Article Text

Download PDFPDF

Comprehensive analysis of CDKN2A (p16INK4A/p14ARF) and CDKN2B genes in 53 melanoma index cases considered to be at heightened risk of melanoma
  1. K Laud1,
  2. C Marian1,
  3. M F Avril2,
  4. M Barrois1,
  5. A Chompret3,
  6. A M Goldstein4,
  7. M A Tucker4,
  8. P A Clark5,
  9. G Peters5,
  10. V Chaudru6,
  11. F Demenais6,
  12. A Spatz7,
  13. M W Smith8,
  14. G M Lenoir1,
  15. B Bressac-de Paillerets1,
  16. French Hereditary Melanoma Study Group*
  1. 1Service de Génétique, Institut Gustave Roussy, 94800 Villejuif, France
  2. 2Service de Dermatologie, Institut Gustave Roussy, 94800 Villejuif, France
  3. 3Département de Médecine, Institut Gustave Roussy, 94800 Villejuif, France
  4. 4Genetic Epidemiology Branch, Division of Cancer, Epidemiology and Genetics, National Cancer Institute, NIH, DHHS, Bethesda, MD 20892, USA
  5. 5Cancer Research UK, London Research Institute, Lincoln’s Inn Fields Laboratories, London, WC2A 3PX, UK
  6. 6INSERM-Université d’Evry, EMI0006, 91034 Evry, France
  7. 7Service d’Anatomopathologie, Institut Gustave Roussy, 94800 Villejuif, France
  8. 8Laboratory of Genomic Diversity, Center for Cancer Research and Basic Research Program, SAIC Frederick, National Cancer Institute at Frederick, NIH, DHHS, Frederick, MD 21702, USA
  1. Correspondence to:
 Dr B Bressac-de Paillerets
 Service de Génétique, Institut Gustave Roussy, 39 rue Camille Desmoulins, 94805 Villejuif Cedex, France; bressac{at}


Objective: Comprehensive analysis of the 9p21 locus including the CDKN2A, ARF, and CDKN2B genes in 53 individuals from melanoma index cases considered to be at heightened risk of melanoma.

Methods and Results: Using a combination of DNA sequencing, gene copy number by real time quantitative PCR, linkage analysis, and transcript analysis in haploid somatic cell hybrids, we found no evidence for germline alteration in either coding or non-coding domains of CDKN2A and CDKN2B. However, we identified a p14ARF exon 1β missense germline mutation (G16D) in a melanoma-neural system tumour syndrome (CMM+NST) family and a 8474 bp germline deletion from 196 bp upstream of p14ARF exon 1β initiation codon to 11233 bp upstream of exon 1α of p16INK4A in a family with five melanoma cases. For three out of 10 families with at least three melanoma cases, the disease gene was unlinked to the 9p21 region, while linkage analysis was not fully conclusive for seven families.

Conclusions: These data reinforce the hypothesis that ARF is a melanoma susceptibility gene and suggest that germline deletions specifically affecting p14ARF may not be solely responsible for NST susceptibility. Predisposition to CMM+NST could either be due to complete disruption of the CDKN2A locus or be the result of more complex genetic inheritance. In addition, the absence of any genetic alteration in 50 melanoma prone families or patients suggests the presence of additional tumour suppressor genes possibly in the 9p21 region, and on other chromosomes.

  • CMM, cutaneous malignant melanoma
  • CMM+NST, melanoma-neural system tumour syndrome
  • dHPLC, denaturing high performance liquid chromatography
  • LFS, Li-Fraumeni syndrome
  • MPM, multiple primary melanoma
  • RQ-PCR, real time quantitative PCR
  • ARF mutations
  • genetic predisposition
  • melanoma prone families

