An Assessment of the CDKN2A Variant Ala148Thr as aNevus/Melanoma Susceptibility Allele

Chandra G. Bertram, Rupert M. Gaut, Jennifer H. Barrett, Elizabeth Pinney,* Linda Whitaker, Faye Turner,Veronique Bataille,* Isabel dos Santos Silva,² Anthony J. Swerdlow,³ D. Timothy Bishop, andJulia A. Newton BishopDivision of Genetic Epidemiology, Cancer Research UK Clinical Center in Leeds, St. James's University Hospital, Leeds, U.K.; *Mathematics,

Statistics and Epidemiology, Cancer Research, London, U.K.; ²Department of Epidemiology and Public Health, London School of Hygiene and

Tropical Medicine, London, U.K.; ³Section of Epidemiology, Institute of Cancer Research, Sutton, U.K.

Melanocytic nevi are the most potent risk factors formelanoma yet identi®ed. Variation in the nevus phe-notype within a population is predominantly genetic-ally determined. Genes that determine nevusexpression may therefore act as low penetrancemelanoma susceptibility genes. Rare germline muta-tions in CDKN2A predispose to melanoma andappear to be nevogenic, although the correlationbetween nevus phenotype and mutation status ispoor. It is plausible that more common CDKN2Avariants may in¯uence both melanoma susceptibilityand nevus susceptibility. Ala148Thr is a G to A mis-sense polymorphism of CDKN2A, which is found in4%±6% of the general population. We have investi-gated the role of Ala148Thr as a low penetrancemelanoma or nevus susceptibility allele in two separ-ate groups of individuals. The ®rst was a sample of488 adults recruited from 179 families of patientswith the atypical nevus phenotype and/or a family

history of melanoma, and the second was a popula-tion-based sample of 599 women. Similar preva-lences of Ala148Thr (4.9% and 5.2%) were found inboth samples but signi®cant variation in the preva-lence of the polymorphism was seen across geo-graphic areas within England. There was noassociation between Ala148Thr status and nevusnumber or history of melanoma, and therefore theresults did not support the hypothesis that theAla148Thr variant is a low penetrance melanoma ornevus susceptibility allele. A signi®cant protectiverole of Ala148Thr on the number of atypical neviwas observed in the family sample (mean of 1 atyp-ical nevus in those with the allele and 3.5 nevi inthose without, p = 0.02). After allowing for potentialconfounders this was not evident in the population-based sample. Key words: Ala148Thr/atypical mole syn-drome/CDKN2A/melanoma/nevi. J Invest Dermatol119:961±965, 2002

Highly penetrant mutations have been found in themelanoma susceptibility gene CDKN2A, at 9p21(Goldstein and Tucker, 1997; Harland et al, 1997).The gene encodes a cyclin-dependent kinase (CDK)inhibitor (p16) (Serrano et al, 1993) that arrests

progression of the cell cycle at the G1 checkpoint prior to S phase.A variety of germline CDKN2A mutations have been described inmelanoma families with variation in the proportion with mutationsbetween centers (Holland et al, 1995; Walker et al, 1995; Fitzgeraldet al, 1996; Sou®r et al, 1998). Mutations have been found in 48%(20 of 42) of families with three or more cases of melanoma in ourseries (unpublished data, Newton Bishop JA).

There are phenotypic risk factors for melanoma such as clinicallyatypical and large numbers of benign nevi (Swerdlow et al, 1986;

Augustsson et al, 1990; Bataille et al, 1996). We use the term the``atypical mole syndrome'' (AMS) to describe this phenotype,which is seen in melanoma families and in the general population(Bataille et al, 1996). Twin studies suggest that variation in nevuscount between individuals is determined largely by genes (Easton etal, 1991; Zhu et al, 1999; Wachsmuth et al, 2001), but the geneticdeterminants remain poorly characterized. It is not known whethergenes that determine common nevi also determine atypical nevi orwhether the latter are under separate genetic control. In view of theobserved associations with risk of melanoma, however, it isplausible that some nevus genes may also act as melanomasusceptibility genes. Putatively genes that in¯uence nevus (andtherefore melanocyte) number could play a role in the causation ofmelanoma, merely by affecting the number of target cells. Nevi andmelanoma could also be causally related to the same genes acting inconjunction with ultraviolet radiation.

