Germline mutations in the p16 and CDK4 genes have been reported in a subset of melanoma pedigrees, but their prevalence is not well known. We searched for such germline mutations in 48 French melanoma-prone families selected according to two major criteria: families with at least three affected members (n = 20) or families with two affected members, one of them affected before the age of 50 (n = 28), and one additional minor criterion. Sixteen different p16 germline mutations were found in 21 families, while one germline mutation, Arg24His, was detected in the CDK4 gene. The frequency of p16 gene mutation in our sample (44%) is among the highest rates yet reported and the CDK4mutation is the second mutation detected in this gene worldwide. In summary, our results show frequent involvement of the p16 gene in familial melanoma and confirm the role of the CDK4 gene as a melanoma-predisposing gene.
Cutaneous malignant melanoma (CMM) incidence is increasing dramatically all over the world. An estimated 40 300 new cases occur each year in the USA, resulting in 7300 deaths (1). In France the estimated annual incidence rate of CMM is between 4/100 000 and 8/100 000 and the number of annual related deaths has been increasing (2). The prognosis of this tumor is highly dependent on the time of diagnosis. Besides known environmental risk factors, such as exposure to sunlight, phenotypes which are likely to be genetically controled, such as skin type, pigmentation, total number of nevi and presence of atypical nevi (AN), play an important role in the etiology of CMM.
High risk families with several affected members were identified in the early 1950s (3). Familial melanoma comprises between 8 and 12% of all melanoma cases (4). Patients with familial melanoma often have red or blond hair and/or phototype I–II (i.e. a propensity to sunburn and inability to tan). Furthermore, they frequently develop multiple primary melanomas, present clinically atypical moles (AN) and are younger at diagnosis than sporadic cases (4–6).
Several important findings in melanoma genetics have led to the mapping of a susceptibility locus at 9p21. Molecular studies showed non-random somatic losses of heterozygosity in sporadic melanoma at several markers on 9p21(7). Cytogenetic studies performed on a patient with eight primary melanomas showed a de novo constitutional rearrangement involving chromosomes 5p and 9p with deletion of 9p21 markers (8). Finally, linkage was found between locus 9p21 and familial melanoma in large pedigrees (9–11). At the 9p21 locus the p16 gene (also called CDK4I, CDKN2 or MTS1) was identified as a gene encoding a protein of Mr 16 000 (p16INK4a) able to bind the cyclin-dependent kinase CDK4 (12). This gene was suspected to be a tumor suppressor gene as it was inactivated in a wide variety of human sporadic primary tumors and cell lines, including melanoma (13,14; reviewed in 15). Indeed, several arguments support the hypothesis that the p16 gene is a tumor suppressor gene. The p16INK4a protein inhibits phosphorylation of retinoblastoma protein (pRB) by the G1 cyclin-dependent kinases CDK4 and CDK6, thereby negatively regulating progression through G1 into S phase of the cell cycle (12,16). Wild-type P16 arrests normal diploid cells in late G1 phase, whereas a tumor-associated mutant of P16 does not (17,18). In addition, mice nullizygous for the INK4a locus (elimination of both p16INK4a and p19ARF) rapidly develop spontaneous tumors whose time of onset is accelerated by carcinogens (19). Moreover, embryo fibroblasts derived from such animals fail to senesce in culture and can be transformed by oncogenic ras alone, suggesting a role for INK4a in cell immortalization (19).
Finally, p16 is the first identified melanoma-predisposing gene. Germline mutations segregate with the phenotype in melanoma kindreds (20,21). So far ∼230 families have been reported in the literature and p16 involvement varies from 0% (0/6) of familial cases in the UK (22) to 50% (9/18) of familial cases in the USA (23). A comparison between studies in order to evaluate the frequency of p16 germline mutations in melanoma families in the world is very difficult because of the high variation in kindred selection criteria (ranging from families with two cases of melanoma to 9p21-linked families) and in the size of the samples studied (from two to 64 families). In addition to melanoma, p16 germline mutation seems to predispose to cancers at other sites, mainly pancreatic cancers (24,25) and possibly head and neck squamous cell carcinoma (26).
