Vitamin D Receptor Polymorphisms Are Associated with Altered Prognosis in Patients with Malignant Melanoma¹

MM⁴ is the most serious cutaneous malignancy, and the prognosis of some tumors is very poor (1, 2). It is predominantly a disease of white-skinned people, and exposure to UV light is thought to be critical, although the relationship between risk and exposure is unclear (2). Other important risk factors for the occurrence of MM include presence of excessive numbers of banal nevi,multiple atypical nevi, fair skin, red hair, and blue or green eyes.

Breslow thickness at presentation remains the most important single prognostic factor for patients with cutaneous MM (3). In general, patients with thin tumors have a much longer survival than those with thick lesions; the 5-year survival rate for lesions <1.5-mm thick is 93%, compared with 67% for 1.5 mm-3.49 mm and 37% for ≥3.5 mm (4). Risk factors for thicker tumors, and hence poorer prognosis, include age at initial presentation and tumor site.

Relatively little is known of the genetic factors that mediate susceptibility to, and outcome of, sporadic MM. Several putatively important genes, including the susceptibility genes melanocyte stimulating hormone receptor (5, 6), glutathione S-transferase GSTM1 (7), and cytochrome P450 CYP2D6 (8, 9) as well as the cancer candidate genes, p16^INK4a and p15^INK4b (10), have been studied, although thus far only the CYP2D6 PM genotype has been associated with increased risk in independent studies.

We propose that the VDR gene may influence susceptibility and outcome in MM. This view is supported by data showing that 1,25(OH)₂D₃ (the hormonal derivative of vitamin D₃ and the ligand of the VDR) has antiproliferative and prodifferentiation effects in VDR-expressing cell types (11, 12, 13, 14). Furthermore,associations have been identified between 1,25(OH)₂D₃ and susceptibility to, and outcome of, systemic malignancies such as breast, prostate, and colon. These include association with both serum vitamin D/1,25(OH)₂D₃levels as well as with polymorphisms in the VDR gene (15, 16, 17, 18, 19).

Similar supportive data exist for MM. Thus, melanocytes and MM cells express the VDR, and 1,25(OH)₂D₃ has an antiproliferative effect in vitro (20, 21). For example, stimulation of tyrosinase activity, a specific prodifferentiation stimulus, has been reported in melanocytes exposed to 1,25(OH)₂D₃(21). In vivo, there is currently little evidence of involvement of vitamin D₃, although low serum levels of 1,25(OH)₂D₃ have been reported in patients with MM (22). The role of sun exposure in MM is unclear. Current literature remains controversial,with most clinicians advocating a causative association between UV exposure and risk, whereas other studies support the view of a possible protective effect of vitamin D (generated at least in part by UV). For example, use of sunscreens is associated with increased MM risk, an all-year tan appears protective, and outdoor occupation appears to demonstrate no association with susceptibility to MM.

Five polymorphic sites have been identified in the VDR. These comprise RFLP in exon 2 (FokI restriction site), the last intron(BsmI and ApaI restriction sites), and an adjacent area of exon 9 (TaqI restriction site) as well as a poly(A) microsatellite length polymorphism in the 3′ untranslated region. The FokI polymorphism results in an altered translation start site and has been shown to be functionally relevant (23). The other four sites demonstrate linkage disequilibrium, and there is evidence to suggest functional consequences of these polymorphisms (24).

Because these data support the view that polymorphism in the VDR gene may be an important determinant of susceptibility and outcome in patients with MM, the aim of the present study was to investigate the relationship between the VDR polymorphisms and susceptibility to and prognosis (as estimated by Breslow thickness) of MM. Because there is no evidence of linkage disequilibrium between the FokI RFLP and the cluster of polymorphisms at the 3′ end of the gene and there is evidence to suggest functional consequences of each of these polymorphic regions, we have concentrated on the FokI polymorphism and a representative example of the 3′ cluster(TaqI RFLP).

