Analysis of pathogenic CDKN2A mutations
The numbers of cases and controls screened at each centre, age at diagnosis (cases) or recruitment (controls), the proportion of cases with a family history of melanoma, and the presence of MPM are shown in Table
1. Information on family history and MPM is based on information available at the time of recruitment.
Table 1
Descriptive characteristics of individuals recruited to studies
Age at diagnosis/recruitment1
| < 40 (%) | 596 (100) | 377 (79) | 232 (31) | 18 (17) | 289 (18) | 61 (12) |
≥ 40 (%) | 0 | 99 (21) | 508 (69) | 91 (83) | 1297 (82) | 438 (88) |
Affected relatives1,2
| 0 (%) | 500 (83.9) | - | 690 (93.2) | - | 1470 (92.7) | - |
1 (%) | 76 (12.8) | - | 34 (4.6) | - | 101 (6.3) | - |
2+ (%) | 20 (3.4) | - | 16 (2.2) | - | 15 (0.9) | - |
Number of melanoma primaries1,2
| 1 | 581 (97.5) | - | 639 (86.4) | - | 1495 (94.3) | - |
2 | 14 (2.4) | - | 80 (10.8) | - | 73 (4.6) | - |
3+ | 1 (0.1) | - | 21 (2.8) | - | 18 (1.1) | - |
A total of 2929 melanoma cases and 1084 controls were screened for mutations in
CDKN2A exons 1α, 1β, 2 and 3 and
CDK4 exon 2. No variants were identified in
CDK4 exon 2. Of the 58
CDKN2A variants identified in this study (Additional file
2) 33 were deemed very likely to be pathogenic mutations (Table
2 and Additional file
3). No pathogenic variants were identified in the control samples. Twelve different pathogenic variants were identified in the Australian cases (14 carriers of pathogenic mutations out of 596 cases in total), 11 in the Spanish cases (19 out of 747) and 17 in the UK cases (31 out of 1586) (Table
2).
Table 2
Pathogenic
CDKN2A
variants identified in proband cases in Australia, Spain and the United Kingdom
5′UTR | c.-34G > T | r.-34_-32 > p.M1 | _ | _ | 1 | 1 | 1 |
Exon 1α | c.9_32del24 | p.A4_10Edel7 | _ | _ | . | . | 3 |
| c.32_33ins9_32 | p.M1_S8dup | _ | _ | 1 | . | 4 |
| c.52_57dup6 | p.20T_21Adup | _ | _ | 1 | . | 1 |
| c.68G > A | p.G23D | _ | _ | . | . | 1 |
| c.71G > C | p.R24P | _ | _ | . | . | 1 |
| c.95T > C | p.L32P | _ | _ | 1 | . | 1 |
| c.104G > A | p.G35E | _ | _ | . | 1 | . |
| c.104G > C | p.G35A | _ | _ | . | . | 1 |
| c.106delG | p.A36RfsX17 | _ | _ | . | 1 | . |
| c.113C > G | p.P38R | _ | _ | . | . | 1 |
Exon 1β | _ | _ | c.45_60dup | p.V22PfsX46 | . | 1 | . |
| _ | _ | c.81C > G | p.I27M | 1 | . | . |
| _ | _ | c.102G > A | p.W34X | 1 | . | . |
| _ | _ | c.193G > A | p.G65S | 1 | . | . |
| _ | _ | c.193G > C | p.G65R | 1 | . | . |
Exon 2 | c.159G > C | p.M53I | c.202G > C | p.D68H | . | . | 5 |
| c.176T > G | p.V59G | c.219T > G | p.S73R | . | 3 | . |
| c.188T > C | p.L63P | c.231T > C | p.A77A | . | . | 2 |
| c.194T > C | p.L65P | c.237T > C | p.A79A | 1 | 1 | . |
| c.202_203GC > TT | p.A68L | c.245_246GC > TT | p.R82L | 1 | . | . |
| c.206A > G | p.E69G | c.249A > G | p.G83G | 3 | . | . |
| c.212A > C | p.N71T | c.255A > C | p.Q85H | . | . | 1 |
| c.228_246del19 | p.L77TfsX62 | c.271_289del19 | p.R90VfsX75 | . | . | 2 |
