Detection of novel quantitative trait loci for cutaneous melanoma by genome-wide scan in the MeLiM swine model



Human cutaneous melanoma is a complex trait inherited in about 10% of cases. Although 2 high-risk genes, CDKN2A and CDK4, and 1 low risk gene, MC1R, have been identified, susceptibility genes remain to be discovered. Here, we attempted to determine new genomic regions linked to melanoma using the pig MeLiM strain, which develops hereditary cutaneous melanomas. We applied quantitative trait loci (QTL) mapping method to a significant genome-wide scan performed on 331 backcross pigs derived from this strain. QTLs were detected at chromosome-wide level for a melanoma synthetic trait corresponding to the development of melanoma. The peak positions on Sus scrofa chromosomes (SSC) were at 49.4 and 88.0 cM (SSC1), 56.0 cM (SSC13), 86.5 cM (SSC15) and 39.8 cM (SSC17), and, on SSC2, at 16.9 cM, in families derived from F1 males only (p < 0.05, except for SSC13, p < 0.01). Analysis of 7 precise specific traits revealed highly significant QTLs on SSC10 (ulceration), on SSC12 (presence of melanoma at birth), on SSC13 (lesion type), and on SSC16 and SSC17 (number of aggressive melanomas) at the respective positions 42.0, 95.6, 81.0, 45.3 and 44.8 cM (p < 0.001 and p < 0.05 respectively at the chromosome- and genome-wide levels). We also showed that MeLiM MC1R*2 allele, which determines black coat colour in pigs, predisposes significantly to melanoma. Interactions were observed between MC1R and markers located on SSC1 (p < 0.05). Taken together, these results indicate that MeLiM swine is a model for human multigenic diseases. Comparative mapping revealed human regions of interest to search for new melanoma susceptibility candidates. © 2006 Wiley-Liss, Inc.

Like for most cancers, the aetiology of cutaneous melanoma (CM) is complex, involving heterogeneous, interacting genetic and environmental determinants. Familial melanoma accounts for about 10% of cases. Mutations in two high-risk genes encoding cell-cycle-regulatory proteins have been shown to cause familial CM, including the cyclin-dependent kinase inhibitor 2A (CDKN2A) located on the Homo sapiens chromosome (HSA) 9p21, and the cyclin-dependent kinase 4 (CDK4) located on HSA12q14.1CDKN2A germline mutations occur in about 40% of families with linkage to 9p21,1, 2 suggesting the existence of additional gene(s) which predispose to melanoma in this chromosomal area. In the CDK4 gene, the Arg24Cys or Arg24His mutations affect the domain binding to the p16 protein, thus preventing regulation of the p16-dependent cell cycle. These mutations have only been identified in a few families world-wide.1 Although mutations in the BRAF gene located on HSA7q34 were found in a high proportion of primary melanoma tumours, BRAF did not appear to be a gene predisposing to familial melanoma,3 however, germline single nucleotide polymorphisms (SNPs) in BRAF were found to be significantly associated with melanoma in German males.4 One low-penetrance gene, the melanocortin receptor 1 gene (MC1R) which has a key role in human pigmentation, was found to code for melanoma susceptibility variants,5 some of which increased the penetrance of CDKN2A mutations.6, 7, 8 Other low-risk genes have been investigated in a few case–control studies, but the results were of low significance.9

To date, few genome scans for human melanoma have been reported. The most significant, with high-density coverage of microsatellite markers, was conducted for human CM in families from which CDKN2A and CDK4 involvement was excluded.10 The results suggested that the human chromosomal region HSA1p22 might contain a novel melanoma susceptibility locus, and this was confirmed in 33 additional pedigrees in the same study. However, further screening of candidate genes for mutations in this region did not reveal any causal mutation.11 High-penetrance genes may be present in other regions; linkage was reported to 1p36 in North American CM/dysplastic nevi pedigrees,12 and to 9q21.32 in Danish pedigrees with CM and ocular melanoma.13

A genome scan using animal models of spontaneous CM can be of great value, not only to guide the choice of candidate genes and chromosomal regions to be studied in humans, but also to support evidence for the possible associations found with these genes before embarking into functional studies. Although very few animal models of spontaneous melanoma exist, CM is inheritable in 3 stocks of miniature swine: the Sinclair swine,14 the Munich Miniature Swine Troll15 and the Melanoblastoma-bearing Libechov Minipig (MeLiM).16, 17 In these pigs, tumours arise spontaneously in utero or in the first months after birth, and then disseminate mainly in the lymph nodes and occasionally in inner organs, including the lung, liver, spleen, gastrointestinal tract, pancreas and kidney. However, most of these tumours regress spontaneously. As in humans, but contrary to mice, swine precursor melanoma cells originate from the basal layer of the epidermis, so that human and swine melanocytes have a comparable microenvironment. The histopathological characteristics of spontaneously occurring swine tumours were also found to resemble those of human melanomas as well documented in Sinclair and MeLiM breed models.17, 18 Previous statistical genetic analyses of hereditary melanoma in these models suggested that 2 or 3 genes affected tumour formation, although none was identified.15, 19, 20

In our preliminary study, the MeLiM model was used to initiate a search for predisposing genes, by conducting a preliminary genome-wide scan on 79 backcross pigs (MeLiM × Duroc), deriving from a unique MeLiM boar. Interval mapping analysis provided evidence for linkage to 2 regions on Sus scrofa chromosome (SSC) 1q25 homologous to the human region HSA9p21 and to 1 region on SSC2p17 homologous to HSA11p15 and 11q13.21 The CDKN2A gene was mapped to the SSC1q25 region22 but was excluded as a major susceptibility gene in this MeLiM family by a haplotype study.23 An association was observed with a MeLiM allele for a marker close to the MC1R gene on SSC6p1.5, but the CDK4 and BRAF genes did not appear to be melanoma susceptibility genes in this MeLiM model.21

In the present study, we undertook a more extensive genome-wide scan, including a total of 331 backcross animals issued from 4 MeLiM pigs, to confirm our initial results and to locate additional regions of potential interest. Quantitative trait loci (QTLs) corresponding to the development of melanoma, a synthetic trait (ST) which combined all phenotypic information, were first detected. Next, 7 specific traits corresponding to precise tumour characteristics were studied, leading to the detection of additional and more significant QTLs. Furthermore, black pigs with the MC1R*2 allele (EMBL accession numbers: MC1R*2, AM231528; MC1R*4, AM231529; MC1R*6, AM231530.) were found to be significantly predisposed to melanoma. All our findings indicate that the development of melanoma in pigs is a multigenic process and comparative mapping revealed human regions of interest in the search for new melanoma susceptibility candidates.


