MATERIALS AND METHODS

Study population

We recruited 108 patients with MM and 105 controls.

Of the MM patients, 29% were incident cases and 71% were prevalent cases. The median time between diagnosis and genetic investigation was 44.6 months. All participants were Caucasian and attended dermatology departments in one of three hospitals in Paris (France), namely Bichat Claude-Bernard, Percy, and Saint-Louis Hospitals. The study population consisted of patients aged from 20 to 80 years with histologically proved MM.

A total of 79 patients sharing features that might underlie genetic predisposition to MM, but without germline mutations in the CDKN2A or Cdk4 genes, were enrolled according to five different criteria:

• sporadic multiple primary melanomas (MPMs), with at least 2 MMs at different sites

• familial melanomas (FAMs), with at least two cases of MM in first or second degree relatives

• MMs associated with another cancer

• MMs diagnosed before the age of 25 years

• MMs considered to be non-photo induced (NPIMs).

NPIMs included melanomas located on non-photo exposed sites, and subungual and acral lentiginous melanomas (ALMs). ALMs are considered to be a particular subtype of MM because, in contrast to other subtypes, they exclusively involve acral skin which is rarely exposed to sunlight, thus suggesting that ultraviolet irradiation is not a major factor in their development.^37,38 Six patients who carried CDKN2A germline mutations were not included (three with FAMs, one with MPM, one with MM before the age of 25 years, and one with NPIM).

Another group of 29 “simple” melanoma patients without any of the above criteria was also enrolled. Patients were not included if immune depressed (HIV or transplantation), or suffering from a genodermatosis predisposing to skin cancer (albinism, Gorlin’s syndrome, or xeroderma pigmentosum).

A control group, without any personal or familial history of skin cancer, was enrolled for the comparison of MC1R allele frequencies. The subjects, aged from 20 to 80 years, were recruited in the same hospitals, as representative of the same demographic area as the patients with MM. Information concerning personal and familial details (parents and grandparents) and country of birth was also recorded. Non-Caucasians were not included in the study.

The Medical Ethics Committee (CCPPRB) approved the study protocol. Informed consent was obtained from all subjects.

Collection of data on risk factors for MM

All participants attended for a standardised personal interview and a total skin examination with a dermatologist, to collect data which was entered onto a preprinted examination sheet.

Information concerning skin (recorded as being either dark, medium, or light), eyes (classified as dark (brown) or light (blue, green, or grey)), and original hair colour (classified as red, blond, light brown, dark brown, or black) was collected. We also recorded naevus body count (< or ⩾50 naevi); presence of solar lentigines; and presence of AMS. AMS was recorded if 50 or more naevi were found, at least three being clinically atypical: 6 mm or more in diameter, variable pigmentation, and indistinct and irregular outline).³⁹ Skin type was assessed according to TB Fitzpatrick’s classification (1988) as follows:

• I, always burns and never tans

• II, always burns and then tans

• III, always tans and sometimes burns

• IV, always tans and never burns.

In addition, for the MM group, location of lesions, age at diagnosis, and histopathological data were collected.

Assessment of UV exposure

A questionnaire was used to determine the history of severe sunburns before and after the age of 15 years, defined as erythema for more than 48 hours or blistering (scored yes/no). Intermittent sun exposure before and after the age of 15 years was defined as UV exposure during the holidays: beachside or sunny vacations were scored as strong exposures. Chronic exposition was evaluated throughout the year, and was scored as nil, light, medium, or strong. Sunscreen use was noted as never, sometimes, or always.