| c.240_253del14 | p.81PfsX56 | c.92_105del14 | p.95TfsX56 | . | . | 1 |
| c.251A > C | p.D84A | c.294A > C | p.R84R | . | . | 1 |
| c.259C > T | p.R87W | c.302C > T | p.P101L | . | 2 | . |
| c.301G > T | p.G101W | c.344G > T | p.R115L | . | 5 | . |
| c.331G > A | p.G111S | c.374G > A | p.G125E | . | . | 1 |
| c.358delG | p.E120SfsX26 | _ | _ | . | 1 | . |
| c.370C > T | p.R124C | _ | _ | . | 2 | . |
| c.379G > T | p.A127S | _ | _ | . | 1 | . |
Intron 2 | c.458-105A > G | p.156_157del | _ | _ | 1 | . | 4 |
Total | | | | | 14 | 19 | 31 |
The prevalence of pathogenic mutations observed for melanoma cases at each centre was similar, with an overall frequency of 2.2% (95% CI: 1.7, 2.7%) for the three populations investigated with no evidence of any differences across the populations (chi-square test, p = 0.63) (Table
3).
Table 3
Number and percentage of individuals for whom pathogenic and low penetrance
CDKN2A
variants were identified
CDKN2A high penetrance mutation | present | cases | 14 (2.3) | 19 (2.5) | 31 (2.0) | 64 (2.2) |
controls | 0 (0) | 0 (0) | 0 (0) | 0 (0) |
wild type | cases | 582 (97.7) | 728 (97.5) | 1555 (98.0) | 2865 (97.8) |
controls | 476 (100) | 109 (100) | 499 (100) | 1084 (100) |
A148T variant | present | cases | 48 (8.1) | 63 (8.4) | 110 (6.9) | 221 (7.6) |
controls | 18 (3.8) | 6 (5.6) | 32 (6.4) | 56 (5.2) |
wild type | cases | 548 (91.9) | 684 (91.6) | 1476 (93.1) | 2708 (92.4) |
controls | 458 (96.2) | 102 (94.4) | 467 (93.6) | 1027 (94.8) |
c.*29C > G variant | present | cases | 163 (27.4) | | 425 (26.9) | 588 (27.0) |
controls | 125 (26.7) | | 133 (27.4) | 258 (26.8) |
wild type | cases | 433 (72.6) | | 1155 (73.1) | 1588 (73.0) |
controls | 351 (73.7) | | 352 (72.6) | 703 (73.2) |
c.*69C > T variant | present | cases | 111 (18.6) | | 291 (18.4) | 402 (18.5) |
controls | 485 (81.4) | | 1289 (81.6) | 1774 (81.5) |
wild type | cases | 78 (16.4) | | 98 (20.2) | 176 (18.3) |
controls | 398 (83.6) | | 387 (79.8) | 785 (81.7) |
The prevalence of pathogenic mutations increased with extent of family history (defined as the number of affected first or second degree relatives) for each of the three populations, rising, in the combined dataset, from 1.5% of cases for those without a family history to 5.2% for those with one affected relative, and to 25.5% for those with two or more affected relatives (Table
4). In Spain and the UK (combined), there was a corresponding increase with the number of lifetime melanoma primaries rising from 1.3% for cases with only a single primary to 6.5% for cases with two primaries and 29.3% for three or more. In Australia, where the eligibility criteria excluded those with a previous invasive melanoma, 2.4% of cases with a single primary at diagnosis had a mutation while none of the 15 cases with MPM (who presented with synchronous primary melanomas or a previous
in situ primary melanoma) had a mutation (Table
4).