CGH, comparative genomic hybridization; CI, cutaneous invasion; cM, centiMorgan; CM, cutaneous melanoma; HSA, Homo sapiens chromosome; LRT, likelihood ratio test; LT, lesion type; MC1R, melanocortin receptor 1; MeLiM, Melanoblastoma-bearing Libechov Minipig; MET, presence of metastasis; NAM, number of aggressive melanoma; NL, number of lesions; NM, nodular melanoma; PAB, presence of melanoma at birth; QTL, quantitative trait locus; RH, radiation hybrid; SSC, Sus scrofa chromosome; SSCP, single-stranded conformational polymorphism; SSM, superficial spreading melanoma; ST, synthetic trait; ULC, ulceration.

Material and methods


The MeLiM pig strain was created and maintained at the Institute of Animal Physiology and Genetics in Libechov (Czech Republic) and was derived from crossings with several breeds (Hormel, Göttingen, Canadian Landrace, Cornwall, Large White and Vietnamese). Within this strain, CM appeared spontaneously, and selective breeding led to a closed stock in which melanoma is inherited and occurs, both in utero and postnatally, in 50% of the progeny.16 The MeLiM strain is a suitable model for human melanoma because of its histopathological similarities. MeLiM animals were imported from Czech Republic and a French herd was bred at LREG (CEA-INRA) in Jouy-en-Josas, France, for physiological and genetic studies.

As shown in Figure 1, 4 affected MeLiM pigs (3 females and 1 male) were mated to 5 healthy Duroc pigs (2 females and 3 males). Nine of the resulting affected F1 pigs (6 females and 3 males) were crossed with 26 healthy Duroc pigs (21 females and 5 males) to produce 331 backcross (BC) pigs which included the 79 BC used in our preliminary genome-scan which derived from F1 pigs: 91016, 81014, 81015.21 Thirty millilitres of fresh blood was collected from each animal for lymphocyte isolation.

Figure 1.

Origin of F1 pigs. Asterisks signify that these 4 MeLiM pigs, which can be traced back to MeLiM brothers H10 and H30, were used for production of F1 animals. Filled squares and circles represent affected animals, and empty squares and circles unaffected animals. Animals with letter D are of Duroc origin.

The F1 and BC families were produced on the experimental farm of Rouillé (INRA, France). Permission for our animal experiments was obtained from the Animal Protection Office of the French Ministry of Agriculture and Forestry (78-16).


Pigs were classified according to their clinical and histopathological characteristics.17 Briefly, they were diagnosed by clinical follow-up over 6 months, except for the 79 BC used in our preliminary study,21 which were only followed up for 3 months. Flat and raised pigmented cutaneous lesions were observed, photographed and recorded, at birth and then once a month. A search for sentinel lymph nodes was performed by palpation. For each piglet of the BC progeny, all raised lesions and each type of flat lesion were histologically analysed according to the human classification, and lesion-bearing animals were autopsied for the detection of inner metastases. All raised lesions were classified as superficial spreading melanoma (SSM) or nodular melanoma (NM), with Clark's level II–V for invasion criteria. Flat lesions were classified as SSM restricted to epidermis, atypical melanocyte proliferation or lentigo.

On the basis of these characteristics, a categorical ST combining all the clinical and histopathological information on melanoma was defined by 5 categories, from absence of melanoma (Phenotype I) to very aggressive forms (Phenotype V), as follows: Phenotype I, pigs with no pigmented lesions; Phenotype II, pigs with only flat lesions which were either histologically benign or could not be histologically analysed because of their large number; Phenotype III, pigs bearing small raised lesions with a slow growth phase (SSM or NM-Clark's levels II–IV) or flat lesions classified as SSM-Clark's level I; these lesions appear during the first 3 postnatal months; Phenotypes IV and V, pigs with one or more large raised or polypoid lesions, often ulcerated and associated with metastasis; these lesions correspond to deeply invasive melanoma (SSM or NM-Clark's level IV–V) and mostly develop in utero or during the first month of postnatal life. Some of them exhibited fast growth (Phenotype V, also called Malignant A), and others, slow growth (Phenotype IV or Malignant B).

Note that for Phenotypes III to V, some pigs had additional lesions corresponding to a lower grade.

We recorded the trait values for sex and coat colour, and for the following specific traits: presence of melanoma at birth (PAB), number of lesions (NL) including all types, number of aggressive Malignant A lesions (NAM), presence of metastasis (MET) defined by 4 categories (no metastasis, adenomegaly detected by palpation, lymph node metastasis confirmed by histology and lymph node metastasis associated with visceral metastasis confirmed by histology) and for the most severe lesion on each animal, histological lesion type (LT) i.e. SSM or NM, cutaneous invasion (CI), graded according to Clark's levels I–V and clinical ulceration (ULC). Two traits (PAB and ULC) were binary; the other specific traits displayed a categorical distribution.

Genotyping data

DNA from each animal was extracted from lymphocytes isolated from 10 ml of fresh blood, and the DNA concentration was adjusted to 10 ng/μl. To cover the entire swine genome, we selected microsatellite markers from panels for pig genome mapping (INRA Toulouse) and tested them first on the F1 parents to establish the allelic distribution and parental information content. Some genotyping data used for the first genome scan20 were combined with the newly obtained data for linkage analysis, depending on the informativity level of the markers. A total of 153 microsatellite markers was selected (supplementary Table I) and the average coverage was estimated at ∼20 cM.

Microsatellite genotyping

Large-scale genotyping was mainly performed on an automated sequencer (Applied Biosystems 3700) at the CRGS Platform of the Toulouse Genopole, France. PCR reactions were run on a Gene Amp System 9700 thermocycler (Applied Biosystems) in 10 μl final volume, containing 2 μl fluorescent end-labelled primers (1 μM each), 2 μl 10× PCR buffer, 1 μl dNTP nucleotide mixed solution (20 μM for each nucleotide), 2.96 μl H2O, 0.2 U of GoTaq® (Promega) and 2 μl DNA(10 ng/μl). The 2 μl of DNA solution and 8 μl of PCR mix solution were distributed among 384 microplates using a TECAN robot (Genesis 200 × 8). Samples were preheated for 5 min at 94°C, run for 35 cycles (94°C for 20 sec, 55°C for 30 sec and 72°C for 30 sec), and heated to 72°C for a final extension step of 5 min. After amplification, PCR products were pooled, and then analysed on the capillary sequencer (Applied Biosystems 3700). PCR fragment size was analysed using GeneScan and Genotyper v3.7 analysis software (Applied Biosystems), and was stored in the GEMMA database.24 γ33P polyacrylamide gel was used to complete the genotyping and also to compare allele identification by these 2 genotyping methods.21 Details about methods used in the different families are described in the supplementary Table I.

Table I. Family Structure of F1 Pigs with Melanoma, Backcrossed with Healthy Duroc Pigs
F1 pig numbersSexBackcross familyCoat colourPhenotype classification1
Number of familiesAverage number of progenyVIVIIIIIISum
  • 1

    Phenotype I, pigs with no pigmented lesions; Phenotype II, pigs with only flat lesions; Phenotype III, pigs bearing small raised lesions with a slow growth phase (SSM or NM–Clark's levels II to IV) or flat lesions classified as SSM-Clark's level I; Phenotypes IV and V, pigs with melanomas; some of them exhibited fast growth (Phenotype V, also called Malignant A), while others slow growth (Phenotype IV or Malignant B).