There is evidence that CDKN2A is nevogenic: the AMS iscommon in families with CDKN2A mutations and mutations havebeen found in dysplastic nevi. 30%±43% of melanomas in multiple-case families, however, occur in people with normal moles(Bergman et al, 1986, 1992; Newton Bishop et al, 1994). Onehypothesis to explain lack of correlation is that there may bedifferential effects of environmental exposures on nevus phenotypeand melanoma risk. An alternative hypothesis is that the nevus

Manuscript received January 25, 2002; revised March 22, 2002; acceptedfor publication April 4, 2002.

Reprint requests to: Dr. Julia Newton Bishop, Genetic EpidemiologyDivision, Cancer Research, UK, Clinical Center in Leeds, Cancer GeneticsBuilding, St. James's University Hospital, Beckett Street, Leeds LS9 7TF,U.K. Email: j.newton-bishop@cancer.org.uk

Abbreviations: AMS, atypical mole syndrome; AN/FM, study grouprecruited in atypical nevus and familial melanoma research program; CDK,cyclin-dependent kinase.

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

961

phenotype is determined by other common genes, variants ofwhich have a low penetrance for melanoma. By bias ofascertainment, family studies may have recruited families withnevi and melanoma attributable to both CDKN2A codingmutations and putative low penetrance nevus/melanoma suscep-tibility alleles.

We report here a study in which we have taken a candidate geneapproach to the identi®cation of nevus genes. Ala148Thr is a G toA missense polymorphism in CDKN2A. It lies outside the criticalfour ankyrin repeat domains and does not appear to affect bindingto CDK4 (Ranada et al, 1995; Likischkis et al, 1996). Ala148Thrhas been excluded as a high penetrance CDKN2A mutation in thatit does not segregate with melanoma in families (Harland et al,1997) but its frequency has been observed to be higher inmelanoma kindreds than in the general population (Hussussian et al,1994; Harland et al, 1997). We chose Ala148Thr secondly becausetwin studies suggest that a melanoma/nevus susceptibility gene inaddition to mutant CDKN2A may lie at 9p21 (Zhu et al, 1999).Ala148Thr was studied ®rst in a sample enhanced for the presenceof the AMS and/or a family history of melanoma, and then in asample of normal adult women. The association betweenAla148Thr and melanoma was studied in the former group, andthe correlation between Ala148Thr and nevus phenotype wasstudied in both groups.

MATERIALS AND METHODS

Subject recruitment Subjects were recruited in two separate studies.The ®rst study group consisted of probands (de®ned below) and theiradult relatives recruited in an atypical nevus and familial melanomaresearch program (AN/FM group). Our group has developed a scoringsystem to de®ne the AMS in which a reproducible score from 0 to 5 isassigned for the presence of various phenotypic criteria (Newton et al,1993; Newton Bishop et al, 2000). Entry criteria to this study were thepresence of an abnormal nevus phenotype (de®ned as an AMS score of 2or more) or the presence of two or more cases of cutaneous melanomain ®rst or second degree relatives, or a case with multiple primaries.Probands (n = 179) were identi®ed by physician referral to the researchprogram. After recruitment he/she was asked to invite other familymembers to participate. 309 family members were recruited, of whom62% were ®rst degree and 21% second degree relatives. The remaining17% were more distant relatives. This sample was therefore heavilyenriched for atypical nevi and/or a personal or family history ofmelanoma (Table I). This opportunistic approach was adopted in viewof the relatively low prevalence of the AMS in the general population(Newton et al, 1993).

