMM and OMM,not metastatic

Increasedinvasiveness oftumors

Broad-spectrum:250±400 nm(60% UVB)

Incremental2.25±6 kJ per m2

tri-weekly for34 wk

Iwamato et al (1991)Kato et al (2000)

Tyrp1-Ret NMRI Yes, OMM Not assessed Not assessed Schmidt et al (1999)Tyr-Hras (G12V) C3H 12% OMM,

no CMMMelanocytehyperplasia

20% naevi andCMM, only onalbinobackground

90% UVB280±340 nm

Bi-weekly for38 wk 5.6±8.06kJ per m2

Broome-Powell et al(1995, 1999)

Tyr-Hras (G12V) mixed Very rarely Melanocytehyperplasia

Not assessed Chin et al (1997)

Mt1-Hgf/Sf FVB Yes, 22%, meanage of onset15.6 mo

Melanocytehyperplasia,CMM,metastatic

No increasein CMM

240±400 nm 2.25±6 kJ per m2

tri-weekly for17 wk

Otsuka et al (1998)Noonan et al (2000)

Mt1-Hgf/Sf FVB As above As above CMM in 80%within 12 mo

240±400 nm 9.2 kJ per m2,1 dose to 4 dold neonates

Noonan et al (2001)

Krt4-Scf C57BL No Epidermalmelanocytosis

Not assessed Kunisada et al (1998)

a% of mice with UV-induced melanomas; see Table III for summary of UV exposure protocols.

786 WALKER AND HAYWARD THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

activity and upregulating CCND1 transcription (Recio andMerlino, 2002). It also affects motility and movement of a widevariety of epithelial cells expressing its RTK, MET. Constitutivelyactive MET has also been implicated in oncogenesis in humans,with some families with hereditary papillary renal carcinomacarrying mis-sense mutations in its tyrosine kinase domain (Schmidtet al, 1997). Moreover, MET is inappropriately expressed in a rangeof tumors such as melanoma, rhabdomyosarcoma, hepatoma, andmammary carcinoma. In the Mt-Hgf/Sf transgenic mice,melanocytes were overabundant in the epidermis, particularly atthe dermal±epidermal junction, and concentrated in nonhairy partsof the body, including the paws, tail, ears, and muzzle. CMMdeveloped in 22% of animals, and 21% of these tumorsmetastasized. The hyperproliferation of melanocytes concentratedthem in locations away from the hair follicles, in the epidermis andbasal layer, the location of melanocytes in human skin. The CMMthat developed in these mice were of dermal origin, however,unlike most human CMM, which are usually of epidermal origin.The dramatic effect on melanocytes in both Mt-Hgf/Sf and Mt-Retanimals seems at odds with the fact that the Mt1 gene promoterseems to be functional in virtually all tissue types, with high levelsobserved in various tissues, including the skin (Takayama et al,1996).

Keratin K4-stem cell factor Mice transgenic for stem cellfactor (Scf) driven by the keratin K4 gene (Krt4) promoter Krt4-Scfalso develop melanocyte dysplasia (melanocytosis) (Kunisada et al,1998). SCF is the ligand for the Kit RTK that regulates themigration and development of primordial germ cells,hematopoietic stem cells, melanocytes, and mast cells (Kunisada et

al, 1998; Lyman and Jacobsen, 1998). Interestingly, the location ofmelanocytes in the skin of adult transgenic animals closelyresembles that of human skin, in that melanocytes tend to staywithin the epidermis, instead of congregating around hair folliclesas in wild-type mice. SCF has this effect on melanocytes viastimulation of the Kit receptor, which in turn can activate theMAPK pathway (Hemesath et al, 1998). Furthermore, as Scfexpression is limited to keratinocytes, this mouse model shows thatsignals from surrounding keratinocytes can dramatically affectmelanocyte localization during development. Although these micedo not develop melanoma, they appear to be a useful mouse modelfor melanocyte localization that mimics human skin.

Tyr-Hras (G12V) Utilizing slightly different versions of themelanocyte-speci®c tyrosinase gene promoter/enhancer sequences,two groups have generated transgenic mice overexpressingoncogenic Hras (G12V). Broome Powell et al (1995) used a2.5 kb fragment containing the tyrosinase promoter and someregulatory elements, whereas Chin et al (1997) used a chimeric5.5 kb fragment containing the same tyrosinase gene promoter butslightly different upstream enhancer sequences (Ganss et al, 1994).The ®rst reported Tyr-Hras transgenics (Broome Powell et al, 1995)were signi®cantly smaller than their wild-type littermates, anddisplayed hyperpigmentation of the snout, feet, and tail. Thetransgenic animals were blind and consistently moved in a circlingor twirling motion, probably caused by severe abnormalities withinthe inner ear. Although these mice only very rarely developedCMM, melanocyte hyperplasia, or melanocytosis was oftenobserved in the dermal and epidermal layers of the skin, innerear, and parts of the brain. Some of the nonmelanocyte

Table II. Knockout/knockin mouse models of melanoma

Genetic modi®cationSpontaneousmelanomasa

Inducedmelanomas

Spontaneous tumors(nonmelanoma)b

Induced tumors(nonmelanoma)b

MEF/culturedtumor cells Reference

Ink4a±/±:Arf±/±, delCdkn2a exons 2and 3

None None 69% at average of29 wk, FS, L, S

90% at 20 wk withDMBA/UVB;c FS,L, SSC

Immortal andtransformedbyactivated Hras

Serrano et al (1996)

Arf±/±, del exon 1b None None 33% at 24 wk;FS, L, SSC, G,S, A

82% at 20 wk withDMBA; SCC, L, S

Immortal andtransformedbyactivated Hras

Kamijo et al (1997)

Ink4a*/*, ``knockinG101 stop mutation''

None Very rare withDMBA, butlimited numbers

17% by 17 mo; L No assessment withDMBA due to limitednumbers

Undergo senescence,not transformed byactivated Hras

Krimpenfortet al (2001)

Ink4a±/±, del exon 1a 2.5% of animals 7% withDMBA

26% by 58 wk; S, L 50% at 23 wk withDMBA; L, A, S

Undergo senescence,not transformed byactivated Hras

Sharplesset al (2001)

Ink4a±/±:Arf±/±:Tyr-Hras (G12V)

60% at 6 mo As for Ink4a±/±/Arf±/±

animalsIn vitro melanomacell growthsuppressedby Ink4a

Chin et al (1997)

