Sunlight and the onset of skin cancer

REVIEWS

H o w do cancers stare) We would prefer to know actual events in human patients, rather than how cancers can be triggered artificially in labomtoW mice. This will be a detective sto W - sifting through availahle human tissue for smoking guns. In addition to identif~ng importam genes, we wish to know the carcinogen, the cell it entered, the structural change in the DNA, the resulting mutations, and the effect of these mutations on cell physiology. Finding a gene in which these events occuned creates a s2,'stem for repeatedly asking questions at the interface of carcinogenesis and genetics, Tools for answering such questions are most powerful for skin cancer, because a great deal is known alzout ukraviolet radiation. Parts of this stow are the subject of specialized reviews t-'i. Here, after briefly describing tile disease, we trace the skin's genetic misfortunes. Skin cancer Sun-exposed skin is heir to three cancers: melanoma; Ixt~l cell carcinoma (BCC); and squamous cell carcinoma (SCC)5. Melanontas, tile most deadly, can arise in young adults. They begin as a radial proliferation of normally nonproliferating melanocytes. Danger enzers widen the radial lesion begins growing vertically, after which metastasis is possible. SCC and BCC are tumors of keratino%'ytes, which are cells that routinely proliferate, differentiate and are shed from the skin, These tumors often appear at the age of 70 years or later. They usually begin on a background of sunMamaged skin, characterized by lost elasticity and individual disordered kevatinocytes 6. Continued sun exposure leads to keratinized reddish patches of actinic kerarosis, with aberrantly differentiating and proliferating ceils. These pre..cancers usually regress, but one in a thousand progresses to SCC; these tumors are often aneuploid and can metastasize, In contrast to this step-wise pro~ession, BCCs seem to arise without precursors, seenlingly front keratinocytes in hair follicles. They are usually diploid and rarely metastasize, dlough they invade Iot-ally. These profound tissue changes are precipitated by photons, quantum packets of light. Sunlight All three cancers correlate with exposure to sunlight. They usually occur in individuals with light skin blondes or red-heads who bum rather than tan - and those living in sunny climates 7. The post-war obsession with the beach has resulted in a surge in skin cancers. In the southern US and Australia, they exceed all other cancers combined and are still increasing. The sun emits radiation ranging from X-ray to ultraviolet to infrared. UV wavelengths all cause skin cancer in mice 8. The most energetic of these, UVC (100-280 am), hms the right wavelength to be directly absorbed by DNA. Less energetic UVA (315-400 nm), found in tanning parlots, is absorbed by other cellular chromophores. Their excitation generates reactive oxygen specie.s, such as the hydroxyl radical, which can cause DNA-strand breaks and chromosome translocations. UVB (280-315 nm) is intermediate, its photons weakly absorbed by the same molecules that tx~st absorb tlVC or UVA. Because adults can reduce their 15sk of precancers by using sunscreen 9, some of the photons leading to cancer

Sunlight and the onset of skin cancer BOUGIAS L BRASH(dougiss~asb~,Ze.~) ?7~ photons of sunlight precipitate a series of geaelic eveHts in skin leoding to cance~, These events to~Ive somatic mutations as well as inherited allele~ Competition belween cell populaticms ensues, as a single mutated cell expands irao a ctoxe. Thus cancer tnvol~es both a singlecell problem and a marq.cell ~'~alera,. In skin cucer, sunlight appears to drive bosh. must act after striking the skin of adults. However, most of the critical sunlight exposure occurs txffore age 18. For example, people who moved from England to Australia as children, but not as adult.,;, acquired the high Australian skin cancer risk lO.lt. Thus, some of the molecular scars left by sunlight are a half-centuw old. This persistence motivated a search for mutations made by sunlight; acute effects of sunlight, such as suppression of immune surveillance, would have long since disappeared. UV ph0toproduets and tJV-b~'.'uced mutations The most frequent DNA photoproducts made by UV join adiacent cytosines or thymines I. UVC and UVB photons are usually absorbed at the 5-6 double bond, allowing it to open. Where two pyrimidines are adjacent, one of two events usually occurs. If Ixxh 5 - 6 bonds open, a ring is formed, creating the 'cyclobutane dimer' (Fig. la). Alternatively, tile double bond of tile 5' pydmidine opens and reacts acrc~s tile exocyclic group of the 3' pyrireidine. After spontaneous rearrangement, a single bond is left between tile two pytimidines. This is the 'pyrlmidine--pyrimidorte (6-4) photoproduct' (Fig. lb) (Ref. 12). Both photoproducts bend the HNA or rotate a base ~o it resembles an alyasic site. Both photoproducts cause mutations. This was shown in ~'cherfcbia coli using the observation that methyl groups prevent (6-4) photoproducts from forming. In d c m - bacteria, which do not methylate cytosines at restriction sites, UVC-induced mutations increase only at the restriction sites13. In human cells, using a DNA repair enzyme to remove cyclohntane dimers from a UVC-irradiated shurde vector before tmnsfection reduces the mutation frequency li. In both organisms, UV-induced mutations are located where one pyrimidine is next to another. Mutations are usually C ~ T (cytosine to thymine), resulting from insertion of A (adenine) opposite tile damaged C dudng subsequent DNA replication. (Thymines in photoproducts are less often mutagenic.) Ten percent of the mutations are CC~qT resulting from replacing both cytosines. This unique specificity of UV mutagen,~is - about 70% C ~ T at dipyrimidines and 10%0C C ~ ' f f - has heen known for many years 15-17. Few other mutagens involve tandem bases, and they primarily make other mutations. C c ~ T r , in particular, is considered to be diagnostic for UN (Ref. 18). These distinctive mutations are the smoking gun. if made by sunlight in a single cell half a centut3, ago, they will still be pre~mt in descendants of that cell.

