Cutaneous Melanoma Subtypes Show Different BRAF and NRAS Mutation Frequencies

Cutaneous melanoma is the most serious form of skin cancer because it metastasizes so readily. Initially, it was regarded as a homogeneous entity with a uniformly poor prognosis, but the advent of specialized pigmented lesion clinics allowed detailed multidisciplinary study of cutaneous melanoma and, as a result, a number of different subtypes with distinct clinicopathologic features were described. The distinctions were based on the preferred site of origin, relative amount of UV light exposure, and duration of preinvasive growth, with four subtypes accounting for the vast majority of cutaneous melanomas. These are superficial spreading, nodular, lentigo maligna, and acral lentiginous melanoma (ALM), comprising ∼70% to 75%, 20% to 25%, 5% to 10%, and 5% of cutaneous melanoma cases, respectively, in White Caucasian populations, with acral melanomas being the most frequent in dark-skinned races (1). More recently, molecular differences between these subtypes have come to light, corroborating the clinicopathologic observations. Compelling evidence for differences comes from patterns of DNA alterations observed in cytogenic and comparative genomic hybridization studies, where distinct alterations have been identified. For example, acral melanomas tend to have frequent genetic loci showing amplifications, some of which are early events that can even be detected in adjacent histologically normal melanocytes (2). An initial study indicated that 66% of melanomas had BRAF mutations, whereas NRAS mutations occurred in many of the remainder (3). BRAF and NRAS genes both encode mitogen-activated protein kinase (MAPK) pathway constituents. This pathway is important in several crucial cellular processes, such as proliferation and differentiation. Subsequent studies have shown that BRAF and NRAS mutation frequencies differ between cutaneous melanoma subtypes (4–6) and in addition BRAF mutation is uncommon in noncutaneous melanomas such as ocular and mucosal types (7–11). A recent study indicated that the degree of sun damage is more predictive of molecular pathology than cutaneous melanoma subtype, which poses a threat to the relevance of histologic subtype in clinical practice (12).

Nevertheless, there is relatively little data focusing on the frequency of BRAF mutations in cutaneous melanoma subtypes and there are only two studies that combine this with analysis of NRAS mutations (5, 13). These were both in Japanese populations where the overall frequency cutaneous melanoma is lower than in Western countries and where acral melanoma is the most common type. This lack of knowledge undermines an emerging goal in cutaneous melanoma research: to determine whether subtypes of cutaneous melanoma have different molecular pathways of tumor progression. Achieving this goal is critical for development of tailored treatment for all of the various forms of cutaneous melanoma. In this context, the aim of this study was to better characterize the frequency of BRAF and NRAS mutations in the four commonest cutaneous melanoma subtypes, the hypothesis being that clinicopathologic differences will be underpinned by molecular differences.

Melanoma tissues. Samples were selected from the archives of the University Hospitals of Leicester NHS Trust with local research ethics committee approval. Cases were identified from a search for melanoma samples over a 5-year period, comprising over 500 cutaneous melanomas, each diagnosis being verified by a dermatopathologist (G.S.). From these samples, melanomas were grouped into one of the four major subtypes. The criteria for these subtypes are well recognized by dermatopathologists, but are liable to some degree of subjectivity. We used the criteria summarized in a published table by Crowson et al. (14). Cases that could not be classified into these subtypes or that were other rarer subtypes were discarded. From 440 remaining samples comprising these major subtypes, 14 superficial spreading malignant melanomas (SSMM), 11 nodular melanomas (NM), 13 lentigo maligna melanomas (LMM), and 21 ALM were selected. The relatively large number of acral melanomas was chosen for statistical purposes to ensure that our finding of a low BRAF mutation frequency was genuine. Cases within each subtype were selected according to consecutive laboratory accession numbers. None of the cases have been reported previously. The mean (and median) Breslow depths were 2.4 mm (1.8 mm), 3.8 mm (3.0 mm), 2.4 mm (1.5 mm), and 3.0 mm (1.8 mm); mean ages were 56.5, 65.5, 70, and 67 years; and male to female ratios were 4:10, 5:6, 5:8, and 10:11 for SSMM, NM, LMM, and ALM respectively.