Ink4a±/±:Arf+/± 8% by 17 mo 50% by 9 mowith DMBA

100% by 17 mo;FS, L, S, A, SCC

P, A, others Krimpenfortet al (2001)

p53 ±/±:Tyr-Hras(G12V)

29% by 17 wket al (2001)

74% death ornonmelanoma tumorsby 17 wk; S, L

In vitro melanomacell growth notsuppressed by Ink4a

Bardeesy

Pten±/±:Ink4a±/±:Arf±/±

7% by 31 wk 100% beginning atweek 7; L, E, S

Increased growth andsusceptibility to Hrastransformation

You et al (2002)

Cdk4R24C/R24C

``knockin'' mutationNone 70% by 20 wk

withDMBA + TPA

93% by 14±16 mo;S, E, A, L

­ severity P, S and15% SCC, withDMBA, TPA

Weakly transformedby activated Hras

Sotillo et al(2001a, >b)

Cdk4±/± None None (diabetes) Rane et al (1999)Integration intounknown locus

100% CMM by12 mo

None Chen et al (1996)Zhu et al (1998)

aPercentage of mice that developed CMM. Blank ®eld = not assessed.bOverall percentage of animals with nonmelanoma tumors, with most common tumor types listed for each model: FS, ®brosarcoma; S, other sarcoma; P, papilloma; L,

lymphoma; A, adenoma; G, glioma; SCC: squamous cell carcinoma; E, endocrine tumors.cWavelengths used, 295±310 nm (UVB); 27 exposures 1±7 kJ per m2 beginning at postnatal days 4±8; strain, C57BL/6.

VOL. 119, NO. 4 OCTOBER 2002 MOUSE MODELS OF MELANOMA 787

abnormalities seen in these animals are due to tyrosinase genepromoter activity that can be detected in parts of the nervoussystem and other neural crest-derived cells (Klein-Szanto et al,1991). Three transgenic lines of similar Tyr-Hras transgenics weregenerated in the laboratory of Chin et al (1997). The highest copynumber overexpresser had a very compromised phenotype similarto that described by Broome Powell et al (1995), but the animalsdied at a young age and the line was not continued. The other twolines were overtly normal, although the occasional animaldeveloped OMM (Chin et al, 1997). As with other transgenicmodels discussed above, activated Ras transgene overexpressionprobably stimulates signaling through the RTK/Ras/Raf/MAPKpathway, or a related signaling cascade that controls proliferation aswell as cell±cell communication and cell positioning with the skin.Thus, these transgenic models also have in common the re-localization and proliferation of melanocytes to producehyperpigmentation of nonhairy skin.

GENE KNOCKOUT AND KNOCKIN MOUSE MODELSOF MELANOMA

Cdkn2a knockouts The ®rst reported Cdkn2a knockout micewere generated using homologous recombination techniques toablate exon 2, which inactivated both the Ink4a and Arf transcripts(Serrano et al, 1996). These animals did not develop melanoma,although they were susceptible to the development of othertumors, particularly ®brosarcomas and lymphomas (Table II).These tumors increased in frequency after treatment with thecarcinogen DMBA. Melanocytes cultured from these animals fail tosenesce, and have decreased pigmentation levels associated with theloss of senescence (Sviderskaya et al, 2002). Furthermore, Ink4a±/±:Arf±/± murine embryonic ®broblasts (MEF) proliferated rapidly,grew in colonies, and could be transformed with activated Hras.The MEFs could also be transformed with equal ef®ciency byactivated Raf, implicating the Ras/Raf/MAPK pathway (Serrano etal, 1996). To assess the effects of deleting the Arf transcript alone,exon 1b of the Cdkn2a locus was ablated in mice (Kamijo et al,1997). The animals had a very similar phenotype to theInk4a±/±:Arf±/± mice, implying that inactivation of Arf, ratherthan Ink4a, may be responsible for the tumor susceptibilityphenotype in the initial model.

Following on from the ®nding that Ink4a±/±:Arf±/± MEF weresusceptible to activated Hras transformation, Chin et al (1997)crossed Ink4a±/±:Arf±/± mice with Tyr-Hras transgenic mice. Theresulting progeny developed CMM spontaneously with a highpenetrance. OMM also occurred, but were less frequent. Thesedata indicate that a further genetic ``hit'' in addition to Cdkn2ainactivation is also needed for melanocyte tumorigenesis; however,the melanomas were only locally invasive and did not metastasize,suggesting that still further changes are necessary for late stagemelanoma progression. A modi®cation of the same strategyinvolved crossing the Ink4a±/±:Arf±/± mice with an ``inducible''Tyr-Hras overexpresser mouse line (Chin et al, 2000), wherebyHras expression could be repressed by the removal of a drug(doxycycline) from the animals' feed. This downregulation resultedin tumor regression in the animals that had already developedmelanomas, implicating Hras in tumor maintenance as well asprogression. This is an important ®nding, with obvious implica-tions for treatment modalities targeting Ras in humans.

The question remained as to whether the functional knockout ofInk4a or Arf was responsible for melanoma susceptibility in mice.We now know that both genes are important. Two groups(Krimpenfort et al, 2001; Sharpless et al, 2001) have generatedspeci®c knockouts of the Ink4a gene in mice. Krimpenfort et alintroduced a stop signal at codon 101 in exon 2, which produced atruncated, unstable Ink4a protein (termed Ink4a*/*), leaving Arfunaffected. The other group, Sharpless et al, deleted exon 1a, alsoleaving Arf expression unaffected. Although results from the twogroups vary somewhat (Table II), they are similar in that theanimals were prone to the development of various tumors, thus

con®rming Ink4a as a bona ®de tumor suppressor. These animals,however, had a slightly lower frequency of spontaneous tumordevelopment than the Arf knockouts, but importantly, both strainswere susceptible to spontaneous melanoma development, albeit atvery low frequency. Carcinogen treatment had little effect onmelanoma incidence. The differences between the results of thetwo studies may re¯ect differences in mouse strains or the strategiesfor ablating Ink4a. Both studies transfected the Ink4a±/± MEFs withHras and found that they were resistant to transformation with theoncogene, in contrast to MEFs from the Ink4a±/±:Arf±/± (Serranoet al, 1996) and Arf±/± (Kamijo et al, 1997) mice, This indicates that,at least in ®broblasts, Ras-induced transformation may be depend-ent on abrogation of Arf function, together with the loss of Ink4a.It is unclear how this may relate to melanoma development untilInk4a±/±/Tyr-Hras and Arf±/±/Tyr-Hras mice are studied.