TIG OCtObER 1997 VOL. 13 NO. 10

REVIEWS

Genetic ehan~ in SCCs

T< >T TO (6-4) Photopradoet To find such mutations, one first has to guess the right gene. Our guess was lyased on a rare skin disease o characterized by lesions that contain human papillomavirus and can progress to SCC. One of the viral proteins binds to and inactivate~ the TP53 tumor suppressor NO CH~ N CH~ protein. It seemed that sunlight might act by mutating the TP53 gene d~rectly. We now know that 7P53 is mutated in over half of all human cancers. The TP53 O O protein has since been found to be a transcription i N... factor, a protein that turns other genes on and off. Over 9(P/~ of the SCCs from the USA have a mutation somewhere in the TP53 tumor suppressor gene 2.3.1'). Fmvp.E1. UV photopaxlucls. The Tr cydobutane pyrimidine The mutations are unusual, occurring only where one dimer and the TC pyrimidine-pyrimidone (6-4) ph~)toproduct. In cytosine or thymine is adjacent to another. About two- the t3,clohutanedilncr, file angle Ix~t,x~n b',t~.,sis less than the thirds of the base changes are C ~ T and several are nonna136~and the DNA-s~andacqui~ a Ixmd. In file t6-4t C C ~ T r (Table 1). Many of the same codons are photoprodu~L the 90° rotatkm of the 3' base makes it re:~emblean mutated as in internal cam:urn, such as colon or blad- abasic site. der, but the base changes are different 2°. The mutations tend to cluster at nine mutation hot.spoLs3, many of order to contribute to cancer, most mutations will make which are sites of slow exdsinn repair of DNA photoamino acid changes that are potentially carcinogenic. products zt. UV-like Trp53 mutations are also found in Tumor mutations therefore tend to reflect the mutation skin tumors from LrV-irradiated mict~z. Correspondingly, spectrum of the carcinogen. Oncogenes are less free mice with a mutated Trp53 gene are more susceptible from phenotypic selection, because they must acquire a to UV-induced skin cancer 2z. Not all genes are tatgeta of novel function for oncogenesis. But when do these sunlight: human skin cancers rarely contain mutations mutations occur? in members of the RAS oncogene family. Observing dipyrimidine C ~ T mutations, including SCC pre-caneers C C i T t ' , allows us to deduce that the mutagen was UV The first clinically apparent lesions are the actinic radiation directly absod~ed by DNA. The culprit is not kemtoses, SCC precancers that begin appearing about UVC, however, because the ozone layer absorbs it com- age 45 in light-skinned individuals. Sunlight-induced pletely. Without an ozone layer, skin cancer would in- C ~ T and C C i T t mutations have been found, as well crease at least 108-fold z3. UVB does penetrate the ozone a.s in the next SCC stage, termed carcinoma in situ3,Z~. In contrast to tile tumors, amino-acid substitutions in layer, the degree depending on the thickne~ of the layer. Its weaker absorption IW DNA k~ads k.*s efficiently to tile actinic keratoses arc spread evenly across tile gene. same UV photoproducts. Some skin minors contain mu- Some TP53 ntutatinns are therefore less likely m tations other t}mn C ~ T , just as seen in UV experiment'~ progress to SCC than others. Actinic keratoses also have using cultured cells to.tT. While these are probably also aIlelic loss at loci in addition to TP53 (Ref. 25), so other due to UV, they arc best considered "non-informative': genes might be involved as well. they are consistent with a UV origin, but are not compelPatients with multiple actinic keratoses often have ling because other physical and chemical carcinogens TP53 mutations in more than one. In these cases, tile also cause Ihem. We have now identific~.t the t~rcinogenic mutation is different in eacfi lesion 24, so ead'= actinic kemtosis records a separate UV pfioton absocption wavelength in humans (UVB), the DNA photoproduct.s Icydobutane dimers and ( 6 - 4 ) photoproductsl, the event. A patient's ~kin is a living 'petri dish' for the types of mutations made ( C ~ T and C C ~ T r ) , and one effects of photons. of the genes mutated by sunlight (a tumor suppressor gene, TP53). TAm~ 1. Mutations in t h e TP53 Rtmor suppressor gene ill squammm cell But do these mutations concarcinomas o f the sldn (representative. subset) 19 trihute to the tumor, or is DNA just an exposure meter l~Jr UV radiPatieQt age ~ r TP53 DNA Base ~ackl ation? A crucial finding is that all (yza~) slta codon sequence change charge mutations change the amino acid (Table 1). Because many base 86 Preauricular 7 TCT C~G/WT Asp~His changes, such as C ~ T at tile dtird 82 AC./WT Giy~Ala ... stop Temple 104/105 GCCT position of a cx:Kton, do not change 69 Scalp 151 C~CC C~A/WT Pro~His tile amino acid, the mutations seen 69 Pro~.~'r Hand 152 CCCCC C~T/0 in skin cancers must have been 80 CC~TI'/WT AsnArg~AsnTrp Nose 247-248 ACCG C~T/0 GlumLy6 Side ~ffFace 258 TfCC selected for. Mere passengers could 56 Cheek 278 TCCT C~T/WT Pro~Ser have been silent or could have con76 Face 285-286 T c c r CC~'IT/Fl GluGlu~GluLys stituted only a portion of the DNA 85 Gln~stop Postauriculsr 317 CCCCA C~T,,~d~r sequencing signal. It is also fortu75 nam that the mutations were found in a tumor ~uppressor gene. Because such genes require inactivation in