Statistics from

Linkage analysis has implicated a gene or genes on human chromosome 9p21 in the inherited predisposition to cutaneous malignant melanoma (CMM).1 To date, three tumour suppressor genes have been identified in this region. CDKN2A and CDKN2B, which presumably arose by tandem duplication, encode structurally similar proteins, p16INK4A and p15INK4b, respectively, that function as inhibitors of the cyclin dependent kinases Cdk4 and Cdk6. However, the CDKN2A locus has the unusual capacity to encode completely distinct proteins from two alternatively spliced transcripts. Whereas the α transcript, comprising exons 1α, 2, and 3, encodes p16INK4A, the smaller β transcript, comprising exons 1β, 2, and 3, specifies a protein designated p14ARF because the exon 2 sequences are translated in an alternative reading frame relative to that used for p16INK4A. Both p16INK4A and p15INK4B are able to cause G1 cell cycle arrest by inhibiting the phosphorylation of the retinoblastoma protein (Rb), while p14ARF can arrest cells in both G1 and G2/M phases via its ability to inhibit MDM2 mediated destruction of the p53 tumour suppressor (reviewed in Sharpless and DePinho2).

Germline mutations of CDKN2A have been found in about 20–40% of families with multiple cases of melanoma and are located in both exon 1α and exon 2.3 In addition to germline mutations that impair the function of p16INK4A, there have been several reports of alterations in non-coding regions of the gene that are clearly associated with melanoma susceptibility.4–8

The status of p14ARF as a tumour suppressor is less clear cut. Although germline mutations in CDKN2A exon 2 have the potential to impair both p16INK4A and p14ARF, a number of studies have shown that removal of the amino acids encoded by exon 2 has no demonstrable effect on the known functions of p14ARF.9,10 Conversely, other studies have suggested that mutations in exon 2 have the capacity to alter the subcellular localisation of p14ARF as well as inactivating p16INK4A.11–13 Moreover, germline alterations affecting p14ARF and possibly p16INK4A were detected in a subset of melanoma-neural system tumour (CMM+NST) families.14–16 Finally, germline mutations restricted to exon 1β have been detected in melanoma prone families or patients.17,18 Although highly suggestive that ARF is also a melanoma susceptibility gene, it has been difficult to obtain unequivocal proof based on functional impairment of p14ARF. Perhaps the strongest evidence comes from mouse models in which p16INK4A and p19ARF genes have been selectively ablated. Mice lacking p16INK4A alone (INK4A−/−; ARF+/+) did not develop melanomas, but when these knockout mice were crossed with mice lacking one copy of ARF (INK4A−/−; ARF +/−) they then had a propensity to develop melanomas.19 The effect of ARF haploinsufficiency in INK4A nullizygotes suggests that the reduced dosage of the ARF is sufficient to contribute to melanoma tumourigenesis in this background.

Interestingly, in one CMM+NST family, a large germline deletion encompassing CDKN2A/ARF and CDKN2B was identified.14CDKN2B is located within about 30 kb centromeric from CDKN2A. Although no specific germline mutations of the CDKN2B gene have yet been reported in familial melanoma kindreds,20–23 somatic point mutations in the CDKN2B gene were described in metastases of a patient affected by a melanoma24 and in a primary melanoma.25

These observations prompted us to undertake a comprehensive analysis of the 9p21 locus and its three genes, CDKN2A, ARF, and CDKN2B, in 53 index cases of melanoma prone families or patients. As well as screening for mutations in exons 1α, 1β, 2, and 3 of the CDKN2A gene, exons 1 and 2 of CDKN2B, and exon 2 of CDK4, we quantified CDKN2A, ARF, and CDKN2B gene copy number by real time quantitative PCR (RQ-PCR).26 We also investigated the 5′UTR region of CDKN2A and screened for the IVS2-105A>G mutation in intron 2 of CDKN2A in all 35 index cases of CMM prone families and patients with multiple primary melanoma (MPM; group A). Finally, we performed linkage analyses in nine families with three melanoma cases (group A), and in six cases showing possible linkage to the 9p21 locus we used long range RT-PCR to search for differentially spliced transcripts that might be indicative of deep intronic mutations.