Total 42  283011847108331

Sequencing and SSCP genotyping of MC1R

We first sequenced the overlapping fragments from the entire coding region of the MC1R gene using the following 3 pairs of primers derived from the swine MC1R sequence (Gene accession number: AF326520). These primers amplified respectively sequence regions 401–1209 (5′-CAAGGAGCCAGGACCAACT-3′ and 5′-AGATGAGCACGTCCATGACA-3′), 1041-1624 (5′-CTGCACTCGCCCATGTACTA-3′ and 5′-AGGGAGAGGTGCAGGAAGA-3′) and 1446-2125 (5′-GCGGTACTGTACGTCCACAT-3′ and 5′-GAGGTCCTGCAGTGAGCAAC-3′). PCRs were performed on 25 ng DNA with 2 mM MgCl2. The optimized PCR conditions were 96°C for 5 min, then (45 sec at 95°C, 45 sec at 61°C, 61°C or 58°C, respectively, for the 3 pairs of primers, and 90 sec at 72°C) for 35 cycles, and finally, 72°C for 10 min. PCR products were then sequenced according to recommended procedures (Genome Express). After sequence analyses, the PCR-SSCP (single-stranded conformational polymorphism) method was used to detect polymorphisms of codons 17 and 22 on 478 animals (MeLiM, F1 and backcross), using the following primers to amplify sequence region 738–910 (5′-CACCTCTGGGAGCCATGA-3′ and 5′-CGTCTGGTTGGTCTGGTT-3′). PCR conditions were 94°C for 5 min, (45 sec at 95°C, 45 sec at 60°C, 45 sec at 72°C) for 35 cycles, and 72°C for 10 min. Seven microlitres of γ33P labelled PCR products were then mixed with 7 μl of loading buffer (final NaOH: 2 M), then heat-denatured at 94°C for 10 min and chilled on ice. Lastly, PCR products were separated on nondenaturing 5% polyacrylamide gel, for 9 hr at 4°C, in 0.5× TBE buffer. The PCR amplification in our families can lead to 2 sizes of fragment, a 173-bp fragment corresponding to alleles MC1R*2 and MC1R*4, and a 175-bp fragment corresponding to allele MC1R*6. Because of the differences in the sequence of these fragments in the 3 MC1R alleles, 3 conformational single-strand doublets can be detected after electrophoretic migration, depending on animal genotype.

Radiation hybrid mapping

Using the pig-hamster radiation hybrid (RH) panel,25 high-throughput comparative RH mapping, focusing on homologous genes between HSA9 and SSC1, was performed with primers designed to amplify swine genomic DNA according to human gene conservative sequences (Kagoshima University and National Institute of Agrobiological Sciences, Tsukuba, Japan). Information regarding the gene primers used is available on request.

Data analysis

All data were imported into and managed by the statistical analysis system SAS 8.2, on the AIX2 platform (SAS Institute). Genotyping data not in the Mendelian inheritance were discarded. The genetic map was evaluated using CRI-MAP,26 and marker order and positions were compared with sex-averaged maps from the U.S. pig genome databases (, and ( and from the Roslin Institute ( database. Marker order and most of the genetic distances between markers were in agreement with the public sex-averaged map, with only slight differences for the 4 marker intervals SW461-SW1466 (SSC2), SW268-S0376, SW1551-S0178 (SSC8) and SW1-SW1349 (SSC9). We therefore used the public map for interval mapping analyses.

In humans, melanoma penetrance might be affected by sex and by skin and hair pigmentation,5 and in pigs, by sex and coat colour.21 Therefore, the ST was first analysed as raw data, then after correction for sex and coat colour (PROC GLM/SAS) and lastly for sex, coat colour and MC1R genotype. For each specific trait, only the results after correction for sex and coat colour are given. PROC FREQ (SAS) was used to test the effects of sex and coat colour on tumour incidence. The general linear model (PROC GLM/SAS) was used to test the effects of individual markers on melanoma occurrence with a regression model27 and, similarly, to test the effects of paired markers. Finally, the transmission-disequilibrium-test was performed by the quantitative trait disequilibrium test (QTDT)28 to trace the parental alleles associated with susceptibility to melanoma.

Models for QTL detection

For QTL detection, interval mapping using maximum likelihood method proposed in livestock, comprising a mixture of full and half-sib families,29, 30 was applied to our data. No crossover interference was assumed. All pedigree information was used to determine the transmission phase of marker alleles: the first generation consisted of MeLiM and Duroc individuals, mated to obtain F1 males and females. The backcross generation was produced by mating F1 with new Duroc individuals. Only the most probable F1 genotype was kept in the calculation, and all genotypes of their mates were estimated with a probability >0.1, so that the likelihood was partially linearized within full-sib families. No assumption was made concerning the number of QTL alleles. The model for a given trait determined by one QTL was

equation image

where yijk is the phenotype of the kth offspring in the family of the ith F1, mated with the jth Duroc (i = 1,…,n; j = 1,…,ni; k =1,…,nij), μ is the averaged family value, μmath image and μmath image are respectively the family means of ith F1 and its jth mate, gSi and gDij are the fixed effects of the QTL genotype of each F1 animal and its mate, respectively (a and −a for homozygotes QiQi and qijqij, and d for heterozygote Qiqij), and eijk is the residual random effect, assumed to be normally and independently distributed in the ith family, with a variance of σmath image. In practice, for each F1 individual, the family mean, residual variance and QTL substitution effect were estimated, and for each Duroc individual, only the family mean and QTL substitution effect.

Three hypotheses were tested for a given trait: the usual null hypothesis “no QTL on the chromosome” (H0), the hypothesis of one QTL on the chromosome (H1), and the hypothesis of two linked QTLs on the chromosome (H2). For each trait, H0versus H1 was tested to exclude the absence of QTL from each chromosome. When necessary, H1versus H2 test was performed for particular chromosomal regions. For the two-QTL model, we used the grid-search method, with QTL effects modelled as gSi = gSip + gSiq and gDij = gDijp + gDijq, where gSip, gDijp, gSiq, gDijq, are fixed QTL effects of F1 animal and its mate at the p and q positions which were considered jointly.

Significance levels were determined empirically by 1,000 Monte Carlo simulations. To take into account missing genotypes, differences in marker density and family structure, only trait values were simulated under the null hypothesis model (normal distribution without QTL). The approximated Bonferroni correction of the Type I error test was applied to the chromosome-wide significance levels to obtain the significance levels for genome-wide linkage.31 Briefly, the significance level for a specific chromosome is proportional to its contribution (r) to the total autosomal genome length. So, the genome-wide significance level, Pgenome-wide = 1 − (1 − Pchromosome-wide)1/r, is used to take account of testing the whole autosomal genome.