Table 4
Association of family history and number of primaries with prevalence of
CDKN2A
mutations
Family history
4
| | | | | | | | | | | |
0 | 7/500 | 1.4 | 1.0 | 13/696 | 1.9 | 1.0 | 20/1470 | 1.4 | 1.0 | 1.5 | 1.0 |
1 | 4/76 | 5.3 | 3.9 (1.1, 13.7) | 3/35 | 8.6 | 4.9 (1.3, 18.2) | 4/101 | 4.0 | 3.0 (1.0, 8.9) | 5.2 | 3.8 (1.9, 7.5) |
2+ | 3/20 | 15.0 | 12.4 (3.0, 52.3) | 3/16 | 18.8 | 12.1 (3.1, 47.7) | 7/15 | 46.7 | 63.4 (21.0, 19.8) | 25.5 | 23.2 (11.3, 47.6) |
No. of primaries
5
| | | | | | | | | | | |
1 | 14/581 | 2.4 | n/a | 8/644 | 1.2 | 1.0 | 20/1495 | 1.4 | 1.0 | 1.3 | 1.0 |
2 | 0/14 | 0 | | 6/80 | 7.5 | 6.5 (2.2, 19.1) | 4/73 | 5.5 | 4.3 (1.4, 12.9) | 6.5 | 5.4 (2.5, 11.6) |
3+ | 0/1 | 0 | | 5/23 | 21.7 | 22.1 (6.6, 74.2) | 7/18 | 38.9 | 46.9 (16.5, 133.5) | 29.3 | 32.4 (14.7, 71.2) |
Compared with cases without a family history, cases with one affected first or second degree relative had a three-fold increased odds of carrying a mutation (OR 3.8; 95% CI: 1.9, 7.5), after adjusting for presence of multiple primaries, age and country, and for those with two or more affected relatives the odds of carrying a mutation was further increased (OR 23.2; 95% CI: 11.3, 47.6). The same trend was apparent for each of the three countries (Table
4).
For Spain and the UK, there was a significant positive association between number of primary melanomas identified in the case and the likelihood of carrying a pathogenic mutation (Table
4). Overall, compared with cases with a single primary melanoma, cases with two primary melanomas had a 5-fold increased odds of carrying a mutation (OR 5.4; 95% CI: 2.5, 11.6), after adjusting for family history, age and country; and for those with three or more primary melanomas the odds of carrying a mutation were much higher (OR 32.4; 95% CI: 14.7, 71.2).
The prevalence of mutations also increased for cases with multiple risk factors (family history plus multiple primaries). For instance, in Spain, mutation prevalence was 8.7% for MPM without a family history, 7.1% for cases with a family history but only one primary, and 17.4% for cases with MPM and a family history (data not shown). A similar pattern was seen for UK cases (7.5%, 5.7% and 45.4%) (data not shown). Table
5 shows, for the 3 countries combined, the prevalence of
CDKN2A mutations, and the OR for finding a mutation for combinations of risk factors compared with the reference category of no family history and a single primary. For persons with two or more affected relatives and a personal history of three or more primary melanomas, the mutation prevalence was 60% and the OR for finding a mutation estimated as 146.6 (95% CI: 23.1, 928.1).
Table 5
Cross-tabulation of mutation prevalence and odds ratios, according to family history and multiple primaries
Prevalence of CDKN2A pathogenic mutation
| | | |
One primary only | 1.1% (27/2495) | 3.8% (7/185) | 20.0% (8/40) |
Two primaries | 3.6% (5/140) | 14.3% (3/21) | 33.3% (2/6) |
Three or more primaries | 26.7% (8/30) | 20.0% (1/5) | 60.0% (3/5) |
Odds ratios
1
and 95% confidence intervals
| | | |
One primary only | 1 (reference) | 3.3 (1.4, 7.7) | 18.9 (7.8, 45.5) |
Two primaries | 3.5 (1.3, 9.3) | 13.1 (3.6, 47.8) | 42.7 (7.5, 244.0) |
Three or more primaries | 33.1 (13.5, 81.5) | 23.8 (2.6, 221.2) | 146.6 (23.1, 928.1) |
Estimation of the expected numbers of cancers based on a multivariate model considering family history, the presence of multiple primaries, age, and sex, but unadjusted for country of recruitment, showed that this model predicted the number of observed carriers well (22.6 expected for Spain (observed = 19), 19.9 expected for the UK (observed = 22) and 12.5 expected for Australia (14 observed)) indicating that the predictors of having a mutation are consistent across populations independently of the baseline incidence rate or the precise method of ascertainment.