Sun damage was assessed in contiguous skin of cutaneous melanoma samples. Cases demonstrating coalescent areas of solar elastosis in the dermis were defined as showing chronic UV damage, cases with noncoalescent areas were defined as nonchronic UV damage and those with absent solar elastosis as minimal UV damage.

Mutation analysis. Mutations in NRAS and BRAF genes were detected using RFLP-PCR or single-strand conformational polymorphism (SSCP) analysis of PCR-amplified DNA from 1 to 5 × 10 μm tissue sections per sample. Enrichment of DNA from melanoma cells was done by removing the tumor tissue from 10 μm sections with a pipette tip. The tumor tissue was identified by comparison with a H&E-stained serial section. Microdissection, DNA extraction, and SSCP and RFLP-PCR analyses were done as described previously (15), except that the PCR primers for NRAS exon 2 SSCP analysis were as follows: forward 5′-CACCCCCAGGATTCTTACAG-3′ and reverse 5′-TCGCCTFTCCTCATGTATTG-3′. The RFLP-PCR primers for BRAF exon 15 analysis were as follows: forward 5′-GGTGATTTTGGTCTAGCTGCA-3′ and reverse 5′-GCTTGCTCTGATAGGAAAATGAG-3′. This primer detects T/A transversions at codon 600 (previously called codon 599) by restriction digestion with BtsI (New England Biolabs, Ipswich, MA). This restriction enzyme digests the wild-type sequence, but not the mutant, thus identifying mutations in codon 600.

Statistical analysis. A χ² test was done for analysis of categorical variables, except for 2 × 2 contingency tables, where Fisher's exact test was used. A Mann-Whitney U test was done for continuous variables. All analyses were two-tailed and a P value of <0.05 was regarded as statistically significant for the a priori (i.e., planned) comparison between cutaneous melanoma subtypes and mutation frequencies, whereas a conservative P value of <0.01 was regarded as significant for all other comparisons. All tests were done using SPSS for Windows, version 12.0.1.

We analyzed the four main clinicopathologic subtypes of cutaneous melanoma for BRAF mutation. Full demographic data are shown in Table 1. Codon 600 in exon 15 of the gene was first assessed using a RFLP-PCR method to detect the T/A transversion that represents the vast majority of alterations seen in cutaneous melanoma. Detected mutations were confirmed by DNA sequencing. SSCP was subsequently used to screen exon 15 of the BRAF gene for alternative sequence alterations in those cases where T/A transversions were absent, but no further mutations were detected. BRAF exon 11 mutations, which are relatively rare in cutaneous melanoma, were not assessed. Examples of RFLP-PCR analysis are shown in Fig. 1A and the BRAF mutation results are summarized in Table 2. The overall BRAF exon 15 mutation frequency was 22 of 37 cases (37.3%). There was a significant difference in BRAF exon 15 mutation frequencies between subtypes, with only 2 of 21 (9.5%) ALM showing this alteration compared with 9 of 14 (64.3%) SSMM, 4 of 11 (36.4%) NM, and 7 of 13 (53.4%) LMM (P < 0.01).

Exposure to UV radiation is the best characterized environmental risk factor for cutaneous melanoma and is one of the factors that distinguish some of the melanoma subtypes. There was a significant difference in the frequency of BRAF mutation between cutaneous melanomas with chronic, nonchronic, and minimal UV exposure, being 10 of 18 (55.6%), 10 of 20 (50%), and 2 of 21 (9.5%), respectively (P = 0.004; see Table 2). However, this reflected the fact that acral melanomas lacked BRAF mutations and are, by definition, from sites that are relatively protected from UV exposure. Repeating this analysis, excluding ALM, showed no significant relationship between UV exposure and BRAF mutation status (P = 0.732). There was no correlation between BRAF mutation status and other clinical variables, namely age, sex, Breslow's depth, or Clark's level, as shown in Table 2.