Intriguingly, in the study of Krimpenfort et al (2001), when theInk4a*/* mice were crossed with Arf hemizygotes, the animalsdeveloped melanomas, the penetrance of which rose from 8% to50% after treatment with DMBA. The effect of Arf haplo-insuf®ciency in an Ink4a±/± background suggests cooperationbetween the Ink4a and Arf pathways in melanoma development.The notion of cross-talk between the two pathways has also beensuggested by Carnero et al (2000), who used anti-sense Ink4a andArf RNA constructs to study the effects of abrogation of bothpathways in MEFs. Their results were consistent with a modelwhere Arf normally regulates both the pRb and p53 pathways.Recent work with mouse melanocytes cultured fromInk4a±/±:Arf±/± mice suggests that Ink4a is necessary for senescence,whereas Arf may be involved in cell death pathways (Sviderskaya etal, 2002). In another study to assess the relative roles of thesepathways (Bardeesy et al, 2001), p53 knockout mice were matedwith Tyr-Hras transgenics, for direct comparison withInk4a±/±:Arf±/±:Tyr-Hras animals reported previously (Chin et al,1997). p53±/±:Tyr-Hras mice also developed melanoma but withlower penetrance (29% vs 60%). Melanoma cells were culturedfrom the mouse tumors and Ink4a reintroduced by retroviraldelivery. As predicted, Ink4a did not suppress cell growth of thep53±/±:Hras lines, whereas in cell lines derived fromInk4a±/±:Arf±/±:Tyr-Hras mice, it ef®ciently inhibited cell growth.Thus Ink4a is not involved in the p53/Ras-induced melanomas,although this does not rule out a role for a compromiseddownstream component in the Ink4a/cyclin D/Cdk4/pRb path-way. These studies suggest that the Arf/p53 pathway can in¯uencethe activity of the Ink4a/pRb pathway in mouse melanocytes, butnot vice versa (Fig 3). ``Cross-talk'' between the two pathwaysmay be mediated at many levels. These may include E2F inductionof Arf transcription and direct binding of Arf to the E2F family oftranscription factors (Bates et al, 1998; Eymin et al, 2001), bindingof Mdm2 to pRb (Xiao et al, 1995), and p53 induction of p21Cip1,which can inhibit the activity of Cdk4/cyclin D (reviewed in Chinet al, 1998) and Cdk2/cyclin E (Mitra et al, 1999). Any of theseevents are alternative mechanisms of pRb pathway inactivation thatmay be capable of substituting for loss of Ink4a.

In summary, Cdkn2a may be involved in mouse melanomadevelopment through mechanisms involving both its Ink4a and Arftranscripts. Hras, probably acting through Raf, is capable ofcooperating with the Arf/p53 pathway to initiate melanomadevelopment in mice, as is inactivation of Ink4a plus carcinogentreatment and inactivation of Ink4a plus a decrease in physiologiclevels of Arf.

Cdk4 knockout/knockin mutants Further support for the roleof the Ink4a/cyclin D/Cdk4/pRb pathway in melanoma comesfrom experiments with mice in which Cdk4 has been targeted byhomologous recombination to either ``knock out'' the gene, or to``knock in'' the activating R24C mutation that is resistant to Ink4ainhibition (Rane et al, 1999). Nullizygous Cdk4 animals were about10% smaller than their wild-type littermates, with the femalesinfertile and males having defective spermatogenesis. The animalsalso suffered from insulin-de®cient diabetes caused by defective

788 WALKER AND HAYWARD THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

pancreatic islet b cell development. On the other hand, the miceexpressing only R24C alleles were about 10% larger than theirlittermates and developed hyperplasia of the endocrine pancreas,but were otherwise normal. None of these animals developedmelanoma; however, they have shed further light on abrogation ofthe pRb pathway(s) in tumorigenesis.

When the Cdk4R24C/R24C mice were studied over a longer timeframe (Sotillo et al, 2001a), it was found that endocrine tumorswere the second most common tumor type encountered(hemangiosarcomas being the most frequent). These endocrinetumors included pancreatic adenomas and pituitary carcinomas.The appearance of pituitary tumors is reminiscent of the phenotypeof the Rb+/± (Jacks et al, 1992; Hu et al, 1994), Cdkn2c±/± (encodingp18Ink4c), and Cdkn1b±/± (encoding p27Kip1) mice (Franklin et al,1998), although these neoplasms generally appeared in theadenohypophysis, not the pars intermedia as in the knockoutmouse models. Interestingly, p27Kip1 expression was nearly alwayslost in the pituitary tumors from the Cdk4R24C/R24C mice. Whenthe Cdk4R24C/R24C and Cdkn1b±/± mice were crossed (Sotillo et al,2001a), progeny with the compound mutant genotype developedpituitary tumors with much decreased latency, and in this case thetumors emanated from the pars intermedia. The role of p27Kip1 inthese tumors is unclear. In most tissues expression of the Ink4a-resistant Cdk4 R24C increased Cdk4 but not Cdk2 kinase activity,and did not induce a signi®cant redistribution of p27Kip1 to Cdk4/cyclin D complexes. Thus, despite evidence that one of thefunctions of Cdk4 is to sequester p27Kip1 and hence upregulateCdk2 activity (Sherr and Roberts, 1999), this does not seem to beat play in these Cdk4 mutant mice. At least in the pituitary, thereappears to be a complex interplay between the Ink4a/cyclin D/CDK4/pRb and p27/cyclin E/CDK2/pRb pathways that is notunderstood.

Exposure of the Cdk4R24C/R24C mice to carcinogens (DMBAand TPA) produced pigmented skin lesions that progressed to naeviand CMM in 70% of animals (Sotillo et al, 2001b). Treatment withDMBA alone produced CMM with decreased penetrance. TheInk4a±/± animals discussed above (Krimpenfort et al, 2001; Sharplesset al, 2001) were not treated with both DMBA and TPA, hence adirect comparison is not possible. Activating Ras mutations wereassociated with carcinogen-induced papillomas on the skin of bothwild-type and Cdk4R4C/R24C animals, but were rarely detected inthe melanocytic lesions from the animals expressing mutated Cdk4.It has yet to be determined whether the carcinogen treatmentinduces other mutations/epigenetic inactivation necessary formelanoma formation, or whether additional mechanisms such assuppression of the immune response or activation of other signalingpathways are involved. To assess the role of various Ink4 inhibitorsin melanoma development, Sotillo et al (2001b) applied the sameDMBA/TPA carcinogen treatment to mice nullizygous for Cdkn2b(encoding p15Ink4b), Cdkn2c (p18 Ink4c), and Ink4a/Arf. Increasedincidence of lymphomas resulting in death by 10±12 wk precludedevaluation of skin lesions on Ink4a±/±:Arf±/± animals. Cdkn2b±/±

animals showed no increase in susceptibility to either papilloma ormelanoma, but the Cdkn1c±/± mice developed signi®cantly moremelanocytic lesions than wild-type controls. The tumors weresimilar to those on the Cdk4R24C/R24C animals, but were lessaggressive. This indicates that p18Ink4c may play a part insuppressing melanoma development in mice. This is unexpected,as, no inactivating CDKN2C mutations have been detected inhuman CMM (Platz et al, 1998). Untreated Cdkn2c±/± mice weresusceptible to adenomas of the pituitary and other tumors, but didnot develop melanoma (Franklin et al, 1998). These results suggestthat there must be a delicate balance of Cdk-inhibitor usagedependent on cell type and environmental stimuli.