The second group comprised 599 adult Caucasian women recruitedfrom the general population in Yorkshire (Leeds and Harrogate) andHertfordshire (St. Albans) in a study investigating nevus prevalence andultraviolet radiation exposure (General Practice group). Ethical approvalwas obtained from ethics committees prior to data collection for bothstudies.

Consenting subjects were examined by a research nurse or dermatolo-gist. All examiners were trained by the same individual (JANB).Previously published interobserver variability data show AMS phenotyp-ing to be reliable between examiners (Newton Bishop et al, 2000) but allanalyses were carried out with adjustment for examiner.

Benign nevi 2 mm or greater in diameter and clinically atypical nevi(de®ned as variably pigmented nevi with an irregular or indistinct marginand diameter of at least 5 mm) were counted in 17 body sites excludingthe posterior scalp, genitalia, palms, soles, and female breasts. Blood wasrequested for DNA extraction. If a blood test was not obtained in theGeneral Practice group a mouthwash was requested (13% of the St.Albans samples and 5% of Yorkshire samples).

Polymerase chain reaction using sequence-speci®c primers (PCR-SSP) The Ala148Thr polymorphism was detected by PCR-SSP.Samples producing a mutant band for the Ala148Thr PCR-SSP testwere then subjected to a restriction digest test to distinguish betweenheterozygous and homozygous mutants. Further details are available fromthe authors on request.

Statistical analysis Statistical analysis was performed using thestatistical package STATA version 6 (Stata Corporation, College Station,TX). Data collected from the AN/FM study were analyzed separatelyfrom those collected from the General Practice group. The relationshipof the Ala148Thr polymorphism with melanoma (and other categoricalvariables) was investigated using c2 tests or Fisher's exact test wherenumbers in some categories were small. The relationship with nevuscount and other continuous variables was examined using t tests.Logarithms of the nevus counts were used in analysis to reduce thepositive skew of the raw data distribution. Untransformed nevus countsare reported in the text for ease of interpretation. Mann±Whitney U testswere used in subset analysis where t tests were not appropriate becauseof small sample size and skewed nevus distributions. As the 488individuals in the AN/FM study were not ascertained independently butconsisted of 179 families, analysis was repeated amongst probands only.Each group was analyzed ®rst in its entirety and then again only amongstsubjects with an AMS score of 2±5. Linear regression was used toinvestigate the effect of different risk factors on nevus count, and logisticregression was used to estimate odds ratios for the risk of melanoma. Inaddition ordered logit (proportional odds) regression was used to estimatethe relationship between the ordinal AMS score (0±5) and Ala148Thrand other risk factors (McCullagh, 1980). Again as the individuals in theAN/FM study were not ascertained independently, possible dependencein the observations due to familial clustering was also allowed for. In thelinear regression analysis this was done by use of a random effects model,which allows variation within and between families to be calculated. Inthe logistic and ordered logit analyses, robust standard errors of parameterestimates were calculated to allow for clustering within pedigrees.

RESULTS

The AN/FM group consisted of 210 (43%) males and 278 femalesaged between 16 and 91 (mean age 42.3 y). The women in theGeneral Practice group were aged 19±46 (mean age 36.5 y). Asexpected, they had signi®cantly lower total nevus counts (mean56.1, range 0±286, compared with a mean of 111, range 0±837, inthe AN/FM group, p < 0.0001). They also had signi®cantly feweratypical nevi (mean 0.3, range 0±7, compared with mean 3.4, range0±42, p < 0.0001). Regression analysis con®rmed that thesedifferences remained signi®cant after adjusting for age and inter-viewer. In the AN/FM group 250 (51.2%) people had atypical nevicompared with 94 (15.7%) in the General Practice group. Thedistribution of AMS scores also differed between the two groups (p< 0.0001).