Letters in bold show mutated ba.'~s. Abbreviations: O, allelic loss; A, deletitm; WT, wiM iype. TIG OCrOBER 1997 VOL. 13 NO. t 0

411

REVIEWS

to normal TP53 protein reveals the arrangemem of mutated cells. Instead of being Scattered in the epidermis, TP53-mutated cells are present as tiny clones 60-3000 cells in size (Fig. 2). Often, these clones ent-ase a hair follicle or form a cone having its apex at the dermalepidermal junction. Clones are most frequent on sunexposed skin. TP53 gene mutations were idemifled in half of 24 clones microdissected and sequenced; most of these were sunlight-induced mutations that changed an amino acid 27,28. l~%~atl~ the antilxJdy method detects most TP53mutations, it reveals the actual mutation frequency in skin. Surprisingly, the frequem3' on sun-exposed skin averages 30 clones cm-L Face and hands contain thousands, with almost one epidemlal cell in 20 having a TP53 mutation. The clonal arrangement reveals that these cancer-prone cells are not sitting still while waiting for additional genetic hits - they are proliferating. Cleady, most such clones never progre~.,;. This hah might be due to regression or to thilure to mutate an additional gene. In mice, regression has been observed. Repeated UV-irmdiation generates Trp~-mutated clones, but after irradiation ends the clones begin to disappear 29.