Figure 3. Pathways that may lead to cell cycle dysregulation in mouse melanocytes. The Ink4a/pRb and Arf/p53 pathways are depicted,together with interactions that may represent cross-talk between the two pathways, which are described in the text. Also depicted are several mitogen-activated Ras pathways that may affect cell cycle progression via upregulation of cyclin D, Cdk4, and/or cyclin E/Cdk2, which ultimately inactivatepRb. Ras cooperates with Arf/p53 to induce melanoma development in mice by an unknown mechanism. These three growth regulation pathways(Ink4a/pRb, Arf/p53, and Ras/MAPK) seem to represent three major ``axes'' of melanoma development in mice.

VOL. 119, NO. 4 OCTOBER 2002 MOUSE MODELS OF MELANOMA 789

Unlike those from Ink4a±/± animals, Cdk4R24C/R24C MEFs weresusceptible to transformation with activated Hras, but in contrastwith Ink4a±/±:Arf±/± MEFs, were not capable of producing tumorsin nude mice Moreover, the ``transformed'' Cdk4R24C/R24C clonesgrew slower than even the nontransformed Ink4a±/±:Arf±/± MEFs,further supporting the notion that defects in the Arf/p53 pathwayare more important for Ras transformation. Components of theArf/p53 pathway were intact in the melanomas from carcinogen-treated mutant Cdk4 mice. Cdk4R24C/R24C:p53±/± mice, althoughnot treated with carcinogens, did not develop CMM (Sotillo et al,2001a), indicating that inactivation of the p53 pathway cannotsubstitute for carcinogen treatment. As in the Ink4a models,however, this work recapitulates the model of melanoma suscep-tibility in humans whereby an inherited mutation of CDKN2A orCDK4 is followed by further environment-driven genetic ``hits''throughout life.

Pten knockouts Homozygous knockout of Pten is embryoniclethal in mice (Di Cristofano et al, 1998; Podsypanina et al, 1999),and MEF derived from the embryos differentiate aberrantly. Pten+/±

mice develop hyperplastic/dysplastic changes in the prostate, skin,and colon, closely resembling the human familial hamartomasyndromes Cowden disease and Bannayan±Riley±Rivulcabasyndrome, which are caused by defects in PTEN. These animals,however, do not develop melanoma, or nervous system tumors,suggesting that Pten is not involved in the development ortransformation of neural crest-derived cells. This notion ischallenged by results from a study by You et al (2002), who haverecently implicated Pten in mouse melanoma development. Theirstudy assessed the effects of Pten heterozygosity in anInk4a±/±:Arf±/± background by crossing the respective knockoutmice. Although the morphology and growth rates of MEFs fromPten+/±:Ink4a±/±:Arf±/± mice were indistinguishable fromInk4a±/±:Arf±/± MEFs, they were more susceptible to trans-formation by activated Hras (increased number of large rapidlygrowing foci). Interestingly, as well as the usual tumors seen in therespective single gene knockouts, the compound knockout micedeveloped a new spectrum of tumors, including invasive CMM andsquamous cell carcinoma. The authors postulated that themelanoma development is due to the inactivation of Ink4a/Arftogether with activation of Ras (by loss of Pten suppression),possibly analogous to that occurring in the mice reported by Chinet al (1997), which also developed melanoma. This is supported bygenetic and functional studies on human PTEN (Tolkacheva andChan, 2000; Tsao et al, 2000), which indicate that both the PTEN/Ras and p16/CDK4/pRb pathways need to be dysregulated formelanoma development. As mentioned previously, Ras signaling isalso mediated though various effector pathways (Hunter, 1997),including the Raf/MAPK, Ras/RalGDS, and Ras/PI3-K/Aktpathways that regulate, among other things, cyclin D/CDK4activity. Interestingly, a new role for PTEN may be the PI3K/AKT-mediated regulation of cyclin E/CDK2 activity bypreventing ubiquitin-mediated degradation of p27Kip1 (Weng etal, 2001). A similar function is also attributed to another Raseffector, RhoA (Hu et al, 1999). This leads to a model for PTEN asan inhibitor of CDK2 as well as CDK4 activity (Fig 3). Cdkn1b±/±

mice, do not develop prostate cancer, but yield a line of mice thatdo so with complete penetrance when crossed with Pten+/± animals(Di Cristofano et al, 2001), underlining the crucial role of Pten,p27KIP1, and Cdk2 in malignant transformation, at least in prostaticepithelial cells. To date, however, a direct role for CDK2deregulation in human melanoma susceptibility has not beenindicated, as melanomas do not seem to harbor activating CDK2mutations similar to those seen in CDK4 (Walker and Hayward,2001), although at the protein level CDK2 is overexpressed insome melanoma cell lines compared with normal melanocytes(Tang et al, 1999).

Integration into an unknown locus An ``accidental''melanoma model (Table II) was created by researchers seekingto overexpress a genomic fragment (clone B) that caused adipoctyes

in culture to differentiate (Chen et al, 1996; Zhu et al, 1998). In oneof the ®ve founder lines generated, transgenic animals developedmetastatic CMM with complete penetrance. The primary lesionswere always initially detected in the skin of the ear or perianalregion, and appeared later on other parts of the body. Althoughsome transgenic mice developed OMM, these always derived fromuveal melanocytes in the choroid, not the RPE. The CMM beganas lesions resembling dysplastic naevi and progressed to invasivemelanomas. All skin lesions derived from melanocytes in thedermis, not the hair follicles, indicating that there may besomething about the microenvironment in follicles that protectsmelanocytes from transformation. The origin of these tumorsdiffers to that of human melanomas, which generally originate fromthe basal layer of the epidermis, not the dermis. These mice areintriguing, given the complete penetrance of CMM resulting fromthe integration of clone B at an unknown location that mayinactivate a TSG, or activate an oncogene that is critically involvedin melanocyte homeostasis. This indicates that there may be otherstrong genetic determinants of melanoma development yet to bediscovered.

