Cutaneous malignant melanoma (CMM) is a potentially fatal form of skin cancer whose etiology is heterogeneous and complex. Major host and environmental risk factors for melanoma include many melanocytic nevi, especially dysplastic nevi, fair pigmentation (skin, eye, and hair), freckling, poor tanning ability, and a tendency to burn after sun exposure (1, 2). Genetic risk factors include germ line mutations in theCDKN2A, CDK4, orMC1R genes.CDKN2A andCDK4 have been designated “high-risk” melanoma susceptibility genes. TheCDKN2A gene is the major known melanoma susceptibility gene. Germ line mutations have been detected in ∼20% of melanoma-prone families with three or more melanoma patients. In contrast, few families with germ line mutations inCDK4 have been identified.MC1R has also been shown to influence melanoma risk, but it is described as a “low risk” melanoma susceptibility gene (3, 4).

MC1R is involved in pigmentation primarily through its binding with α-melanocyte-stimulating hormone (5).MC1R is very polymorphic, with >65 nonsynonymous alleles identified to date (6, 7). Three variants (R151C, R160W, andD294H) designated red hair color orRHC variants have been repeatedly shown to be associated with red hair color, poor tanning ability, pale/fair skin color, and extensive freckling (8-10). Most other variants (designatednon-RHC orNRHC) have a weaker or no association with red hair (10, 11). Several studies conducted in generally fair-skinned populations of Northern European origin have shown that risk of melanoma is higher amongMC1R variant carriers than among noncarriers, with the strongest effects observed for carriers of multiple variants (9, 12, 13).

MC1R has also been shown to be a risk factor for melanoma in families segregatingCDKN2A mutations. A study of 15 AustralianCDKN2A mutation–carrying families with nine different mutations (14) and a study of 101p16-Leiden–mutation carriers from six Dutch families (15) both showed that the presence ofMC1R variants increased the frequency/penetrance of melanoma amongCDKN2A mutation carriers. TheMC1R-melanoma association was primarily related to theR151C variant in the Dutch families and to the threeRHC variants in the Australian families. There was also an inconsistent reduction in age at melanoma diagnosis associated with the presence of at least oneMC1R variant; this age reduction was observed in the Australian study sample but not in the Dutch families. Further studies are needed to confirm and refine the findings from the Australian and DutchCDKN2A families. The objective of the current study was to evaluate the relationship betweenMC1R and melanoma risk in 16CDKN2A melanoma-prone American families with extensive clinical and epidemiologic risk factor data.

Families were recruited for this non–population-based family study if there was a history of invasive melanoma in at least two first-degree relatives. The families were referred by health-care professionals or through self-referrals. Written informed consent was obtained prior to participation under an Institutional Review Board–approved protocol. All family members willing to participate in the study were clinically evaluated. Variables recorded during the clinical examination included the type and total number of nevi, extent of freckling, skin complexion, evidence for solar injury, and hair and eye color. In addition, a self-administered questionnaire obtained information on sun-related variables such as the skin's reaction to acute and chronic sun exposure (i.e., tanning ability). The subjects for this study were drawn from 16 families in which aCDKN2A mutation had been previously identified (16). The families had the following 12 mutations:1_8dup8, L16R, M53I, R58X, N71S, R87P, S56fs (c.167_197del31), c.240_253del14, P75fs (c.225_243del19), G101W (n = 3),V126D (n = 3), andc.IVS2+1 G>T. The families have been followed prospectively from 4 to 26 years starting in the 1970s. All melanoma diagnoses were confirmed by review of histologic materials, pathology reports, medical records, or death certificates. Total numbers of primary melanomas were recorded for each melanoma patient.