Detection of MC1R variants

Genomic DNA was isolated from peripheral blood leukocytes of all participants by routine methods. The MC1R coding sequence was amplified by PCR with two overlapping couples of primers: MC1R-F1–5′-CAG CAC CAT GAA CTA AGC AGG ACA CCT G -3′ and MC1R1-IR-5′-CCA GCA TAG CCA GGA AGA AGA CCA CGA G -3′, and MC1R-F2–5′-TGG GTG GCC AGT GTC GTC TTC AGC A -3′ and MC1R2-R-5′-AAG GGT CCG CGC TTC AAC ACT TTC AGA G -3′, (respective sizes of PCR products, 671 and 610 bp). The PCR reaction mixture comprised 150 ng genomic template DNA, 5 μl 10X PCR buffer, 25 mM MgCl₂, 5 μM each of dGTP, dTTP, dATP, and dCTP, 22.5 pmol each of PCR primer, and 2 U AmpliTaq (Perkin Elmer; Courtabocuf cedex, France), in a total volume of 50 μl. Samples were denatured for six minutes at 96°C, and passed through 35 cycles of amplification, consisting of 30 seconds of denaturation at 95°C, 30 seconds of primer annealing at 60°C, one minute of elongation at 72°C, and final elongation for three minutes at 72°C. The amplifications were carried out in 0.5 ml tubes (Perkin Elmer).

Sequence analysis

DNA samples for sequencing were obtained by PCR as described above. Sequence analysis was performed on an ABI-Prism 3100 automated DNA sequencer using 10 ng PCR purified products and Big-Dye Terminator Cycle Sequencing kits (Perkin Elmer), according to the manufacturer’s instructions.

Statistical analyses

Statistical analyses were carried out using SAS software release 8.2 (SAS Institute, Cary, NC, USA). ANOVA or χ² analysis (Fisher’s exact test when necessary) were used to compare clinical and genetic characteristics, as well as UV exposure parameters, between MM patients and controls.

For clinical analysis, the usual MM risk factors were compared between cases and controls.

For genetic analysis, the only MC1R polymorphisms retained were:

• those previously shown to be loss of function variants by functional studies^{22–26,41,47}

• those associated with MM risk^8,13,30

• those associated with RHC phenotype²⁹

• those predicted to be possibly damaging by the Polyphen informatics program (http://tux.EMBL-Heidelberg.DE/ramensky/).

This approach was conservative and allowed us to include all patients and controls for the analysis. Pairwise linkage disequilibrium between the different MC1R polymorphisms was also studied, using the EH (Estimation of Haplotypes) program. First, univariate analyses were used to compare genetic risk factors (MC1R genotype) between MM patients and controls. Secondly, multiple logistic regression analysis (using a stepwise procedure) was performed to take into account potential confounders among clinical risk factors, such as skin type, eye and hair colour, number of naevi, AMS, dorsal lentigines, and UV exposure. Odds ratios (ORs) were calculated with 95% confidence intervals. All significance levels reported were two-sided and set at p<0.05.

Comparisons were carried out of age at onset between:

• MM patients carrying MC1R variants v MM patients that do not carry MC1R variants

• and MM patients carrying MC1R variants v MM patients carrying germline CDKN2A mutations.

Comparisons were performed using the ANOVA test.

The percentage population attributable risk (PAR%) associated with the most frequent MC1R variants was calculated as previously described.⁴⁸

RESULTS

Composition of patient population

The final series for analysis comprised 213 participants: 108 patients with MM and 105 controls. The percentage of individuals that were born in France, with both parents born in France, was 70% in the patients’ group, and 60% in the control group. Other places of birth included European countries (Switzerland, Germany, Greece, Spain, Portugal, Italy, Slovenia, Croatia, and Turkey) or the Maghreb (Tunisia, Algeria, Morocco). The 108 patients were categorised into two different groups.

The first group was composed of 79 patients suspected of genetic predisposition to MM, including the following pathologies:

• 18 FAMs, three of which harboured the typical AMS phenotype

• 17 sporadic MPMs

• 11 MMs with an additional cancer (two mammary adenocarcinomas, one papillary thyroid carcinoma, one prostate carcinoma, one uterine carcinoma, one carcinoma of the colon, four non-melanoma skin cancers, one meningioma)

• 20 MMs diagnosed before the age of 25 years

• 14 NPIMs (one on the anus, two on the colon, four on the buttocks, two on the genitals, one on the scalp, and four ALMs (three subungual and one on the sole of the foot)).

The second group comprised 28 simple sporadic MMs without any of the above pathologies.