A second susceptibility gene, CDK4, was also recently characterized. A unique mutation, Arg24Cys located in exon 2 of the gene, was found to segregate in two of 31 families (6%) in which no p16 germline mutation could be detected (21). The resulting mutated protein, which was first described as a tumor-specific antigen in a human sporadic melanoma, specifically prevents binding of CDK4 protein to P16 (27). The CDK4 gene does not seem to be frequently involved in familial melanoma, as in a recent article no CDK4 germline mutations were found among 38 families studied (28).
In order to determine the respective involvements of p16 and CDK4 in melanoma-prone families we conducted a mutational analysis in 48 families who attended several dermatology departments in France.
Among the 48 selected families 20 had at least three melanoma-affected members and 28 had two melanoma-affected relatives, one of them being diagnosed before the age of 50 years. Twenty one families included at least one member with multiple primary melanomas, nine families included at least one case of pancreatic cancer and some families had several of these criteria (see Table 1). In the present sample all affected members had CMM except one case diagnosed with choroid melanoma. Given the small number of affected family members, a linkage analysis was not performed.
Probands were initially screened for germline mutation of the p16 gene. The entire coding sequence including adjacent splice junctions was amplified in four parts corresponding to the three exons of the gene. SSCP analysis revealed abnormal migrating bands in 23 patients. DNA sequence analysis revealed 16 different mutations in the coding sequence in 21 patient samples and an identical mutation in the non-coding sequence in two patient samples. Among the 16 mutations five were located in exon 1 and 11 in exon 2 of the gene. Fourteen were single missense mutations, one was a double substitution and one a duplication of a 6 bp motif. All these germline mutations were located within the 9–131 region of P16 protein, which plays an important role in inhibition of cyclin D1/CDK4 kinase activity in vitro and is correlated with the ability to bind CDK4 (29). In two patients we also detected a single substitution in the non-coding sequence of exon 1 (5′-UTR), 33 bp upstream of the ATG translation initiation codon.
The pathogenicity of each mutation was assessed by studying segregation of the mutation in families. However, this was not always feasible because of too small a number of living cases in families and therefore several additional criteria were used (see Table 2). These criteria included reference to previously published germline or somatic mutations, location of the mutation in one of the four ankyrin repeats of the protein (29), location in a codon which is conserved between murine and human p16INK4a protein sequences (30) and location in a codon which appears to be essential for protein activity (18,31–35). Moreover, SSCP analysis of PCR products from 100 unrelated individuals was also carried out and no shift was detected, therefore ruling out that these mutations might be frequent polymorphisms.
Among the 16 germline mutations located in the p16 gene coding sequence five mutations detected in 10 patients have previously been characterized in melanoma-prone families: Arg24Pro, detected in two apparently unrelated families, was found to segregate in one Australian family (36); Gly35Ala has also been detected in an Australian family (37); Met53Ile has been described in two unrelated Australian families (37) and more recently in an American family (28); Gly101Trp seems to be a mutational hot spot, since it has been reported in several families (20,23–25) and was detected in five unrelated families in our study; Val126Asp was reported in four families (20,23,25). These latter two mutations, Gly101Trp and Val126Asp, in addition to their association with familial melanoma, were found to be temperature sensitive for binding to CDK4 and CDK6 in vitro and for inhibition of cyclin D1-CDK4 in a reconstituted pRb kinase assay (35). Co-segregation of Met53Ile and Gly101Trp germline mutations with melanoma were studied in three families and these mutations were found in all melanoma cases.
Six other mutations were previously described as somatic mutations in various cancers. The Leu16Pro mutation is located in the first ankyrin repeat at an amino acid conserved between human and mouse p16 gene sequences. It was reported in a biliary tract carcinoma (38) and is present in the two melanoma patients of one affected family. The Gly23Asp mutation is also located in the first ankyrin repeat at an amino acid conserved between human and mouse p16 gene sequences and has been previously detected in a pancreatic carcinoma (39). This mutation was present in one out of three melanoma cases and in three of six first degree unaffected relatives. Therefore, in the absence of functional studies it is not possible at present to conclude that Gly23Asp is a disease-related mutation. Another mutation is caused by a 6 bp insertion at nt 57, which is an in-frame duplication of codons 18 and 19. The same mutation has been described in a pancreatic carcinoma cell line (40). The Ala57Val mutation is located in the second ankyrin repeat and has previously been described in a case of acute lymphoblastic leukemia (41). The Val59Gly mutation has been described in a sporadic melanoma (42) and is located in a conserved codon of the second ankyrin repeat. The Arg99Pro mutation has been characterized in a melanoma cell line (43).