MC1R genotyping was conducted at the National Cancer Institute, Frederick, MD employing PCR amplification of the 951 bp coding region ofMC1R, either in its entirety or in smaller overlapping segments, followed by complete direct sequencing of the amplicon(s). The coding region ofMC1R was amplified from genomic DNA extracted from patient blood samples using two sets of M13-tagged PCR primers MC1R_1F: 5′-GTA AAA CGA CGG CCA GTG AAG ACT TCT GGG CTC CCT C-3′; MC1R_IIIR: 5′-GGA AAC AGC TAT GAC CAT GGC GTG CTG AAG ACG ACA CT-3′; and MC1R_IVF: 5′-GTA AAA CGA CGG CCA GTG TGC TGT ACG TCC ACA TGC T-3′; MC1R_IVR: 5′-GGA AAC AGC TAT GAC CAT GCT CTG CCC AGC ACA CTT AAA-3′. The underlined region of the primer is specific to the target DNA. The reaction mix for PCR amplification included 1× PCR buffer (Invitrogen high-fidelity PCR buffer), 1.5 mmol/L MgSO₄, 175 nmol/L each pair of primers, 50 nmol/L each of the four deoxynucleotide triphosphates, and 1 unit of HiFi Platinum Taq polymerase (Invitrogen, Carlsbad, CA). All PCR products were processed prior to sequencing. All products from two regions of PCR were sequenced with ABI prism BigDye terminator cycle sequencing kit 1.0 (Applied Biosystems, Inc.) on ABI3700 sequence analyzer using sequence primers 1F: 5′-GCT CCC TCA ACT CCA CC-3′; IR: 5′-GAA GAC GAC ACT GGC CAC-3′ and M13F: 5′-GTA AAA CGA CGG CCA GT-3′; M13R: 5′-GGA AAC AGC TAT GAC CAT G-3′, respectively. All sequences were analyzed and variants were detected using Mutation Surveyor (SoftGenetics Inc., PA) and sequence analysis software package developed at the Laboratory of Molecular Technology, National Cancer Institute.

Initially, we evaluated eachMC1R variant individually comparing 1+ variant to the consensusMC1R sequence (i.e., wild-typeMC1R). Because manyMC1R variants were too rare to examine their individual associations with melanoma risk after adjustment for major melanoma risk factors (i.e.,CDKN2A status, nevus/pigmentation factors—see below), we also used the followingMC1R variables in the analyses: carriers of anyMC1R variant compared with wild-typeMC1R; carriers of multiple (1, 2+) variants compared with the consensus sequence; carriers of 1NRHC variant, 2+NRHC variants, 1RHC variant, 2+RHC variants, or carriers of bothRHC andNRHC variants compared with wild-typeMC1R.

For purposes of this study, the measure of association between melanoma risk and the clinical, genetic, and environmental variables was the odds ratio (OR). Point estimates and 95% confidence intervals (CI) of adjusted ORs were calculated using logistic regression analysis as implemented in the EPICURE package (17).

We assessed the association of pigmentation and nevus characteristics with all nonsynonymousMC1R variants combined using χ² and Fisher exact tests in the unaffected relative and spouse controls separately (Stata 8.2; ref.18). Dysplastic nevi, hair color, eye color, skin complexion, freckling, solar injury, and tanning ability were all associated withMC1R. We also evaluated the ORs between these same factors and melanoma risk. Because of the relatively small number of cases, we created summary factors that combined the covariates showing the strongest associations with bothMC1R variants and melanoma risk. A three-category nevus factor was created by combining dysplastic nevi (absent, indeterminate, present) and total numbers of nevi. Similarly, a three-category pigmentation factor was developed by combining skin complexion (medium/dark, pale/fair) and extent of freckling (none/few, moderate, many).

We conducted two logistic regression analyses (17). The first analysis conditioned on family membership using the entire data set (72 melanoma cases, 245 unaffected relative, and 78 spouse controls). We also conducted an unconditional logistic regression analysis on the subset of confirmedCDKN2A mutation carriers (69 cases and 72 unaffected relative controls). All analyses were adjusted for age as a continuous variable. Sex had no effect on risk of melanoma and therefore was excluded from all analyses (data not shown). For the conditional logistic regression analysis, three models were examined: univariate (with age adjustment); adjustment forCDKN2A status and age; and adjustment for age,CDKN2A, and the pigmentation/nevus factors. For the unconditional analysis ofCDKN2A mutation carriers, two models were evaluated: univariate (adjusted for age) and adjustment for age and pigmentation/nevus factors.

We examined the distribution ofMC1R variants in multiple primary melanoma (MPM) compared with single primary melanoma (SPM) patients using Fisher exact test as implemented in StatXact 4 (19). We also estimated the median ages at diagnosis of initial melanomas in all melanoma patients and in MPM and SPM patients separately. The nonparametric Jonckheere-Terpstra test was used to investigate the hypothesis of no differences among the ages at diagnosis of melanoma according to numbers or numbers/types ofMC1R variants against the alternative that the ages at diagnosis decreased as the numbers or numbers/types ofMC1R variants increased.

