Melanoma is one of the most aggressive human malignancies. The mean survival time for metastatic patients is 6 months when standard chemotherapy is used or up to 12 months if used in combination with the BRAF enzyme inhibitor vemurafenib. The incidence of melanoma has been increasing dramatically in the Caucasian population since the 1950s [1].

In Poland the number of affected individuals has increased by 60% over the past 11 years and has reached over 2,600 new cases identified annually. The etiology of melanoma is complex, involving both genetic and environmental factors. Significant heritability of this malignancy was observed in large twin study [2].

Family studies provide strong support for heterogeneous mechanisms involved in a polygenic inherited susceptibility to melanoma that includes both inherited high-risk and lowrisk alleles. To date only one major melanoma susceptibility gene: CDKN2A has been identified, responsible for only up to 20% of melanoma-prone patients [3]. Mutations within other malignant melanoma (MM) high-risk genes, such as CDK4, ACD, CXC, TERT, TERF2IP, BAP1, or POT1 are extremely rare and to date have been identified only in few families worldwide. Because germline mutations in CDKN2A are present in a minority of melanoma patients it is expected that additional genes will be discovered with next generation sequence technology that focuses on whole-exome or wholegenome sequencing. Onset of MM at an early age is suggestive of a hereditary predisposition. However, the heritability of early-onset MM remains unexplained even in a larger proportion of apparent sporadic cases than among familial cases. Thus far in the literature there appears to be a paucity of data from whole exome sequencing among early-onset MM cases.

Given the unique homogenous structure of the Polish population, we undertook an exome screen to determine if there existed one or more novel high/moderate risk gene(s) for MM, which may be associated with specific founder effects. Herein, we performed whole-exome sequencing (WES) on a cohort of 96 unrelated Polish patients with melanoma diagnosed under the age of 40 years. All of whom had previously been screened for the present of causative CDKN2A-variants and found not to harbor any. We studied both recessive and dominant inheritance patterns by analyses of both sporadic and familial early onset MM cases. We evaluated protein truncating DNA alterations only and selected genes for further study based on the relevance of the gene to carcinogenesis.

In the subsequent verification step, we evaluated recurrent truncating mutations identified by WES in large case-control study of 1,200 unselected MM patients and 6,200 control subjects from Poland.

Using GATK algorithm, we detected 2,272,282 indels in the total data set for our cohort of early-onset melanoma patients, after prioritization we identified 1,012 new and 218 previously reported frameshift mutations (523 new variants in sporadic early-onset cases and 489 in familial cases; 137 previously reported variants in sporadic cases and 81 in familial cases). Using SCALPEL algorithm, we detected 2,297 variants (1,151 of them in sporadic early-onset cases, 1,146 in familial cases). These comprised 3′ UTR and 5′ UTR variants (452 and 56, respectively), 228 downstream variants, 88 in frame deletions, 40 intergenic variants, 1,000 intron variants, 11 intron variants and NMD transcript variants, 55 intron variants and non-coding transcript variants, 57 non-coding transcript exon variants and noncoding transcript variants, five regulatory regions variants, three splice acceptor variants, 60 splice region variants and intron variants, two transcription factor variants, one splice region variant and intron variant and noncoding transcript variant, 161 upstream gene variants and 74 frameshift mutations.

The frameshift mutations were selected for further interrogation as they were most likely to reveal new genes of interest with respect to melanoma development.

All the frameshift mutations were filtered by examining the quality control values, minor allele frequencies and intense gene databases/literature data analysis supported by the ToppGene algorithm. We selected 21 potentially deleterious variants in 20 genes, which after confirmation by Sanger sequencing were evaluated in the association study. These included the genes VRK1 (a serine/threonine kinase actively expressed in dividing cells), MYCT1 (proposed to regulate certain MYC target genes), DNAH14 (involved in microtubule motilities), CASC3 (functions in nonsense mediated mRNA decay), MS4A12 (associated with inhibition of cellular proliferation of colon cancer cells), PRC1 (a key regulator of cytokinesis), WWOX (acts as a tumour supressor, expression is associated with apoptosis), CARD6 (encodes a caspase recruitment domain that may be involved in apoptosis), EFCAB6 (encodes a protein which directly binds the oncogene DJ-1 and androgen receptor), EXO5 (involved in DNA repair following Uv-light irradiation), CASC3 (known as cancer susceptibility 3 and involved in nonsensemediated mRNA decay), STK33 (a serine threonine kinase involved in MAP kinase signaling), SAMD11 (transcription coactivator activity and PH domain binding required to maintain the transcriptionally repressive state of many genes), CPNE1 (involved in calcium mediated intra-cellular processes), CABLES1 (a cyclin dependent kinase binding protein that is involved in p53/p73 induced cell death), LEKR1 (structural constituent of ribosome), NUDT17 (most likely mediates the hydrolysis of some nucleoside disphosphate derivatives), CASP8AP2 (plays a regulatory role in Fasmediated apoptosis), RRP 15 (a ribosomal protein subunit), and CREB3L3 (member of the basic-leucine zipper family and the AMP-dependent transcription factor family, linked to acute inflammatory response and hepatocellular carcinoma). Mutations in VRK1, MYCT1, DNAH14, MS4A12, PRC1, CARD6, STK33, EFCAB6, EXO5, and SAMD11 genes were found among early onset cases; CREB3L3, CNDP2, CABLES1, RRP 15, and CASP8AP2 genes in familial cases; CASC3, WWOX, CPNE1, LEKR1, and NUDT17 in both subgroups.

