A Large French Case-Control Study Emphasizes the Role of Rare Mc1R Variants in Melanoma Risk

Background. The MC1R gene implicated in melanogenesis and skin pigmentation is highly polymorphic. Several alleles are associated with red hair and fair skin phenotypes and contribute to melanoma risk. Objective. This work aims to assess the effect of different classes of MC1R variants, notably rare variants, on melanoma risk. Methods. MC1R coding region was sequenced in 1131 melanoma patients and 869 healthy controls. MC1R variants were classified as RHC (R) and non-RHC (r). Rare variants (frequency < 1%) were subdivided into two subgroups, predicted to be damaging (D) or not (nD). Results. Both R and r alleles were associated with melanoma (OR = 2.66 [2.20–3.23] and 1.51 [1.32–1.73]) and had similar population attributable risks (15.8% and 16.6%). We also identified 69 rare variants, of which 25 were novel. D variants were strongly associated with melanoma (OR = 2.38 [1.38–4.15]) and clustered in the same MC1R domains as R alleles (intracellular 2, transmembrane 2 and 7). Conclusion. This work confirms the role of R and r alleles in melanoma risk in the French population and proposes a novel class of rare D variants as important melanoma risk factors. These findings may improve the definition of high-risk subjects that could be targeted for melanoma prevention and screening.


Introduction
The incidence of cutaneous melanoma, the most lethal type of skin cancer, is increasing in western countries, doubling every ten years [1]. Melanoma is a complex disease that arises through multiple etiological pathways [2]. Ultraviolet radiation exposure is the main environmental cause, and pigmentation characteristics such as light skin, hair and eye colour, and high number of nevi have also been identified as melanoma risk factors [3].
Among pigmentation genes, MC1R, which is the most studied, is associated with human skin pigmentation and melanoma susceptibility. MC1R, the receptor for -melanocyte stimulating hormone ( -MSH), is a G protein coupled receptor with seven transmembrane domains that regulates the relative concentration of brown-black eumelanin and red-yellow pheomelanin [18]. Eumelanin has been shown to reduce the accumulation of DNA photodamage and to protect melanocytes from UV-induced apoptosis. Pheomelanin is, on the contrary phototoxic, generating oxidative stress by the production of reactive oxygen species [19][20][21].
Numerous association studies have demonstrated the important role of MC1R variants in melanoma predisposition [8,11,28,29]. The influence of MC1R variants on melanoma risk has also been reported but there are several discrepancies in the different published works [4,[29][30][31]. In addition, to date there is no conclusion on the role of rare MC1R variants in melanoma. In this large French case-control study, we investigated the role of different classes of MC1R variants in melanoma risk, focussing particularly on the role of rare MC1R variants.

Studied Populations.
A cohort of 1131 Caucasian melanoma patients was recruited between 2002 and 2008 from the dermatology departments of all university-affiliated hospitals in Paris (the MelanCohort). The main characteristics of the patients have been previously described [32]. Melanoma was sporadic in 784 patients (69%), including 81 patients (7%) who had multiple melanomas and 229 patients (20%) who had familial melanoma (at least 2 cases in first-or second-degree relatives, including the proband). Among the familial and multiple sporadic cases, 8.5% of patients carried mutations in the CDKN2A or CDK4 gene.
The control group comprised 869 ethnically matched skin cancer-free blood donors recruited from the EFS (Etablissement Français du Sang) in Bichat and Saint-Louis hospitals over the same period. These subjects have previously been described and used as controls in case-control studies [32][33][34].
Data from patients and controls regarding sex, age, ethnic origin, date of birth, anatomoclinical data, pigmentation characteristics (hair, eye, and skin colours), skin type (Fitzpatrick classification I to IV), nevus count (<10, 10-50, 51-100, and >100), and heavy freckles (yes/no) were collected in a standard document. Ancestry was investigated through birth location of parents and grandparents, and only those with a Caucasian ancestry were retained for the study. The pigmentation characteristics of patients and controls are summarized in Supplementary  PCR products were verified on a 2% agarose gel and purified by EXOSAP-IT (USB Corporation, OH, USA). Sequencing reaction was performed on 8900 Fast Thermal Cycler (Applied Biosystems), using 10 ng of purified PCR products and the Big-Dye Terminator Cycle Kit (Applied Biosystems). Sequence analysis was performed with an ABI-Prism 3130 automated sequencer (Applied Biosystems) and read with SeqScape software v2.5 (Applied Biosystems).

Classification of MC1R
Variants. The functional impact of numerous MC1R variants has been assessed in previous studies [47] (Supplementary Table 2). Some variants lead to poor MC1R expression, due to endoplasmic reticulum (ER) retention or aberrant trafficking from ER to Golgi [50]. Other variants result in a diminished functional receptor due to lower affinity for -MSH, reduced coupling with cAMP, or decreased ability to stimulate cAMP production [47,[50][51][52][53].
In order to predict the impact on protein function of MC1R variants, we used in silico prediction tools, SIFT (http://sift.jcvi.org/), SNPs3D (http://www.snps3d.org/), and PolyPhen (http://genetics.bwh.harvard.edu/pph/). Rare variants, which were defined by an allele frequency <1% and predicted to be damaging by at least 1 of 3 prediction tools, were predicted to be damaging ( ) variants, while variants without any damaging effect were regarded as nondamaging ( ) variants. Variants that had a clear functional impact (i.e., nonsense or frameshift mutations) were classified as damaging ( ) variants.

