Number of Nevi and Early-Life Ambient UV Exposure Are Associated with BRAF-Mutant Melanoma

Melanoma risk factors include increased number of melanocytic nevi, solar UV exposure, including high levels in childhood, and fair phenotypic characteristics (1-4). An increased number of nevi is a very strong risk factor for the occurrence of melanoma (1). Ambient UV irradiance in early life (birth and 10 years of age) has been found by Kricker et al. (5) to be one of the strongest sun-related risk factors for multiple primary melanomas, and these results should theoretically apply to risk of all melanomas (6). Phenotypic factors identified as associated with increased melanoma risk include sun sensitivity, freckles, blond or red hair, and light-colored eyes (3). However, the epidemiology of melanoma is complex (7), a variety of histologic subtypes exist, and the presence of different genetic alterations (8) indicates that melanoma is a heterogeneous disease. A better understanding of melanoma heterogeneity and its relationship to risk factors should assist efforts toward improving prevention.

BRAF mutations have been reported to occur in ∼50% of melanomas, predominately in or around codon 600 (accession number NM_004333.2) in exon 15 with the remainder in exon 11 (9). NRAS mutations have been reported to occur, exclusive of BRAF mutations, in ∼15% of melanomas, mainly in codons 12/13 and 61 in exons 2 and 3 (NM_002524; ref. 10). Most of these mutations have been shown to activate the RAS–RAF–mitogen-activated protein kinase/extracellular signal-regulated kinase kinase–extracellular signal-regulated kinase cell signaling pathway and have been implicated in cancer initiation and progression (11, 12).

The objective of this study was to investigate the relationships of BRAF and NRAS somatic mutations with phenotypic susceptibility and ambient solar UV exposure, including childhood exposure. These relationships were examined for 214 population-based first invasive primary cutaneous melanoma cases from North Carolina in the year 2000 who were interviewed with regard to their risk factors. Erythemal UV irradiance, derived from residential history and satellite-based measures, was used to estimate ambient solar UV exposure. Tumors were screened for BRAF and NRAS mutations using single-strand conformation polymorphism analysis and DNA sequencing.

The participants in this study were from one center (North Carolina) in the Genes, Environment, and Melanoma Study (5, 13, 14). Eligible subjects, identified by the population-based North Carolina Cancer Registry, were diagnosed with a first incident cutaneous invasive melanoma between January 1 and December 31, 2000, and resided within 42 counties of North Carolina. All of the dermatologists and hospitals in the 42 counties were responsible for reporting new cases. North Carolina dermatologists were notified of the study and reporting procedures and asked where their histopathology specimens were being sent. The study protocol was approved by the Institutional Review Board of The University of North Carolina at Chapel Hill. Physician approval was sought before contacting ascertained cases, and each participant or their guardian provided informed consent. Of the 582 eligible patients at this study site, 285 (49%) participated. We conducted analyses comparing participants and nonparticipants using information collected from the Central Cancer Registry. There were no significant differences between these groups based on age, gender, histologic type, tumor site, or Breslow thickness.

Participants completed a self-administered questionnaire and 1-h telephone interview regarding demographic characteristics, medical history, and phenotypic factors. Using a glossy-colored guide to aid in differentiating between nevi and other skin lesions, subjects were asked to have the nevi on their backs counted by a family member or friend, and, for the purposes of this article, back nevi counts were categorized based on tertiles in the wild-type group. The presence of nevi was also assessed, as in Marrett et al. (15), by asking the subject to select which of four whole-body diagrams corresponded most closely to his or her nevus density, and the two diagrams representing the most nevi were combined to form the highest category.