All MM cases (n = 316) were of Northern European Caucasian extraction, originally presented between January 1994 and December 1997 and attended the Dermatology Departments at the Leicester Royal Infirmary, North Staffordshire Hospital or Royal Cornwall Hospitals, between 1996 and 1997. All tumors were histologically diagnosed as in situ or invasive MM. Lentigo maligna and lentigo maligna melanoma were not included, in view of the biological singularity of lentigo maligna. Patients with acral tumors or those with MM and other malignant pathologies (cutaneous or internal)were also excluded. We attempted to recruit all eligible patients,although some were randomly lost in busy clinics. None of the subjects approached refused to participate. This cohort comprises ∼80% of all eligible patients and represents a typical sample of MM patients presenting to dermatologists in the participating centers. The controls(n = 108) comprised randomly recruited, hospital-based Northern European Caucasians attending these Dermatology departments with basal cell papillomas and without clinical or histological evidence of malignancy. Subjects with a history of inflammatory pathology were also excluded. The study was performed with local Ethical Committee approval, and informed consent was obtained from all of the individuals recruited.

Cases and controls were interviewed by a dermatologist (J. E. O.,J. T. L., A. G. S., or P. W. B.). The following demographic and clinical data were recorded: patient age at presentation, gender, skin type in terms of propensity to sun burning and tanning using the Fitzpatrick classification (25), eye and hair color at age 21 years, tumor site, and Breslow thickness. Breslow thickness (defined as the vertical thickness of the tumor from the granular layer of the epidermis to the deepest part of the melanoma) was determined by specialist pathologists. On the basis of Breslow thickness, patients were divided into five categories: in situ, <0.75 mm,0.75–1.49 mm, 1.5–3.49 mm, and ≥3.5 mm. Table 1 shows the distribution of these clinical parameters in the total case group. As also indicated in Table 1, complete clinical data could not be obtained from all patients because of insufficient time in busy clinics (74–95% for TaqI and 72–92% for FokI genotyped cases).

All genotyping assays were performed by workers who were unaware of the clinical status of individual subjects. PCR assays to identify VDR genotypes included one DNA sample (selected at random) of known genotype for each batch of eight samples of unknown genotype, at least one homozygous variant DNA (tt or ff) as a control for restriction enzyme digestion, one negative control (no DNA), and molecular weight markers. Approximately 15% of all patient DNA samples were re-assayed on at least one occasion, and the genotype assignment was confirmed. All assignments were validated by an independent, blinded observer examining the agarose gels. DNA was extracted from peripheral blood (5 ml; collected into EDTA) using standard phenol-chloroform methods. PCR- RFLP based assays were used to identify alleles containing the exon 2 (FokI) and exon 9(TaqI) variants. Primers were selected based on the methods of Gross et al. (Ref. 26; FokI) and Spector et al. (Ref. 27; TaqI) with modifications. The exon 2 wild-type (F) and variant(f) alleles were identified using primers 5′-AGCTGGCCCTGGCACTGACTCTGCTCT-3′ and 5′-ATGGAAACACCTTGCTTCTTCTCCCT-C-3′ to amplify a 265-bp product. Amplification of template DNA was performed in an incubation mixture(total volume, 50 μl) comprising 20 pmol each of primer, 200μ m deoxynucleotide triphosphates, 1.5 mm MgCl₂, and 1 unit of Taq polymerase in buffer containing 10 mm Tris-HCl⁻ (pH 9.0), 50 mm KCl, and 0.1% (w/v) Triton X-100. The PCR conditions were: initial denaturation (94°C for 3 min), followed by 30 cycles of denaturation (94°C for 30 s), annealing (60°C for 30 s), and extension (72°C for 30 s), followed by a final extension at 72°C for 5 min. PCR products were then digested with FokI (37°C for 20 h), and the products were examined after electrophoresis in 2% agarose gels. The F allele was refractory to digestion, whereas f was identified by fragments of 196 and 69 bp. The TaqI wild-type(T) and variant (t) alleles were identified using the forward primer from Spector et al. (27),5′-CAGAGCATGGACAGGGAGCAAG-3′, and a novel reverse primer,5′-CGGCAGCGGATGTACGTCTGCAG-3′, to amplify a 345-bp PCR product. The PCR conditions were as for the FokI RFLP. PCR products were then digested with TaqI (65°C for 20 h), and the products were examined after electrophoresis in 2% agarose gels. The T allele was refractory to digestion, whereas twas identified by fragments of 260 and 85 bp. We attempted to obtain genotype data from all samples, but in some earlier cases, DNA was exhausted or refractory to amplification.

