|Home | About | Journals | Submit | Contact Us | Français|
In a hospital-based case-control study of 805 non-Hispanic whites with cutaneous melanoma and 841cancer-free age-, sex- and ethnicity-matched control subjects, three VDR polymorphisms (i.e., TaqI, BsmI, and FokI) were genotyped using blood samples collected between 1994 and 2006. We tested the hypothesis that the haplotypes and combined genotypes of these polymorphisms were associated with melanoma risk by interacting with known risk factors. Haplotypes t-B-F (adjusted odds ratio [OR], 0.52; 95 percent confidence interval [CI], 0.34–0.80) and t-B-f (adjusted OR, 0.51; CI, 0.27–0.94) were associated with a reduced risk when compared with T-b-f. The combined genotypes Tt+tt/Bb+BB/Ff+ff (adjusted OR, 0.69; CI, 0.52, 0.90) and Tt+tt/Bb+BB/FF (adjusted OR, 0.58; CI, 0.43, 0.78) were also associated with reduced risk, whereas the combined genotype TT/Bb+BB/Ff+ff genotype (adjusted OR, 2.35; CI, 1.13, 4.98) was associated with increased risk when compared with TT/bb/Ff+ff genotypes. On multivariate analysis, only the TaqI polymorphism was an independent risk factor, while the FokI polymorphism interacted with skin color (p = 0.029), moles (p = 0.017), and first-degree relatives with any cancer (p = 0.013) in modifying risk. Together, these findings suggest that VDR polymorphisms may directly effect or modify the risk associated with known melanoma risk factors. Larger, population-based studies are needed to replicate our findings.
Vitamin D is involved in a variety of biological processes including bone metabolism, immunomodulation, and regulation of cell proliferation and differentiation 1–4. Vitamin D is also known to have a potential protective effect against cancers, including cutaneous melanoma 5, 6, a lethal skin cancer that has an increasing incidence in the United States over the last 30 years 7. Vitamin D exerts its tumor-suppressive effects by binding to the vitamin D receptor (VDR). A ubiquitously expressed intracellular polypeptide that belongs to the steroid/retinoid receptor superfamily of nuclear receptors, VDR specifically binds to 1,252D3 and interacts with target cell nuclei 8. The VDR protein is overexpressed in malignant melanocytes responsive to vitamin D’s antiproliferative effects 2. Several studies have suggested that VDR polymorphisms may alter the functions of genes involved in cell division and adhesion 2, 9, thus implicating such polymorphisms in melanoma development 10.
Located on chromosome 12q12-q14, VDR contains at least five promoter regions 11, eight protein-coding exons, and six untranslated exons, all of these regions bejng alternatively spliced 12. VDR at least has 196 single nucleotide polymorphisms (SNPs) (http://egp.gs.washington.edu/data/vdr/vdrxx.csnps.txt), of which 64 lie in the promoter region, 32 in the 3′ and 5′ untranslated regions, and 2 synonymous and 2 nonsynonymous SNPs in the coding region. FokI (exon 2, rs10735810), BsmI (intron 8, rs1544410), and TaqI (exon 9, rs731236) are the three most frequently investigated SNPs for their associations with various cancers 13–15. Some studies also addressed gene-environment interactions since environmental factors (e.g., sunlight) can influence vitamin D production 16.
Although ultraviolet light plays an important role in melanoma development 17, 18, only three studies to date have assessed the associations between this factor, VDR polymorphisms, and melanoma risk. Moreover, no studies have examined whether VDR polymorphisms modulate the risk associated with other established melanoma risk factors 19–21. Therefore, we conducted a relatively large case-control study of non-Hispanic whites (i.e., 805 patients with melanoma and 841 cancer-free controls) to determine whether the haplotypes and combined genotypes of VDR polymorphisms TaqI, BsmI, and FokI are associated with melanoma risk and whether these polymorphisms can modify the risk associated with known melanoma risk factors.
The study protocol was approved by The University of Texas M. D. Anderson Cancer Center institutional review board, and written informed consent was obtained from all subjects. The subject recruitment has been described previously 20, 22, 23. In brief, all patients with newly diagnosed, histologically confirmed 24 and untreated cutaneous melanoma, who were referred to M. D. Anderson Cancer Center between May 1994 and February 2006, were recruited as case subjects. Because most of the patients (~99 percent) were non-Hispanic whites, the few minority subjects who were recruited were excluded from analysis. Although there were no restrictions on patient age or tumor stage, only those patients who were free of metastases or other cancers and agreed to donate a blood sample were included in the present study. Approximately 85 percent of eligible patients recruited for this study agreed to participate. Cancer-free control subjects were recruited during the same period from among cancer-free visitors to M. D. Anderson Cancer Center who were accompanying patients to our outpatient clinics, were not seeking medical care, and were not related by blood to the patients. Approximately 90 percent of eligible control subjects recruited for this study agreed to participate. The control subjects were matched by frequency to case subjects by age (±5 years), sex, and ethnicity.