Index case selection and control groups

The patients in this study were enrolled through the dermatology department of the Institut Gustave Roussy and different oncogenetics or dermatology departments from all over France.

Group A

This group comprised 36 cases of CMM, confirmed by pathological reports, that were considered to have a high probability of being hereditary based upon the following inclusion criteria: (a) families with at least three affected members (n = 15); (b) families with two melanoma cases, one of them being affected by at least two melanoma (n = 7); and (c) patients affected by at least three melanoma (n = 14). The patients tested were index cases in melanoma prone families or MPM patients. These probands had been pre-screened for CDKN2A (exon 1α, 2, and 3) and CDK4 exon 2 coding sequences, by either single strand conformation polymorphism (SSCP) or denaturing high performance liquid chromatography (dHPLC), and selected because found negative for germline mutation. In addition, the proband (FG7617) from an American multiplex melanoma family (family AN), ascertained by NCI, that included five patients with melanoma including three patients with multiple melanoma tumours, was selected for this study because 9p21 genotyping results were suspicious for a germline deletion. Two first cousins from family AN (FG7617 and FG7381) appeared homozygous for markers D9S974, D9S1748, and D9S171. Furthermore, the alleles that appeared homozygous in FG7381 were inherited from her mother (fig 1).

Group B

This group consisted of 12 melanoma cases, confirmed by pathological reports, with at least one second degree relative affected by a tumour from the CMM-NST syndrome (OMIM 155755).

Group C

These were five melanoma cases, confirmed by pathological reports, associated with families with cancer aggregations that do not strictly fit Li-Fraumeni syndrome (LFS) criteria but include paediatric tumours from the LFS spectrum (sarcoma or medulloblastoma).27 The patients selected have no germline mutation in the p53 gene coding sequence.

Control group

A total of 202 DNA samples were prepared from lymphocytes of patients without a familial cancer history and sequenced for CDKN2A/ARF (exon 1α, 1β, 2, and 3) and CDK4 (exon 2).

dHPLC and sequence analyses

Genomic DNA was extracted from peripheral blood lymphocytes using the QIAamp DNA Blood Mini Kit (Qiagen, Valencia, CA) according to the manufacturer’s instructions. We screened for germline mutations in CDKN2B (exon 1 and 2), ARF (exon 1β), and CDKN2A (exon 1α, 2, and 3) and CDK4 (exon 2) by dHPLC analysis, an automated heteroduplex detection method.28–30 PCR amplification was performed in a 20 μl reaction with 100 ng genomic DNA, 1×HotStar Taq DNA polymerase buffer including 1.5 mM MgCl2 (Qiagen), 1 U of HotStar Taq DNA polymerase (Qiagen), and 4 pmol of each primer. For CDKN2A exon 1α, 1.25 M betain (Sigma, St Louis, MO) was added to the PCR reaction mix. Primer sequences and dHPLC conditions are described in appendix A. For each sample, amplification reactions were performed using a touch down protocol: initial denaturation step at 95°C for 10 min; two cycles (30 s at 95°C, 30 s at 66°C, 30 s at 72°C); two cycles (30 s at 95°C, 30 s at 64°C, 30 s at 72°C); two cycles (30 s at 95°C, 30 s at 62°C, 30 s at 72°C); 40 cycles (30 s at 95°C, 30 s at 60°C, 30 s at 72°C). Heteroduplex analyses were carried out as described in Laud et al.31 Samples displaying abnormal profiles were subsequently sequenced on both strands with the Big Dye Terminator sequencing kit (Perkin Elmer Applied Biosystems, Foster City, CA) according to the manufacturer’s instructions.