Statistical
Analysis. Statistical analysis was performed by using PASW software version 18. The level of significance for all tests corresponded to an alpha error rate of 5%. All odds ratios (OR) were calculated with 95% confidence intervals.   To assess the association of and variants with melanoma, we used Fisher's exact test, the number of haplotypes without any variants alleles (wild-type and synonymous variants) being considered as reference. Rare variants were thereafter classified into two subgroups according to functional prediction (predicted to be damaging ( ) or nondamaging ( )), and their effects on melanoma risk were also tested.
A multivariate analysis adjusted for hair and eye colours, skin type, and nevus count was carried out to investigate the independent effect of and alleles on melanoma risk. values and their corresponding OR were calculated with logistic regression. Due to the interdependency of all our tests and the magnitude of our results, no correction for multiple testing was performed.
Finally, in order to investigate the respective role of MC1R protein domains in melanoma, the number of and variants in different domains was compared between patients and controls. Each protein domain was determined by ExPASy Bioinformatics Resource Portal (http://expasy.org/) using UniProtKB Q01726.2 as query.

Characterization of MC1R Variants.
By sequencing the entire MC1R coding sequence, we found 79 MC1R variants: 2 nonsense, 3 frameshift, 53 missense, and 21 silent variants ( Table 1), 9% of which were localized in the extracellular portion of the receptor, 65% in the transmembrane domains, and 25% in the cytoplasmic domains ( Figure 1).
To investigate whether the effect of the two main MC1R variant categories on melanoma susceptibility was independent of pigmentation traits, we conducted a multivariate analysis including the main clinical melanoma risk factors

Impact of the Different MC1R Protein Domains.
In order to study the impact of MC1R protein domains on melanoma predisposition, the different classes of MC1R variants ( , , , and ) were positioned on the different protein domains (Figure 1).
Most MC1R variants were localized in six receptor domains: transmembrane 1, 2, 3, and 7, intracellular 2, and C terminal domains. Very few variants were located in the extracellular portion. Among these domains, variants located in the intracellular domain 2 and in the transmembrane domain 7 had the highest impact on melanoma risk (OR = 2.75 [2.22-3.40] and OR = 2.48 [1.67-3.69]).
The repartition of , , , and variants in each protein domain indicated in Table 4 showed that 63% of variants were located in four domains (intracellular 2 and transmembrane 2, 5, and 7) whereas only 18% of variants were located in these domains ( < 0.0001). Importantly, three of these four domains also contain at least one variant, suggesting an important role of these domains in MC1R function and pathogenicity.

Discussion
MC1R variants are usually classified into two main categories, RHC ( ) and non-RHC ( ), according to their association or not with the red hair colour phenotype [21,24,27,55]. For the past ten years several association studies have demonstrated the importance of alleles on melanoma predisposition [28,29,42,56,57]. However, the influence of rare MC1R variants on melanoma predisposition has been poorly investigated which prompted us to study in detail the role of these variants in the French population.
In this study, we confirmed the association of most alleles and melanoma risk with strengths that were close to those observed in previous studies [28,29,42,56,57].
We also showed a clear association of frequent MC1R alleles, especially V60L, V92M, and I155T, with melanoma. In an early meta-analysis neither of them was found to be associated with melanoma [55], whereas, in a more recent and larger meta-analysis, both were associated with melanoma [57]. V60L is a loss of function variant with reduced coupling to the cAMP signalling [50,53] and V92M has a lower affinity for -MSH than wild-type MC1R and a decreased ability to stimulate the production of cAMP [23]. These functional data argue for an association of theses variants with melanoma risk.
Even though the OR of alleles was much higher than that of alleles, their PAFs were very close (15.8% for and 16.6% for ) suggesting that the impact of alleles on melanoma seems to be important in the French population. Notably, in our work, the individual PAFs of V60L and V92M (6.5% and 4.5%) were very close to that of alleles R151C and R160W (7.4% and 5.7%). Our results are different from that published by Williams et al., in which the PAF of variants is only 7.4. This may be due at least, in part, to the high allelic frequency of variants observed in French melanoma patients (0.43) (versus 0.33 in Williams meta-analysis) [57]. This further emphasises the role alleles in melanoma risk according to the ethnical background of the population studied. In addition, both and alleles remained associated with melanoma risk in multivariate analyses, suggesting that they exert a role in melanoma risk independently of their effect on pigmentation, as previously suggested.
In this work we identified 25 variants that have not been reported before [22,23,31,47], further underscoring the highly polymorphic character of MC1R. According to Gerstenblith's work, the proportion of rare MC1R variants varies across populations and within Caucasian populations [23]. In our study, the proportion of rare variants (87%) was close to that described in Scherer's work (74% in Germany and 78% in Spain) [31]. Interestingly, rare variants predicted to be damaging ( variants) were associated with melanoma as strongly as alleles (OR = 2.38 [1.38-4.15]) whereas variants not predicted to be damaging ( ) had no effect on melanoma. In addition, PAF of variants (1.6%) was higher than two variants (R142H and D84E), which seems to indicate that the contribution of this subgroup in melanoma predisposition should be taken into consideration, at least in the French population.
Yet, there is a limitation in our work: the absence of functional studies concerning the potential effects of these novel MC1R variants, notably on -MSH binding, receptor cellular localisation, and cAMP signalling.
Finally, the majority of variants were located in the same domains as the alleles (intracellular 2 and transmembrane 2, 5, and 7, Table 4 and Figure 1). It had been shown before that there was a similar localization of variants in the German, Spanish, and Italian populations [29,31] suggesting an important role of these domains in the receptor's function.

Conclusion
In this large study we confirmed the role of MC1R alleles in melanoma susceptibility and clearly showed that MC1R alleles also significantly increase the risk of melanoma. In addition, we defined the role of rare MC1R variants and proposed a novel class of variants that are strong melanoma risk factors. These findings might help in the definition of high-risk melanoma subgroups that could be targeted for melanoma prevention.