Erythemal UV irradiances were calculated as wavelength-integrated spectral irradiance between 250 and 400 nm, weighted by the Commission Internationale de l'Eclairage erythemal sensitivity function (16), which describes the stronger skin-reddening capacity of light at shorter wavelengths. The tropospheric UV-visible model (17) was used to calculate the irradiances as a function of solar zenith angle, ozone column, and surface elevation, after the method of Madronich (18). The model used a discrete ordinates method (19) and a pseudospherical correction (20). Corrections for variations in the Earth-Sun distance and for cloud cover (21) were applied. Ozone column and cloud reflectivity data were obtained from the satellite-borne Total Ozone Mapping Spectrometer (22-24), for November 1978 to June 2000.

For each participant, the location of their home at birth and each decade up to age at diagnosis was recorded. Location-specific erythemal UV dose values from the 1978 to 1989 climatology were applied to all participant exposure dates before 1990, and values from the 1990 to 2000 climatology were applied to exposure dates from 1990 onward. These decadal values were integrated to give lifetime total potential exposure. As in Kricker et al. (5), early-life UV irradiance was defined as the average of UV irradiance at birth and age 10 years, and average annual lifetime UV irradiance was calculated as the lifetime total divided by age.

Participants provided informed consent to obtain diagnostic slides and melanoma recuts. Tumors were reviewed by one dermatopathologist (K.B.). Of the 285 cases, 16 were excluded after histopathologic review based on ineligible diagnoses (12 in situ, 1 mucosal and 1 metastatic melanoma, and 2 dysplastic nevi). Eligible cases (N = 269) were classified according to criteria proposed by Clark et al. (25) and McGovern et al. (26). Although in many cases, only biopsies of the melanoma were available for review, the biopsies typically contained the majority of the tumor, except for cases of lentigo maligna melanoma (LMM), but the findings were representative. The diagnosis of LMM was based on the presence of a predominant lentiginous growth pattern of skin with some evidence of sun damage (mild to moderate or severe solar elastosis). Chronic sun damage (CSD) was scored using the 0 to 3+ multipoint scale and dichotomized into non-CSD (level 0 to 2−) and CSD (level 2 to 3+) melanomas, as in Landi et al. (27).

Of the 269 participants eligible after slide review, we obtained tissue recuts for 245 of 269 (91%). Of tissues collected, 223 of 245 (91%) had sufficient tissue on the recuts to proceed with analysis, and, of these, 214 of 223 (96%) had successful PCR amplification and sufficient high-quality DNA for complete mutational analysis of BRAF exons 11 and 15 and NRAS exons 2 and 3. Using information collected from study participants, we compared participants whose tumors were successfully analyzed for BRAS and NRAS mutations versus participants whose tumors were not analyzed. There were no significant differences based on age, gender, histologic type, site, Breslow thickness, number of back nevi, nevus density, propensity to tan, childhood freckles, hair color, eye color, estimated lifetime average ambient UV, or estimated early-life ambient UV.

When indicated, because of small tumor size or admixture of nonmalignant cells, tumor cells were selectively procured using laser capture microdissection under the supervision of a dermatopathologist (P.G.). Tumor DNA was prepared as previously described (28) and analyzed for mutations in BRAF exons 11 and 15 (including codons 466 and 600) and in NRAS exons 2 and 3 (including codons 12, 13 and 61), using single-strand conformational polymorphism (SSCP) analysis and radiolabeled sequencing of SSCP-positive samples and 10% of negatives. Radiolabeled, rather than automated fluorescent, sequencing of PCR products was carried out because it was more sensitive in tumor samples with mixed populations of mutant and normal cells (29, 30). All mutations were independently confirmed, and mutational artifacts were eliminated by sequencing a separately amplified aliquot of DNA. Furthermore, we directly compared SSCP and sequencing in the BRAF and NRAS fragments for a series of 40 specimens. In no case did SSCP fail to detect a mutation that was observable by sequencing. In addition, in the 10% of SSCP-negative samples that we sequenced, we did not observe a failure of SSCP to detect mutations. Experimental details are in Supplementary Information.