Screening of the CDKN2A promoter and the IVS2-105A>G deep intronic mutation was only performed on 36 index cases of the melanoma families (group A). We amplified the CDKN2A promoter region by PCR using the conditions described above. In order to screen for the IVS2-105A>G mutation (located in intron 2), the amplification reactions were performed using a touch down protocol with the following profile: initial denaturation step at 97°C for 15 min; six cycles (1 min at 97°C, 30 s at 68°C, 1 min at 72°C); six cycles (1 min at 97°C, 30 s at 66°C, 1 min at 72°C); six cycles (1 min at 97°C, 30 s at 64°C, 1 min at 72°C); 22 cycles (1 min at 97°C, 30 s at 60°C, 1 min at 72°C). PCR products were sequenced with the Big Dye Terminator sequencing kit (Applied Biosystems) according to the manufacturer’s instructions. The primers used are described in appendix A.


The gene copy number of CDKN2B, ARF exon 1β, and CDKN2A exon 1α was estimated by real time quantitative PCR as described by Barrois et al.32 The primers and fluorogenic probes used are described in appendix A. PCR reactions were carried out in a volume of 50 µl, using 25 μl of TaqMan Universal PCR Master Mix (2×) (PE Biosystems, Foster City, CA) for all probes except the ARF exon 1β probe for which we used 5 μl of the TaqMan PCR Core Reagent Kit (10×) (PE Biosystems) complemented with 5% glycerol. Each PCR was performed with 1.25 U of Taq polymerase, 25 ng of DNA, 20 pmol of each primer, and 10 pmol of the fluorogenic probe. Each sample was analysed in triplicate. The thermal cycling conditions were as follows: activation of Taq polymerase at 95°C for 20 min and 40 cycles at 95°C for 15 s and 60°C for 1 min for CDKN2A and ARF and 65°C for CDKN2B. As positive controls, we used the HL60 cell line which is haploid at the 9p21 locus (CDKN2B, ARF, and CDKN2A) and the F615 cell line which is haploid for ARF only (kindly provided by Juliette Moor and Julia Newton-Bishop, CRUK Cancer Medicine Research Unit, St James’s University Hospital, Leeds, UK).

Characterisation of the germline deletion

The germline deletion was identified by long range PCR using peripheral blood lymphocyte DNA from patient no. 7617 and the GeneAmp XLPCR kit (Applied Biosystems). Primers were designed to amplify a large region (22.2 kb) encompassing the putative deletion (appendix A). The deletion breakpoints were then mapped by digesting the long range PCR product with BamH1 and SpeI. Another set of primers was then designed to amplify a shorter fragment (8899 bp) which encompassed ARF exon 1β (appendix A). The different fragments obtained were gel purified with a Qiagen purification kit and sequenced using a Big Dye Terminator sequencing kit.

Linkage analyses and statistical methods

Linkage studies were performed on nine melanoma prone families, displaying at least three melanoma cases, as described by Auroy et al33 using four microsatellite markers on chromosome 9p21 (D9S736, D9S1749 (CDKN2A), D9S942, and D9S1748). The primer sequences are available through the Genome Database ( Linkage analyses were carried out using the LINKMAP program of the LINKAGE package.34 The disease locus was moved across the following fixed map: D9S736 – 0.003 cM – D9S1749 – 0.011 cM – D9S942 – 0.00003 cM – D9S1748. LOD scores were calculated in each of the families studied assuming a dominant model for the disease gene with a disease allele frequency of 0.0001. Reduced penetrances in males and females were assumed according to preliminary analyses (Florence Demenais, unpublished data). A LOD score greater than 3.0 indicates evidence for linkage, while a LOD score less than –2 indicates evidence against linkage. No conclusions can be drawn if the LOD score is between –2 and 3.