ROLE OF UVR IN MOUSE MELANOMADEVELOPMENT

Experiments testing different protocols in terms of intensity,timing, and duration of exposure have been used to assess therole of UVR in CMM development in genetically engineeredmice. Tyr-SV40E mice developed melanoma with an increasedfrequency and decreased latency compared with untreated animalsafter exposure to UVR using several different treatment regimens(Kelsall and Mintz, 1998). Studies using the high expressertransgenic animals were impractical due to the very high mortalityrates from OMM. Using low susceptibility Tyr-SV40E lines,however, Klein-Szanto et al (1994) and Kelsall and Mintz, 1998)showed that melanocytic naevi and melanoma development couldbest be promoted with a series of UVR treatments over 5 dbeginning at postnatal day 4. High intensity, but chronic exposuresto adult animals, however, have also been successful in inducingCMM in Tyr-Hras mice, which do not generally develop thisneoplasm (Broome-Powell et al, 1999). Using a very similarregimen (Table III), Kato et al (2000), reported that Mt1-Ret-induced melanocytic tumors tended to become more invasive afterthe animals were exposed to UVR.

The importance of treatment regimens was further underlined bystudies using the Mt-Hgf/Sf overexpresser mice. A UVR treatmentregimen exposing adult transgenics to long-term chronic doses(Table III) had little effect on melanoma formation, but didincrease the frequency of nonmelanoma skin tumors (Noonan et al,2000). When other protocols were employed, however, it wasdiscovered that with a single erythemal UVR dose to 3.5 d oldneonates, the frequency of CMM was dramatically increased, andlatency decreased, compared with the chronic dose to adult animals(Noonan et al, 2001). The neonatal treatment was much moreeffective at inducing CMM despite the total dose being 30-fold lessthan that given to adult mice.

Surprisingly, UVR treatments on mice carrying mutations in thehuman melanoma susceptibility genes (Ink4a±/±:Arf±/± andCdk4R24C/R24C) did not result in melanoma development(Serrano et al, 1996; Sotillo et al, 2001b), despite use of anexperimental protocol consisting of neonatal exposure beginning asearly as postnatal day 4±8 in one of the studies (Serrano et al, 1996).The Ink4a±/±:Arf±/±:Tyr-Hras animals were not exposed to UVR todetermine whether this had any effect on tumor latency orpenetrance.

Normal mice do not develop melanomas even when exposed tochronic UVR treatments (Gallagher et al, 1984), so historically themouse has not been a good model for UVR-induced melanoma.UV treatment protocols used on various genetically modi®ed micediscussed here have been varied, but have generally used broad-spectrum radiation consisting of 70±90% UVB, with a smaller

790 WALKER AND HAYWARD THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

portion of UVA to simulate solar UV light, which is a major riskfactor for melanoma development in humans. Generally, the mostsuccessful protocols have used exposures that begin soon after birth,at about day 4. This was successful for the transgenic linesoverexpressing SV40 Tag, and Hgf/Sf. It is puzzling, however, thatin the mouse models recapitulating the germline mutations inhuman melanoma-prone families (Ink4±/±, Arf±/±, Cdk4R24C),UVR treatments were unable induce CMM, despite neonataltreatments of the Ink4±/±:Arf±/± animals. This could mean thatUVR-induced melanomas in mice may result from the degree of``priming'' of melanocytes for transformation, by, for example, theoveractivity of SV40 Tag and Hgf/Sf during embryonic andneonatal development. Arti®cial overexpression of these oncogenesprobably dysregulates multiple pathways and is more severe and``unnatural'' than mutations of Cdkn2a or Cdk4, therefore makingmelanocytes in the ``overexpresser'' transgenic mice more suscep-tible to UV-induced transformation. The success of neonatal,compared with chronic UVR exposure, however, ®ts nicely withhuman epidemiologic studies where a high level of childhood sunexposure is a strong determinant of melanoma risk (Whiteman et al,2001). One explanation for this phenomenon in mice is that amuch higher proportion of melanocytes in neonates seem to belocated in the epidermal layer, whereas in adult mice melanocytesare mainly localized to hair follicles, and thus possibly shielded to agreater extent from UVR. Alternatively, it is known thatmammalian neonatal skin contains a higher proportion ofmelanoblasts, and incompletely differentiated melanocytes, thatmay be less well equipped to deal with UVR insults (Erickson,1993) than melanocytes in adult skin.

CONCLUSIONS

By analysis of melanoma-prone families and sporadic tumors, theinvolvement of several genes in human melanoma development hasbeen con®rmed. To date, the most important of these areCDKN2A, CDK4, and PTEN. Studies analyzing the function ofthese genes and their protein products are generally performedin vitro where conclusions may often only represent one componentof activity and interactions within the intracellular milieu. Ourunderstanding of the molecular mechanisms of melanoma predis-position, however, have been greatly improved by modelingvarious pathways in the mouse. The Cdkn2a locus encodes twototally different proteins, Ink4a and Arf. How the two apparentlydistinct pathways, of which Ink4a and Arf are components,cooperate to increase tumor susceptibility is puzzling, as there islimited evidence for Arf involvement from genetic studies ofmelanoma susceptibility and progression in humans. There is ampleevidence, however, of communication between the two pathways.In addition, doubts about the importance of PTEN in melanomatumorigenesis have been quashed by the ®nding of cooperationbetween the Ink4a/Arf and Pten pathways in mouse melanomadevelopment. We also know from mouse models that Ras pathwayactivation is very important in melanoma development, eitherthrough direct activation of Ras (e.g., Hras G12V), or via activationof Ras-effector pathways by other oncogenes (e.g., Ret, Hgf/Sf).From this information we can postulate models for melanomadevelopment that can be tested using various transgenic andknockout mice. On the other hand, assuming that signals througheach pathway proceed in a linear manner may be too simplistic, andultimately, we may have to envisage a web of interacting pathways,centered around pRb and p53, both of which may need to bedysregulated for melanoma development. Finally, experiments withgenetically modi®ed mice have begun to give us some indicationsof the intensity and timing of UV exposure that may be mostresponsible for melanoma development. At ®rst glance, mice donot appear to be a good model for melanoma as they do notnormally develop the neoplasm. But recent results using geneticallymodi®ed animals have pointed melanoma researchers in someunexpected directions.