The variants in these genes were further investigated in those families where trios were available to determine if they were recessively or dominantly inherited. The results of this aspect of the study revealed that there were some genetic variants that were homozygous in the proband and in one or other of the parents. Further evaluation revealed that these variants were unlikely to be associated with disease since either the unaffected parents were both homozygous for the allele and the proband heterozygous or the proband was homozygous for the variant and one of the parents were homozygous and the other heterozygous for that specific allele. Genotyping of the parents of homozygous carriers of identified mutations revealed no case of homozygous proband and both parents being the carriers of the mutations. Together, this suggested that those variants were unlikely to be associated with any change in melanoma risk (Table 1).

Additional evidence about the pathogenicity of the identified variants was provided by a replication study where a set of 1,200 melanoma patient DNA samples was compared to a similarly large series of control DNA samples taken from healthy subjects (Table 2).

The results revealed that most of the changes were neutral but although we found no statistically significant differences between cases and controls, 10 genes cannot be excluded as being associated with melanoma risk—due to genetic heterogeneity and the reduced penetrance of this disease. A greater level of confidence in the pathogenicity of the variant was observed if the homozygous state was associated with a greater difference in risk compared to being in the heterozygote state. Unfortunately, many of the variants occurred within the study populations at much lower frequency to be reliably associated with disease risk. Potential candidates included WWOX, VRK1, CREB3L3, EFCAB6, CARD6, EXO5, MS4A12, CASC3, and STK3. From this list of genes, the most promising were VRK1, CREB3L3, EXO5, and STK33 as there was a tendency of decreasing risk of melanoma for heterozygote and homozygote carriers of VRK1 and EXO5 and an increasing risk of disease for CREB3L3 and STK33.

From 2008 to 2013, the approach of finding melanoma susceptibility genes centered on genome-wide association studies (GWAS). The results of several of these studies aimed at identifying susceptibility loci for melanoma have been reported [9-12]. The results of these studies have reported new susceptibility loci for melanoma, but the effect sizes reported from these GWAS have been modest with odds ratios at best being ~1.5. None of the GWAS has thus far led to the adoption of a clinical test for a newly discovered genetic locus.

Low-penetrance variants cannot be categorically associated with familial clusters, for many reasons that include molecular interactions between low/moderate penetrance genes, which could potentially be responsible for early-onset disease.

Recently, WES has been developed which overcomes many of the problems of GWAS [6]. It is estimated that the protein coding regions of the human genome (exons) constitute about 85% of disease-causing mutations [13]. Wholeexome and whole-genome studies have identified numerous genes somatically altered in melanoma tumors and highlighted a higher mutation load in melanoma compared with those in other cancers, such as, TP53, NF1, and ARID2 [14,15]. However, to date only a few new genes/cancer susceptibility alleles in melanoma-prone families have been identified using whole exome or whole genome sequencing. Germline mutations of all of the genes identified to date are found in a very small proportion of families that can not account for the growing frequency of the disease. The list includes the TERT promoter mutation found among German familial melanoma patients [16], a POLE1 missense mutation detected in one family [17], ACD with mutations in six families and four families carrying TERF2IP variants, which included nonsense mutations in both genes) and point mutations that co-segregated with melanoma [18]; finally a RAD51B nonsense mutation was identified in one family [19]. WES also identified a BAP1 splice mutation occurring in a family with uveal MM [20]. Direct Sanger sequencing of this gene revealed germline mutation among early-onset uveal patients [21].

Given the homogeneity of the Polish population we expected to find a limited number of recurrent founder mutations. However, we found large numbers of germline DNA alterations which appeared causative as they were predicted to result in truncated protein products, but none appeared to be novel high-risk melanoma gene in either sporadic familial early-onset patients. We observed known recurrent truncating mutations in genes predisposing to other cancers, such as NOD2 (3020insC), CHEK2 (1100delC, IVS2+1G>A, del5395), and NBS1 (657del5). Their distribution among melanoma patients was not significantly different to that observed amongst healthy controls (data available upon request). Although evaluation of somatic mutations was not the subject of the study, it would be very interesting to compare the results from melanoma skin and blood samples. Increased frequency of somatic mutations could point at melanoma-susceptibility candidate genes.

Taken together the results from this study with respect to the Polish population suggests that there is extensive genetic heterogeneity within melanoma families from Poland that can not be explained by a few recurrent highly penetrant causal variants. Large multi-center studies are needed for further investigation of possible involvement of VRK1, CREB3L3, EXO5, and STK33 genes in melanoma development. Additionally to reveal the genetic association with new mutation with relation with early-onset melanoma long, prospective follow-up study of the families with identified DNA variants is needed.