Diploid to haploid conversion

Haploid converted clones from six index cases of families possibly linked to 9p21 locus (table 1) were created by GMP Genetics by a technique originally described by Yan et al.35 Hybrid cells were maintained in DMEM with high glucose including 10% FBS, 0.5 mg/ml Geneticin, 1×HAT, and penicillin-streptomycin, according to the manufacturer’s instructions. In order to verify that each haploid converted clone contained only one allele, we compared its haplotype with that of its parental lymphoblastoid cell line using the microsatellite markers D9S1749 and D9S942. The primer sequences are available through the Genome Database ( and PCR conditions are described in Auroy et al.33

RT-PCR analysis of CDKN2A transcripts

Total RNA was extracted from cell lines that were haploid for chromosome 9 using the Tri-Reagent kit (Sigma). To eliminate genomic DNA contamination, RNA samples were treated with DNase in a final volume of 100 μl, under the following conditions: 20 μg total RNA, 40 U RNase inhibitor (Invitrogen, Carlsbad, CA), 40 U RNase-free DNase I (Roche Diagnostics, Meylan, France), 5 mM MgCl2, and 5 mM Tris-HCl pH 7.5. The DNase reaction was performed at 37°C for 1 h, followed by phenol/chloroform extraction and precipitation with 3 M sodium acetate and ethanol.

The reverse transcription reaction was performed using 2 μg of total RNA with 100 U Superscript II reverse transcriptase (Gibco-BRL Life Technologies, Carlsbad, CA) and 50 pmol of random primers in a final volume of 20 μl, for 50 min at 42°C. A 1 μl sample of each reverse transcription reaction product was amplified by long range PCR with the Long Template PCR System kit (Roche Molecular Biochemicals, Mannheim, Germany) and using primers specific for the p16INK4A and p14ARF cDNAs (appendix A). Reaction conditions were as described in the Expand Long Template PCR System protocol, with an annealing temperature of 68°C and system 3.


A comprehensive analysis of the 9p21 locus was performed on our three groups of patients (table 1). The first group (36 cases) comprised CMM kindreds and patients with MPM; the second group consisted of 12 families with CMM+NST; and the third group of five families had features of LFS, including a melanoma case, but without p53 germline mutations.

A germline deletion affecting ARF exon 1β

We investigated whether the 9p21 locus was affected by germline deletions by quantifying the copy number of the CDKN2B, CDKN2A, and ARF genes by real time PCR. In all index cases tested from the three clinical groups, we observed two copies of CDKN2A and CDKN2B (table 1). However, with a probe located close to exon 1β of p14ARF (184 bp) we found only a single copy in two melanoma affected patients (FG7617 and FG7381) from a family with five cases of CMM (fig 1A). As the probe is 19 kb proximal to exon 1α of CDKN2A, these findings suggested the presence of a germline deletion affecting ARF but not CDKN2A or CDKN2B. This was confirmed by haplotype analysis on patients FG7617 and FG7381 with microsatellite markers spanning the 9p21 region. Two markers were potentially homozygous: D9S1748, which is 300 bp distal to ARF exon 1β, and D9S974 which is 7 kb distal to ARF exon 1β and 13 kb proximal to CDKN2A exon 1α (fig 1B).

To characterise the deletion in more detail, we performed long range PCR amplification of genomic DNA from patient FG7617 and a normal control using primers that encompass p14ARF exon 1β and the two microsatellite markers D9S1748 and D9S974. The expected 22 kb product was observed with both DNA samples plus an additional 14 kb band specific for patient F7617 (data not shown). Mapping of this shorter DNA fragment with the restriction enzymes SpeI and BglII gave an indication of the boundaries of the deletion and new primers were designed to locate the exact breakpoints. Sequencing of the resulting PCR product showed that the germline deletion extended for 8474 bp beginning at 196 bp upstream of the initiation codon of p14ARF exon 1β and ending at 11233 bp upstream of p16INK4A exon 1α (fig 1C,D). An identical deletion occurred in patient FG7381.