In addition, most of the animals in this swine model developed melanoma in utero or in the perinatal period, suggesting that some aberrant epigenetic or imprinting effect might be responsible for melanoma. Consequently, QTL detection depending on parental origin (sire versus dam) was performed on SSC2, which harbours an imprinting region.


Family study

F1 animals were produced from 4 closely-related affected MeLiM pigs (B143, 7, 9025 and B310), whose origins could be traced back to MeLiM brothers H30 and H10, and healthy Duroc animals (91291D, 75015D, 51761D, 91299D and 270D, Fig. 1), so that affected F1 pigs were expected to be heterozygous for potential susceptibility loci. Table I shows that, in all, 331 backcross animals (42 families), with 7.9 animals per family (range: 5.0–10.5), were generated by crossing 9 affected F1 pigs with 26 healthy Duroc animals for this genome-wide linkage study. Alleles of melanoma susceptibility loci were expected to segregate in this backcross generation. Thirty percent of the F1 animals and 17.5% of the backcross animals developed melanoma (aggressive phenotypes IV and V), suggesting a complex melanoma inheritance with incomplete penetrance. The tests on the frequencies (PROC FREQ statistical analysis system, SAS) revealed that the melanoma occurred independently of the sex but not of coat colour as pigs with a black or grey coat developed significantly more melanoma than red pigs (p < 0.05).

QTL mapping of the synthetic trait

To search for genetic linkage between genomic regions and melanoma development, the interval mapping method was applied to the genome scan performed with 153 microsatellite markers on 331 backcross animals. First, the single-QTL model was applied to a ST corresponding to the development of the tumours, after correction of the phenotypic data for sex and coat colour. QTLs were detected, at a significant chromosome-wide level only, on SSC1 (49.4 and 88.0 cM), 15 (86.5 cM) and 17 (39.8 cM) at a significant level (p < 0.05), and on SSC13 (56.0 cM) at a very significant level (p < 0.01, Fig. 2). On SSC1, a QTL was also detected with raw data (p < 0.05), with a maximum likelihood ratio test (LRT) at 96.0 cM, suggesting multiple locus segregation on this chromosome. In addition, a significant QTL was revealed on SSC2 at position 16.9 cM (p < 0.05), in families derived from F1 males only. Three additional QTLs were detected at a suggestive chromosome-wide significance on SSC2, 7 and 14 (p < 0.1). The maximum LRTs and their positions are shown in Table II. Significant contributions of F1 individuals to the LRT are indicated, showing only one to three significant QTLs for some families, because of either differences in the informativity of genetic markers and the number of animals in each family or to different locus segregations.

Figure 2.

Likelihood ratio test (LRT) statistic profiles. LRT statistic profiles for QTL detection (1-QTL model) on the synthetic trait (raw data and data corrected for sex and coat colour) on SSC1 (a), SSC13 (b), SSC15 (c), SSC17 (d), and SSC2 (e). A significant QTL on SSC 2, after statistical adjustment, was only detected with animals derived from F1 males. Thresholds are indicated at chromosome-wide level. Triangles indicate microsatellite marker positions.

Table II. Interval Mapping Results for the Synthetic Trait in the Backcross Population
SSCPosition (cM)Nearest markerTest statisticP1Significant QTLs at F1 family level
  • 1

    Chromosome-wide significance (**, 1%; *, 5%; +, 10%). These QTLs are not significant at genome-wide level after Bonferroni correction.

  • 2

    Analysis of families from F1 males.

149.4SW165384.9*  * *    
188SW102080.6*    *    
29.9SW262370.5+    *    
74.0S002572.8+    *    
1356.0SW244893.0**        *
1446.0SW170972.4+  **      
1586.5SW93679.4*      *  

A grid search for two linked QTLs conducted on SSC1 for the ST, corrected for sex and coat colour, showed a pair of QTLs at 31.0 and 87.0 cM, with a very significant LRT (120.4, p < 0.01, Fig. 3a).

Figure 3.

Two-QTL model detection on SSC1, without (a) or with (b) correction for the MC1R genotype. (a) Arrow indicates that two QTLs at 31 and 87 cM are located on SSC1, using the synthetic trait corrected for sex and coat colour effect, and without correction on the MC1R genotype. (b) With further correction for MC1R genotypes, one pair of QTLs at 77 and 85 cM was mapped. Horizontal x-and y-axes: chromosomal genetic positions in centiMorgan; vertical z-axis: likelihood ratio test value.

QTL mapping of specific traits

Seven precise specific traits were used with the aim of detecting more significant QTLs underlying different aspects of melanoma development, including clinical ulceration (ULC), presence of melanoma at birth (PAB), histological lesion type (LT), number of aggressive melanomas called “Malignant A” (NAM), presence of metastasis (MET), cutaneous invasion (CI) and the total number of lesions (NL). Application of the single-QTL model to these 7 traits revealed 5 QTLs, respectively on SSC10 (ULC), 12 (PAB), 13 (LT) and 16 and 17 (NAM) at a highly significant chromosome-wide level (p < 0.001, Table III and Fig. 4) and at a significant genome-wide level after approximated Bonferroni correction (p < 0.05). Five other QTLs were respectively found on SSC1 (NAM), 14 (LT), 4, 12 and 14 (ULC), at a very significant chromosome-wide level (p < 0.005) and a suggestive genome-wide level (p < 0.1). For those 7 specific traits, 21 additional QTLs were detected at a significant chromosome-wide level (p < 0.05 and p < 0.01). Some QTLs for different traits were located close to each other on the same chromosome (e.g., PAB and ULC on SSC12; NAM and CI on SSC1), suggesting that close or pleiotropic gene(s) might segregate for those traits.

Figure 4.

Likelihood ratio test statistic profiles of significant QTLs identified for 4 specific traits (1-QTL model). (a) Presence of melanoma A at birth: only 1 QTL was detected for this trait, at one end of SSC12 and at a 1‰ chromosome-wide significance level, which corresponds to 5% genome-wide significance. (b) Lesion type: 2 genome-wide significant QTLs, at 5% and at a suggestive 10% level, on SSC13 and SSC14, respectively. (c) Number of melanoma A tumours: two 5% genome-wide significant QTLs on SSC16 and SSC17, respectively, and 1 suggestive 10% genome-wide significant QTL on SSC1. (d) Ulceration: one 5% genome-wide significant QTL on SSC10, and 3 other suggestive 10% genome-wide significant QTLs on SSC4, SSC12 and SSC14, respectively.

Table III. Quantitative Trait Loci Detection of Specific Traits
TraitSSCPosition (cM)Nearest markerTest statisticP1
  • 1

    Chromosome-wide significance: * 5%, ** 1% and *** 1‰.

  • 2

    Genome-wide significance: 5% (after Bonferroni correction).

  • 3

    Suggestive genome-wide significance: 10% (after Bonferroni correction).