In Spain, there was a strong association between age at diagnosis and presence of a mutation, with carriers of a pathogenic mutation being diagnosed at age 35.9 years on average compared with 50.1 years for the cases not found to carry a pathogenic mutation in CDKN2A (p = 0.0002, data not shown). There was no similar association in the Australian (33.9 years vs. 32.7 years, respectively; p = 0.27) and UK samples (51.0 years vs. 53.8 years, respectively; p = 0.26), but the limited age range of the Australian study precluded further analysis. Within the Australian and UK samples, there was no evidence that mutation prevalence differed by gender (data not shown) but for Spain, 3.9% of male cases were carriers compared with 1.6% of female cases (p = 0.06 Fisher’s exact test).
The frequency of pathogenic mutations was similar for cases with melanomas at different body sites (head and neck; trunk; limbs; other) for Australia and Spain. For the UK, there was some evidence that cases with tumours on the trunk were more likely to carry a mutation than cases with tumours at other sites (Fisher’s exact test p = 0.04, data not shown). There was also some weak evidence that thinner tumours were associated with CDKN2A mutations for the UK samples (p = 0.05, data not shown), but not for the Australian or Spanish samples. For the UK, 3.3% of tumours with Breslow ≤1 mm had CDKN2A mutations, compared with 1.2% of those >1 mm.
Analysis of lower penetrance CDKN2A variants
The low penetrance variant p.A148T was observed at an overall frequency of 7.6% for melanoma cases, and 5.2% for controls (p = 0.008) (Table
3) (p = 0.17 and 0.57 for homogeneity of prevalence among controls and cases, respectively). Crude ORs for risk of melanoma for the three countries were; Australia: 2.2 (95% CI: 1.3, 3.9); Spain: 1.6 (95% CI: 0.7, 3.7); and UK: 1.1 (95% CI: 0.7, 1.6). Analysis of all studies together suggested that presence of a variant increased melanoma risk (OR 1.5; 95% CI: 1.1, 2.0) (p = 0.009, data not shown). As the Australian cases were all <40 years of age, we examined whether the positive finding for Australia related to an effect of age at onset. Analysis from the UK supported this suggestion with ORs of 1.6 (95% CI: 0.7, 3.7) and 0.9 (95% CI: 0.6, 1.5) within the age groups ≤50 and >50 years respectively, with evidence of different effect sizes (p = 0.001). Conversely, the Spanish data suggested a stronger association for older onset cases: crude ORs were 0.9 (95% CI: 0.3, 2.4) for ≤50 years and 4.8 (95% CI: 0.6, 36.1) for >50 years, but the evidence for a difference was weak, p = 0.5. The carriage of the p.A148T variant was not associated with MPM (p = 0.14) or family history (p = 0.60).
The 3′ UTR variant c.*29C > G was observed at an overall frequency of 27.0% for cases and 26.8% for controls (Table
3). The 3′UTR variants were not analyzed for the Spanish sample, as they were located outside the limits of the SSCP screen employed. Crude ORs for risk of melanoma for Australia and the UK were 1.06 (95% CI: 0.81, 1.39) and 0.97 (95% CI: 0.77, 1.22) respectively. The 3′ UTR variant c.*69C > T was observed at an overall frequency of 18.5% for cases and 18.3% for controls. Crude ORs for risk of melanoma for Australia and the UK were 1.17 (95% CI: 0.85, 1.61) and 0.89 (95% CI: 0.69, 1.15) respectively.