We have described five new mutations: Ser56Ile, Leu62Pro, Ala68Leu, Asn71Lys and Leu97Arg. All of them are located in the second exon of the gene, within the second and third ankyrin repeats of the protein. Leu62Pro, Ala68Leu and Asn71Lys are present in at least three melanoma-affected relatives in the families (see Table 1) and can therefore be considered as disease-related mutations. Ser56Ile and Leu97Arg are present in two melanoma-affected first degree relatives, therefore no conclusion can be made about their relation with melanoma.
Finally, a transversion (G→C) was detected at nucleotide position −33 of the non-coding cDNA sequence, 5′ upstream of the ATG translation initiation codon, in two unrelated families. This suggests the existence of a mutation affecting the regulatory region, i.e. the promoter. Analysis of co-segregation of this mutation with disease showed its presence in all three melanoma cases in one family, but only in one melanoma patient out of three in the second family. In order to detect a putative transcription defect lymphoblastoid RT-PCR analysis was performed in one patient from the first family. This patient carried a well-known polymorphism, Ala148Thr in exon 2. Therefore, transcripts encoded by each allele could be distinguished by a PCR digestion assay with SacII. In this assay transcripts of both alleles were detected at the same level (data not shown). This result was confirmed by sequencing exon 2 of the p16 gene on genomic DNA and cDNA extracted from a lymphoblastoid cell line established from this patient. Sequences showed heterozygous G/A at nt 442 (Ala148Thr) in the cDNA, indicating that both alleles were equally transcribed (data not shown). Taken together, these results exclude a deleterious transcriptional effect of the –33 G→ C substitution. This mutation is possibly a rare variant (this G→C mutation at –33 was not detected in 100 unrelated controls).
A unique disease-related germline mutation, Arg24Cys, located in exon 2 of the CDK4 gene was previously described in two unrelated melanoma families who do not carry a p16 germline mutation (21). In order to detect mutations in exon 2 of the CDK4 gene SSCP analysis was performed in the 48 probands. An abnormal migrating band was detected in one sample, corresponding to a missense substitution at codon 24, changing an Arg to a His (Fig. 1). Analysis of the corresponding family led to detection of the same mutation in two melanoma-affected patients, a third being an obligate carrier (patient II-4 in Fig. 1). Furthermore, SSCP-based DNA analysis of 100 unrelated controls failed to find the same shift. These results, taken together with the fact that this mutation leads to a different amino acid substitution at the same Arg 24 codon, which is directly involved in binding to P16 and P15 proteins (27), suggest that Arg24His is likely to be a disease-related mutation.
Association of germline mutations with clinical characteristics
When screening p16 and exon 2 of CDK4 for germline mutations without prior family selection based on linkage analysis we detected germline mutations in 22 of 48 cases (46%) (Table 1). We found germline mutations more frequently in families with at least three affected members (criterion 2) than in families with only two melanoma cases, one of them diagnosed before the age of 50 (criterion 1), the molecular diagnostic efficiency being 70 (14 of 20 cases) and 28% (8 of 28 cases) respectively. The frequency of germline mutation detection in families with two first degree relatives (both criteria 1 and 5) is low (1 of 13, 7%). Conversely, we found a high frequency of p16 germline mutations in families including at least one relative with multiple melanomas (criterion 3): in 17 of 21 (81%) families with criteria 1 or 2 and 3 and in 10 of 10 (100%) families with criteria 2 and 3.
Among the nine families initially selected because of a case of pancreatic cancer in a first or second degree relative of the proband (criterion 4) four had a p16 mutation, while five had not. However, it is important to note that all four families with a p16 mutation also included a patient with primary multiple melanomas.
Among the 22 melanoma probands with AN 12 (55%) had a germline mutation, whereas 10 (45%) had none (data not shown). Among the 25 probands without AN nine (38%) had a germline mutation and 16 (62%) had none. Therefore, the distribution of p16 germline mutations is not significantly different in melanoma cases with and without AN (P = 0.27).