MC1R and melanoma predisposition

We previously observed an association between melanoma development in our MeLiM breed and the microsatellite S0035 located at 7.5 cM on SSC6, 2–3 cM downstream from the MC1R gene.21 To test the relationship between MC1R and melanoma, we first sequenced this gene on DNA from 6 animals representing the 5 different alleles of the marker S0035 detected in the breed. The sequences of the entire coding regions were compared to the full swine MC1R sequence and to the MC1R sequence of the European Wild Type Boar (EWB, Table IV). We found 3 MC1R alleles, MC1R*2, MC1R*4 and MC1R*6, as previously described,32 except that in all three, we observed an additional polymorphism in the second codon (Pro2, CCC to CCT). The MC1R*2 allele corresponded to black animals and to the S0035 alleles 1, 4 and 5, associated with black coat colour and melanoma. As more MeLiM and F1 families than backcross families segregated for black and red coat colour, we genotyped 478 animals for S0035 and MC1R, including all available MeLiM, F1 and backcross families. The quantitative transmission disequilibrium test (QTDT) showed highly significant associations with raw data between both MC1R*2 and S0035-1 alleles and melanoma development (P = 3 × 10−6 and 7 × 10−5 respectively). However, after stratification for sex and coat colour effects, only MC1R*2 which contained the Val95Met and Leu102Pro substitutions was significantly associated with melanoma (P = 4 × 10−2), suggesting a direct role for MC1R or a close gene in melanoma susceptibility.

Table IV. MC1R Gene Polymorphisms1
MC1R alleleS0035 alleleCoat colourBreedCodon
  • 1

    MC1R alleles were determined by sequencing and were compared in MeLiM and Duroc pigs, versus European wild boar (EWB).32MC1R alleles 2, 4 and 6 in the melanoma herd (MeLiM, F1 and backcross families) were determined after analysis of PCR amplification of codons 17 and 22 by single-stranded conformational polymorphism (SSCP) method.

MC1R*21,4,5BlackMeLiM- - T- - A A - -- C -- - C  - - A
MC1R*62RedMeLiM- - T + CC   A - -  
MC1R*42,3RedDuroc- - T      - T -A - -

The frequency of melanoma according to MC1R genotypes (Fig. 5) revealed that all the 17 pigs homozygous for the MC1R*2 allele developed melanoma. Animals heterozygous for this allele also developed more melanoma than the others, mainly when MC1R*2 was associated with the MeLiM allele MC1R*6 instead of the Duroc allele MC1R*4. These results also suggest that MC1R or a closely linked gene plays an important role on the development of melanoma and that a close linked recessive gene may also cooperate. Application of the single-QTL model to the MeLiM, F1 and backcross families as a whole, for the ST, revealed a QTL at the MC1R position (p < 0.01) with only raw data, suggesting that pigmentation also affects melanoma development (data not shown).

Figure 5.

Association between the risk of melanoma and the MC1R genotype. The figure represents the frequency of melanoma according to MC1R genotypes. The synthetic trait value (y-axis) is plotted against each MC1R genotype combination of alleles 2, 4 and 6, as defined in Table IV (x-axis).

Furthermore, when the general linear model (GLM) was used to search for 2 marker interactions between MC1R and other microsatellite markers over the whole genome, interactions were observed between MC1R and the markers S0316 and SW1653 located respectively at 33.4 and 49.4 cM on SSC1 (p < 0.05). When the ST value was corrected, not only for sex and coat colour, but also for MC1R genotype, the QTL at 49.4 cM on SSC1 was no longer detected, but a new QTL at 84.0 cM was mapped (Fig. 6). In addition, application of the two-QTL model to these phenotypic data revealed a pair of QTLs at positions 77.0 and 85.0 cM on SSC1 (LRT: 109.0, p < 0.001, Fig. 3b).

Figure 6.

Likelihood ratio test statistic profiles for significant QTLs detected on SSC1 (1-QTL model). The different coloured curves correspond to the different QTLs detected on SSC1. Arrows indicate the significant peak positions for each trait. One asterisk: significant at 5% chromosome-wide level. Two asterisks: 1% chromosome-wide significance. Blue curve: one QTL for cutaneous invasion located at 85 cM (5% chromosome-wide level). Green curve: a QTL for tumour number of melanoma A located at 78 cM, (1% chromosome-wide level). Red dotted curve: a QTL identified for raw data at 96 cM (5% chromosome-wide level). Light blue curve: a QTL at 49 cM detected after correction for sex and coat colour (5% chromosome-wide level). Violet curve: a QTL at 84 cM detected when data were further corrected by MC1R genotype (5% chromosome-wide level).

Potential candidates for melanoma susceptibility

We used large-scale comparative mapping to select potential candidate melanoma susceptibility genes in homologous genomic regions between human and swine. As 3 potential QTL regions might exist on SSC1, which contains the HSA9p21 human counterpart, massive RH mapping was done in swine for the genes lying on HSA9, to establish the relative positions of orthologous genes between HSA9 and SSC1. We found that 2 distinct regions on SSC1 were related to HSA9p21 (Fig. 7). One region located at about 81.0 cM (SSC1q25), which might contain 2 melanoma QTLs, and the other, located at about 56 cM, which is included in the region displaying significant linkage with ST.

Figure 7.

Comparative mapping between the HSA9 and QTL locations on SSC1. Massive RH mapping in swine of genes lying on HSA9 revealed that genes homologous to HSA9p21 genes are present in 2 genomic regions on SSC1, at about 56 and 80 cM, corresponding to 2 statistically significant peaks (p < 0.01). The vertical dashed and dash-dotted lines represent the significant thresholds at 5 and 10%, respectively.

Comparative mapping was then done for other swine QTL regions by scanning the literature and available public genome mapping databases for humans and swine (see Table V). The human chromosomal regions corresponding to the swine regions surrounding the main QTLs (peak position ±15 cM), as well as the breakpoints in QTL regions on pig chromosomes 1, 2, 13 and 16, are listed in Table V. Among the potential candidate genes located in the swine QTL human counterparts, we selected, on the OMIM site, the genes involved in melanocytic neoplasia and other malignancies, tumour progression, cell cycle regulation and DNA repair. As swine melanoma originates from neonatal neural crest-derived melanocytes (NC-M), we also selected genes involved in childhood tumours and in the signalling pathways of melanocyte development connected with the NC-M survival and terminal differentiation. This analysis revealed that the human counterparts of some swine QTL regions such as HSA 9p21 and 9q21, 1p36, 3p25 and 16q24 were previously shown to be related to melanoma. Similarly, some candidate genes have already been stated or suggested to play a role in human CM, like the genes CDKN2A, MC1R, MITF, and TP53. Lastly, potential candidates may lie on other human regions, mainly on HSA 1q32-q44, 3p26-q22, 5q34-q35 or 18p11, 17p13-q12 and 20p12-q11, for both ST and the different specific traits which are governed by various loci.