Statistical analysis was undertaken using the Stata software package (version 5.0; Stata Corp., College Station, TX).χ ² tests were used to test for homogeneity between and within cases and controls (28). Because some frequencies were small, the StatXact-Turbo statistical package (version 3; Cytel Software Corp., Cambridge, MA) was used to obtain exact significance levels (Ps). Logistic regression analysis was used to examine differences between cases and controls while simultaneously correcting for imbalances in age and gender. Logistic regression was also used to examine differences in genotype frequencies between cases stratified by Breslow thickness, while correcting for age at presentation and gender. Significant associations of combined genotypes (e.g., ttff) were only accepted if they remained significant in the presence of the main effects(i.e., a model including ttff, tt, and ff). If the significance of the combined genotype disappeared, this would suggest that the factors were acting independently, and the significance of the combined effect was driven by the strength of either (or both) of the component factors. Associations of Breslow thickness were confirmed using linear regression after transformation of thickness values to normality and correction for age at presentation and gender. Because some tumors were in situ (0 mm thick), transformation was performed using the formula: ln (Breslow thickness + 1).

Three hundred and sixteen patients with MM (mean age ± SD,53.3 ± 16.7 years; 67% female) and 108 controls (mean age ± SD, 55.7 ± 19.7 years; 50% female) were recruited.

Table 2A shows the allele frequencies of TaqI and FokI alleles in controls and MM cases. All allele frequencies conformed to Hardy-Weinberg equilibrium. The Fallele was significantly less common in MM cases than controls(P = 0.029; OR, 0.69; 95% CI, 0.50–0.96). Table 2Bshows the relationship between TaqI and FokI genotypes in cases and controls. No significant correlations between genotypes at the two sites were identified in either controls(P = 0.365) or cases (P = 0.847),suggesting that the two polymorphisms did not demonstrate linkage disequilibrium. Table 1 shows frequencies of FokI and TaqI genotypes in controls compared with MM cases. There was a decreased proportion of individuals with the FokI FF genotype in cases versuscontrols. Thus, for FF versus other FokI genotypes, the uncorrected OR for MM was 0.60(P = 0.026). The findings remained significant after correction for age and gender using multivariate logistic regression(OR, 0.59; P = 0.029). The estimated risk reduction attributable to the FF genotype was 23.7% (95% CI,1.2–51.3%).

Table 1 shows the frequency of VDR genotypes in the MM cases stratified by patient characteristics. There was no significant association between genotype frequencies and tumor site, skin type, or eye color. However, the ff and tt genotypes were significantly less common in MM patients with red hair than in patients with other hair colors (P = 0.021,χ ²₁ = 5.31 and P = 0.040, χ²₁= 4.21, respectively).

Patients were categorized by Breslow thickness (Table 1). In tumors≥1.5-mm thick, for TaqI, there was an increased proportion of the tt genotype (P = 0.047), but there was no obvious effect for FokI (P = 0.701). Homozygosity for variant alleles at either FokI and TaqI loci (tt or ff genotypes) was associated with an increased proportion of tumors ≥3.5 mm thick,although this did not achieve statistical significance (tt: P = 0.105; OR, 2.84; and ff: P = 0.266; OR, 1.99, uncorrected).

The effects of combinations of the TaqI and FokI polymorphisms are shown in Table 3. There was an association of ttff combined genotype with thicker tumors, using either≥1.5 mm (P = 0.065) or ≥3.5 mm (P <0.001) as the cutoff. These results retained similar significance,particularly for tumors ≥3.5-mm thick, after correction for potential confounding factors (age, gender, and tumor site). Thus, the mean Breslow thickness in patients with the ttff genotype combination was 2.9 mm compared with 1.1 mm in patients with other genotype combinations. This association was further confirmed using linear regression analysis, which showed that the ttffgenotype was correlated with Breslow thickness (P =0.002, transformed to normality and corrected for age and gender). Significant associations were also identified between Breslow thickness and combinations of genotypes including the genotypes Ttffand ttFf, although these were less effective at predicting Breslow thickness.