A germline missense mutation in ARF exon 1β

We did not detect germline mutations in exons 1α, 2, and 3 of CDKN2A and exons 1 and 2 of CDKN2B in any of the patients tested (table 1). We did however find one heterozygote germline mutation in exon 1β of p14ARF in a patient (II-3) who developed a cutaneous melanoma at the age of 45 (fig 2). The G>A missense mutation (fig 2B) results in the substitution of glycine by aspartic acid at codon 16 (G16D). This individual belongs to an astrocytoma-melanoma syndrome family with two cases of CMM (fig 2A). Subsequent investigation of this family showed that the proband’s brother (II-2), who developed CMM at the age of 50, is an obligate carrier of the G16D mutation because his daughter tested positive for the mutation. Unexpectedly, the sister (II-1), who developed breast cancer and a meningioma, does not carry the G16D mutation, casting some doubt on its association with cancer predisposition. However, as the mutation was not observed in 202 control individuals, it cannot be regarded as a common polymorphism (data not shown).

To try to ascertain the functional relevance of the G16D mutation, we substituted the relevant nucleotide in p14ARF cDNA by site directed mutagenesis, and examined the ability of the altered protein to interact with MDM2 and to block the MDM2 mediated degradation of p53. It has been reported that the amino-terminal region of p14ARF is largely responsible for the interaction with MDM2, and for the localisation of p14ARF in the nucleolus.37,38 However, a form of p14ARF that lacks the amino-terminal 20 residues is still able to function normally.38 As shown in fig 2C, the G16D variant, expressed as a fusion protein with GFP, retained the ability to interact with MDM2 as judged by co-immunoprecipitation with a p14ARF specific antibody. Moreover, the G16D variant and wildtype p14ARF showed the same ability to prevent the MDM2 mediated turnover of p53 in a standard transient transfection assay (fig 2D). At a gross level, therefore, the G16D variant is functional, but the available assays are not suitable enough to detect subtle changes in p14ARF activity.

An indirect indication of the significance of G16D would be if a “second hit” occurred somatically in the wildtype allele of ARF in the tumour cells of the affected patients. However, we did not detect loss of heterozygosity or somatic mutation in exon 1β in the primary melanoma of individual II-3 or in the metastases of individual II-2 (data not shown).

Analysis of non-coding regions of 9p21

As the majority of the CMM prone families (group A) showed no evidence of germline alterations in the coding exons of CDKN2A, CDKN2B, and ARF, we considered the possibility that the underlying germline defect could be located in non-coding regions (promoter, intron) and affect the expression, structure, or stability of the respective transcripts. We therefore screened for the presence of the known recurrent deep intronic mutation (IVS2-105A>G) and for the −34G>T mutation in the CDKN2A promoter. Neither mutation was detected in any of the index cases from group A (table 1).

9p21 linkage analysis

We then analysed linkage to the 9p21 locus in nine families with three or more melanoma cases (group A) using the microsatellite markers D9S736, D9S1749 (CDKN2A), D9S942, and D9S1748 (table 1). For one of the nine families (no. 10279), the disease gene was unlinked to the 9p21 region (LOD score <−2). For one family (no. 3324), there was a suggestion of independence between the disease gene and the 9p21 region (−1.8<LOD score<−1.2), while the results for the seven other families (nos. 10339, 3026, 2535, 9849, 1403, 9834, 3284) were inconclusive (−1.2<LOD scores<0.7). It should be noted that in a tenth family (family no. 571) for which linkage at 9p21 locus was not performed in the present study, the disease gene was linked to the recently identified 1p22 locus.36

Haploid cell analysis of CDKN2A/ARF transcripts

In six out seven families for which linkage analysis was not conclusive (group A), we were able to obtain lymphoblastoid cell lines and generate somatic cell hybrids carrying separate 9p21 alleles (through GMP Genetics). Diploid to haploid conversion facilitates the detection and interpretation of abnormal transcripts. In each case, we selected two haploid clones carrying distinct alleles of chromosome 9, as judged using the D9S1749 and D9S942 markers (data not shown). Long range RT-PCR performed for both p16INK4A and p14ARF transcripts revealed only normally sized bands and therefore no aberrant transcripts.