AN/FM group Of the 488 individuals screened in this study,Ala148Thr G to A polymorphisms were found in 24, giving aprevalence of 4.9%. No Thr/Thr homozygotes were detected,which is in accordance with Hardy±Weinberg expectations(expected number 0.3). 145 individuals (30%) had had cutaneousmelanoma. Three of these individuals also had had an ocularmelanoma. Amongst those with melanoma, Ala148Thr G to Apolymorphisms were found in seven (4.8%). Genotype frequenciesin the melanoma subgroup were also in Hardy±Weinbergequilibrium. No association between Ala148Thr status and sex

Table I. Criteria met by probands in the AN/FM group

AMSScore

0 casesmelanomain family

1 casemelanomain familya

2 ormore casesmelanomain familya Total

0 0 2b 6 81 0 1b 7 82 4 10 13 273 19 31 28 784 19 24 10 535 2 3 0 5

44 71 64 179

aA melanoma in the family could include a melanoma in the proband or in a re-lative of the proband.

bMultiple melanoma primaries in the proband.

962 BERTRAM ET AL THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

was apparent (p = 0.78) nor between subjects studied by differentexaminers (p = 0.71).

Univariate analyses were carried out ®rst in the overall group,then again in probands alone (as these represent a sample ofindependent individuals, in order to eliminate the effect of familialclustering), and ®nally in all subjects with AMS scores of 2±5. Forsimplicity the results presented here are those of the overall groupunless stated otherwise. No association was found betweenAla148Thr and melanoma status (p = 0.95), nor was any associationfound with age at diagnosis amongst melanoma cases (mean age 43in cases with Ala148Thr and 42 in those without, p = 0.87).Multivariate logistic regression adjusting for age, sex, examiner, andrandom family effect con®rmed that Ala148Thr status was a poorpredictor of melanoma status (odds ratio 1.14 when baseline equalshomozygote wild-type, robust standard error 0.53, p = 0.78). Theresult was similar when no adjustment was made for age and sex.Although 44 out of 145 melanoma cases had multiple primaries,only one of these had a germline Ala148Thr polymorphism.Binomial probability con®rmed that this frequency was notsigni®cantly different from the frequency of 4.9% seen in theoverall population (p = 0.72). No association was found with familyhistory of melanoma amongst probands [odds ratio 0.31 (95%con®dence interval 0.06, 1.52) when baseline equals homozygotewild-type, p = 0.15].

No association was found between Ala148Thr and total numberof nevi in the overall group [mean count 112 if wild-type (Ala/Ala), 98 if variant (Ala/Thr), p = 0.97 from t test of logged values](Table II) or in subset analysis as described earlier. In the overallgroup multivariate linear regression adjusting for age, sex, mela-noma status, and examiner and treating family as a random effectshowed that total nevus count varied signi®cantly with age,melanoma status, and examiner. No signi®cant effect of Ala148Thrstatus was shown on the log of the total nevus count, however(Table III). These results were similar when age and sex wereexcluded from the model.

In the overall group the Ala148Thr polymorphism was associ-ated with the presence of fewer atypical nevi (mean values 3.5 ifwild-type compared with 1.0 if variant, p = 0.03 from t test oflogged values). This ®nding persisted in the subsets of those with anAMS score of 2±5 and the independent probands (Table II).

Multivariate linear regression in the whole group, adjusting for age,sex, melanoma status, and examiner and treating family as a randomeffect, con®rmed that the relationship between the log of theatypical nevus count and Ala148Thr status remained after adjust-ment for these factors (Table III). Again the results were similarwhen age and sex were excluded from the model.

No association was found with AMS score in the overall group(test for trend, p = 0.58) or in subset analysis as described earlier.Ordered multivariate logit regression adjusting for age, sex,melanoma status, examiner, and random family effect using theoverall group showed that although AMS score varied signi®cantlywith age, melanoma status, and examiner, it was poorly predictedby Ala148Thr status (Table III). Results were similar when ageand sex were excluded from the model.