We have postulated that polymorphism in the VDR gene is important in MM. This hypothesis is supported by data showing that 1,25(OH)₂D₃ inhibits cell proliferation (12, 13) and stimulates differentiation (11, 14) and apoptosis (29) in several cell types expressing the VDR. There is evidence that 1,25(OH)₂D₃ has an anticancer effect in several systemic cancers such as breast (30), prostate (31), colon (32),leukemia (33), and kidney (34). Furthermore, in vitro studies have demonstrated that 1,25(OH)₂D₃ inhibits growth of cultured malignant cells (11, 12, 13, 14, 34) and inhibits experimental carcinogenesis (35, 36). In vivo,decreased mean serum levels of 1,25(OH)₂D₃ or its precursors have been reported in carcinoma of the breast (15), prostate (17), and colon (19). More recently, polymorphisms of the VDR have been reported associated with cancer of the breast [FokI and poly(A) site RFLP; Ref. 16) and prostate [BsmI and poly(A) site RFLP; Refs. 18 and 37).

Data for MM are similar, although more limited. Normal (38) and malignant melanocytes (20)express the VDR, and 1,25(OH)₂D₃ has been shown to inhibit normal (21) and malignant melanocyte (20) growth in vitro. In a study of 1,25(OH)₂D₃ serum levels in MM patients, lower levels were found compared with controls, although this did not achieve statistical significance (22).

In our study, homozygosity for the wild-type (F)allele at the FokI restriction sites was associated with a reduced risk of MM, with a risk reduction attributable to the FF genotype estimated at 23.7%. Furthermore, the proportion of F alleles was significantly lower in the case group compared with controls. The number of controls, however, was relatively small, and larger cohorts would be required to reduce the risk of both type 1 and type 2 errors. In this initial study, we have used hospital-based controls. Selection of control subjects is always difficult, and although the use of “normal” volunteers or blood donors would reduce the risk of potential bias because of occult associations with other disease processes, since they are generally not examined by a clinician, the possibility of undetected malignant or inflammatory pathologies cannot be excluded. By use of hospital-based controls, it was possible to focus only on controls who were clinically free of other malignant or inflammatory pathologies. Furthermore, the control genotype frequencies were similar to those described in other studies (18, 26, 27, 39), supporting the view that our control group is representative of the normal population.

The FokI RFLP has been reported previously to be associated with breast cancer (16), where the FF genotype was associated with a decreased risk of ∼50% in certain racial groups. The poly(A) polymorphism (classified into long, L,or short, S) has been associated with altered risk of breast (16) and prostate (18) cancer. In breast cancer, LL and LS alleles were also associated with a ∼50% reduction in risk (16). However, in prostate cancer, the presence of L, whether in the heterozygous (LL) or homozygous (LS) state, was associated with a 4–5-fold increased risk of prostate cancer (18, 37). Because the TaqI restriction site is in strong linkage disequilibrium with the poly(A) polymorphism(T demonstrates linkage disequilibrium with L;Ref. 39), the findings in breast cancer are comparable with our findings in MM, although our data on the TaqI RFLP did not achieve statistical significance.

Our data also identified an association between VDR genotypes and red hair in patients with MM. There was insufficient hair color data on our control subjects to examine whether this was a general phenomenon. Although the mechanism for this association is not known, other studies have identified links between polymorphism at other loci and hair color in MM (6, 8). These data suggest that the molecular route by which patients with red hair develop MM may differ from patients with other hair colors, supporting the view that these patients represent a high risk subgroup. However, these data require confirmation in independent studies, including in control individuals.

