In this study, 363 tag SNPs in 65 pigmentation-related gene regions were successfully genotyped in 590 MM cases and 507 controls in discovery phase I. In addition, 624 MM cases and 789 controls were genotyped for the top six MM-associated SNPs in validation phase II. Thus, a total of 1214 MM cases and 1296 controls have been evaluated in this study. Among 363 SNPs in 65 pigmentation gene regions, allele frequencies were highly correlated between the HapMap (CEU) and the Spanish samples (R2
0.85). Despite of 143 SNPs with minor allele frequencies different from those published in HapMap, none remained significant after multiple testing (Bonferroni correction). Although the published HapMap MAFs should probably not be taken as definitive (at least for a Southern European population), overall, the HapMap CEU sample is representative of the Spanish population. Therefore, the HapMap data can be used in assessing the appropriateness of SNPs for inclusion in association studies even though we are analyzing genes with putative phenotypic implications.
From discovery phase I, p-values below 0.05 were obtained for 30 SNPs and, of these, ten final candidate SNPs, based on an arbitrary p-value threshold of 0.01, were identified. We were encouraged to follow a validation phase due to two SNPs (rs35414 and rs2069398) remaining statistically significant after Bonferroni correction. After testing six of these ten SNPs in an independent Spanish series with 624 MM cases and 789 controls, one SNP (rs35414) on an intronic region of the SLC45A2 gene significantly associated with melanoma was confirmed (p<0.0001). Three additional SNPs in TYR, SILV/CDK2 and ADAMTS20 (rs17793678, rs2069398 and rs1510521 respectively) had an overall p-value<0.05 when considering the whole data set (1214 MM cases and 1296 controls). However, in validation phase II, data for these three SNPs did not reach statistical significant results per se.
Recent GWAS have unveiled association between SNPs or genetic variants in MC1R
, and pigmentation as well as melanoma. These findings emphasize the contribution of pigmentation to melanoma predisposition and tumourigenesis through gene-environment interactions 
The top predisposing MM-associated gene in our study has functions related to melanosomal formation, maturation and transportation. The SLC45A2
gene (MIM#606202) encodes a protein that acts, presumably, as a membrane-associated transporter 
exhibits structural homology to plant sucrose-proton symporters and probably directs the traffic of melanosomal proteins and other substances to the melanosomes 
mutations cause pigmentation variation in several organisms 
. In humans, pathogenic mutations in SLC45A2
lead to type IV oculocutaneous albinism (OCA4) 
mutations disrupt tyrosinase processing and trafficking at the post-Golgi level 
. Other variants located on the promoter and in exonic sequences on the SLC45A2
gene have been shown to be significantly associated with dark hair, skin, and eye pigmentation in Caucasian populations 
. Previous results from our group and others show that the coding variant c.1122C>G (p.F374L, rs16891982) has been associated with protection against melanoma susceptibility in Southern European populations 
. In addition, a recent study by Duffy and cols. (
) has detected protective effects of SLC45A2
through three SNPs in that gene: rs16891982, rs35391 and rs28777. This last study concludes that the two noncoding SNPs, rs35391 and rs28777, are in LD with each other and are not significant melanoma risk predictors in a multivariate logistic regression models after adjusting for rs16891982 (p.F374L). They confirm that rs16891982 is the SNP most strongly associated with melanoma risk in their population. These three SNPs are placed in the middle region of the SLC45A2
gene, a segment of the gene that, according to HapMap data, does not belong to any block structure (probably due to their low minor allele frequencies in the Caucasian population where they were genotyped) (Figure S1
The results obtained in this study support the protective role of SLC45A2
and propose two novel intronic SNPs associated with MM predisposition. Both of these SNPs, rs35414 and rs35415, are located inside intron 1 of the SLC45A2
gene. rs35415 (p-value
0.0004 in phase I) was selected as the tag of the LD block located at the 5′ end of the gene, while rs35414, studied in phase I and validated in phase II (global p-value
0.0001), is located in a region without apparent LD block structure (Figure S1
). However, both SNPs (rs35414 and rs35415) are in strong linkage disequilibrium with each other (R2
0.95) (Figure S1
). Both constitute the second most common haplotype of the gene (43%), also conferring protection against melanoma (p-value
0.0001). Thus, the association of this 5′ end region to MM predisposition suggests a putative involvement of unknown functional variants, in LD with these SNPs, involving regulatory elements in the promoter region of the SLC45A2
Furthermore, the protective role of the rs35414 variant in SLC45A2 is reinforced due to its capability to modulate the MM risk conferred by the MC1R locus.