Table V. Putative Candidate Genes for MeLiM Melanoma Selected Through Swine-Human Comparative Mapping1
SSCTrait2Marker and QTL positions (cM)3Nearest markers or genesSwine cytogenetic band4Human homologous cytogenetic band5Human physical position (Mb)6Potential candidate genes7Functional involvement
  • 1

    Quantitative trait loci (QTLs) are shown by chromosome. Only porcine chromosome regions bearing at least two QTLs or one highly significant QTL (p < 0.001) are shown.

  • 2

    Traits are defined below, except for ST1 and ST2, which are synthetic traits respectively analysed by the one-QTL and two-QTL models. NAM, number of aggressive melanoma A tumour; CI, cutaneous invasion; LT, histological lesion type; ULC, ulceration; PAB, presence of melanoma at birth; MET, metastasis; EP, extreme phenotype (I and IV+V only, data not shown).

  • 3

    QTL positions are in bold type, as well as their significance level (see Tables I and II). Marker positions were retrieved from the MARC/USDA site. The following genes were located on the swine RH map: TEK, MTAP1 and TYRP1 (Fig. 7), CDKN2A,22 MC1R21 and IGF2 (3′UTR contains the microsatellite marker SWC9 at 1.1 cM).33

  • 4

    The cytogenetic positions of the swine genes and markers are derived from the INRA sites ( or the multispecies comparable table (

  • 5

    Data for human cytogenetics were isolated from the Ensembl database, after the position in Mb had first been determined by Robic et al.,34 or retrieved from Rattink et al.35

  • 6

    Corresponding HSA location in Mb derived from Robic et al.34 ( or from the Ensembl Human Genome Browser (, and from Demeure et al.36

  • 7

    Potential candidate genes and their disease involvement were derived from the OMIM site (, except for SRG and SNAI3, which were respectively described by Yuan et al.37 and Katoh et al.38

  • 8

    This QTL was mapped using all available animals, including the MeLiM, F1 and backcross families (data not shown).

1 23.5SW317 6q25.1149    
     6q24.1146.6GRM1604473Glutamate receptor metabotropic 1Melanocytic neoplasia
  29SW1332 6q24.2145.4    
 ST1, EP49.4*SW1653 6q1265.67    
  ∼55.8TEK (Fig.7) 9p21.227.15    
  58.5SW1430 6q14.181.1    
  60.2SW1668 1843    
  67.6SW2185 6q14.181    
  67.6SWR702 1544-5436    
  73S0331 18q22.367.8    
 ST2, NAM78**        
  79.4SW2073 15q22.260.9    
   EST(AR021B04) 1p36.2210.536    
     1p36.22 SRGYuan et al. 2005Survival-related geneNovel cell survival gene controlling apoptosis and tumorigenesis
  79.4SW1846 14q21-q2240-5336    
  ∼81CDKN2A1q259p21.321.96CDKN2A600160Cyclin- dependent kinase inhibitor 2AMelanocytic neoplasia
  ∼82.4MTAP1 (Fig.7) 9p21.321.85MTAP1156540Methylthioadenosine phosphorylaseInactivation in several carcinomas
  83.2S0142 9p22.316.3    
  ∼83.7TYRP1 (Fig.7) 9p2312.69    
 ST2, CI85*        
  93.9SW1462 9q21.276.8    
2    11p15.5 WT2194071Multiple tumour-associated chromosome region 1Wilms tumours
     11p15.50.52HRAS190020V-HA-RAS Harvey rat sarcoma viral oncogene homologBladder cancer; few mutations in human melanoma
     11p15.51.97H19103280H19 geneImprinting disrupted in childhood and adult cancers
  1.1IGF2 (near SWC9)2p1711p15.52.1IGF2147470Insulin-like growth factor 2Imprinting disrupted in childhood and adult cancers
     11p15.42.86CDKN1C600856Cyclin dependent kinase inhibitor 1CBeckwith–Wiedemann syndrome
     20q11.2231.73E2F1189971E2F transcription factor-1Cell cycle regulation, tumour suppressor in mice
     11q12.260.8DDB1600045DNA damage-binding protein 1DNA repair
     11q13.164.33MEN1131100Multiple endocrine neoplasia type 1Endocrine neoplasia; few mutations in human melanoma
     11q13.165.17SIPA1602180Signal-induced proliferation associated gene 1Tumour metastasis candidate
     11q13.265.86BRMS1606259Breast cancer metastasis suppressor1Breast cancer metastasis suppressor
     11q13.266.9RAD9A603761RAD9ADNA repair
     11q13.369.17CCND1168461Cyclin D1Cell cycle progression, amplification or overexpression in several cancers
  24SWR783 11q12-q1335     
6ST14.5*8MC1R6p1516q24.388.5MC1R155555Melanocortin 1 receptorGenetic risk factor for melanoma and nonmelanoma skin cancer.
     16q24.388.3CDK10603464Cyclin-dependent kinase 10Cell cycle regulation
     16q24.388.2FANCA607139Fanconi anaemia, complementation group ADNA repair
     16q24.387.5GAS 11605178Growth arrest-specific 11Expressed in mice during growth arrest
     16q24.387.2SNAI3Katoh et al., 2003Snail homolog 3, DrosophilaTranscriptional repressor family implicated in carcinogenesis and embryogenesis
7 4.0S00257p13      
   EDN17p12-p136p24.112.4EDN1131240Endothelin 1Mitogenic and melanogenic factor; regulation of melanoma invasion promoters
10 42S0351 1q44241.2    
12  TP5312q12-q1417p13.17.5TP53191170Tumour protein p53, Li-Fraumeni syndromeCell cycle regulation DNA repair; predisposition to early-onset melanoma
     17q11.226.55NF1162200Neurofibromatosis type ICases of melanoma in NF1 patients
13  MITF13q23-q243p14.2-p14.169.95MITF156845Microphthalmia-associated transcription factorNC-M survival and terminal differentiation (amplified in melanoma)
     3p25-p22 OVCAS1607893Ovarian cancer, epithelial, susceptibility toOvarian cancer susceptibility loci
     3p25.38.94RAD18605256RAD18 homolog, S. cerevisiaeDNA repair
     3p25.39.77OGG16019828-oxoguanine DNA glycosylaseDNA repair
     3p25.310.16VHL608537von Hippel-Lindau tumour suppressorPredisposition to several neoplasia
     3p25.310.08FANCD2227646Fanconi anaemia, complementation group D2DNA repair
    13q23-q263p25.212.64RAF1164760v-raf-1 murine leukaemia viral oncogene homolog 1RAS/MAPK pathway
     3p25.2 UBM2606661Melanoma, uveal, susceptibility to, 2Uveal melanoma susceptibility Locus
     3p25.113.86WNT7A601570Wingless-type MMTV integration site family, member 7AWNT-β-catenin pathway
   XPC13q31-q323p25.114.17XPC278720Xeroderma pigmentosum, complementation group CDNA repair
     3p25 ST11602011Suppression of tumorigenicity 11, pancreasPancreatic endocrine tumour suppressor
   TFDP213q323q23143.22TFDP2602169Transcription factor Dp-2, E2F dimerization partner 2Cell cycle regulation-interaction with E2F transcription factor
  61.1SW960 3p26.15.2    
  71.7SW968 3p12.378.7    
  79.3SW955 3q13.11104.6    
15  GDF815q232q32.2190.7    
     2q33189.4DIRC1606423Disrupted in renal carcinoma 1Disrupted in familial clear cell renal cancer
     2q33.1201.9CASP8601763Caspase 8Loss leads to neuroblastoma metastasis
     2q33202.7FZD7603410Frizzled homolog 7, DrosophilaFamily of receptors for WNT proteins WNT-β-catenin pathway
  76SW1945 2q33.2204.7    
     2q33.3208CREB1123810cAMP responsive element binding protein 1Major nuclear target of the cAMP pathway
     2q33.3208.5FZD5601723Frizzled homolog 5, DrosophilaWNT-β-catenin pathway
     2q35215.4BARD1601593BRCA1-associated RING domain 1Potentially involved in breast cancer predisposition
     2q35216.8XRCC5194364X-ray repair complementing defective repair in Chinese hamster cells 5DNA repair
 ST1, EP86.5*        
     2q35219.56WNT6604663Wingless-type MMTV integration site family member 6Overexpressed in cervical and colorectal cancer cell lines
   PAX315q252q36.1222.9PAX3606597Paired box gene 3Melanocyte terminal differentiation
16  S039016q21      
     5q35.2176.4FGFR4134935Fibroblast growth factor receptor 4Cell motility and cancer progression
     5q35.1172.1DUSP1600714Dual specificity phosphatase 1Inhibition of MAPK activation
     5q35.1170.8FGF18603726Fibroblast growth factor 18Cell mitogenesis and survival
  44.2SW977 5q35.1169.7    
  44.8SW1454 18p11.314.4    
  46.9SW262 5q34161.1    
17 32S029617q13-q1420p12.36.7    
     20p12.35.04PCNA176740Proliferating cell nuclear antigenCofactor for DNA polymerase delta; fidelity of mammalian DNA replication
     20q11.2231.7E2F1189971Transcription factor E2F1Cell cycle regulation, tumour suppressor in mice
  56.4SW1920 20q11.2232.9    