Clinical characteristics

The clinical characteristics of all patients with MM and of control subjects are summarised in table 1. The strongest risk factors identified for MM were a mole count >50 (p<0.0001), presence of AMS (p  =  0.0004), presence of fair skin (p<0.0001), and dorsal lentigines (p<0.0001). Other pigmentation characteristics (light eye and hair colours), as well as severe sunburns before and after the age of 15 years, were lower predictors of MM risk (table 1). In this sample, skin type I or II, although more frequent in MM patients, was not significantly associated with MM risk.

Ultraviolet exposure

Univariate analysis showed that severe sunburns before and after the age of 15 years (p  =  0.03 and 0.05, respectively), intermittent exposure during holidays after the age of 15 years (p  =  0.03), and absence of sunscreen use (p  =  0.04) were associated with MM risk, as in previous studies (table 1). Multiple analyses, after controlling for other variables, confirmed this association for two parameters— absence of sunscreen use, and severe sunburns after the age of 15 years (data not shown). In this study, cumulative sun exposure, as determined by chronic exposure during the week during occupational activities, was not identified as a risk factor for MM (data not shown).

Frequency of MC1R variants and effect on MM risk

We identified 16 non-synonymous MC1R variant alleles, of which 12 have shown experimental and/or predicted loss of function effects, and/or have been previously associated with MM risk and/or RHC (see Materials and Methods and table 2). Among these, two previously unreported MC1R variants (Arg142Cys and Tyr298His) were observed, each in one (differing) MM patient.

Functional variants were much more common in the MM group, with 73/108 patients (67.6%) carrying at least one MC1R variant, compared with only 33/105 control subjects (31.4%) (p<0.0001, table 3). In addition, there was a gene dosage effect of MC1R variant on MM risk (ORs, respectively, 4.3 for one variant and 6.78 for two variants; Mantel Haenszel trend test p<0.0001).

Not all individual MC1R alleles conferred the same MM risk (table 2). Only three variants (Val60Leu, Arg151Cys, and Arg160Trp) were significantly more frequent in the MM group, with the strongest risk being for Arg151Cys (OR 6.26). The nine other MC1R variants (Val122Met, Arg142His, Arg142Cys, Ile155Thr, ins86, Ser83Pro, Asp84Glu, Asp294His, and Tyr298His) were not individually associated with MM risk, although it should be noted that for Asp294His statistical results were nearly significant (p  =  0.058).

MC1R variants allelic frequency observed in incident cases (genetic investigation within one year after MM diagnosis) was not statistically different from that observed in prevalent cases (36.8% v 45%; p  =  0.25, Fisher’s exact test).

The PAR% of MM associated with the main MC1R variants is indicated in table 4. Val60Leu has the highest PAR, followed by Arg151Cys and Arg160Trp.

Effect of MC1R on the age of onset of melanoma

No effect of MC1R genotype on the age of onset of melanoma was observed: median age at diagnosis of MM patient carriers of MC1R variants was 46.17 v 39.88 years for the MM group without variant (p  =  0.087; standard deviation 17.11 v 17.17 using the ANOVA test). Median age at diagnosis was lower for patients carrying CDKN2A mutations (35.5 years, range 23–64), but this was not statistically significant (p  =  0.23, ANOVA test), presumably because of the small number of CDKN2A mutated patients studied.

Distribution of MC1R variants in the different MM subgroups

Table 5 shows the distribution of MC1R variants among the five MM subgroups. The highest MC1R allelic frequencies were observed in FAM and MPM subgroups (50%), whereas the lowest were observed in the NPIM group (28.5%). However, there was no statistical difference in MC1R allelic frequencies between the different subgroups (p  =  0.37), or between the simple MM group and the group with more stringent genetic criteria (FAM, MPM, NPIM, MM <25 years of age, MM associated with another cancer) (p  =  0.35).

Effect of MC1R on pigmentation characteristics

In order to study the effects of MC1R on pigmentation characteristics and naevus count, cases and controls were grouped together. Statistical analysis confirmed previous observations, in that MC1R variants were significantly associated with light hair colour (red, blond, light brown, p<0.0001); skin type I–II (p  =  0.0015); solar lentigines (p  =  0.0019); and either light or medium skin colour (Mantel Haenszel trend test, p  =  0.047). In addition, MC1R variants were also associated with light eye colour (p  =  0.027). Conversely, MC1R variants were not associated with naevus count (p  =  0.19) (data not shown).

