Cutaneous malignant melanoma (CMM; MIM #155601) comprises about 1% of all cancer cases in European populations, but its mortality is over 2% because of the propensity of this tumour to metastasise. Surveillance of CMM-prone individuals and populations should decrease the morbidity and mortality associated with this disease.

In approximately 5–12% of all melanoma cases, there is a family history of the disease, defined as having one or more affected first degree relatives.¹ Germline mutations of CDKN2A, located on chromosome 9p21, cosegregate with affected cases in an estimated 20% of melanoma families worldwide (reviewed by Goldstein²). CDKN2A encodes two different tumour suppressor proteins: the CDK4/6 cell cycle inhibitor p16^INK4A from exons 1α, 2, and 3; and the MDM2 inhibitor p14^ARF from exons 1β, 2, and 3. The frequency of discovering a CDKN2A mutation within a family increases with the number of affected relatives, ranging from ∼15% in families with two affected members, to 20–40% in those with three, and 70% in those with four or more (unpublished data from Hogg’s laboratory; Kefford and Mann¹). In addition, mutations in the CDK4 gene^3,4 and polymorphisms in the MC1R gene⁵ have been implicated rarely in melanoma predisposition.

To date, mutations in the two major predisposition genes, CDKN2A and CDK4, account for melanoma susceptibility in less than half of all melanoma families,⁶ and the genetic bases of the disease in the remaining families have not been found. Gillanders et al recently provided strong evidence for another melanoma predisposition locus at chromosome 1p22.⁷ Current research in melanoma genetics is hindered by the allelic and locus heterogeneities, which render whole genome linkage analysis difficult. In addition, the sample sizes required to study the effects of weak or modifier alleles in a genetically heterogeneous population are large.⁸

To address the problem of both allelic and locus heterogeneities, we have focused on populations that have undergone genetic bottlenecks and are genetically isolated. We hypothesise that familial melanoma patients within such populations arose from a limited number of founders and thus display a restricted number of melanoma predisposition alleles. These alleles are in linkage disequilibrium with adjacent markers. By searching for melanoma associated alleles and genotypes, we can isolate the predisposition loci and subsequently determine the variant or variants responsible.

Ashkenazi Jews originated in the Near East, and subsequently migrated to eastern and central Europe. Throughout history, the Ashkenazi Jewish population has experienced numerous bottlenecks⁹ which led to highly altered allele frequencies relative to the diverse ancestral population. This genetic drift is evident from the high incidences of multiple genetic disorders, attributable to founder mutations,^10,11 and the population bottlenecks and the practice of endogamy have also decreased genetic variability.^9,12

The incidence of melanoma in Ashkenazi Jews is relatively high (table 1), but no CDKN2A mutations have been found in any Ashkenazi Jewish families.¹³ This population is therefore considered to be a promising setting for the identification of novel melanoma susceptibility genes.⁶ In this report we present the results of our search for CDKN2A mutations and polymorphisms, as well as 9p21 and 1p22 alleles and genotypes that are associated with melanoma, in our collection of Ashkenazi Jewish melanoma patients.

METHODS

Recruitment of patients and controls

Fifty eight Ashkenazi Jewish patients with malignant melanoma (table 2; fig 1) were recruited to this study as described by Yakobson et al.¹³ Twenty of these reported a family history of the disease, defined as having at least one first degree relative affected by melanoma. All tumours were confirmed to be malignant melanoma by hospital pathologists. Genomic DNA was extracted from peripheral blood of all 58 patients.

The 86 control DNA samples were a kind gift from Dr Steve Narod (Toronto). They were obtained from individuals of Ashkenazi Jewish background who had no personal history of cancer, but may have had a family history of breast or ovarian cancer.

Germline CDKN2A and CDK4 sequencing

For all samples, we sequenced exons 1α and 2 of CDKN2A as described by Liu et al,¹⁴ and exon 3 as described in Loo et al.¹⁵ We also amplified exon 1β using primers p14E1b-f and p14E1b-r (table 3), and sequenced using the same primers. We amplified exon 3 of CDKN2A using primers p16E3TMF and p16E3TMR (table 3) and the following amplification profile: 95°C for three minutes; 38 cycles of 95°C for 21 seconds, 66°C for 16 seconds, 72°C for 51 seconds; and 72°C for 22 minutes.

To screen for the two known melanoma associated mutations in CDK4^3,4 in patient samples, we amplified exon 2 of the gene using primers CDK4E2f and CDK4E2r (table 3), and sequenced the polymerase chain reaction (PCR) products using the same primers.

Genotype analysis at the 9p21 and the 1p22 chromosomal regions

To determine whether the chromosomal locus 9p21 and 1p22 are associated with melanoma, we carried out genotype analysis to search for alleles shared as a result of founder effect. The microsatellite markers used were D9S1749, 974, 942, 1748, 1604, and 171 for 9p21; and D1S435, 2779, 236, 2664, and 2726 for 1p22 (fig 2). We obtained primer sequences for all markers from the Genome Database (http://www.gdb.org/). We added the sequence GTTTCT to the 5′ of the forward primer of D9S1749 according to the principles of the “PIGtailing” technique.¹⁶ This increased the clarity of the genotypes by ensuring that non-templated adenylation by Taq polymerase occurs to 100% of labelled PCR products. We carried out PCR (annealing at 58°C) and genotyping as described by Loo et al.¹⁵

Statistical analyses

For each marker, alleles were analysed if their frequencies were more than 10% in either patients or controls, and all other alleles were grouped together. We then compared the allele frequencies in the patients v the controls for each marker. In examinations of genotypic associations, genotypes made up of alleles with frequencies of more than 10% in either patients or controls were analysed, and all other genotypes were grouped together. We then compared the genotype frequencies in the patients v the controls for each marker.