In the present study, we reported the first significant genome-wide scan for melanoma susceptibility in animals. This study combined the data acquired in our preliminary study21 of the MeLiM swine model conducted on 79 backcross pigs with data obtained for 252 additional swine. We therefore increased the number of families and also paid special attention to the phenotypes, using more precisely defined specific traits in order to improve power of QTL detection.

We detected four significant QTLs for the development of melanoma, defined by a ST combining all clinical and histological data, on chromosomes 1, 13, 15 and 17, irrespective of the sex of the affected F1 pigs. They were significant at the chromosome-wide level only, the QTL on SSC13 being the most significant (p < 0.01). A fifth QTL was detected on SSC2, when data from affected F1 males only were used. Analysis of the data for 7 more precise traits (PAB, LT, NL, NAM, CI, ULC, and MET) revealed 31 additional QTLs. Five of these QTLs corresponding to 4 different traits (ULC, PAB, LT and NAM) on chromosomes 10, 12, 13, 16 and 17, were highly significant at chromosome-wide level and reached genome-wide significance.

Compared with our previous analysis for the ST,21 we observed almost the same LRT profile on SSC1 in the present study, whereas the peak detected on SSC2 decreased in significance. The previous association on SSC6 in the MC1R region was also confirmed by the finding that MC1R*2 allele was related to melanoma development. QTLs for ULC were observed on SSC7 (73 cM) and SSC8 (60 cM) in the regions where associations were found with melanoma in the preliminary study.

A large number of QTLs was detected here but only 5 reached genome-wide significance. We used our QTLMAP program to analyse a mixture of full-sib and half-sib families.30 This program is designed to analyse continuous normally distributed traits in familial populations. The traits studied here were hardly within the normality hypothesis; ST (classes I to V) and some specific traits were categorical, and some were binary (PAB and ULC), usually with an unbalanced number of progeny in each class. Specific models have been previously developed within the QTLMAP software to account for such distributions of the data, using data transformations and threshold models.39 However, when the power of these models was compared with that of the usual Gaussian models, the general conclusions agreed for robustness of the Gaussian model. In the present study, data transformations of melanoma traits (not shown) were explored but did not significantly improve the analyses, thus justifying the choice of robust Gaussian models, although this choice may have reduced the general power of the analysis. The relatively small number of highly significant QTLs may also be due to the fact that, as in human models for melanoma occurrence, the traits analysed here are still complex and correspond to various physiological processes, especially the major trait (ST) which is a global symptom, interpreted as the result of numerous interacting and multigenic determinisms, from the time of disease onset to the severe stage. Such traits might therefore be affected by several QTLs. Analyses of traits specific for the different development phases of melanoma indeed led to higher significance, thus refining the individual phenotypes linked to the QTLs detected for ST. In addition, the recent origin of the MeLiM population, a hexahybrid cross obtained in the 1980s, appeared to be a source of the diversity of the haplotypes segregating in the F1 families. This diversity allows the detection of more QTLs but might cause the dispersion of heterozygous loci in different families, thus reducing the general power for the detection of individual loci. This reduction might be amplified by differences in family size and marker informativity. This is very similar to the multigenic inheritance observed in human populations exhibiting genetically linked melanoma. To overcome those limits, new genetic markers will be tested in the regions harbouring the most significant QTLs and new families will be produced, to obtain new informative recombinations and refine the locations.

Among the 5 most significant QTLs detected in this study, the QTL for ULC on SSC 10 (42 cM) lies in a region homologous to the human 219-241 Mb area on HSA1q41-q44 (Table V). At this time, we did not find any potential candidate on the OMIM site. The QTL for PAB on SSC12 (95.6 cM) was mapped in the SSC12q12-q14 region containing the gene coding for the most common tumour suppressor p53, already suggested as being involved in predisposition to early-onset melanomas in humans.1 This candidate is in contradiction with the fact that the SSC12q12-q14 region is included in the region of chromosome gain (SSC12q) identified by comparative genomic hybridization (CGH) analysis on both SSM and nodular melanomas in our pigs.40 However, this chromosomal gain correlated with the DNA amplification observed on HSA17q in human neuroblastoma, a childhood neoplasm that also arises from neural crest-derived cells.41 The QTLs for LT and CI observed around 81 and 71 cM, respectively, on SSC13q are in agreement with the recurrent loss of SSC13q36-q49 restricted to nodular melanomas previously observed in our CGH study.40 This observation reinforces the idea that this region plays an important role in tumour invasion, since the growth of nodular melanoma, contrary to SSM, starts both vertically and superficially. Note that the two QTLs may be governed by the same gene. Around 56 cM, also lies the most significant QTL for ST, underlining again the importance of this region for the development of melanomas in our pigs. The human counterpart for the 3 QTLs lies from HSA3p26 to 3q22 and harbours not only the melanoma uveal susceptibility locus UBM2, but also numerous other potential candidate loci involved in cancer susceptibility, DNA repair, cell cycle regulation and signalling pathways suggested to play a role in melanoma (Table V). For example, MITF, the master melanocyte regulator involved in the regulation of pigment genes and the melanocyte lineage survival, was recently shown to function as an oncogene in primary melanocytes bearing the V600E mutation in BRAF and MITF amplification in melanoma was found to correlate with a decrease in patient survival.42 Finally, two QTLs highly significant were detected for NAM on SSC16 (45.3 cM) and SSC17 (44.8 cM). The regions harbouring these QTLs on SSC16 and SSC17 correspond respectively to the HSA5q34-q35 (or HSA18p11.31) and to the HSA20p12-q11 regions. Both of them contain potential candidates listed in Table V. Like those on SSC13, additional QTLs were also detected on SSC17 in the region between 35.8 and 52.8 cM, including QTLs for the development of melanoma, ST and EP, MET and CI, thus underlining the interest of this region and that of the corresponding human region.