We undertook a comprehensive survey of the 9p21 region in a series of 53 melanoma prone families and individuals, including 36 index cases that in a routine screen had shown no detectable germline mutations in the coding exons of CDKN2A or CDK4 (exon 2). Using a combination of DNA sequencing, quantitation of gene copy number by real time PCR, LOH analyses of microsatellite markers, and transcript analyses in haploid somatic cell hybrids, we found no evidence for germline alterations in either the coding or non-coding domains of CDKN2A and CDKN2B. However, two cases showed germline abnormalities that specifically affected ARF.

The first was the G16D missense mutation in exon 1β, detected in a CMM+NST family. Although the pedigree of this family implies that the mutation was present in two brothers who developed CMM, it was not present in a sister who developed meningioma and breast cancer. Association with melanoma-neural system tumour predisposition therefore remains ambiguous, as does the impact of the mutation on p14ARF function. G16D lies within the most conserved region of p14ARF and in a domain that interacts with MDM2,37 but the mutation had no discernible effect on the ability of p14ARF to bind to HDM2 or to stabilise p53, at least as assessed using the available assays. As these assays rely on transient co-expression of exogenous proteins, they do not allow us quantitative measurements of relatively subtle changes. Analysis of the primary melanoma and metastasis from the two brothers provided no evidence for a second somatic event affecting p14ARF exon 1β. However, we have not excluded the possibility that the wildtype ARF allele has been silenced by promoter methylation, as described in other studies.39,40 In conclusion, either the G16D mutation represents a p14ARF loss of function mutation and this family does not have hereditary CMM+NST assuming that the meningioma is a sporadic tumour or the G16D mutation represents a rare variant without functional consequence and the combination of CMM+NST is caused by some other unknown genetic defect.

The second alteration, found in a family with five documented CMM cases, was a large germline deletion that encompassed p14ARF exon 1β. By mapping the deletion breakpoints we found that it ended >11.2 kb upstream of the initiation codon of p16INK4A and was therefore highly unlikely to affect the expression of p16INK4A. There has been an ongoing debate as to whether the combined loss of p14ARF and p16INK4A function is responsible for the CMM+NST syndrome. Large deletions encompassing both CDKN2A and ARF and a splicing mutation that affects processing of both the p16INK4A and p14ARF RNA transcripts have been described in three CMM+NST families.14,15 The deletion we describe here only affects ARF exon 1β and occurred in a melanoma prone family without evidence of NSTs. Our findings therefore suggest that germline deletions specifically affecting p14ARF are not responsible for NST susceptibility. Predisposition to CMM+NST could either be due to complete disruption of the CDKN2A locus or be the result of more complex genetic inheritance. Our data also reinforce the hypothesis that ARF is indeed a melanoma susceptibility gene as argued in other studies.17,18

What explanation can be proposed for the other 50 cases in which we failed to find germline alterations by any of the methods used. Seven families from group A (nos. 10339, 3026, 2535, 9849, 1403, 9834, 10279) gave equivocal results, leaving open the possibility of linkage to 9p21. Our survey would not have revealed single nucleotide changes or small deletions or insertions in non-coding regions which have not been previously recorded. Such changes could in principle affect gene expression, for example by altering the binding of protein complexes that determine chromatin conformation in response to various stimuli. Finally, several previous studies have proposed that there are additional tumour suppressor genes in the chromosome 9p21 region that are implicated in melanoma. LOH studies have suggested at least two loci, one telomeric to IFNA and one centromeric to D9S17141 (reviewed in Pollock et al42), and there are potential candidate genes that may warrant consideration, such as TUSC1 (tumour suppressor candidate 1)43 and MTAP (methylthioadenosine phosphorylase).44 It is noteworthy that there are recent reports of MTAP being deleted or silenced in melanoma45 as well as pancreatic cancer.46