After giving informed consent, all subjects completed a self-administered questionnaire that elicited information on demographic factors (e.g., age, sex, education, and income), ethnicity, medical history, family history, and sunlight exposure history (i.e., tanning ability, lifetime number of severe sunburns, and freckling in the sun as a child) 25. Then, each subject was interviewed in person to assess his or her host characteristics (e.g., natural hair, eye, and skin color) as well as self-reported skin conditions (e.g., color, moles, and pigmented nevi). After each interview, a sample of blood (30 mL) was drawn from the subject and collected in a heparinized tube.
Genotyping was performed as follows. First, 1 mL of each whole blood sample was centrifuged to isolate a leukocyte cell pellet from the buffy coat fraction. Genomic DNA was extracted from the pellet, purified using a DNA blood mini kit (Qiagen, Valencia, CA), and assayed for purity and concentration by spectrophotometry (i.e., absorbance at 260 nm and 280 nm). Next, DNA fragments of VDR containing the TaqI 26, BsmI 27 and FokI 10, 19 polymorphisms were amplified by polymerase chain reaction, subjected to restriction fragment length polymorphism analysis, and sequenced (Figure I). Approximately 10 percent of samples were genotyped a second time; the repeat genotyping results agreed completely with the initial results.
The χ2 test was used to evaluate case-control differences in the frequency distributions of selected demographic variables, known risk factors, and each allele and genotype of the VDR polymorphisms. Because skin color was self-assessed on the screening questionnaire on a scale of 1 (light) to 10 (dark), skin colors were categorized as fair (1 or 2), brown (3 or 4), , or dark (5–10); the aim was to obtain similar numbers of observations in each stratum to facilitate further stratification analysis. Some subjects did not provide information about some variables (e.g., hair color, eye color, skin color, tanning ability, number of sunburns, freckling, pigmented nevi, and family history of skin cancer); the missing variables for those subjects were treated as missing data on multivariate analysis. The linkage disequilibrium for each SNP of interest (i.e., TaqI, BsmI, and FokI) was calculated, and the polymorphism haplotypes for each subject were reconstructed on the basis of the known TaqI, BsmI, and FokI genotypes. Because of the potential effect of locus-locus interactions of the polymorphisms on melanoma risk, associations between risk and the haplotypes and combined genotypes of the three polymorphisms were also evaluated.
Crude and adjusted odds ratios [ORs] and associated 95 percent confidence intervals (CIs) were determined by univariate and multivariate unconditional logistic regression analyses. Multivariate adjustments were made, where appropriate, for age, sex, and other known risk factors. Odds ratios, CIs, and p values for interactions and trend tests were obtained from multivariate logistic regression models.
The null hypotheses of multiplicative gene-gene interactions were tested, and departures from multiplicative interaction models were assessed empirically. A more-than-multiplicative interaction was suggested when OR11 > OR10 * OR01 28. To assess potential departures from a multiplicative model, interaction terms between variables were modeled according to standard unconditional logistic regression techniques. Finally, to determine whether the main effect of the VDR polymorphisms was independent of other known risk factors, selected variables were included in the multivariate logistic regression analyses of data from only those subjects who completely answered their questionnaires 29.
Two models were fitted. The first model included age, sex, and the three polymorphisms of interest, the aim being to control for any potential effects due to associations among the polymorphisms. The second model was to exclude the polymorphism that showed no statistically significant association with risk in the first model and then include all other known risk factors, the aim being to assess further the independent effects of the polymorphisms. A p value of ≤ 0.05 was considered statistically significant. All tests were two-sided and were performed using SAS software (version 9.13; SAS Institute, Cary, NC).
The initial analysis included all cases (n=805) and controls (n=841). The two groups had similar age (p = 0.37), sex (p = 0.16), education (p = 0.99), and household income (p = 0.35) (Table I). The similar age and sex distributions implied adequate frequency matching. Because some subjects did not completely answer their questionnaires, the numbers of subjects in some risk factor strata were less than the total number of subjects in the study. Nevertheless, our results were consistent with previous findings by others 30–32. Except for skin color (p = 0.16), the frequencies of known melanoma risk factors were significantly higher among cases than among controls and were associated with 1.55- to 7.78-fold increased melanoma risk (Table II). Subjects with these risk factors were placed into dichotomized groupings for further stratification and assessment of interactions in multivariate logistic regression analyses.