We performed a mutation analysis of 48 melanoma-prone families, ascertained through well-defined clinical criteria and without prior linkage analysis. We characterized 16 p16 germline mutations present in 21 of 48 (44%) families. This frequency of p16 germline mutations in 48 French families (44%) is in marked contrast to the low frequencies reported by other investigators. Their studies were comparable in size of the population-based cohort: 0% (0 of 38) p16 germline mutations were found in American and Australian families (20), 7.8% (5 of 64) in Swedish melanoma families (44) and 17.8% (5 of 28) in American families (28). This discrepancy may have two explanations. Firstly, the proportion of melanoma kindreds with 9p linkage may differ widely between populations. Secondly, our selection criteria are more stringent than those used in other population-based studies and may decrease the probability that melanoma families may in fact be a cluster of sporadic cases.
Among the 16 different p16 germline mutations detected in this study five (Arg24Pro, Gly35Ala, Met53Ile, Gly101Trp and Val126Asp) had previously been reported in familial melanoma (20,23,28,36,37). Among these, Arg24Pro and Gly101Trp mutations were detected in two and five apparently unrelated families respectively. Analysis of 9p21 microsatellite markers is needed in order to know if families carrying the same germline mutation share a common haplotype suggestive of a founder effect, as shown in Swedish and Dutch families (45,46), or if these are hot spot mutations. Six other p16 germline mutations (Leu16Pro, ins619–20, Gly23Asp, Ala57Val, Val59Gly and Arg99Pro) have been published as somatic mutations in various sporadic cancers or cell lines. Five mutations (Ser56Ile, Leu62Pro, Ala68Leu, Asn71Lys and Leu97Arg) are newly described. All of them were shown to segregate with the malignant disease phenotype. Our finding of a majority of missense mutations raises the difficult issue of discriminating between deleterious germline mutation and polymorphism. In the absence of functional tests, such as studying the ability of mutant P16 proteins to bind in vitro to CDK4 protein (47), we can only make presumptions based on concordant observations.
We also detected one novel CDK4 germline mutation, present in 1 of 48 (2%) families. Despite the low frequency of CDK4 germline mutations in the French population-based cohort studied here, this result is important, since it is the second worldwide CDK4 mutation to be detected.
We found no p16 or CDK4 germline mutations in 54% of families. In the absence of linkage data four hypotheses may explain such an observation. Firstly, other susceptibility genes may exist. Linkage to 1p36 has been reported in North American melanoma kindreds (48). The 1p locus appears to contribute more to combined trait CMM/AN families than to CMM families alone (49). The presence of atypical moles is considered as a marker for increased risk of developing cutaneous malignant melanoma. In our study we detected p16 germline mutations in a similar proportion of melanoma cases with (55%) and without AN (45%). Our results are consistent with the reporting of several AN cases with melanoma but without the 19 bp deletion of the p16 gene transmitted throughout a family (50). They suggest that AN phenotype is not directly related to p16 mutation and that another gene(s) may be involved in this trait. Further analyses are needed to investigate the relationship between AN and melanoma and the underlying genetic susceptibilities.
Secondly, another susceptibility gene located in the 9p21 locus may be involved, since no mutation was found in a few families linked to the 9p21 locus (20,23,37,51). The p15 gene (MTS2), which also encodes a CDK inhibitor, is at this locus. At present no germline mutations have been detected in 9p21-linked melanoma kindreds, therefore it is unlikely that p15 could be a melanoma susceptibility gene (51,52). p19ARF is another candidate gene which has recently been characterized. Its ectopic expression has been shown to induce G1 and G2 cell cycle arrest (53). P19ARF protein, encoded by the INK4a locus, is generated by an alternative exon (exon 1β) and an alternative reading frame of exon 2 of the p16 gene (53). Nevertheless, its involvement in familial melanoma is not obvious, since no abnormalities in the p19ARF gene have yet been detected by mutational analysis in 38 melanoma-prone families (28,51). Furthermore, mutations of exon 2 of p19ARF have been shown to result in cell cycle arrest, whereas corresponding mutations within p16 are associated with a complete loss of its activity (19). The hypothesis of another gene mapping to the same region has recently been emphasized by the discovery, in sporadic melanoma, of deletions of 9p21 locus markers without anomalies of the p16 gene (54).
Thirdly, 9p21-linked melanoma families may carry undetected p16 gene germline mutations, such as those altering gene transcription or transcript stability/translation efficiency.
