In the last 5 yrs, genome-wide association studies (GWAS) have emerged as an effective approach to conclusively identify regions of the genome that harbor common genetic variants associated with risk of specific cancers. Over 200 regions have been reported that are associated with risk of two dozen different cancers (Chung and Chanock, 2011). Nearly all of the common single nucleotide polymorphism (SNP) markers associated with cancer appear to have low effect sizes, namely, an estimated odds ratio of <1.5 in the discovery sets with minor allele frequencies (MAF) greater than 10%. So far, with rare exception, the signals map to regulatory regions in known genes or intergenic regions, thus raising the specter of different mechanisms contributing to cancer risk. In only a small fraction, does more than one cancer map to the same region. There are <10 so-called ‘multi-cancer susceptibility regions’, including the intergenic regions of 8q24 and 11q13 but also a region on 5p15.33 (which harbors the telomerase, TERT gene) (Chung and Chanock, 2011).

Before the age of GWAS, linkage and candidate gene analysis were employed to discover highly penetrant mutations of either CDKN2A or CDK4 in <50% of melanoma families ‘loaded’ with multiple affected individuals (Meyle and Guldberg, 2009). Recently, the use of next-generation sequencing technology has re-invigorated the search for rare alleles both in families and in special populations. In a notable proof-of-principle, two parallel groups employed candidate gene and next-generation sequencing approaches to identify an uncommon non-synonymous coding SNP in the MITF gene associated with melanoma risk that results in a subtle alteration of the transcriptional profile of MITF targets (Bertolotto et al., 2011; Yokoyama et al., 2011).

The use of GWAS to identify common susceptibility alleles for melanoma risk has met with success thus far, initially reporting a handful of new loci with small effect sizes. Melanoma GWAS pinpointed regions of the genome that harbor plausible candidate genes on 20q11.22 (ASIP- also associated with red hair), 11q14.3 (TYR- previously associated with pigmentation), two intergenic regions on 10q25.1 and 22q13.1 as well as common variants in two regions known to harbor rare mutations, 9p21.3 (adjacent to MTAP and flanking CDKN2A) and 16q24 (MC1R) (Bishop et al., 2009; Brown et al., 2008; Duffy et al., 2010; Teerlink et al., 2012). Further work has shown that the TYR locus on 11q14.3 is also associated with risk of basal cell carcinoma (Gudbjartsson et al., 2008), but interestingly, an allele on the multi-cancer susceptibility locus of 5p15.33 (TERT-CLPTM1L) appears to confer susceptibility to melanoma but provide protection against basal cell carcinoma (Rafnar et al., 2009; Stacey et al., 2009).

As investigators transition from conducting separate studies to performing meta-analysis and combined scans, the prospects for discovery have improved, as several models have suggested (Park et al., 2010; Yang et al., 2010). Recently, new GWAS reports detailing novel melanoma susceptibility loci are being reported in Nature Genetics and Human Molecular Genetics (Amos et al., 2011; Barrett et al., 2011; Macgregor et al., 2011). With larger sample sizes created by meta-analyses across studies, the investigators have doubled the number of regions conclusively associated with melanoma risk. In one study, a second stage GWAS confirmed prior findings for 9p21.3 (MTAP/CDKN2A), 16q24.3 (MC1R), and 20q11.22 (ASIP) but extended to new discoveries for three new loci on 2q33.1 (CASP8), 11q22.3 (ATM), and 21q22.3 (MX2) (Barrett et al., 2011). A coding variant in CASP8 has previously been associated with breast cancer risk, whereas the second gene is ATM, an established cancer gene (Cox et al., 2007; Goldgar et al., 2011). A second study has revealed a new susceptibility locus on 1q21.3, which harbors several intriguing candidate genes, such as ARNT and SETDB1, each of which has been implicated in other cancers (Macgregor et al., 2011). A third study provided further confirmation of known loci (Amos et al., 2011), but also reported a promising risk association with 15q13.1 (HERC2/OCA2), a region that has been mapped to eye and skin color (Duffy et al., 2010; Sturm et al., 2008); notably, 50% of variability in eye color is associated with rs12913832 (Amos et al., 2011).

Now, that the catalog of common variants associated with melanoma risk has begun to expand, several points are worth considering. First, many of the regions track with hair, eye color, and nevus counts, consistent with epidemiological observations made decades ago. Second, common variants are mapping to genetic regions previously shown to harbor highly penetrant mutations, thus providing strong evidence for differential effects of distinct mutations/SNPs. Third, new regions not associated with pigment, nevus count, or hair color are emerging, pointing toward other mechanisms that could contribute to melanoma risk.

While some hoped that GWAS would provide near-complete explanations for the genetics of complex diseases (like most cancers), it is now clear that GWAS have been successful in discovering a fraction of the genetic contribution to distinct cancers, which may have different architectures underlying the component of genetic susceptibility. Although there is a debate about the search for the missing heritability for complex traits and diseases (Manolio et al., 2009), larger sample sizes are needed to continue to discover additional common alleles (Park et al., 2010; Yang et al., 2010) as well as uncommon and rare alleles.

Large meta-analyses come at a price for discovery and leave many new regions to be identified. Out of necessity, the first generation of melanoma GWAS has had to relax epidemiological rigor with respect to study design, for instance using ‘convenient’ controls as well as combining different types of melanoma, leading to possible misclassification and reduction in power. In this regard, the next generation of GWAS may benefit from incorporating subtypes and better exposure measures into study design. In this regard, it is likely that a number of regions have been missed thus far and worth pursuit in new studies.

With the discovery of new regions harboring susceptibility alleles, a major effort needs to be focused on fine mapping risk loci using the tools of the 1000 Genomes Project, together with regional resequencing to develop a comprehensive catalog of variants per region, many of which should emerge as candidates for subsequent biological studies. Eventually, the biological plausibility of these loci can be only tested in experimental settings (Chung and Chanock, 2011). This becomes even more evident as nearly all GWAS signals have mapped to non-coding regions in the genome.

Lastly, the discoveries by GWAS have derived from the collaboration of large epidemiological consortia and geneticists to scan the genome in search of regions associated with melanoma risk. In a sense, this effort has yielded a new paradigm for biology, one that will require extensive follow-up to understand the underlying basis for the many susceptibility alleles and begin to challenge the field to develop novel approaches toward more integrated analyses. The follow-up has to be focused on the uniqueness of each region – which is opposite from the paradigm established by GWAS, scanning across common SNPs. First, each region should be dissected individually, and the functional analyses will have to account for the specifics of each genomic region. Eventually, sets of directly associated variants will need to be analyzed for interaction. In this regard, it is daunting to invoke the term ‘causal’ for any one of many variants that can have a small effect (and are neither sufficient nor necessary). Still, the promise of GWAS is yet to be realized fully until the basic biology of the discovered regions are thoroughly investigated and then integrated into a more comprehensive analysis that will incorporate environmental exposures and somatic alterations in the cancer genomes.


  1. Top of page
  2. References