Although no additional SNPs were replicated in the second phase of the study, three of them had statistical significance when the entire data collection was taken as a whole. Therefore, we consider them as quite relevant for further validations in other populations.
SILV (MIM#155550) melanosomal matrix protein represents a melanoma specific antigen recognized by tumour infiltrating cytotoxic T lymphocytes 
. It is also known as PMEL17, GP100 or ME20 
and it is crucial for proper formation and maturation of melanosomes. In stage II melanosomes, processed SILV protein aggregates to form fibrils, to which presumably the eumelanin pigment is attached 
. It was found to be orthologous to the mouse silver
locus. In this study, SNP rs2069398 was selected in order to represent the 5′end of the SILV
gene. Interestingly, the SILV
locus is flanked quite closely on each side by CDK2
at distances of 829 and 193bp, respectively, placed head to head with CDK2
in such a way as to allow sharing or overlapping promoter elements 
. rs2069398 is located in exon1 of the CDK2
gene, resulting in the synonymous polymorphism c.84G>A, p.Glu28Glu. The significance of this gene arrangement and the putative functionality of this SNP are yet to be tested.
The human tyrosinase gene (TYR
, MIM#606933) is another melanosomal membrane-bound enzyme that catalyzes the first two rate-limiting steps in the melanin biosynthesis pathway 
. Mutations in the TYR
gene cause the most severe form of oculocutaneous albinism OCA1 (TYR
, MIM 606933, OCA1, MIM 203100) 
. MM associated SNP rs17793678 is located in intron 1 and was selected as the tag for one of the five most frequent haplotypes formed at the promoter LD block in this gene (19 kb). Interestingly, TYR
has previously been associated with melanoma in GWAS studies 
performed in Caucasian populations other than our South European sample.
Finally, the ADAMTS20 protein (A d
etalloprotease domain with t
pondin type-1 motifs, MIM#611681) is a member of a family of secreted metalloproteases that can process a variety of extracellular matrix components and secreted molecules. Adamts20
mutations in belted
) mice cause white spotting of the dorsal and ventral torso, indicative of defective neural crest-derived melanoblast development and thus implicating metalloproteases in skin pigmentation 
. A role for Adamts20
in melanoblast survival has been described 
. In our study, SNP rs1510521 was selected as the tag of the most frequent haplotype of the large LD block at the 3′ end of the gene (107 kb), containing the majority of the gene region with the exception of the first three exons. Additional studies are needed in order to clarify the role of ADAMTS20
in MM susceptibility.
genes have already been associated with MM 
, it seemed biologically plausible that genetic interactions would be detected between protective SLC45A2
variants and, similarly, between risks variants within TYR
. Indeed, both effects were observed when summing up rare alleles as it has been described for other interactions among pigmentation genes 
. On one hand, a great reduction of risk was detected when rare alleles at SLC45A2
genes were combined (OR
0.0001). On the other hand, an increased risk, although marginally significant, appeared with the combination of rare alleles at TYR
0.088) (See ).
We were not able to clearly detect any other gene association with MM in our study. However, several candidate genes that may be of interest to other pigmentation and/or MM studies are present in the list of the 30 SNPs with p<0.05. Unfortunately, some genes already known to be associated with pigmentation were not selected due to reasons such as large gene size, lack of suitable tag SNPs within the gene, genes lying in high copy number regions, etc. We acknowledge the fact that our sample set comes from different hospitals distributed throughout different Spanish provinces. However, a recent work by Laayouni and cols. using a 300K-SNP Illumina array demonstrates lack of genetic heterogeneity within different Spanish regions including the Basque country 
In summary, we conducted a two-stage MM susceptibility case-control study (candidate SNP selection and replication). First, we screened 363 tag-SNPs in 65 pigmentation-related gene regions in a large Spanish case-control series. Then we validated observed associations in an independent Spanish series, making this the largest sporadic MM susceptibility study in the Spanish population up to now. One SNP (rs35414) on an intronic region of the SLC45A2 gene was associated with MM risk after adjustment for multiple testing in phase I and II, indicating that common variation in SLC45A2 is associated with protection from the disease. Nevertheless, further studies are welcomed to add weight to our conclusions. We also show that carrying this protective variant decreases the risk of developing MM in bearers of two or more mutations in MC1R, a well-known low penetrance MM predisposing gene. Moreover, three additional variants in TYR, SILV/CDK2 and ADAMTS20 (rs17793678, rs2069398 and rs1510521 respectively) had an overall p-value<0.05, emphasizing the importance of pigmentation genes on MM risk.