Finally, the existence of a possible chromosome 9p imprinter gene has recently been discussed; methylation of the p16 gene 5′-CpG island led to transcriptional silencing of the gene in 11 of 42 primary tumors (non-small cell lung cancer, head and neck squamous cell carcinoma and glioma) (55,56). Whether such an imprinter gene may predispose to familial melanoma is currently unknown.
In this study we found a p16 germline mutation in 81% (17 of 21) of families having one member with multiple primary melanomas and in 100% (10 of 10) of families with at least three melanoma cases, one of them having multiple primary melanomas. These results suggest that every melanoma-prone family including a patient with multiple primary melanomas should be screened for p16 germline mutation.
Several observations suggest that p16 is a tumor suppressor gene predisposing to melanoma and also to pancreatic carcinoma. Epidemiological studies show that pancreatic cancer is over-represented in melanoma kindreds (57) and recent publications have shown an increased risk of this tumor in melanoma-prone families with p16 germline mutations (24,25). In addition, a pancreatic cancer developed by one carrier of a 19 bp germline deletion in the p16 gene showed loss of the wild-type allele (50). In the present study four of nine families including pancreatic cancer-affected members harbored a p16 germline mutation. However, given the small size of our sample we cannot conclude that an increased risk of pancreatic cancer is associated with p16 germline mutations. In addition, all four families with a case of pancreatic cancer and a p16 germline mutation also included patients with primary multiple melanomas.
Genetic susceptibility to melanoma can now be assessed in some familial cases. However, in most cases germline mutations are point mutations whose consequence on P16 protein biological function is unknown. Therefore, functional studies are needed in order to verify P16 protein loss of function associated with germline mutations. Once done, these findings should enable physicians to identify those subjects in melanoma families who are at high risk of melanoma and require careful surveillance.
Materials and Methods
Family and patient selection
Fifty one families were reviewed from the Institut Gustave Roussy (IGR) familial melanoma directory and selected upon the following two criteria: (1) families with two affected members, one of them being affected before the age of 50 or (2) families with at least three affected members. A minimum of one of the following three subcriteria was added: (3) one affected member with multiple primary melanomas, (4) a case of pancreatic cancer in a first or second degree relative or (5) two affected first degree relatives. Pathology reports were obtained for at least two affected members of each family. In order to confirm the melanoma diagnosis, histopathological samples were reviewed at the IGR for all initially tested patients, and for all affected members in half of the families. Three cases were reclassified as dysplastic nevi instead of melanoma and the three corresponding families were therefore excluded.
All families were of Caucasian origin and living in France, with the exception of one family living in Italy.
For each patient included in this study clinical information was obtained from medical records: age at diagnosis of melanoma, tumor location, histopathological classification and Breslow thickness, diagnosis of multiple primary melanomas or other cancer and presence of clinically atypical nevus syndrome or AN (as defined by at least 50 nevi >2 mm in diameter and including at least one atypical nevus).
Selection of affected family members for screening
One affected member from each of the 48 melanoma families (which was designated as the proband) was chosen based on three criteria: (i) availability of genomic DNA extracted from peripheral blood mononuclear cells, (ii) early age of onset of melanoma and (iii) multiple primary melanomas. Criteria (ii) and (iii) were to minimize the chance of screening sporadic melanoma cases within the families.
Each of the three p16 exons were amplified by PCR from genomic DNA using the following primers: exon 1, p16 F1 (5′-GAAGAAAGAGGAGGGGCTG-3′) and p16 R1 (5′-GCGCTACCTGATTCCAAITC-3′); first part of exon 2, p16 F2 (5′-GGGGCTTGTGTGGGGGTCTG-3′) and p16 R2-1 (5′-CAGCACCACCAGCGTGTC-3′); second part of exon 2, p16 F2-2 (5′-GACCCCGCCACTCTCACC-3′) and p16 R2-2 (5′-GTGCTGGAAAATGAATGCTCTG-3′); exon 3, p16 F3 (5′-CGGTAGGGACGGCAAGAGAG-3′) and p16 R3 (5′-CCTGTAGGACCTTCGGTGACTGA-3′). Only one part of exon 2 of the CDK4 gene was analyzed using the published primer sequences CDK4-2AF (5′-GCTGCAGGTCATACCATCCT-3′) and CDK4-2AR (5′-CTCTCACACTCTTGAGGGCC-3′) (21).