We would like to thank the National Health and Medical Research Council of

Australia, and the Queensland Cancer Fund, for supporting our work.

REFERENCES

Ali S, DeCaprio J: Cellular transformation by SV40 large T antigen: interaction withhost proteins. Semin Cancer Biol 11:15±23, 2001

Atkas H, Cai H, Cooper GM: Ras links factor signalling to the cell cycle machineryvia regulation of cyclin D1 and Cdk inhibitor p27. Mol Cell Biol 17:3850±3857,1997

Bahuau M, Vidaud D, Jenkins RB, et al: Germ-line deletion involving the INK4locus in familial proneness to melanoma and nervous system tumors. Cancer Res58:2298±2303, 1998

Bardeesy N, Bastian B, Hezel A, Pinkel D, DePinho R, Chin L: Dual inactivation ofRB and p53 pathways in RAS-induced melanomas. Mol Cell Biol 21:2144±2153, 2001

Bates S, Phillips A, Clark P, Stott F, Peters G, Luwig R, Voudsen K: p14ARF links thetumour suppressors RB and p53. Nature 395:124±125, 1998

Beermann F, Hunziker A, Foletti A: Transgenic models for tumors of melanocytesand retinal pigment epithelium. Pigment Cell Res 12:71±80, 1999

Bradl M, Klein-Szanto A, Porter S, Mintz B: Malignant melanoma in transgenicmice. Proc Natl Acad Sci USA 88:164±168, 1991

Broome Powell M, Hyman P, Bell OD, et al: Hyperpigmentation and melanocyticdysplasia in transgenic mice expressing human T24 Hras gene regulated by amouse tyrosinase promoter. Mol Carcinogen 12:82±90, 1995

Broome-Powell M, Gause P, Hyman P, Gregus J, Lluria-Prevatt M, Nagle R,Bowden GT: Induction of melanoma in TPras transgenic mice. Carcinogenesis20:1747±1753, 1999

Cantley LC, Neel BG: New insights into tumor suppression: PTEN suppresses tumorformation by restraining the phosphoinositide 3-kinase/AKT pathway. ProcNatl Acad Sci USA 96:4240±4245, 1999

Carnero A, Hudson J, Price C, Beach D: p16INK4A and p19ARF act in overlappingpathways in cellular immortalisation. Nature Cell Biol 2:148±155, 2000

Castellano M, Pollock PM, Walters MK, et al: CDKN2A/p16 is inactivated in mostmelanoma cell lines. Cancer Res 57:4868±4875, 1997

Chen S, Zhu H, Wetzel WJ, Philbert MA: Spontaneous melanocytosis in transgenicmice. J Invest Dermatol 106:1145±1151, 1996

Chin L, Pomerantz J, Polsky D, et al: Cooperative effects of INK4a and ras inmelanoma susceptibility in vivo. Genes Dev 11:2822±2834, 1997

Chin L, Pomerantz J, DePinho R: The INK4a/ARF tumor suppressor: one gene-two products-two pathways. Trends Biochem Sci 23:291±296, 1998

Chin L, Tam A, Pomerantz J, et al: Essential role for oncogenic Ras in tumourmaintenance. Nature 400:468±472, 2000

Di Cristofano A, Pesce B, Cordon-Cardo C, Pandol® P: Pten is essential forembryonic development and tumour suppression. Nat Genet 19:348±355, 1998

Di Cristofano A, De Acetis M, Koff A, Cordon-Cardo C, Pandol® P: Pten andp27Kip1 cooperate in prostate cancer tumor suppression in mouse. Nat Genet27:222±224, 2001

DiLiberti JH: Inherited macrocephaly±hamartoma syndromes. Am J Med Genet79:284±290, 1998

Dracopoli NC, Fountain JW: CDKN2A mutations in melanoma. Cancer Surveys26:115±132, 1996

Easty D, Bennett D: Protein tyrosine kinases in malignant melanoma. Melanoma Res10:401±411, 2000

van Elsas A, Zerp S, van der Flier S, et al: N-ras oncogene point mutations induced bysun exposure in primary cutaneous melanoma. Am J Pathol 149:883±893, 1996

Erickson C: From the crest to the periphery. control of pigment cell migration andlineage segregation. Pigment Cell Res 6:336±347, 1993

Eymin B, Karayan L, Seite P, Brambilla C, Brambilla E, Larsen CJ, Gazzeri S: HumanARF binds E2F1 and inhibits its transcriptional activity. Oncogene 20:1033±1041, 2001

Franklin D, Godfrey V, Lee H, et al: CDK inhibitors p18INK4C and p27KIP1 mediatetwo sperate pathways to collaboratively suppress pituitary tumorigenesis. GenesDev 12:2899±2911, 1998

Gallagher C, Can®eld P, Greenhoak G, Reeve V: Characterization and histogenesisof tumors in the hairless mouse produced by low-dosage incrementalultraviolet radiation. J Invest Dermatol 83:169±174, 1984

Ganss R, Montoliu L, Monaghan AP, Schutz G: A cell-speci®c enhancer farupstream of the mouse tyrosinase gene confers high level and copy number-related expression in transgenic mice. EMBO J 13:3083±3093, 1994

He J, Allen JR, Collins VP, Allalunis-Turner MJ, Godbout R, Day RS 3rd, JamesCD: CDK4 ampli®cation is an alternative mechanism to p16 gene homozygousdeletion in glioma cells lines. Cancer Res 54:5804±5807, 1994

Hemesath T, Price ER, Takemoto C, Badalian T, Fisher D: MAP kinase links thetranscription factor microphthalmia to c-kit signalling in melanocytes. Nature391:298±301, 1998

Hirobe T: Structure and function of melanocytes: Microscopic morphology and cellbiology of mouse melanocytes in the epidermis and hair follicle. HistolHistopathol 10:223±237, 1995

Hu N, Gutsmann A, Herbert DC, Bradley A, Lee WH, Lee EY: Heterozygous Rb-1delta 20/+mice are predisposed to tumors of the pituitary gland with a nearlycomplete penetrance. Oncogene 4:1021±1027, 1994

Hu W, Bellone CJ, Baldassare JJ: RhoA stimulates p27 (Kip) degradation through itsregulation of cyclin E/CDK2 activity. J Biol Chem 274:3396±3401, 1999