Persistence of MM risk after stratification on pigmentation characteristics and UV exposure

Clinical and epidemiological data indicated further investigation of the association between MC1R functional variants and MM risk. Table 6 shows the significant persistence of MM risk according to MC1R variants after stratification for the different melanoma associated clinical risk factors (skin types I or II, medium or light skin, red or light hair, light eye colour). These results strongly suggest that MC1R variants are an independent risk factor for the development of MM.

In addition, MC1R MM risk also persisted after stratification on classical UV exposure risk factors (severe sunburns and high intermittent UV exposure, both before and after the age of 15 years), suggesting that MC1R confers an MM risk independently of UV exposure (table 6). Interestingly, in multiple logistic regression analyses, which take into account all these potential confounders, the presence of MC1R variants was the second most important MM risk factor risk (odds ratio 4.52, 95% confidence interval 2.15–9.52), after a high naevus count (odds ratio 12.66, 95% confidence interval 4.81–33.33, table 7).

DISCUSSION

Numerous studies have shown that MC1R is highly polymorphic in the Caucasian population in the British Isles, Holland, and Australia,^{12,18,28,42–44} and that some MC1R variants are associated with the risk of MM and non-melanoma skin cancer in populations of Celtic and/or Germanic origin (Australia, Scotland, Ireland, and the Netherlands).^8,13,33,45

In this study, we have shown a strong association between functional MC1R variants and MM in the French population, which is of a different geographical origin. This confirms and highlights the role of MC1R in genetic predisposition to MM.

MC1R variants were previously found to be associated with RHC^18,19,27,28 and sun sensitivity.⁴⁶ In our study, MC1R functional alleles were also associated with medium or light skin colour, skin types I and II, fair hair colour (red, blond, and light brown), and presence of solar lentigines, further demonstrating that MC1R plays a crucial role in human pigmentation and response to UV radiation in a French population. However, it should be noted that the individual allelic frequencies of the main MC1R functional variants in our control population (table 2) were much lower than those reported in previous studies.^8,13,45 This difference in allele frequencies might due to the particular characteristics of the French population, which has a darker complexion and a different genetic ancestry from the other populations studied (that were mainly of Celtic origin). In addition, we observed a very low MC1R variant frequency in control subjects with skin type I/II (23%) (table 6). As it has been shown that pigmentation is a complex process under the control of at least 50 different loci in mice,⁴⁹ it seems reasonable to assume that other polymorphisms (localised in pigmentation genes different from MC1R) also play an important role in the genetic control of human pigmentation and could explain our result.

Two large studies reported that MC1R variants increase MM risk, with a mean odds ratio varying between 2 and 4, depending on whether one or two variants were present.^8,31 It is difficult to compare our data with the previous studies, as the coding sequence of MC1R was completely examined in only one of them.¹³ However, we confirmed the effect of MC1R variants on MM risk, increasing the risk by a factor of four when one variant was present, and by seven when two variants were present (table 3). The MC1R related MM risks were nearly the double those previously reported,^8,13 a result most likely due to an increased frequency of the MC1R consensus genotype in our control population.

Separate analyses of individual MC1R variant alleles showed that three variant alleles (all of which are common) were significantly associated with an increased risk of MM (table 2). Arg151Cys and Arg160Trp showed the highest risk, as in previous studies.^8,31 However, Val60Leu, which was not or was weakly associated with MM risk in Australia and the Netherlands,^8,31 appeared here to be the most frequent variant and to be associated with an important MM risk (odds ratio 3.99, 95% confidence interval 2.03–7.82). Two variants, Asp84Glu and Asp294His alleles, that were previously associated with an important MM risk,^8,30,31 were not here significantly associated with MM. This could probably be related to the size of our effectives, as their low frequency could have masked their effects on MM risk. We found no association with MM risk for the frequent variant Val92Met, in agreement with a previous report.³¹

There is a discrepancy concerning the effect of this substitution on MC1R function. One study⁴⁰ suggested that the Val92Met substitution reduced the binding affinity of MC1R for α-MSH, whereas another found that melanocytes homozygous for Val92Met responded dose dependently to α-MSH with stimulation of cAMP formation, tyrosinase activity, and proliferation, suggesting that this polymorphism did not represent a loss of function of MC1R.⁴⁷

The PAR% of MM associated with the main MC1R variants (table 4) was much higher than that associated with CDKN2A mutations, which are very rare in most series of unselected melanoma.⁵⁰

The allelic frequency of MC1R variants was the highest in MM patients with the most stringent genetic criteria (multiple and familial MM subgroups, table 5). However, our groups were too small to draw definitive conclusions, and we observed no statistical difference between the different MM subgroups; these results need to be assessed with a higher number of patients.