General Practice group None of the 599 women had had amelanoma. Thirty-one (5.2%) were found to have genotype Ala/Thr. No Thr/Thr homozygotes were detected, which is inaccordance with Hardy±Weinberg expectations (expected numberis 0.4). A signi®cant difference in Ala148Thr frequency was seenbetween groups studied by different examiners (p = 0.03). Thisre¯ects a regional difference, as the frequency in St. Albans [2.2% (5out of 225)] was signi®cantly lower than that found in Yorkshirewomen [7.5% (26 out of 348)], whereas the difference between thetwo Yorkshire interviewers' samples was not signi®cant.

No association was seen between the Ala148Thr polymorphismand the total number of nevi (mean count 56 if wild-type, 68 ifvariant, p = 0.23 from t test using logged values) either in the wholegroup or when restricting AMS scores to 2±5 only (Table II).Multivariate linear regression adjusting for interviewer also failed toshow an effect of Ala148Thr status on the logarithm of the totalnevus count (p = 0.29, Table III). Atypical nevi were uncommon.In univariate analysis no association was seen between Ala148Thrstatus and atypical nevus count in the whole group (mean values0.32 if wild-type compared with 0.13 if variant, p = 0.31 from t testbased on logged values). When restricting the analysis to womenwith AMS scores of 2±5, however (Table II), those with thepolymorphism had signi®cantly fewer atypical moles than thosewithout (p = 0.02 based on Mann±Whitney test). Of 11 womenwith the rare variant and a high AMS score, none had any atypicalmoles. In contrast, in women with an AMS score of 2±5 and

Table II. Nevus counts according to Ala148Thr status in the two study groupsa

Wild-typeMean (standard deviation)

G to A polymorphismMean (standard deviation)

p-value(two-sided t test)

Mann±WhitneyU test

AN/FM groupAMS 0±5 (n = 488) (n = 464) (n = 24)

Log total nevus count 4.17 (1.23) 4.18 (1.13) 0.97 0.97Log atypical nevus count 0.89 (1.03) 0.42 (0.69) 0.03 0.03

AMS 2±5 (n = 295) (n = 279) (n = 16)Log total nevus count 4.91 (0.68) 4.77 (0.49) 0.41 0.25Log atypical nevus count 1.37 (1.03) 0.59 (0.78) 0.004 0.004

Probands AMS 0±5 (n = 179)b (n = 169) (n = 10)Log total nevus count 5.00 (0.84) 4.62 (0.84) 0.18 0.13Log atypical nevus count 1.44 (1.04) 0.62 (0.77) 0.02 0.02

Probands AMS 2±5 (n = 163)b (n = 155) (n = 8)Log total nevus count 5.17 (0.57) 4.93 (0.59) 0.26 0.30Log atypical nevus count 1.55 (1.02) 0.69 (0.83) 0.02 0.02

HEALTHY WOMEN groupAMS 0±5 (n = 598) (n = 567) (n = 31)

Log total nevus count 3.71 (0.89) 3.90 (0.94)c 0.23 0.14Log atypical nevus count 0.16 (0.41) 0.09 (0.24) 0.31 0.55

AMS 2±5 (n = 118) (n = 107) (n = 11)Log total nevus count 4.60 (0.63) 4.63 (0.51) 0.89 0.91Log atypical nevus count 0.46 (0.65) 0 (0) 0.02 0.02

aIn all cases 1 is added to the nevus count before taking logarithms to avoid zeros.bProbands only are considered to remove the effect of family clustering.cn = 30: total nevus count missing for one woman.

VOL. 119, NO. 4 OCTOBER 2002 ALA148THR AND NEVUS/MELANOMA SUSCEPTIBILITY 963

without the variant the mean number of atypical moles was 1.0 andonly 64% had no atypical moles. After adjusting for interviewer(and hence also for region) this difference was no longer statisticallysigni®cant (p = 0.22, Table III).