Previous data indicates that many cutaneous melanoma that lack BRAF mutation harbor NRAS mutation. Thus, the low frequency of BRAF mutations in ALM could be due to an increase in NRAS alterations. We therefore looked for NRAS mutations in cases where BRAF exon 15 mutation was absent. Only exon 2 of the NRAS gene was analyzed (using SSCP) because this is the commonest site for alterations in cutaneous melanoma, with codon 61 representing a mutation hotspot. The results are shown in Fig. 1B and C and Table 2. Nine of 19 (47.4%) ALM that lacked BRAF exon 15 mutation had an NRAS exon 2 mutation, although the frequency may be higher, because in three cases PCR with the NRAS primers was unsuccessful, despite several attempts. We also looked for NRAS exon 2 mutations in the SSMM, NM, and LMM cases that lacked BRAF exon 15 mutations. They showed 2 of 5 (40%), 6 of 7 (87.1%), and 1 of 6 (16.7%) mutations, respectively. All NRAS exon 2 mutations were at codon 61, being either Q61R (CAA→CGA) or Q61K (CAA→AAA) alterations. The frequency of NRAS exon 2 mutations in these cases showed no statistically significant differences (P = 0.091). The frequency of BRAF exon 15 and NRAS exon 2 mutation was combined to determine the overall frequency in SSMM, NM, LMM, and ALM. This was 11 of 14 (78.6%), 10 of 11 (90.9%), 8 of 13 (61.5%), and 11 of 21 (52.4%), respectively (P = 0.114). This indicates that when NRAS mutations are accounted for, cutaneous melanoma subtypes show no statistically significant difference in the overall frequency of BRAF exon 15 and NRAS exon 2 mutations. Forty of 59 cases (67.8%) harbored one or other mutation. Because these mutations are predicted to activate the MAPK pathway, the latter seems to be important in all of the four common subtypes of melanoma. The overall frequency of BRAF exon 15 and NRAS exon 2 mutations showed no significant correlation with patient sex, Breslow depth, Clark's level, or sun damage (including an analysis that excluded ALM, P = 0.184), but there was a correlation with patient age, namely mutant samples were from younger patients, at the 5% significance level, but not at the more conservative 1% level used for unplanned comparisons, as described in Materials and Methods (see Table 2).

This study extends current knowledge of the prevalence of BRAF and NRAS mutations in cutaneous melanoma. Few studies have looked at both mutations in a range of cutaneous melanoma subtypes. This study shows that the BRAF mutation frequency differs between cutaneous melanoma subtypes. However, our key finding is that the overall BRAF/NRAS frequency is not significantly different among cutaneous melanoma subtypes, most notably because ALM, which have low frequency of BRAF mutation, have a high level of NRAS mutations. Although this may support the notion that cutaneous melanoma subtypes follow different roads to malignancy, both BRAF and NRAS mutations have a common effect: activation of the MAPK pathway. This study therefore shows that activation of this pathway has an important role in all major subtypes of cutaneous melanoma, but the precise way of achieving this varies. However, it should be noted that our study did not incorporate a stratified selection strategy to account for sun damage, and so this confounding effect is not addressed. This is an important consideration when our data are interpreted.