More significantly, we have identified significant associations between VDR genotypes and outcome in patients with MM. Thus, our data suggest that VDR polymorphism is a better determinant of outcome in MM than of its initiation. Melanoma depth is well recognized as an important prognostic indicator with respect to risk of metastatic disease and survival (40). In general, for both restriction sites, the proportion of thick MMs (with either ≥1.5 or≥3.5 mm cutoff) increased with increasing number of variant alleles. The effect of VDR genotypes on Breslow thickness was markedly increased when the two polymorphic sites were considered together. Thus, the combined ttff genotype was associated with tumors ≥1.5-mm thick but particularly those ≥3.5-mm thick (P <0.001; Table 3). We also corrected the data for the potential confounding effects of gender, tumor site, and age at presentation because thicker tumors are associated with male gender,head/neck tumor site, and older age. The association of the ttff genotype remained significant, suggesting that the effect on Breslow thickness is independent of these factors. Similar results were obtained with other genotype combinations, although the magnitudes of the effects were smaller, suggesting that the heterozygote genotypes were of intermediate importance in determining Breslow thickness. Similarly, in carcinoma of the prostate, poly(A)microsatellite variants are reported to be associated with more advanced disease (37). In addition, low serum levels of 1,25(OH)₂D₃ have been implicated in metastatic rather than in situ disease in prostatic cancer, suggesting an impact on tumor progression rather than development (17).

The polymorphism at the FokI restriction site (T-C transition) produces an ATG start codon resulting in translation initiation 10 bp upstream and therefore the production of a lengthened protein of 427 amino acids (26, 41). The Fallele (restriction site absent, ACG), which results in a shorter protein, has been shown to be more effective at activating the transcription of a VDR reporter construct (23), thereby indicating that the polymorphism is functionally significant. The cluster of polymorphisms at the 3′ end of VDR, which includes TaqI, are in mutual tight linkage disequilibrium and a representative, Bsm, is known to be in linkage disequilibrium with the poly(A) microsatellite (39). It has been suggested that the length of the poly(A) repeat affects mRNA stability or is tightly linked to a further functionally significant site (24). The net effect of the ff and the tt polymorphisms can be envisaged as a reduction in the cellular effect of 1,25(OH)₂D₃ and therefore a growth advantage of the melanocytes. This conclusion is supported by the increased effect of combined homozygosity, which would be expected to have a more profound effect on the VDR protein.

These data support the hypothesis that the VDR genotype has a significant role in determining tumor occurrence and behavior in MM and indicate a role for vitamin D in melanoma cell cycle control and differentiation in vivo. There is evidence of a blocking effect of 1,25(OH)₂D₃ at the transition from G₁ to S phase of the cell cycle via several mechanisms, such as stimulation of the CDK inhibitory proteins, P21 (42), which contains a VDR response element (43), and P27 (44), and inhibition of cyclin D₁ (45). It is of interest that the other reported genetic changes associated with melanoma also have an impact at the G₁ to S-phase check point. These include mutations of the CDK inhibitor genes, p16^INK4a (46, 47, 48) and p15^INK4b (10), and mutations in the CDK4 gene (49, 50). This point in the cell cycle may therefore be of pivotal importance in the development and/or progression of MM.

It has been argued above that the effect of the VDR polymorphisms, reported here, is a functional, cellular deficiency of 1,25(OH)₂D₃. Decreased serum levels of 1,25(OH)₂D₃have been reported in certain cancers (15, 17, 19),including MM (22). Furthermore, some studies suggest that vitamin D deficiency, resulting from decreased cutaneous production from solar irradiation, may be contributory to the development of carcinoma of the breast (51), prostate (52),and colon (53). The role of sun exposure in MM is,however, more complex. On the one hand, it is firmly held by the majority of clinicians that solar radiation is causative in MM, and on the other, there is the potential deleterious effect of vitamin D deficiency, a cause of which is lack of sun exposure. Inhibition of MM cell growth in vitro by 1,25(OH)₂D₃(20), the effect of VDR polymorphisms described in this study, and low serum levels of vitamin D (22) implicate a possible role of vitamin D deficiency in MM pathogenesis. In addition,a protective effect of vitamin D (produced at least in part by sun exposure) might explain previous ambiguous results, such as the increased incidence of melanoma associated with sunscreen use (54), the protective effect of an all-year tan (55), and the lack of an association of MM with outdoor occupation (56). The effect on vitamin D status in these circumstances warrants further investigation, particularly in the climates where these associations have been reported.