The goal of the study was to compare the etiologic profiles of tumors characterized by distinct molecular pathologic features (i.e., the presence of BRAF or NRAS mutations versus the absence of these mutations) using a case series design (31). In such a design, significant differences in the distribution of risk factors between pathologic subgroups provide evidence of etiologic heterogeneity. Because BRAF and NRAS mutations were exclusive of each other in all samples, cases were analyzed according to three subclassifications, those containing either a BRAF mutation (BRAF+), an NRAS mutation (NRAS+), or neither mutation (wild-type).

The cutoff points for UV exposures were chosen based on the distribution in our sample and previous findings (5) that early-life UV, which was the strongest sun-related risk factor distinguishing cases and controls in the entire Genes, Environment, and Melanoma study, showed an elevated odds ratio (OR) for risk of melanoma. For both lifetime and early-life UV, low UV was defined as the lowest tertile and high UV was defined as the combined upper two tertiles of exposure for the wild-type group.

The ANOVA test was used for comparison of mean ages across mutational subtypes. ORs and accompanying 95% confidence intervals (95% CI) were calculated in logistic regression models in SAS (SAS Institute, 1989) with adjustment for age (continuous). Trend tests were accomplished by modeling each variable as a single quantitative covariate. All significance tests were two-sided and a P value of 0.05 was considered statistically significant. In addition, models were fit to estimate the independent effects of age at diagnosis, early-life UV exposure, and number of back nevi in their association with BRAF and NRAS mutations. ORs for BRAF and NRAS were estimated using separate logistic regression models. Such an approach is simpler and leads to equivalent results compared with polytomous logistic regression modeling (32). The BRAF and NRAS models were reanalyzed excluding LMM and acral lentiginous melanoma (ALM).

Study cases had a mean age at diagnosis of 52.2 years and were 54.7% male. The study included a high percentage (56.1%) of thin (Breslow thickness <0.75 mm) melanomas. The percentages of histologic subtypes were 79% superficial spreading melanoma, 4.7% nodular melanoma, 10.3% LMM, 0.9% ALM, 1.4% nevoid or spitzoid melanoma, and 3.7% unclassifiable melanoma. The composition of our cases is consistent with other recent population-based studies, reflecting a downward trend for tumor thickness and a trend toward more superficial spreading melanomas over time (33, 34).

BRAF and NRAS mutations were found in 92 of 214 (43.0%) and 29 of 214 (13.6%) melanomas, respectively (Table 1), and were exclusive of each other, with a low probability of this distribution occurring by chance (P < 0.001). Of the BRAF mutations, 88 of 92 (95.7%) were in exon 15 in and around codon 600, whereas the other 4 of 92 (4.3%) were in exon 11. Of these BRAF mutations, 98.6% have been associated with in vitro enhancement of downstream extracellular signal-regulated kinase activity (9, 35, 36). The majority of BRAF mutations found in the invasive melanomas analyzed were located at codon V600. Of the V600 mutations, 26.5% (22 of 83) are double base pair (tandem) substitutions and one is a complex deletion-insertion (VKS600-602DT). The V600 mutations are not considered classic UV signature mutations. The other BRAF mutations listed in Table 1, which were found at very low frequencies, are opposite pairs of pyrimidine dimers. The NRAS mutations, which were all at codon 61, are known to activate several downstream effectors of RAS, including RAF, phosphatidylinositol 3-kinase, and RalGDS (37). A few synonymous BRAF and NRAS mutations were found (see Supplementary Information) and were considered negative in the analyses.

Clinicopathologic characteristics were categorized by mutational subtype, and BRAF+ and NRAS+ cases were analyzed in reference to wild-type cases (Table 2). The mean ages at diagnosis were 47.3 years for BRAF+, 62.1 years for NRAS+, and 53.9 years for wild-type cases (P < 0.001). As age was clearly associated with mutational status, all further analyses were age adjusted. No significant association was found for gender. Although LMM was less likely to contain BRAF mutations before age adjustment (data not shown), there was no statistically significant association of histologic subtype with mutational status after age adjustment. Breslow thickness ≥0.75 mm was much more common in BRAF+ cases (54%) and NRAS+ cases (59%) than in wild-type cases (29%).