Hunter T: Oncoprotein networks. Cell 88:333±346, 1997

VOL. 119, NO. 4 OCTOBER 2002 MOUSE MODELS OF MELANOMA 791

Hussussian CJ, Struewing JP, Goldstein AM, et al: Germline p16 mutations in familialmelanoma. Nat Genet 8:15±21, 1994

Iwamoto T, Takahashi M, Ito M, et al: Aberrant melanogenesis and melanocytictumour development in transgenic mice that carry a metallothionein/ret fusiongene. EMBO J 10:3167±3175, 1991

Jacks T, Fazeli A, Schmitt EM, Bronson RT, Goodell MA, Weinberg RA: Effects ofan Rb mutation in the mouse. Nature 359:295±300, 1992

Kamb A, Shattuck-Eidens D, Eeles R, et al: Analysis of the p16 gene (CDKN2) as acandidate for the chromosome 9p melanoma susceptibility locus. Nat Genet8:23±26, 1994

Kamijo T, Zindy F, Roussel M, et al: Tumour suppression at the mouse INK4a locusmediated by the alternate reading frame product p19ARF. Cell 91:649±659,1997

Kato M, Takahashi M, Akhand AA, et al: Transgenic mouse model for skin malignantmelanoma. Oncogene 17:1885±1888, 1998

Kato M, Liu W, Akhand AA, Hossain K, Takeda K, Takahashi M, Nakashima I:Ultraviolet radiation induces both full activation of ret kinase and malignantmelanocytic tumor promotion in RFP-RET-transgenic mice. Invest Dermatol115:1157±1158, 2000

Kelsall SR, Mintz B: Metastatic cutaneous melanoma promoted by ultravioletradiation in mice with transgene-initiated low melanoma susceptibility. CancerRes 58:4061±4065, 1998

Klein-Szanto A, Bradl M, Porter S, Mintz B: Melanosis and associated tumors intransgenic mice. Proc Natl Acad Sci USA 88:169±173, 1991

Klein-Szanto AJ, Silvers WK, Mintz B: Ultraviolet radiation-induced malignant skinmelanoma in melanoma-susceptible transgenic mice. Cancer Res 54:4569±4572,1994

Krimpenfort P, Quon K, Mooi W, Loonstra A, Berns A: Loss of p16Ink4a conferssusceptibility to metastatic melanoma in mice. Nature 413:83±86, 2001

Kumar R, Sauroja I, Punnonen K, Jansen C, Hemminki K: Selective deletion ofexon 1 beta of the p19ARF gene in metastatic melanoma cell lines. GenesChrom Cancer 23:273±277, 1998

Kunisada T, Lu SZ, Yoshida H et al: Murine cutaneous mastocytosis and epidermalmelanocytosis induced by keratinocyte expression of transgenic stem cell factor.J Exp Med 187:1565±1573, 1998

Lyman S, Jacobsen S: c-kit ligand and Flt3 ligand: stem/progenitor cell factors withoverlapping yet distinct activities. Blood 91:1101±1134, 1998

Maehama T, Dixon JE: PTEN: a tumour suppressor that functions as a phospholipidphosphatase. Trends Cell Biol 9:125±128, 1999

Mintz B, Silvers WK: Transgenic mouse model of malignant skin melanoma. ProcNatl Acad Sci USA 90:8817±8821, 1993

Mitra J, Dai C, Somasundaram K, El-Deiry W, Satyamoorthy K, Herlyn M, EndersG: Induction of p21 (WAF1/CIP1) and inhibition of Cdk2 mediated by tumorsuppressor p16 (INK4A). Mol Cell Biol 19:3916±3928, 1999

Mulligan LM, Kwok JB, Healey CS, et al: Germ-line mutations of the RET proto-oncogene in multiple endocrine neoplasia type 2A. Nature 336:458±460, 1993

Noonan FP, Otsuka T, Bang S, Anver M, Merlino G: Accelerated ultravioletradiation-induced carcinogenesis in hepatocyte growth factor/scatter factortransgenic mice. Cancer Res 60:3738±3743, 2000

Noonan F, Recio J, Takayama H, et al: Neonatal sunburn and melanoma in mice.Nature 413:271±272, 2001

Otsuka T, Takayama H, Sharp R, et al: c-Met autocrine activation inducesdevelopment of malignant melanoma and acquisition of the metastaticphenotype. Cancer Res 58:5157±5167, 1998

Paramio JM, Navarro M, Segrelles C, Gomez-Casero E, Jorcano JL: PTEN tumoursuppressor is linked to the cell cycle control through the retinoblastomaprotein. Oncogene 18:7462±7468, 1999

Penna D, Schmidt A, Beermann F: Tumors of the retinal pigment epitheliummetastasise to the inguinal lymph nodes and spleen in tyrosinase-related protein1/SV40 T antigen transgenic mice. Oncogene 17:2601±2607, 1998

Petronzelli F, Sollima D, Coppola G, Martini-Neri ME, Neri G, Genuardi M:CDKN2A germline splicing mutation affecting both p16 (ink4) and p14 (arf)RNA processing in a melanoma/neuro®broma kindred. Genes ChromosomCancer 31:398±401, 2001

Platz A, Hansson J, Ringborg U: Screening of germline mutations in the CDK4,CDKN2C and TP53 genes in familial melanoma: a clinic-based populationstudy. Int J Cancer 78:13±15, 1998

Podsypanina K, Ellensen L, Nemes A, et al: Mutations of Pten/Mmac1 in mice causesneoplasia in multiple organ systems. Proc Natl Acad Sci USA 96:1563±1568,1999

Pollock P, Trent J: The, genetics, of, cutaneous, melanoma. Clin Lab Med 20:667±690, 2000

Pollock P, Walker G, Glendening M, Que Noy T, Bloch N, Fountain J, HaywardN: PTEN inactivation is rare in melanoma tumours but occurs frequently inmelanoma cell lines. Melanoma Res, in press, 2002

Quevedo W, Fleischmann R: Developmental biology of mammalian melanocytes. JInvest Dermatol 75:116±120, 1980

Randerson-Moor JA, Harland M, Williams S, et al: A germline deletion of p14 (ARF)but not CDKN2A in a melanoma-neural system tumour syndrome family.Hum Mol Genet 10:55±62, 2001

Rane S, Dubus P, Mettus R, Galbreath E, Boden G, Reddy E, Barbacid M: Loss ofCDK4 expression causes insulin-de®cient diabetes and CDK4 activation resultsin b-islet cell hyperplasia. Nature Genet 22:44±52, 1999