No association was found between Ala148Thr and AMS score(test for trend, p = 0.3). Ordered multivariate logit regressionadjusting for interviewer also con®rmed that Ala148Thr status wasa poor predictor of AMS score (p = 0.4, Table III). No adjustmentwas made for age in these analyses as age was not related to nevusphenotype in this group.

DISCUSSION

The two groups studied here were chosen in order to study theeffect of Ala148Thr within a population enriched for an abnormalnevus phenotype and/or family history of melanoma and secondlywithin a sample chosen from the general population.

Both the AN/FM and the General Practice populations studiedhere revealed prevalences of Ala148Thr (4.9% and 5.2%) similar tothose previously identi®ed in general population samples(Hussussian et al, 1994; Kamb et al, 1994). We found no evidencefor Ala148Thr as a melanoma susceptibility gene. As the Ala148Thrpolymorphism is relatively rare, a very large sample size would beneeded to rule out a small effect on melanoma risk but the AN/FMstudy would have over 80% power to detect a relative risk of 3associated with the polymorphism (signi®cance level 0.05).

These data do not support the hypothesis that Ala148Thr is anormal nevus susceptibility allele. For the AN/FM group adifference in log mole count of just over half a standard deviationbetween those with and without the polymorphism could bedetected with 80% power. Translating this into raw mole counts,the geometric mean in noncarriers was 67, and the abovecalculation roughly corresponds to detecting a doubling or halvingof the geometric mean mole count for noncarriers. In the largerGeneral Practice group a difference of just less than half a standarddeviation could be detected. Thus, as mole counts are highly

variable, this study cannot rule out a modest effect of thispolymorphism.

Our results do suggest a possible protective effect of Ala148Thron atypical nevi within a population with an abnormal nevusphenotype, however, and this ®nding persisted in subset analysis.The result was weakly re¯ected in the General Practice groupamong those with a high AMS score. Among 11 women with theAla148Thr polymorphism and an otherwise abnormal nevusphenotype, no atypical moles were seen, whereas in those withwild-type genotype one-third had one or more atypical moles. Inthis group, however, the difference was no longer statisticallysigni®cant after adjusting for interviewer. A joint analysis of 1054subjects from both study groups showed a protective effect (0.22fewer atypical moles in those with the polymorphism, 95%con®dence interval 0.04, 0.41; p = 0.02) after adjustment for age,sex, interviewer, melanoma, and study. When the analysis wascon®ned to the 393 subjects from either study with an AMS scoreof 2 or more, the difference was more marked [0.51 (0.15, 0.87)fewer moles, p = 0.005].

Although the protective effect observed was unexpected andmay be a false positive result, it is biologically plausible that apolymorphism could have a protective effect instead of a deleteri-ous effect on atypical nevus development. It is predicted thatpolymorphisms in genes will result in proteins that are functionallynormal or possibly mildly functionally impaired. Ala148Thr is inlinkage disequilibrium, however, with a promoter variant (±493 A/T) (Harland et al, 2000) and promoter variants couldconceivably result in elevated p16 levels in the cell. It is known thatp16 induces cell senescence (Uhrbom et al, 1997) and is stronglyexpressed in benign nevi (Keller-Melchior et al, 1998). Atypicalnevi are characterized by a failure of the melanocytes to senesce andit is therefore of interest that these data suggest a possible protectiveeffect on the development of atypical nevi.

The considerable difference in polymorphism frequency be-tween two centers 190 miles apart was surprising. Systematicgenotyping failure was excluded as a cause as all genotyping was

Table III. Multivariate regression analyses in the AN/FM group (n = 488) and General Practice group (n = 598)a

Parameters in the model Coef®cient (standard error) p-value

Log total nevus count (AN/FM group)Age 0.08 (0.01) < 0.001Age-squared ±0.001 (0.0001) < 0.001Sex ± females vs males ±0.06 (0.09) 0.52Melanoma ± present vs absent 0.65 (0.10) < 0.001Ala148Thr variant vs wild-type 0.05 (0.20) 0.81