The results of this study show the critical importance of MAPK pathway signaling in cutaneous melanoma, which is not surprising given the role of that pathway in cell growth (16). However, our data reveals that the way of activating the pathway differs between subtypes, possibly reflective of the differences in site and/or sun exposure. It is unclear why NRAS should be the favored mechanism for MAPK pathway activation in ALM compared with other cutaneous melanoma types. UV exposure is the best known environmental risk factor for cutaneous melanoma, but ALM occurs in a relatively UV-protected sites, such as the hairless skin of palms, soles, and nail bed. Thus, differences in UV exposures could account for the differences in BRAF and NRAS mutation frequency. However, the relationship between cutaneous melanoma occurrence and UV exposure is complex because there is no clear relationship between dose and response (4). Thus, BRAF mutations are generally more common in intermittently exposed sites, such as the trunk, than sites of chronic damage, such as the face (4). Furthermore, BRAF mutations are seen in some internal malignancies, such as thyroid papillary carcinoma, colorectal carcinoma, pancreatic carcinoma, and borderline ovarian tumors, demonstrating that there are other factors apart from UV exposure that lead to BRAF mutation. One possibility is that inflammation and resulting oxidative damage could be the theme that links the occurrence of BRAF mutation at diverse sites, with UV damage being one of several ways that an inflammatory response can be generated. Also, the commonest alteration in BRAF, a T/A transversion, is not a signature change of UV damage, although there is evidence that psoralen plus UVA (i.e., PUVA) can induce this alteration (17). Likewise, the NRAS mutations that we found do not show a UV damage signature. An alternative possibility underlying the variable pattern of mutations is that there are intrinsic differences between melanocytes from different regions of the body. Perhaps subtle variations in gene expression and local environment mean that acral melanocytes get greater advantage from NRAS rather than BRAF mutation. Alternatively, because acral melanocytes have extremely little exposure to the carcinogenic effects of UV, it may be critical that they progress by the least number of molecular “hits.” In this respect, mutations in NRAS would be more advantageous than BRAF. This is because NRAS acts proximally to BRAF in the MAPK pathway and has additional effects on other cellular pathways, especially activation of the AKT pathway, which seems to be important in cutaneous melanoma (18, 19). Thus, NRAS mutation simultaneously activates two key pathways at a stroke. This multipathway effect is supported by a recent study that reported a melanoma model in human skin, where the presence of NRAS mutations rendered mutations in BRAF or the AKT pathway unnecessary, whereas BRAF mutations required additional events to activate AKT (20). However, according to this argument, one might also expect NRAS mutations to be relatively more common in SSMM than LMM because they also have relatively low levels of UV exposure. On the contrary, these cases have frequent BRAF mutations. However, this type of cutaneous melanoma is particularly associated with benign common nevi (21). The majority of common nevi have a BRAF mutation (22) and thus the high prevalence of mutations in both lesions may reflect the fact that common nevi and SSMM share a related tumorigenic pathway, as suggested by a proportion of SSMM arising ex nevus, a phenomenon that is relatively uncommon in some other types of cutaneous melanoma (21). This idea of distinct tumorigenic pathways is supported by recent studies on a group of melanocytic tumors with a characteristic morphology, Spitz nevi and Spitzoid melanoma, which seem to progress via a BRAF/NRAS-independent mechanism (23).

The results of the present study differ from those of Takata et al. (5) and Sasaki et al. (13). Sasaki et al. identified BRAF mutation in 4 of 8 SSMM, 5 of 15 ALM, none of 6 NM, 5 LMM, and 1 mucosal melanoma. Takata et al. looked at only 13 primary acral lesions, finding one BRAF mutation. Neither study identified an NRAS mutation in primary ALM. By contrast, the present study found 11 of 21 mutations in ALM, of which two were in BRAF. This discrepancy could have arisen because of inherent differences in melanoma pathogenesis in Japanese and United Kingdom populations. Also, the frequency at which mutations are detected is dependent on the detection method (24). Both previous studies used direct sequencing, the least sensitive method. Lang et al. (25) analyzed BRAF mutation (but not NRAS mutation) in cutaneous melanoma subtypes, and found 2 of 13 in ALM, in agreement with our data. They found the prevalence of mutation was highest in SSMM, followed by NM, LMM, and ALM. In our data, there is a similar trend except for LMM, where we found a much higher frequency of mutations. Also, their overall frequency of BRAF mutations in all cutaneous melanomas was relatively low (13 of 52; 25%). The discrepancy is due to our higher frequency of mutations in LMM and SSMM. This is possibly due to differences in the mutation detection techniques, although their methods are sensitive (24). Another potential bias is that the cutaneous melanomas used by Lang et al. were relatively thick (mean 6.3 mm versus 2.9 mm). Lang et al. also found that BRAF mutation frequency was independent of sun damage, whereas we found a significant relationship. This is likely to be because our sample was biased toward ALMs. Indeed, this significant effect disappeared when ALMs were excluded from our analysis (P = 0.732). Lang et al. also used a different criterion for assessing sun damage, namely site of origin rather than a direct assessment of solar elastosis in surrounding tissue. This strategy leads to unverified assumptions about the degree of solar damage and is likely to be less reliable than direct assessment of sun damage that we used.