Apoptosisand cellular proofreading

Fmtr~ 2, TP53 immunoposkive dories in whole-muum preparations of human epidermis The epidermis was separated ft~m the de~is, fixed and immunostained filr wild-~.'pe TP53 prt~tein. ElevatedTP53 u~ually¢ormspund~ to a mutation in the TP53gene, which was confirmed by mlclodimection and DNA scxluencing tat Clime of ~P53-mumted kenainucTtes: (hi clone surrounding a hatr follicle.(c) Chine vie~ ed from the side in a Ihree-dimensiunal con[oual inicn~setJpyintagc The :lpeX of tile clone lies :it the dennal--epidermal junction (:lrmwL (Repr~xluced with permi~qon from ReL 27.) Mutations in normal sldn Cancer is believed to result from multiple genetic hit~. Normal individuals would then have mutated cells containing fewer hita than needed for cancer. Several 7P53codons were found mutated in sun-exposed human skin at frequencies of one mutated gene per 106 cells or higher 2426. The three
To I~am how inactivating TP53 co:-tributes to skin cancer, we need to know its function. As a transcription factor, it tnms on or off the expre.'~inn of genes involved in the cell cycle, DNA synthesis, DNA repair, and programmed cell death ~. TP53 is part of a DNA damage rcsi~-mse, I~ecoming more stable after a ceil is irradiated with UV or y-rays. The signal for La/-induction of TP53 and evil death was recently found to originate from photoproducts in actively transcribed genes :u. Without TP53, culrtlred colon cells have a fi,a~lbld higher ~pontaneous mutation rate 32. In skin, TP53 also regulates cell death. DennatologisLs know that skin overexposed to sunlight contains unusual keratinocytes, dubbed 'sunhum cells'. These have pycnotre nuclei and an incensely eosinophilic cytoplasm (Fig. 3at, a morpholc,gy wpical of apoplosis. More conelusively, they contain nie hullmm'k of apoptosis, DNAstrand breaks IFig. 3b). Generating these apoptotic cells requires TP53. In Trp53 knockout mice, the sunburn cell I~cqucncy after UVB is an order of nla~itude lower than in wild-type: the heterozygote is intermediate 24. ThErefore, the skin recognizes and eliminates aberrant kemtinocytes - a supplement to the protection afforded by routinely shedding keratinocytes during differentiation. UVA may Ix: as important for apoptosis as U~,rB, because it is mote prevalent. The elimination mechanism has been termed 'cellular proofreading' because the mistake is erased rather than repaired :~:~.The term also draws attention to the fact that apoptosis is just an ellector step. The sensor that recognizes that a cell is aberrant is, perhaps, more interesting. For example, abnomlal cell Qa'les caused by inadive retinohlastoma protein lead to TRP53dependent apoptosis in the lens and choroid plexus of transgenic miee3~. When pregnant mice are X-irradiated, Trp53is required for the apoptotie resorption of embryos that otherwise would he born malformed33. Capitalizing on the cell's own sensor for recognizing cancer-prone cells might be a mort" precise route to cancer therapy than bludgeoning the cell with toxic drugs.

TIG OCTOBER1997 VOL. 13 NO. 10 412

REVIEWS

(n]

Fmtrg~ 3. Sunburn cells are apnptc~tic. (a) Pycnntic nuclei anti intensely eosinnphilic ofloplasm of two sambum cells in routine skin tarrowst. (b] Cells :','ill] DNA-~tmndbreaks (pitlkk after end-labeling with fluorescent nucluotidcs in tile batl~ paraffin see'tit;n (amm,M. BCCs and othergenetic changes BCCs are a little different. Although nearly all BCCs have overly-stable TP53 protein, only half the tumors have mutations in the TP53 structural gene itself3. These lumors o[len have sunlight-induc.,.'d mutations in both alleles, a striking demonstration of the mutagenicity of sunlight. The mutations in !lie remaining tumors evidently lie in another member of the TP53 pathway. An exciting devektpment has been the emet'genee of a new, apparently BCC-specifie, pathway of tu;nor suppressor genes and oncogenes. The starting point ~.,~lSGorlin syndrorne, an autusomal dominant d ~ r d e r in whirl1 patients have multiple BCCs as well as jaw cysts and tiny pits in the palms of the hands. Positional cloning revealed the gene to be PTCH {'patched'), a homolog of the D~sopbila gene ptc (Ref. 35). It is the tmnsmembrane receptor for the Hedgehog signal transduction pathway, whose downstream targets include 1~';/:-8 and the WNTfamily. F K M is lost not only in the BCCs of Gorlin patients, but also in their jaw cysts - evidently clonal developmental defects that arise in the

same molecular way tumors do. Sporadic BCCs also have PTCH mutations, mo~t of which are UV-like, indicating the gene as another genetic target ior sunlight. An inhibitory ligand for dae PTCH receptor is the secreted protein Sonic hedgehog. Transgenic mice overexpressing the Shh gene develop many fc~atures of Gorlin syndrome, and an SHH mutation has been identifitxt in a BCC (gel. 36). Ahhough melanomas rarely involve TP53, the tumor suppressor gene CDKN2A (or 'pl6") is mutated in several familial melanoma kindreds. It encodes an iohibitor of CDK4, a ceil-cycle regulator. Mos~, melanoma cell lines have mutations in GT]KN2A, often C ~ T at dipyrimidine sites or C C I T T (Refs 37. 38). In primaD' melanomas, mutations are rarer and the relation to UV less clea# 9. The gene encx)ding J3-catenin, a signaling protein, has mutations in cell lines that might also be UV-induced 40.