In addition to the seven families which gave equivocal results, haplotype analysis revealed that at least three of the families (nos. 3324, 571, 10279) showed no evidence for a linkage to chromosome 9p21, suggesting that there are additional melanoma susceptibility genes. A recent genome-wide scan for linkage has been performed in a set of families with three or more CMM cases originating from Australia. Results provided evidence for linkage to the 1p22 region, strongest in families with the earliest mean age at diagnosis, giving therefore significant evidence of a novel melanoma susceptibility gene located at 1p22.36 Linkage to 1p36 has also been reported in North American melanoma kindreds.47 However, it should be mentioned that no melanoma susceptibility genes have been identified at these loci to date.


Table AI

 Primer sequences for PCR and sequencing, primers and probe sequences for RQ-PCR, primer sequences for PCR long range, and dHPLC conditions

Table 1

 Molecular analysis at 9p21 locus

Figure 1

 (A) Pedigree of the CMM prone family AN (group A). Black symbols indicate melanoma affected patients and stripes indicate individuals with other types of cancers. The tumours and genotypes are indicated below the symbol and the number of melanomas is indicated in brackets. Stars indicate the patients who have been tested. (B) 9p21 locus map indicating microsatellite markers, p15INK4B, p14ARF, and p16INK4A coding exons, and location of TaqMan probes. (C) 9p21 locus sequences of a normal control and of patient FG7617 showing the 8474 bp deletion breakpoints. (D) Mapping of the 8474 bp deletion breakpoints at 9p21 locus.

Figure 2

 (A) Pedigree of the CMM+NST family no. 10754. Black squares indicate melanoma affected patients and stripes indicate the individual with other cancers. The tumours and genotypes are indicated below the symbol and the number of melanomas is indicated in brackets. Stars indicate the patients who have been tested. (B) p14ARF exon 1β partial sequences of a normal control (top) and of index case III-3 displaying the G16D heterozygote germline mutation. (C) Functional analysis of the G16D variant of p14ARF: MDM2 binding and p53 stabilisation by the G16D variant of p14ARF. U2OS cells were transfected with vectors encoding MDM2 and GFP-ARF or the G16D derivative as indicated. GFP alone was used as a control. After 48 h, lysates were immunoprecipitated with a polyclonal antibody against p14ARF (JR14), fractionated by SDS-PAGE in a 12% gel, and immunoblotted for MDM2 and GFP. There was no difference in the amount of MDM2 co-precipitated with wildtype or mutant p14ARF. (D) U20S cells were transfected with the indicated combinations of p53, MDM2, and GFP-ARF plasmids. Samples (50 μg) of total protein were fractionated by SDS-PAGE and immunoblotted for p53 and GFP. The ability of MDM2 to promote degradation of p53 (lane 3) is blocked by the presence of either wildtype (lane 4) or the G16D variant of p14ARF (lane 5).


We thank Blandine Passinge-Dumas, Danièle Pham, and Antoine Piriou for technical assistance.



  • * The following French Hereditary Melanoma Study Group members participated in this study: Valérie Arigon, Bertrand Bachollet, Françoise Boitier, Jean Pierre Cesarini, Liliane Demange, Nicolas Janin, Pascal Joly, and Michel Longy

  • Published Online First 3 June 2005

  • This work was supported by a grant from the Association pour la Recherche contre le Cancer (ARC), Subvention libre 1998, a Programme Hospitalier de Recherche Clinique (PHRC) Regional 2001 (AOR 01 091). KL is a recipient of a postdoctoral fellowship from Institut Gustave Roussy (CRC 2001). CM was a recipient of a Marie Curie fellowship (contract QLGA-1999-50406)

  • Competing interests: none declared

  • Note added in proof: After submission of this manuscript, the results of a new mutation scanning by direct sequencing were obtained for all familial cases from group A (n = 14) except for family FG7617 carrying the deletion. We detected a CDKN2A germline mutation for index case no. 1772, c.104G>T, p.Gly35Val

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.