As shown in Table 3, tumors from patients with increasing number of back nevi were progressively more likely to harbor a BRAF mutation (P = 0.006). Similarly, using nevus density diagrams, patients who reported higher nevus densities had a progressively increased odds of having a BRAF+ melanoma (P = 0.009). Patients were more likely to have an NRAS+ melanoma with increasing number of nevi, when assessed by back nevus counts and by nevus density diagrams, although the trends were not statistically significant.

The ability to develop a tan was modestly associated with both BRAF+ and NRAS+ tumors, but the trends were not statistically significant. Childhood freckling and hair and eye color were not significantly associated with mutational subtype.

As shown in Table 4, melanomas on chronically exposed (head, neck, or extremities) sites were less likely to harbor BRAF mutations (OR, 0.5; 95% CI, 0.3-1.0) or NRAS mutations (OR, 0.4; 95% CI, 0.2-0.9) than those on intermittently exposed (trunk) sites. Chronically exposed sites, when redefined as head, neck, lower arms, and lower legs, remained less likely to harbor a BRAF mutation (OR, 0.5; 95% CI, 0.3-0.9) or NRAS mutation (OR, 0.5; 95% CI, 0.2-1.3). Evidence of CSD, as assessed by histologic solar elastosis, was less evident for tumors associated with BRAF mutation before age adjustment (data not shown); however, this comparison was not significant after age adjustment.

BRAF+ cases were characterized by high estimated lifetime ambient UV irradiance (OR, 2.0; 95% CI, 1.0-4.0) and high estimated early-life ambient UV irradiance (OR, 2.6; 95% CI, 1.2-5.3), whereas NRAS+ tumors were similar to wild-type tumors with respect to this factor. By decadal age, BRAF+ cases were associated with estimated high UV irradiance at ages 0, 10, and 20 years, but a positive association was not evident for the decade years from age 30 years onward. By contrast, NRAS+ cases showed a nonsignificant association with high UV at age 50 years (OR, 2.5; 95% CI, 0.7-8.5), which persisted at age 60 years (OR, 2.0; 95% CI, 0.4-9.8).

We examined the independent effects of age, number of back nevi, and early-life UV irradiance in models analyzing BRAF+ and NRAS+ cases compared with wild-type cases (Table 5). In the BRAF model, adjusting for all three factors, number of back nevi, and high early-life UV irradiance remained significantly associated with BRAF mutation, while the association with age was borderline. In the NRAS model, older age remained significantly associated with NRAS mutation, and number of back nevi had a nonsignificant positive association for >14 versus 0 to 4 nevi. Early-life UV exposure was not associated with presence of an NRAS mutation. Reanalyses controlling for Breslow depth in both the BRAF and NRAS models gave similar results (data not shown).

The BRAF and NRAS models were reanalyzed excluding LMM, which is generally considered to be related to high levels of chronic sun exposure (38, 39), and ALM, which may be unrelated to sun exposure. In the BRAF model, the association of back nevi with BRAF mutation was unchanged, whereas the association with high early-life UV irradiance was stronger (OR, 3.7; 95% CI, 1.6-8.7). Excluding LMM and ALM from the NRAS model, the association of back nevi with NRAS mutations was stronger (5-14 nevi: OR, 1.7; 95% CI, 0.5-5.7; >14 nevi: OR, 2.4; 95% CI, 0.7-7.7).