Comparative genomic hybridization data suggest that there are important differences between ALM and SSMM. Our finding of low BRAF mutation frequency in ALM compared with other types of cutaneous melanoma supports this difference. Comparative genomic hybridization shows that ALMs have frequent and early gene amplifications that are often at loci containing known oncogenes such as cyclin D1, CDK4, and HRAS, although some loci contain unknown targets (2). Bastian et al. (26) reported, in the proceedings abstract from the XIIth annual Pan-American Society for Pigment Cell Research conference, a low overall frequency of NRAS mutations, but made no specific mention of NRAS mutation in acral lesions. It is difficult to make systematic comparisons of our findings to the limited amount of information presented in an abstract. However, in a full study presenting similar data, Curtin et al. (12) show that the underlying molecular pathology of melanoma is strongly correlated with UV damage, which was a stronger predictor of genetic changes than subtype. They reported that the frequency of BRAF mutations was significantly different on chronic versus nonchronic sun-damaged skin when acral lesions were discounted. By contrast, a comparable analysis using our data, where ALMs were excluded, showed that sun damage and mutation status were independent. This is because the LMMs in our study, which all had chronic sun damage, had a relatively high BRAF mutation frequency. This same discrepancy also distinguishes our findings from those of Maldonado et al. (4), who also assessed mutations in relation to solar damage. There are several possibilities for this specific discrepancy. It is possible that the mixture of cutaneous melanoma subtypes that were assigned to the chronic sun damage category was different. There are also inherent problems in the precise way that that chronic sun damage is defined because this assessment is somewhat subjective. Additionally, LMM mutation frequencies in any of these studies could be skewed by purely stochastic events. Despite these differences, there is a broadly similar theme in our, Maldonado et al.'s, and Curtin et al.'s studies: All found the highest proportion of cutaneous melanomas with BRAF/NRAS wild-type status in those from acral and chronically sun damaged skin, whereas those from nonchronic sun-damaged skin had the lowest frequency. Curtin et al. additionally identified common, focused CDK4 and CCND1 amplifications in cutaneous melanomas from acral and chronically sun damage sites, respectively, at the expense of BRAF and NRAS mutations. Taken together, these findings would be consistent with cutaneous melanomas from all three studies having relatively frequent amplifications in cutaneous melanomas from acral and chronically sun damaged skin compared with nonchronically damaged skin. However, this latter point is somewhat speculative because it assumes that the findings of Curtin et al. can be generalized, given that neither we nor Maldonado et al. did CCND1/CDK4 gene dosage analyses. Notably, both our and Curtin et al.'s data emphasize the critical need for cell cycle deregulation in all cutaneous melanoma subtypes. Our study emphasizes MAPK pathway activation, whereas Curtin et al. emphasize both MAPK pathway activation and amplification of downstream targets, such as CCND1 and CDK4.

In summary, the findings of the present study refine our understanding of the role of BRAF and NRAS mutation in cutaneous melanoma. Our principal finding was to show the importance of MAPK pathway activation in each of the major cutaneous melanoma subtypes, although other studies provide evidence that amplification of downstream MAPK pathway targets may also be important. An emerging and critical aspect of current research is to use molecular pathology to generate new treatments, notably drugs targeted at relevant kinases (27). This study directly informs future use of these new treatments, but more work will be needed to further clarify the relative importance gene amplifications versus BRAF/NRAS mutation and also to assess whether cutaneous melanoma subtype or degree of sun damage are better predictors of molecular pathology.