Total PCR reaction volume was 20 µl, including 50–100 ng genomic DNA template, 30 pmol each primer, 200 µM dNTP (Pharmacia Biotech), 5% DMSO, 1× PCR buffer, 0.5 U Taq polymerase (PE Applied Biosystem) and 0.1 µl [α-33P]dCTP (Amersham). Final MgCl2 concentration was 1 mM for exon 1, the first part of exon 2 and the second part of exon 2 and 1.25 mM for exon 3 of the p16 gene and exon 2A of the CDK4 gene. PCR conditions for exons 1 and 3 of p16 and exon 2 of CDK4 were as follows: one initial denaturing step at 95°C for 5 min; 35 cycles of 95°C (30 s), 55°C (30 s) and 72°C (30 s); a final extension step of 10 min at 72°C. Exon 2 was amplified using the same conditions but with an annealing temperature of 60°C. For SSCP PCR products were diluted as previously described (58) and were electrophoresed on two Hydrolink MDE gels (FMC Bioproducts), with either 8% glycerol at room temperature or without glycerol at +4°C. Gels were run at 8 W, either for 14 h at room temperature or for 12 h at 4°C, dried and autoradiographed.
DNA and RNA sequencing
Exons were sequenced (both strands) by either manual or automated methods using the PCR primers. Products from 50 µl PCR reactions were either sequenced manually with a T7 sequencing kit (Pharmacia Biotech) as previously described (59) or purified using MicroSpin S-400 HR columns (Pharmacia Biotech) and sequenced using either the Dye Terminator Cycle sequencing Ready Reaction kit or the ABI prism™ dRhodamine Terminator Cycle sequencing Ready Reaction kit (PE Applied Biosystem) on an automated sequencer 377 (PE Applied Biosystem).
Codons ate numbered according to the complete p16 coding sequence (60).
Allele-specific expression analysis
Total RNA was isolated and purified with the modified guanidinium phosphate buffer method using RNA-B™ (Bioprobe) from lymphocytes. Two micrograms of total RNA were reverse transcribed using 50 pmol oligo(dT), 25 pmol random hexamers, 20 U RNase inhibitor, 200 µM deoxynucleoside triphosphate (Pharmacia Biotech) and 50 U avian myeloblastosis virus enzyme (PE Applied Biosystem) according to the manufacturer's instructions. PCR was performed with primers p16 E2F (5′-CCGCCACTCTCACCCGAC-3′) and p16 R3. Total PCR reaction volume was 50 µl, including 1.5 µl cDNA template, 7.5 pmol each primer, 200 µM dNTP (Pharmacia), 3 mM MgCl2, 1× PCR buffer, 1.25 U Taq Gold polymerase (PE Applied Biosystem), 5% DMSO. PCR was processed as follows: one initial denaturing step at 95° C for 5 min; 40 cycles of 95°C (30 s), 60°C (30 s) and 72°C (30 s); a final extension step of 10 min at 72°C. Identical PCR conditions were used to amplify genomic DNA with the exception of the primers, which were p16 F2-2 and p16 R2-2.
Restriction enzyme analysis of cDNA (RT-PCR products) and gDNA (PCR products) was performed using SacII (Pharmacia Biotech). Digested products were electrophoresed on a 1% agarose gel and visualized after ethidium bromide staining.
Sequencing of cDNA (RT-PCR products) and gDNA (PCR products) was performed using the ABI prism™ dRhodamine Terminator Cycle sequencing Ready Reaction kit (PE Applied Biosystem) as mentioned above.
We thank the family members, particularly those who funded this study, and Dr Uhart. We also thank Drs Jean Feunteun and Mehmet Ozturk for critical reading of the manuscript and Josyane Le Calvez and Danièle Pham for technical assistance. Nadem Soufir is the recipient of a fellowship from the Ligue Nationale contre le Cancer. This work was supported by grants from the Ligue contre le Cancer des Yvelines et de la Seine Maritime, the French Ministry of Research (ACC-SV2/1A027A), the AP-HP (CRC96011), the PHRC FM94/091, the PHRC IDF94/014 and the Fondation Sanofi Vaincre le Mélanome.
multiple atypical nevi
cutaneous malignant melanoma
single strand conformation polymorphism