Whiteman et al. (40) proposed that cutaneous malignant melanoma arises through at least two different causal pathways, one associated with aberrant nevus growth and another with chronic sun exposure. BRAF+ melanoma, due to its association with intermittently sun-exposed site (41), has been hypothesized to identify one of these pathways (42). Our study further develops the divergent pathway hypothesis by finding differences in the distribution of risk factors between subgroups based on BRAF and NRAS mutational status, providing evidence of etiologic heterogeneity. Our data indicate that BRAF+ melanoma cases are characterized by young age at diagnosis, increased number of nevi, and high early-life ambient solar UV exposure, whereas NRAS+ cases are characterized by late age at diagnosis, increased number of nevi, and later-life ambient UV exposure compared with wild-type cases.

The specific association of BRAF+ melanomas with high childhood ambient solar UV exposure, which is apparent from the data in this report, lends further support to the suggestion that error-prone replication of UV-induced DNA damage could underlie the acquisition of BRAF mutations in melanocytic tumors (43). Although the most common V600 mutations are not immediately recognized as UV signature mutations, the role of UV-induced DNA damage cannot be excluded at this point (43). It is possible that BRAF mutations are produced in melanocytes in childhood as a result of early-life UV irradiance, and these altered cells progress over time to melanoma. This may account for the younger mean age at diagnosis of BRAF+ compared with NRAS+ melanomas. Nevi, which also frequently contain BRAF mutations (44) and have been found to be increased in number with high early-life UV exposure (45), could be causal intermediates for some melanomas and/or arise in parallel. Factors other than or in addition to ambient UV irradiance, such as pattern and timing of sun exposure, phenotypic susceptibility, and/or site-specific populations of melanocytes may influence melanoma location (46, 47).

The differential age association for BRAF+ and NRAS+ melanomas found in our study with 214 subjects is consistent with other studies despite modest sample sizes and differences in analytic approaches for detecting BRAF mutations. A study of 219 subjects using pyrosquencing (48) and another study with 69 patients using direct sequencing (49) found similar associations with age. Two studies (27, 50), each with fewer than 100 cases, did not find BRAF+ melanomas to be significantly associated with number of nevi or fair phenotypic characteristics. In contrast, we found BRAF+ melanoma to be significantly associated with number of nevi, possibly due to greater statistical power for finding this association in our study. Similar to these other studies, our results do not support a strong association of BRAF mutations with fair phenotypic characteristics, but a modest association of BRAF+ and NRAS+ tumors with ability to develop a tan cannot be excluded.

Concordant with other studies, we found BRAF+ melanomas to be inversely associated with chronically sun-exposed sites (41, 48, 51), histologic solar elastosis (reported as CSD; refs. 8, 41), and LMM (52, 53), although the latter two factors were not significantly associated after age adjustment. Our finding of an association of NRAS mutations with truncal location contrasts with previous reports of NRAS+ melanomas occurring more frequently on habitually than intermittently sun-exposed sites (10, 54), associated with chronic occupational exposure (10), and occurring with approximately equal frequencies for CSD and non-CSD anatomic sites (8). However, our population-based study included a higher percentage of thin and superficial spreading melanoma tumors than previous studies and two of these prior studies (10, 54) were done before the discovery of BRAF mutations, which would be expected to comprise part of their reference groups.

In our study, we found NRAS+ tumors to be similar to wild-type tumors with respect to estimated early-life and lifetime ambient UV irradiance. However, when examined by decade, NRAS+ melanomas were positively associated with high ambient UV irradiance at age 50 and 60 years, indicating that occurrence of these cases could be influenced by adult sun exposure nearer to the time of diagnosis, which might explain the older mean age at diagnosis of NRAS+ cases. In addition, our data indicate that a nevus-prone tendency may be important for occurrence of NRAS+ cases, which is seemingly consistent with previous findings that nevi can contain NRAS mutations (55).

Our results are concordant with migration studies that determined melanoma risk to be highest in those exposed to sunlight in early life, even if exposed for a relatively brief period (4). Our data further indicate that one influence of early-life sun exposure may be specifically to increase the risk of BRAF+ melanoma. Our finding that early-life UV irradiance and number of nevi are independent predictors of BRAF mutation indicates that a nevus-prone tendency also influences the risk of BRAF+ melanoma. Others are investigating genetic susceptibility for nevus production (56) and that susceptibility in combination with solar UV exposure may increase risk of BRAF+ melanoma.