Clonalexpansion:the many-cell problem How do the TPS~mutsted clones in normal skin arise from a single mutated celP An apoptosis-resistant cell is

~unbum 0o{I

l

TP53 induced ....

Tissue

epidermis

NormaJ

Cell

ksralinOcyt o

DNA

Norm=lDNA

U V B

~

0; l

Outof sun: cloneregresses Ftct~ 4. A m(Jdel for Aenetic and cellular evems in tile 0nset of hutllan squamous cell car¢inonla of IlK' skin, Sunlight ~.TCat~ cycklhutane dimers and (6-4) photoprcKlucls in I)NA, snnle of whM1 cause muLationsin the ll'~J lunior suppreK~orIlene. One of ti~e cellular phenotypus of a 1P$fi mutation, resistance to gtlnliRhl-intluced apopto,~is, allows rel×'att'd sunlight t'xl~}sure Io .',elect t~)r /PC3 mutmed culls. One ~}fthese cells might aRain incur Ihe firs1dlbt't or sunlight, mulagellusls. (Adapt~-J from Itel- 24.t 'FIG OCI'OBEK t997 VOL | 3 No. | 0 413

REVIEWS

more likely to survive a DNAMamaging trip to the beach. Yet, failure of cellular proofreading has a worse consequence. Because the mutated cell's normal neighbors still undergo apoptosls when damaged, they will leave room for the mutam to expand ¢lonally into other stem-cell compartments. Thus sunlight might act as a selection pressure favoring the clonal expansion of TP53.mutated cells (Fig. 4), each beach visa giving the clone a nudge. Sunlight appems to act in t*xo ways: mutating genes and then, afterwards, selecting for clonal expansion of mutated cells. This model is supported, but not proven, by the observation that Trp53-mutated clones in mice regte~ in the absence of UV exposure zg. Similarly, in humans the TP53-mutated clones are largest in sun-exposed areas 27. CIonal expansion actually contributes more mutant cells than the initial mutagenic effect of sunlight. In chronically sun-exposed skin, the aggregate size of large TP53mutated clones (the ones most clearly dependent on exposure) exceeds the total of the more numerous small clones -~7. An expanding cell-death defective clone will be more likely to accumulate additional mutations (Fig. 4). First, it presents a larger target for future UV exposures. Second, it is a target in which a greater [ractinn of cell, can survive UV irradiation to acquire an additional mutation in TP53 or another eerie. Geneticists might he surprimd hy the numerical impact of a cellular mechanLcm such as donal expansion. Mutations after low UV doses occur in mammalian gen~s at a frequent.T of 10-~ par cell generation or lesM7. Thus, the probability of nlutaling both alleles of two parllcular tumor suppressor genes is 10"-20. Because the number of prohferating keratino~Ttes in human skin ir~ ~bout 106 cm -2, with about 0.1 m z of skin exposed, ahout 10-ll cells will be quadruply mutated. Even after 100 cell generations, only one in 1000 million people would have a skin minor. For comparison, tile lilt'time expectancy of skin cancer in sunny climates is actually 20°/0 or more. This discrepancy cannot easily be resoh,ed by invoking mutator phenotypes, and would grow worse if additional genes are involved. Cinnal expansion, however, can easily incr~ase tile likeliltood oi" each subsequent mutation a thousandfold. Apoptusis is a relatively high-frequency physiological event that t-an facilitate the expansion of many TP53mutated cells simuhancou.sly. The next mutation can then be quite rare, because only one cell in one of the clones must he hit. Clonal expansion makes multiplegenetic-hit cancer feasible. The lessons sunlight has taught us - about mutations and expanding clones are, thus, likely to illuminate the remaining catalog of tumors whose onset occurs in the dark, out of sight. Aekaowkdgements This work is supporied by the National C2qc~r Institute, American Cancer Society, Swiss National Foundation and Swiss Cancer League, the Munson Foundation, and Hull and Swebelius Cancer Research Awards.