The similar LRT profiles observed on SSC1 in our 2 separate studies and the fact that several QTLs lie in a swine region corresponding to several human counterparts displaying linkage with human melanoma (1p36, 9p21 and 9q21), reinforce the confidence in true QTL segregation on this chromosome, despite the low level of significance. The results with the single-QTL model revealed 2 significant areas around 49.4 and 88 cM, suggesting the presence of at least 2 QTLs on this chromosome. Then, analysis with the two-QTL model, the search for interactions between MC1R and chromosome markers, and the correction for the MC1R genotype suggested the existence of at least 3 QTLs on SSC1, one in the 33.4–49.4 cM region interacting with MC1R, and two others at positions 77 and 85 cM, respectively. The overall results on SSC1 are shown on Figure 6. Analysis of specific traits led to a model in which one or both these QTLs might determine the relationship of NAM to tumour development and that of CI to the invasion, 2 traits which showed significant QTLs with the single-QTL model. The 77–85 cM region contains the CDKN2A gene, but in the affected MeLiM parental strain, this gene was excluded as a major susceptibility gene in the first family by haplotype analysis of the 2 cM area surrounding CDKN2A,23 suggesting the existence of other susceptibility gene(s) in this chromosomal region. The possible presence of additional melanoma suppressor genes around the CDKN2A gene has already been suggested in humans, as a large proportion of families without a CDKN2A mutation displayed linkage to the 9p21 region. Evidence for the presence of additional tumour suppressor gene(s) in this region was provided both in familial analyses and in tumours, as well as in in vitro experiments using microcell hybrids.43, 44, 45 Our comparative mapping, using all available data, suggests that potential tumour susceptibility candidates such as the MTAP gene, located at about 82 cM (Fig. 7), might be present on SSC1. The human MTAP gene is deleted in a subset of carcinomas,46 acts as a tumour suppressor in a breast cancer cell line47 and also seems to have a role in hepatocellular carcinoma development and invasiveness.48 In humans, the survival-related gene SRG located at 1p36.22, a region also linked to melanoma in certain human families,12 controls apoptosis and tumorigenesis.37 This gene, which is probably located at about 79 cM on SSC1 (Table V), might be a melanoma-susceptibility candidate in both humans and swine.

We found in this study that after stratification for sex and coat colour, the MC1R*2 allele, which is dominant for black coat colour, was associated with melanoma, suggesting that MC1R is directly involved in its development. This result is in agreement with the fact that MC1R variants in humans might also increase the risk of melanoma independently of UV exposure and other clinical risk factors linked to pigmentation,49, 50 although the underlying mechanism is still unknown. In human melanocytes, α-MSH via its binding to MC1R stimulates cAMP formation, pigmentation and proliferation.51 The swine MC1R*2 allele contains the amino acid substitution Val95Met which corresponds to the human substitution Val92Met. In humans, this substitution was first suggested to reduce the binding affinity of MC1R for α-MSH.52 However, human melanocytes homozygous for Val92Met have also been shown to increase significantly the response to α-MSH, suggesting that it did not induce a loss of function.53 The MC1R*2 allele contains also the amino acid substitution Leu102Pro suspected to induce constitutive activation of the MC1R receptor in animals with dark coat colour.54, 55 If that is indeed the case, a constitutive cAMP level might be sufficient to upregulate the genes involved in cell survival via the targets of the cAMP signalling pathway in black swine,56 and thus explain the association observed in our model between black coat colour or MC1R*2 and predisposition to melanoma. In humans, the common MC1R variants Arg151Cys and Arg160Trp that are highly associated with red hair and fair skin57 and confer high melanoma risk, are not “loss of function” receptors. A recent study revealed that they retained considerable signalling capacity,58 arguing for a direct role of MC1R independently of pigmentation that might be common in individuals with dark or light complexion. In humans, MC1R variants also modify predisposition to melanoma in segregating families with CDKN2A mutations.6, 7, 8 MC1R interactions detected in our study, with microsatellite markers on SSC1 located at 33.4 and 49.4 cM may also increase the effect of the alleles segregating for the QTLs on SSC1. In this study, only one marker, S0644 (∼81 cM), located within CDKN2A, has been genotyped in the MeLiM, F1 and backcross families. Interaction of MC1R and this marker was detected in these families (data not shown); however, new markers have to be tested to target the location of the highest interaction. The fact that red pigs also develop melanoma indicates that MC1R, or a close gene, is not a major gene, but acts probably as a modifier gene. We indeed cannot rule out the possibility that MC1R was a marker of other neighbouring gene(s) that play a role in carcinogenesis or in cell cycle regulation such as Gas11, SNAI3 or CDK10 (Table V).

To conclude, we found putative QTLs underlying the development of melanoma in swine. These results confirm and greatly extend those of our preliminary study. Comparative mapping revealed that some of these swine QTL regions are homologous to human regions known to be involved in human cutaneous or uveal melanoma such as 1p36, 3p25, 9p21, 9q21 and 16q24, whereas others correspond to additional regions that should be investigated in human melanoma research, mainly on HSA 1q32-q44, 3p26-q22, 5q34-q35 or 18p11, 17p13-q12 and 20p12-q11. Conventional fine mapping and large-scale tumour expression analyses combined with a candidate approach will be the next step. The availability of the swine whole genome sequence59 expected by the end of 2006 will greatly help to increase the density of markers, including single nucleotide polymorphisms, in the regions of interest, and to improve fine comparative mapping and detection of melanoma susceptibility candidates, both in humans and swine.


The authors are grateful to Mr. P. Bacon, Mr. F. Andréoletti and all the SESP staff for animal care. We are also grateful to Mr. Bussière of SEIA, Rouillé, France, for his assistance in swine artificial insemination. As a post-doctoral fellow, Z.Q.D was supported by fellowships from Région Ile-de-France and Foreign affairs departments from INRA and CEA. We thank Dr. Gérard Frelat for constant and active support for this project.