Log atypical nevus count (AN/FM group)Age 0.03 (0.01) 0.03Age-squared ±0.0004 (0.0001) 0.002Sex ± females vs males ±0.20 (0.08) 0.01Melanoma ± present vs absent 0.59 (0.10) < 0.001Ala148Thr variant vs wild-type ±0.43 (0.19) 0.02

AMS score (0±5) (AN/FM group)Age 0.11 (0.03) < 0.001Age-squared ±0.001 (0.0003) < 0.001Sex ± females vs males ±0.17 (0.18) 0.35Melanoma ± present vs absent 1.16 (0.17) < 0.001Ala148Thr variant vs. wild type 0.31 (0.44) 0.48

Log total nevus count (General Practice group)Ala148Thr variant vs wild-type 0.18 (0.17) 0.29

Log atypical nevus count (General Practice group)Ala148Thr variant vs wild-type ±0.01 (0.07) 0.88

Log atypical nevus count (those with AMS score of 2±5 in General Practice group)Ala148Thr variant vs wild-type ±0.21 (0.17) 0.22AMS score (0±5) (General Practice group)

Ala148Thr variant vs wild-type 0.32 (0.36) 0.38

aEach subheading describes the outcome variable and the study group. All analyses are also adjusted for examiner and in the AN/FM group for family clustering as descri-bed in the text.

964 BERTRAM ET AL THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

carried out in Leeds. There was a higher rate of mouthwash samplesfrom St. Albans. We therefore re-analyzed the data excluding allmouthwash samples, but found that, although the failure rate wasnow the same in the two centers' samples, there was still asigni®cant difference in the prevalence of the A148T mutation(2.4% compared with 7.2%) suggesting that this difference was real.The differences in frequency of a common polymorphism betweentwo areas separated by a relatively small distance clearly hassigni®cance for design of association studies, with respect to theselection of controls.

We have found no evidence that Ala148Thr is a melanoma ornevus susceptibility allele.

An interesting protective effect of this variant on atypical neviwas suggested, however, which needs to be further explored. Amarked difference in prevalence in the polymorphism within theUK was demonstrated, which has implications for the design ofassociation studies in general.

This study was funded by the Cancer Research, UK and the NHS R&D

Executive. We are grateful to all clinicians who referred individuals to the AN/FM

study and to all those individuals who kindly participated. We are grateful to Dr. J.

Apps and colleagues of the Street Lane Practice, Leeds, Dr. M. Blanshard and

colleagues of the Parkbury House Surgery, St. Albans, and Dr. J. Bradshaw and

colleagues of Eastgate Surgery, Knaresborough, who assisted us greatly in asking

women in their practices to take part in the study. We also wish to thank the

following for their help with subject recruitment and data entry: M. Glover, K.

Grif®ths, R. Wachsmuth, M. Chan, J. Frazer, and M.Swanwick.

REFERENCES

Augustsson A, Stierner U, Rosdahl I, Suurkula M: Common and dysplastic naevi asrisk factors for cutaneous malignant melanoma in a Swedish population. ActaDerm Venereol 71:518±524, 1990

Bataille V, Newton Bishop JA, Sasieni P, Swerdlow AJ, Pinney E, Grif®ths K,Cuzick J: Risk of cutaneous melanoma in relation to the numbers, types andsites of naevi: a case control study. Br J Cancer 73:1605±1611, 1996

Bergman W, Palan A, Went LN: Clinical and genetic studies in six Dutch kindredswith the dysplastic naevus syndrome. Ann Hum Genet 50:249±258, 1986

Bergman W, Gruis NA, Frants RR: The Dutch FAMMM family material. Clinicaland genetic data. Cytogenet Cell Genet 59:161±164, 1992

Easton D, Cox G, Macdonald A, Ponder B: Genetic susceptibility to naevi ± a twinstudy. Br J Cancer 64:1164±1167, 1991