Strengths of this study include population-based case ascertainment, a quantitative observationally based approach to measuring ambient UV irradiance instead of using a proxy such as latitude, standardized histopathology review, and rigorous screening methodology designed for sensitive detection of all known BRAF and NRAS oncogenic mutations. Caveats of this study are that we do not know whether the results can be generalized to other geographic areas, and our findings pertain only to cutaneous melanoma, as ocular and mucosal melanomas were excluded. Also, all participants were Caucasian and only two had ALM, so the mutational profile of acral melanomas or melanomas in other racial groups could not be addressed. Individual-level sun exposure was not examined in this analysis and could differ for subjects living in areas of high and low ambient UV. Furthermore, additional somatic genetic alterations that occur in melanomas (57) were not assessed in this study. Mutations and homozygous deletions in CDKN2A, a potential UV target, were associated with BRAF/NRAS mutation in one study (58).

We note as a limitation that the association of NRAS mutations with nevi and later-life sun exposure were not statistically significant; however, this may simply be due to low statistical power, as the number of NRAS+ cases was relatively small. In addition, several sources of bias may have affected this study. We anticipated and attempted to decrease reporting bias, which is thought to be common for melanoma (59), by notifying all North Carolina dermatologists of the study and reporting procedures. In addition, our consent rate was high compared with other studies that collected biological samples. When we compared participants with nonparticipants, there were no significant differences for demographic characteristics, tumor size, or the key risk factors under study. Our block collection and PCR amplification rates were high, minimizing additional sources of selection bias that can occur in molecular epidemiology studies. Recall bias is unlikely to be a major factor because the analyses were based on residential locations rather than more complex sunlight-exposure behaviors.

Self-reported back nevus counts may have resulted in misclassification; however, the instrument we used had a correlation coefficient of 0.57 compared with dermatologist review of photographic documentation (60), and nevus density diagrams resulted in similar associations in our data. Although our finding of an association between BRAF and NRAS mutations and thicker melanoma could have been affected by selective difficulties in analyzing thin melanomas, we attempted to minimize this possibility by using laser capture microdissection for all small samples. In addition, our high PCR amplification and analysis rate (96%) ensured that a high percentage of tumors were screened for mutations. Furthermore, the associations in our BRAF and NRAS models did not change when we adjusted for Breslow depth.

Our results suggest that the factors driving the development of melanoma vary for different mutational subtypes. Our data indicate that both early-life UV exposure and nevus propensity contribute to occurrence of BRAF+ melanoma, whereas nevus propensity and later-life sun exposure influence the occurrence of NRAS+ melanoma. From a public health perspective, the association of BRAF+ melanoma with early-life solar UV exposure provides additional support for an emphasis on childhood sun protection in melanoma prevention programs. In addition, our study indicates that the erythemal UV index calculated by the National Weather Service (61) may be useful for predicting when sun exposure should most be avoided for melanoma prevention.

We thank Urvi Mujumdar at Memorial Sloan Kettering Cancer Institute for assistance with data management; Sasha Madronich at the National Center for Atmospheric Research (Boulder, CO) for his assistance with model estimation of the erythemal UV irradiances (the National Center for Atmospheric Research is operated by the University Corporation for Atmospheric Research under National Science Foundation sponsorship); Dr. Marila Cordeiro-Stone for useful discussion regarding the manuscript; the two anonymous reviewers for helpful comments on the manuscript; and the following National Cancer Institute–designated core laboratories of the University of North Carolina Lineberger Comprehensive Cancer Center for their support: The Tissue Procurement and Analysis Facility, under the direction of Dr. Lynn G. Dressler, and The Microscopy Services Laboratory's Laser Capture Microdissection Facility, under the direction of Dr. C. Robert Bagnell.