Ten nonsynonymous and five synonymous (i.e., silent)MC1R variants were detected in the 395 subjects sequenced forMC1R.Table 1 shows the number of cases, unaffected relative controls, and spouse controls with each of the nonsilent variants observed. The five silent variants found (R34R, A166A, A240A, I264I, andT314T) were excluded from all analyses. The most frequent variants observed wereV60L, R160W, andR151C. As has been previously observed in other studies, there was a strong association between theRHC variantsR151C, R160W, andD294H and red hair color. Seventy-seven percent of the subjects (20 of 26) with red hair had at least twoRHC variants. Also, only 29% of the subjects (8 of 28) without red hair had twoRHC variants. These eight subjects had primarily brown or light brown hair. In contrast, few subjects withV60L, V92M, I155T, orR163Q had red hair color.

Table 1 presents ORs between melanoma risk and individualMC1R variants. After conditioning on family membership and adjusting for age, there were significant associations between melanoma and the presence ofR151C orR160W. Unconditional analyses restricted toCDKN2A mutation carriers showed significant associations between melanoma risk and allMC1R variants evaluated except for86insA andI155T, but with wide confidence intervals.

Table 2 shows the associations between melanoma risk and selectedMC1R covariates.Table 2A presents the ORs and 95% CIs for all three conditional analysis models evaluated. For the three analyses, the presence of at least twoMC1R variants was significantly associated with melanoma [OR, 5.6 (2.1-14.7); OR, 20 (5-80); and OR, 6.1 (1.2-29.7), respectively]. AnyMC1R variant, the number of variants, and types of variants also showed significant but imprecise associations with melanoma when we adjusted for age only or age andCDKN2A status.R151C andR160W also showed significant associations with CMM after adjustment for both age andCDKN2A status [OR, 11.3 (1.4-93.3) and OR, 9.1 (1.6-52.4), respectively].Table 2B shows the number of cases and unaffected relative controls who wereCDKN2A mutation carriers and results from the unconditional subset analysis ofCDKN2A mutation carriers. There were significant associations between melanoma risk and all three summaryMC1R variables examined after adjustment for age only. In addition, after adjustment for age and the pigmentation/nevus factors, there were significant associations between melanoma and multipleMC1R variants [OR, 7.3 (1.6-33.2)] as well as suggestive associations with the presence of at least twoNRHC variants [OR, 7.1 (1.0-49.4)] or the presence of bothRHC andNRHC variants [OR, 5.7 (1.0-32.2)]. These analyses were, however, based on relatively small numbers that resulted in wide confidence intervals. It was not possible to fully evaluate the number ofRHC variants. Specifically, seven cases and no controls had twoRHC variants.

Table 3 shows the number ofMC1R variants in MPM and SPM patients withCDKN2A mutations. There were three SPM patients who were notCDKN2A mutation carriers; these patients were excluded from the MPM-SPM evaluations. There were significant differences in the distribution ofMC1R variants between MPM and SPM patients. All 40 MPM patients had at least oneMC1R variant; 65% of MPM patients versus only 17% of SPM patients had at least twoMC1R variants (P < 0.0001). MultipleNRHC variants and presence of bothRHC andNRHC variants were more frequent in MPM versus SPM patients. Variation in other major melanoma risk factors including freckling, hair color, eye color, tanning ability, total nevi, or dysplastic nevi did not explain the differences inMC1R covariates in MPM versus SPM patients (data not shown).

Table 4 shows the median age at first melanoma diagnosis for SPM, MPM, and allCDKN2A mutation–carrying CMM patients according to the number ofMC1R variants or numbers/types ofMC1R variants. For all 69 patients combined, there was a statistically significant decrease in median age at diagnosis as the number ofMC1R variants increased (P = 0.010) even consideringRHC andNRHC variants separately (P = 0.003). This reduction in median age at CMM diagnosis in allCDKN2A mutation–carrying melanoma patients was primarily because of a significant decrease in age at diagnosis in MPM patients. No significant reduction in age at melanoma diagnosis was observed for SPM patients (Table 4).

We examined the association betweenMC1R variants and melanoma risk in 16 melanoma-prone American families withCDKN2A mutations. Similar to what has been observed in otherCDKN2A mutation–carrying melanoma-prone families (14, 15, 20), we observed a significant association between increased numbers ofMC1R variants and melanoma risk even after adjustment for major melanoma risk factors. In addition, comparison of MPM and SPM patients revealed striking differences in the distributions ofMC1R variants in these two groups of patients. There were also significant differences in median ages at melanoma diagnosis according to numbers and/or types ofMC1R variants in allCDKN2A mutation–carrying melanoma patients and MPM patients.