Genotype data were assessed for deviations from the Hardy–Weinberg Equilibrium separately for patient and control groups using the HWE program from the URL: http://linkage.rockefeller.edu/ott/linkutil.htm#HWE. To test whether specific alleles or genotypes were associated with the disease, the genotype data were analysed using the χ² test (EpiInfo6 from http://www.cdc.gov).

RESULTS

Table 2 presents the descriptive statistics for the patient and control groups in the present study. The statistically significant difference seen between the sex distributions in the patient and control groups is probably because the control group consisted primarily of individuals seeking genetic testing for breast cancer susceptibility, a disease that predominantly affects women. As melanoma predisposition is not inherited in a sex restricted manner, the difference in sex distribution is unlikely to affect subsequent analyses. Twenty of the 58 patients (34%, including four patients from two families) reported a family history, with clustering patterns consistent with an autosomal dominant mode of inheritance (fig 1). The remaining 66% of the patients, though reported as sporadic cases, may still possess a genetic predisposition. Their “family history” of the disease may be unknown owing to dispersion of relatives, or is simply not observed if the penetrance is low and the family size is small.

We searched for melanoma associated mutations in the CDKN2A and the CDK4 genes, but did not find any such variants in our patient population. We did not find a statistically significant difference (data not shown) between the frequencies of the C500G (rs11515) and C540T (rs3088440) polymorphisms, previously shown to be associated with melanoma.^17,18

As the frequency of CDKN2A coding mutations does not account for all 9p21 linked families in other populations studied, we sought for a difference in allele and genotype frequencies between patients and controls, which may be indicative of candidate non-coding variants present in linkage disequilibrium with genetic markers in 9p21. The allele and genotype distributions for all six markers (fig 2) did not deviate from the HWE in either the patient or the control groups. Table 4 shows 9p21 allelic and genotypic frequency comparisons between patient and control groups, along with the calculated heterozygosities. Of the markers where the analyses were available, none showed allelic or genotypic associations with melanoma.

Allelic and genotypic associations at the 1p22 locus

Gillanders et al⁷ provided strong evidence that a novel melanoma susceptibility locus is present at 1p22. We therefore searched for associations between 1p22 and melanoma cases in the Ashkenazi Jewish population. We undertook the same types of analyses for 1p22 as we did for 9p21, but found no significant allelic or genotypic associations with melanoma (table 5).

DISCUSSION

Multiple lines of evidence support a genetic basis for melanoma in Ashkenazi Jews. First, the incidence of cutaneous malignant melanoma in the Jewish population differs from that found in non-Jewish populations (table 1). Although differences in incidence rates among ethnic groups may reflect different environments and habits, they can also suggest that genetic backgrounds play a significant role in melanoma predisposition. Second, familial clustering suggests the segregation of mutant predisposition alleles. In our patient group, 20 of 58 patients (34%; including four patients from two families) reported a family history of the disease, with some having multiple affected relatives (fig 1). Third, some individuals—such as the four (7%) in our series—are afflicted with multiple primary melanomas, or develop melanoma at an early age, observations that are also consistent with an underlying genetic predisposition.¹⁹ The early age of diagnosis of melanoma for some of the patients (for example, two patients diagnosed at age 22 in fig 1) possibly reflects accelerated tumour formation kinetics.

Despite the evidence pointing to the cosegregation of genetic variants with melanoma predisposition in the Ashkenazi Jewish population, no such variants have so far been uncovered. In an earlier study, Yakobson et al¹³ found two CDKN2A mutations, Gly101Trp and Gly122Val, in two Israeli families, but neither is of Ashkenazi Jewish origin. Later, Yakobson et al²⁰ uncovered a Val59Gly mutation in four melanoma pedigrees. The concordance of the microsatellites at this chromosomal region shows that the mutation in these four families is probably derived from one Mediterranean non-Ashkenazi founder. The genetic basis for melanoma susceptibility in the Ashkenazi Jewish population thus remains unknown and is the focus of the present study.

Analyses of mtDNA haplogroups¹⁸ and the non-recombining regions of the Y chromosome¹² have shown that a major genetic drift has occurred in the history of the Ashkenazi Jewish population. Our strategy of taking advantage of the linkage disequilibrium between markers that resulted from recent founder effects should have enabled us to define genetic associations with melanoma at the 9p21 and 1p22 candidate loci. However, we were unable to find any allelic or genotypic associations between either of the two single nucleotide polymorphisms (exon 3 of CDKN2A) or the 11 microsatellite markers (9p21 and 1p22) and melanoma. There are several possible reasons for our findings; for example, it is possible that even within the Ashkenazi Jewish population, numerous melanoma predisposition variants exist, and our sample size did not allow us to find a definitive association that is statistically significant; furthermore, the melanoma predisposition alleles for melanoma are not located in the 9p21 or 1p22 chromosomal regions for the Ashkenazi population. These two possibilities are not mutually exclusive.

Investigation by Catalin et al (Catalin M, personal communication) on 24 Ashkenazi Jewish families with malignant melanoma and neural system tumour syndrome (MMNST) has also found no predisposing mutations in CDKN2A. These results, together with ours (82 families in all) show that CDKN2A is very unlikely to be the melanoma susceptibility locus for the Ashkenazi Jewish population. Further investigation should uncover a novel locus (or multiple loci) or environmental causes that underlie melanoma susceptibility of Ashkenazi Jews. The knowledge will allow the identification and surveillance of at-risk families, and thus decrease the morbidity and mortality associated with this disease.