There is a debate as to whether the MC1R MM risk is independent or not of other melanoma clinical risk factors (skin type, mole count, and eye, hair, and skin colours). Initially, MC1R variants were shown to increase MM risk only in individuals whose darker complexions would normally be considered protective, the association between MM and MC1R variants disappearing in persons with light skins.⁸ More recently, the role of MC1R was also demonstrated in individuals with skin types I and II and light hair colour in the Netherlands population.³¹ We further investigated whether the association between MC1R variants and MM is independent of clinical risk factors in our French population, and showed the persistent effect of MC1R variants on MM risk after stratification for the different clinical risk factors (table 6). This shows that MC1R variants and pigmentary characteristics are independent MM risk factors, and confirms and extends the results of the Dutch study.³¹ Furthermore, the effect of MC1R variants also persists after stratification on UV exposure parameters, particularly in the absence of severe sunburns (table 6) that are classically considered to be important risk factors for melanoma formation.⁵¹

The exact mechanisms underlying the increased risk to individuals carrying MC1R variants of developing MM are not known. First, recent investigations suggest that the quality of melanin, rather than its quantity, determines skin cancer risk.⁵² In fact, increase of red yellow phaeomelanin in MC1R variant melanocytes could contribute to mutagenesis and confer susceptibility to MM and non-melanoma skin cancer by generating free radicals following UV exposure. In addition, melanocytes with non-functional MC1R exhibit a pronounced increase of sensitivity to the cytotoxic effect of UV radiation compared with melanocytes expressing functional MC1R.⁴⁷

Secondly, MC1R variants could affect some UV induced cellular protective effects. The cell cycle inhibitor p16^INK4a, that plays a major role in genetic predisposition to familial melanoma and sporadic MM formation, appears to have an important role in a cell cycle checkpoint in skin after UV exposure.^53,54 As it has been shown that this UV epidermal induction of p16^INK4A is induced by α MSH via MC1R,⁵⁵ the presence of MC1R variants could alter this effect, thereby reducing its tumour suppressor function and providing a physiopathological basis for the formation of melanoma cells. An additional hypothesis is that UV irradiation, in the presence of MC1R variants, induces an important oxidative stress with formation of reactive oxygen species that have been shown to lead to a decrease of interaction between p16^INK4A and Cdk4 proteins, thus promoting carcinogenesis.⁵⁶

Thirdly, in our study, MC1R variants were also present in 57% of patients with non-photo induced melanoma, further suggesting that MC1R could act on MM risk via non-pigmentary mechanisms. One of these mechanisms could be an autocrine effect of α-MSH on melanoma cells. It has recently been demonstrated that α-MSH significantly reduces the growth and progression (via decreasing fibronectin binding) of wild type MC1R melanoma cells, but has no effect on melanoma cells with MC1R variants.⁵⁷

Finally, we showed that in multiple logistic regression analysis taking into account all MM clinical and genetic risk factors, MC1R variants were identified as the second MM risk factor following a high naevus count (table 7). This result inclines us to consider the MC1R genotype to be an important MM marker risk, at least equal to eye colour, hair colour, or skin type in MM risk evaluation.

In conclusion, we have shown that carrying particular MC1R variants increases MM risk in France. The association between MM and MC1R variants persists once skin, hair, and eye colours, skin type, mole count, and AMS phenotypes are taken into account. In addition, MC1R variants were identified as the second MM risk factor following a high naevus count. If our data could be confirmed in larger series, this would suggest that assessing MC1R status, in association with clinical risk factors, might be useful in identifying high risk groups to be targeted for prevention.