Fitzgerald MG, Harkin DP, Silva-Arrieta S, et al: Prevalence of germ-line mutations

in p16, p19ARF and CDK4 in familial melanoma: analysis of a clinic basedpopulation. Proc Natl Acad Sci 93:8541±8545, 1996

Goldstein AM, Tucker MA: Screening for CDKN2A mutations in hereditarymelanoma. J Natl Cancer Inst 89:676±678, 1997

Harland MH, Meloni R, Gruis N, et al: Germline mutations of the CDKN2 gene inUK melanoma families. Hum Molec Genet 6:2061±2067, 1997

Harland M, Holland EA, Ghiorzo P, et al: Mutation screening of the CDKN2Apromoter in melanoma families. Genes Chrom Cancer 28:45±57, 2000

Holland E, Beaton S, Becker T, et al: Analysis of the p16 gene, CDKN2, in 17Australian melanoma kindreds. Oncogene 11:2289±2294, 1995

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

Kamb A, Shattuck-Eidens D, Eeles R, et al: Analysis of the p16 gene (CDKN2) as acandidate for the chromosome 9p melanoma susceptibility locus. Nature Genet8:22±26, 1994

Keller-Melchior R, Schmidt R, Piepkorn M: Expression of the tumor suppressorgene product p16INK4 in benign and malignant melanocytic lesions. J InvestDermatol 110:932±938, 1998

Likischkis R, Sarcevic B, Kennedy C, Warlters A, Sutherland RL: Cancer-associatedmissense and deletion mutations impair p16INK4 CDK inhibitory activity. IntJ Cancer 66:249±254, 1996

McCullagh P: Regression models for ordinal data. J R Stat Soc B 42:109±142, 1980Newton JA, Bataille V, Grif®ths K, et al: How common is the atypical mole

syndrome phenotype in apparently sporadic melanoma? J Am Acad Dermatol29:989±996, 1993

Newton Bishop J, Bataille V, Pinney E, Bishop DT: Family studies in melanoma:identi®cation of the atypical mole syndrome (AMS) phenotype. Melanoma Res4:199±206, 1994

Newton Bishop JA, Bradburn M, Bergman W, et al: Teaching non-specialist healthcare professionals how to identify the atypical mole syndrome phenotype: amulti-national study. Br J Dermatol 142:331±337, 2000

Ranada K, Hussussian CJ, Sikorski RS, et al: Mutations associated with familialmelanoma impair p16 INK4 function. Nature Genet 10:114±116, 1995

Serrano M, Hannon GJ, Beach D: A new regulatory motif in cell-cycle controlcausing speci®c inhibition of cyclin D/CDK4. Nature 366:704±707, 1993

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

Swerdlow AJ, English J, MacKie RM, O'Doherty CJ, Hunter JAA, Clark J, Hole D:Benign melanocytic naevi as a risk factor for malignant melanoma. Br Med J292:1555±1560, 1986

Uhrbom L, Nister M, Westermark B: Induction of senescence in human malignantglioma cells by p16INK4A. Oncogene 15:505±514, 1997

Wachsmuth RC, Gaut RM, Barrett JH, et al: Heritability and gene±environmentinteractions for melanocytic nevus density examined in a U.K. adolescent twinstudy. J Invest Dermatol 117:348±352, 2001

Walker GJ, Hussussian CJ, Flores JF, et al: Mutations of the CDKN2/p16INK4 genein Australian melanoma kindreds. Hum Molec Genet 4:1845±1852, 1995

Zhu G, Duffy DL, Eldridge A, et al: A major quantitative-trait locus for mole densityis linked to the familial melanoma gene CDKN2A: a maximum-likelihoodcombined linkage and association analysis in twins and their sibs. Am J HumGenet 65:483±492, 1999

VOL. 119, NO. 4 OCTOBER 2002 ALA148THR AND NEVUS/MELANOMA SUSCEPTIBILITY 965