Recio JA, Merlino G: Hepatocyte growth factor/scatter factor activates proliferation

in melanoma cells through p38 MAPK, ATF-2 and cyclin D1. Oncogene21:1000±1008, 2002

Rizos H, Puig S, Badenas C, et al: A melanoma-associated germline mutation in exon1b inactivates p14ARF. Oncogene 20:5543±5547, 2001b

Schmidt A, Tief K, Yavuzer U, Beermann F: Ectopic expression of ret results inmicrophthalmia and tumors of the retinal pigment epithelium. Int J Cancer80:600±605, 1999

Schmidt EE, Ichimura K, Reifenberger G, Collins VP: CDKN2 (p16/MTS1) genedeletion or CDK4 ampli®cation occurs in the majority of glioblastomas. CancerRes 54:6321, 1994

Schmidt L, Duh FM, Chen F, et al: Germline and somatic mutations in the tyrosinekinase domain of the MET proto-oncogene in papillary renal carcinomas. NatGenet 16:68±73, 1997

Sears R, Nevins J: Signalling networks that link cell proliferation and cell fate. J BiolChem 277:11617±11620, 2002

Serrano M, Lee H-W, Chin L, Cordon-Cardo C, Beach D, DePinho R: Role ofINK4a locus in tumour suppression and cell mortality. Cell 85:27±37, 1996

Sharpless N, Bardeesy N, Lee K-H, et al: Loss of p16Ink4a with retention of p19Arf

predisposes mice to tumorigenesis. Nature 413:86±91, 2001Sherr C, Roberts J: CDK inhibitors. positive and negative regulators of G1-S phase

progression. Genes Dev 13:1501±1512, 1999Silvers WK, Mintz B: Differences in latency and inducibility of mouse skin

melanomas depending on the age and anatomic site of the skin. Cancer Res58:630±632, 1998

Sotillo R, Dubus P, Martin J, de la Cueva E, Ortega S, Malumbres M, Barbacid M:Wide spectrum of tumours in knock-in mice carrying a Cdk4 proteininsensitive to INK4 inhibitors. EBMO J 20:6637±6647, 2001a

Sotillo R, Garcia J, Ortega S, Martin J, Dubus P, Barbacid M, Malumbres M:Invasive melanoma in Cdk4-targeted mice. Proc Natl Acad Sci USA 98:13312±13317, 2001b

Sou®r N, Avril MF, Chompret A, et al: Prevalence of p16 and CDK4 germlinemutations in 48 melanoma-prone families in France. Hum Mol Genet 7:209±216, 1998

Sviderskaya EV, Hill SP, Evans-Whipp TJ et al: p16 (Ink4a) in melanocytesenescence and differantiation. J Natl Cancer Inst. 94(6):446±454, 2002

Takahashi M, Iwamoto T, Nakashima I: Proliferation and neoplastic transformationof pigment cells in metallothionein/ret transgenic mice. Pigment Cell Res5:(2)344±347, 1992

Takayama H, LaRochelle W, Anver M, Bockman D, Merlino G: Scatter factor/hepatocyte growth factor as a regulator of skeletal muscle and neural crestdevelopment. Proc Natl Acad Sci USA 93:5866±5871, 1996

Tang L, Li L, Tron V, Trotter M, Ho V: Expression of cell cycle regulators in humancutaneous malignant melanoma. Melanoma Res 9:148±154, 1999

Tolkacheva T, Chan AM: Inhibition of H-Ras transformation by the PTEN/MMAC1/TEP1 tumor suppressor gene. Oncogene 19:680±689, 2000

Tsao H, Benoit E, Sober AJ, Thiele C, Haluska FG: Novel mutations in the p16/CDKN2A binding region of the cyclin dependent kinase-4 gene. Cancer Res58:109±112, 1998

Tsao H, Zhang X, Fowlkes K, Haluska FG: Relative reciprocity of NRAS andPTEN/MMAC1 alterations in cutaneous melanoma cell lines. Cancer Res60:1800±1804, 2000

Van Elsas A, Zerp SF, Van der Fliers S: et al: Relevance of ultraviolet-induced N-rasoncogene point mutations in development of primary cutaneous melanoma.Am J Pathol 149:883±893, 1996

Van Weering D, Bos J: Signal transduction by the receptor tyrosine kinase Ret.Recent Results Cancer Res 154:271±281, 1998

Walker GJ, Flores JF, Glendening JM, Lin AH, Markl ID, Fountain JW: Virtually100% of melanoma cell lines harbor alterations at the DNA level withinCDKN2A, CDKN2B, or one of their downstream targets. Genes Chrom Cancer22:157±163, 1998

Walker GJ, Hayward NK: No evidence of a role for activating CDK2 mutations inmelanoma. Melanoma Res 11:343±348, 2001

Wellbrock C, Gomez A, Schartl M: Signal transduction by melanoma oncogenicreceptor tyrosine kinase Xmrk in melanoma formation of Xiphophorus. PigmentCell Res 10:34±40, 1997

Weng LP, Brown JL, Eng C: PTEN coordinates G(1) arrest by down-regulatingcyclin D1 via its protein phosphatase activity and up-regulating p27 via its lipidphosphatase activity in a breast cancer model. Hum Mol Genet 10:599±604,2001

Whiteman D, Whiteman C, Green A: Childhood sun exposure as a risk factor formelanoma. A systematic review of epidemiological studies. Cancer CausesControl 12:69±82, 2001

Xiao Z, Chen J, Levine A, Modjtahedi N, Xing J, Sellers W, Livingston D:Interactions between the retinoblastoma protein and the oncoprotein mdm2.Nature 375:694±697, 1995

You MJ, Castrillon DH, Bastian BC, et al: Genetic analysis of Pten and Ink4a/Arfinteractions in the suppression of tumorigenesis in mice. Proc Natl Acad Sci USA99:1455±1460, 2002

Zhang Y, Xiong Y, Yarbrough WG: ARF promotes mdm2 degradation andstabilizes p53; ARF-INK4a locus deletion impairs both Rb and p53 tumoursuppression pathways. Cell 92:725±734, 1998

Zhu H, Reuhl K, Zhang X, Botha R, Ryan K, Wei J, Chen S: Development ofheritable melanoma in transgenic mice. J Invest Dermatol 110:247±252, 1998

Zuo L, Weger J, Yang Q, et al: Germline mutations of the p16 binding domain ofCDK4 in familial melanoma. Nat Genet 12:97±99, 1996

792 WALKER AND HAYWARD THE JOURNAL OF INVESTIGATIVE DERMATOLOGY