Proc. 1,136-142 4 Kmemer, K.H. (19971 Pixy:. ,'VAILAcad. SoL U. S. A. 94, 11-14 5 DeVita, V.T. Hellman, S. and llt~senberg. S.A., eds (1997) Cance~ Priltciffles and Practice of Oncolo,~;

Lippincett-Raven 6 Gildlres,t, B.A., ed. 11995) Pbolodamage, Blackwell Sci~qL'e

7 Udrach, E (1984) in Topics in PbotomedicinP(Smith, K.C., ed.), pp. 67-104. Plenum Pre~s 8 de Grui]L F.R. and Fofl~es. P.D. 11995) BioFA~ays17, 65!~69 9 Thompson, S.C.,Jolley. D. and Marks, R. 119931Ak,wEngL J. Met/. 329, 1147-1151 10 Marks, R., Jolley. D., Leclsas, S. and Foley, P. 119901bled. .L Aifst. 152, 62~o6 11 Kricker, A.. Amxstrong, BK., English, D.R. and Heenan, PO. 119911hU.J Cancer48, 650-662 12 Mitchell, D.L and Naim, ILS. (1989} photodmm. PhotobioL 49, 805~419 13 Glickrmln, B,W. el at ( 19861 ps~c. Natl. Acad. ScL U. S. A. 83, 6945-6949 14 Brash. D.E. et aL ( 19871proc. A~aa.Acad. ScL U. S. A. 84, 3782-3786 15 Miller,J.H. 119831AmltL Rev. C-e~mt.17, 215-238 16 LebkowskL J.S., Clancy, s., Miller,J.H. and Calos. M.P. 11985) Prec. NatL Acad, Sci. tL S. A. 82, 8606~8610 17 McGregoL W.G. et aL 119911Mnl. Cell. Biol. 11, 1927-1934 18 Hutchinson, F. 119941Murat. Res. 309.11-15 19 Brash. D,E. et al. 119911proc. NaIL Acad. Sci. O~S. A. 88, 1012@10128 20 Greenblatt. M.S.. Bennett. W,P,, Holl~tein. M. and Harris. C.C. 11994) CancerRes. 54, 4855~i878 21 "E3rnaleai. S. and Pfeifer. G.P. (19941 Science263, 1430-1438 Z2 Li, G., Ho, V.C., Berean, K. and Tron, V.A. t19951 Cancer Res 55. 2070-2074 23 Frederick. J.E.. Snell, H.E. and Hay~vo~xl,E.K. i19891 Photochem. Photoblol. 50, 443-450 24 Ziegter, A. et at. ( 19941Natule 372, 773-776 25 Rehman. I,. Quinn, A.G.. Healy, E. and Rees, J.L 119941 Lancet 344.7~kq--789 26 Naka?~lWa.H. et ~IL(1~14) I'roc. Null, Acad. Sci IL S. A 91,360-364 27 Jonason, A.S. et aL (19961 P~c. Natl. Acad Sci. U S A. 93. 14025--14029 28 Pen. Z.P. et aL 119961On~xJ,qene 12, 76%773 29 Berg. R.J,W. el al. (19961 Prec. Natl. Acad. SoL U. S. A. 93. 274-278 30 Levine, A J 119971Cell88, 325-331 31 I.jungman, M. and Zhang. F. 119961OHcoge;lel3. 82.~,451 32 Havre, P.A. et al. 119951Cancer Res. 55. 4420~4424 33 Brash, D.E. (19961 Nat. Med. 2, 525-526 34 White, E. 119941A'altttt,371.21-22 35 Cmilani, M.It. et aL 119961~tt. Ge,uet. 14, 78-81 36 Ore, AE. etaL 115~71Science276, 817-821 39" Liu, Q. etal. 11995) OncogenelO. 1061-1067 3 8 Pollock) P.M., Yu, F., Par~uns, P.G. and Hayward, N.K. (1995) Oncogene 11, 663-(K~ 39 Gnfis. N.A. et al. 111.1951Am.J. PatboL 146.1199-1206 40 Rublnfe]d, B. et aL 119971Science 275,1790-1792

Refercnc~

b E . Brash is Oz the Departments of Therapeutic Radiology and Genetics. and the Yale Comprehet~ive Call¢¢r Center, Yale School of Medicine, 15 York Slreet/HRT~09, New Haven, CT 065204~040, USA.

I Brash, D.E. (19881 Photochem. Pbotobiol. 48. 59-66 2 Nataraj, A.J.. Trent. J.C and Ananthaswamy. H.N. i1995) Photochem. PhotobioL 62, 218-230 3 Brash, D.E. etal. (19961J. Invest, DermatoL Synlp.

TIG Oc2roREg 1997 VOL. 13 No. 10

414