To the best of our knowledge, this is the first study ofMC1R variants inCDKN2A mutation carriers that examined the relationship between MPM and SPM patients from the same study sample. The MPM findings observed here are further supported by a small Italian study of 14 MPM patients without a positive family history for melanoma; Peris et al. (21) detectedMC1R variants in 11 of 12 patients with nonfamilial MPM, a much higher frequency relative to that previously reported in other populations (22). Two of the patients withMC1R variants also hadCDKN2A mutations as well as red hair color. The authors suggested that the results might represent an example of the effects of gene-gene interaction on disease risk (21). The current study with thrice the number of MPM patients plus 29 SPM patients, all withCDKN2A mutations, suggests that the presence of multipleMC1R variants is associated with the development of multiple melanoma tumors in patients withCDKN2A mutations. Although the small sample size precludes full evaluation of this association, the dampening of the complex host risk with sun-related factors (i.e., freckling/multiple nevi/dysplastic nevi) hints at the possible importance of sun exposure. Additional studies are needed to confirm these findings and to explore the mechanisms that may contribute to this relationship.

The Australian and Dutch studies ofMC1R variants in melanoma-prone families withCDKN2A mutations showed inconsistent differences in age at melanoma diagnosis. In the Australian study, mean age at melanoma diagnosis decreased significantly from 58.1 to 37.8 years with the presence of one or moreMC1R variants (14). In contrast, the Dutch study showed no such reduction in age at diagnosis; in fact, the mean age at melanoma diagnosis was 40 years in melanoma patients with noMC1R variants and 42 to 45 years in patients with two or moreMC1R variants (15). The current study revealed a significant decrease in median age at melanoma diagnosis as the overall number ofMC1R variants increased and when looking at the number ofRHC andNRHC variants separately. However, this association resulted from melanoma patients with >1 melanoma tumor (i.e., MPM patients). That is, among the 29 patients with only one melanoma tumor, there was no significant association betweenMC1R variants and age at melanoma diagnosis. It is possible that differences in the number of MPM versus SPM patients between the Dutch and Australian studies may have contributed to the inconsistent results observed in these two studies. Alternatively (or in addition), differences in the types or distribution ofCDKN2A mutations across the two studies—nineCDKN2A mutations in the Australian study versus one founder mutation (p16-Leiden) in the Dutch study—might have influenced the ages at melanoma diagnosis and/or the development of MPM tumors. Finally, distribution of major melanoma risk factors including relative amounts of sun exposure and the skin's reaction to sun exposure may have differed between the two studies. Further studies are needed to evaluate the age association betweenMC1R and numbers of melanoma tumors (and sun exposure).

The current study was limited by the small number of confirmed mutation carriers. The small size precluded adjustment for family membership in theCDKN2A mutation carrier subset analysis. In addition, it was not possible to examine individualCDKN2A mutations orCDKN2A mutations classified according to their type, location, or effect on the p14ARF protein. Also, it was difficult to evaluateMC1R variants separately. In addition, even though significant associations between melanoma risk and multipleMC1R variants were observed after adjustment for major melanoma risk factors, the odds ratio estimates were imprecise with wide confidence intervals. Finally, although all family members were invited to participate in the study, differential inclusion of mutation carriers, deceased melanoma cases or relatives with certain exposures could influence the results. It is difficult, however, to predict whether the odds ratios would be decreased or increased by this potential participation bias. In conclusion, this study of 16 melanoma-prone American families withCDKN2A mutations adds to the growing literature of studies demonstrating a relationship between multipleMC1R variants and melanoma risk. The study also provides new directions for research to further explore the differences in the distribution ofMC1R variants and ages at melanoma diagnosis observed in MPM versus SPM patients. Studies with much larger sample sizes and extensive epidemiologic, clinical, and genetic risk factor data will be required to investigate these relationships further.

We are indebted to the participating families, whose generosity and cooperation have made this study possible. We also acknowledge the contributions to this work that were made by Virginia Pichler, Laura Fontaine, Mary Wells, and Deborah Zametkin. We thank Benjamin Hulley for analytic support and Rashida Williams for help with manuscript preparation. This research was supported in part by the Intramural Research program of the NIH, NCI, DCEG.