Allele and genotype frequencies of the polymorphisms of interest are presented in Table III. Genotype distributions among controls were consistent with the Hardy-Weinberg equilibrium (p = 0.49 for TaqI, p = 0.31 for BsmI, and p = 0.64 for FokI). TaqI alleles t and BsmI alleles B were significantly less frequent among cases than among controls (0.370 vs. 0.429 [p < 0.01] and 0.394 vs. 0.431 [p = 0.03], respectively), whereas the FokI allele f was more frequent, though not significantly so (0.378 vs. 0.356 [p = 0.20]). This suggested that t, B, and F might protect carriers against melanoma or T, b, and f might put them at risk. Moreover, the t and B genotypes (i.e., Tt+tt and Bb+BB) were consistently less frequent among cases than among controls (p < 0.05 for both) and were associated with a significantly lower melanoma risk (i.e., a protective effect) for Tt+tt vs. TT genotypes (adjusted OR [CI], 0.72 [0.58, 0.90] and 0.68, [0.56, 0.83] and Bb+BB vs. bb genotypes, respectively (Table III). In contrast, the f genotypes (i.e., ff+Ff) were significantly more frequent among cases than among controls and were associated with a significantly greater melanoma risk than was the FF genotype (adjusted OR [CI], 1.25 [1.03, 1.53]) (Table III).
TaqI, BsmI, and FokI polymorphisms were in linkage disequilibrium (t and B alleles: D′ = 0.918, R2 = 0.855, p < 0.001; t and F alleles: D′ = 0.039, R2 = 0.001, p < 0.001; B and F alleles: D′ = 0.027, R2 = 0.001, p < 0.001), suggesting a potentially joint effect of the haplotypes of the three VDR polymorphisms on melanoma risk. Eight hypothetical haplotypes were estimated based on the observed genotypes (Table IV). However, the overall distributions of these haplotypes did not significantly differ between cases and controls (p = 0.381). When the Tbf haplotype was used as the referent (the T, b, and f alleles being putatively associated with increased melanoma risk), the haplotypes tBF and tBf were both associated with a significantly reduced melanoma risk (adjusted OR [CI], 0.52 [0.33, 0.79]) and (0.51 [0.27, 0.94], respectively) (Table IV). This suggested that the tB haplotype was protective, regardless of the f allele’s presence or absence.
When the putative risk genotypes (i.e., TT, bb, and ff+Ff) were combined and used as the referent (TT/bb/ff+Ff), only the Tt+tt/Bb+BB/Ff+ff and Tt+tt/Bb+BB/FF genotypes were associated with a significantly reduced melanoma risk (adjusted OR [CI], 0.69 [0.52, 0.90] and 0.58 [0.43, 0.78], respectively), whereas the TT/Bb+BB/Ff+ff genotype was associated with a significantly increased risk (adjusted OR [CI], 2.35 [1.13, 4.98]). Together, these findings suggested that the Tt+tt/Bb+BB genotypes were protective, consistent with the effect of the tB haplotype, and that the Bb+BB genotypes were not protective in the presence of the TT genotype (Table IV).
Because the FokI variants were associated with increased melanoma risk and the TaqI and BsmI variants with reduced risk, subjects bearing the protective TaqI and BsmI variant genotypes were further stratified by the Ff+ff and FF genotypes and all known melanoma risk factors (Table V). In the Ff+ff subgroup, the Tt+tt genotypes were associated with a significantly lower risk of melanoma than was the TT genotype, provided the carriers of the Tt+tt genotypes were old, male, and blue-eyed; had not freckled in the sun as a child; or had no pigmented nevi. In contrast, the protective Bb+BB genotypes were associated with a significantly lower risk than was the bb genotype only if the carriers were old (Table V).
In the FF subgroup, the Tt+tt genotypes were more likely to be associated with reduced risk in carriers who were young, female, black- or brown-skinned, black- or brown-haired, or non-blue-eyed; had poor tanning ability, had ≥ 1 lifetime sunburn with blistering, had a childhood history of freckling in the sun, had no moles or pigmented nevi, or had no family history of cancer. The same was generally true of the Bb+BB genotypes, except that being young was not a risk factor. Further tests for interaction were significant for age (p = 0.03 for TaqI and p = 0.01 for BsmI) among subjects carrying the Ff+ff genotype and for sex (p = 0.05 for BsmI) and hair color (p = 0.05 for TaqI and p = 0.03 for BsmI) among subjects carrying the FF genotype. However, these findings may have been due to chance since multiple tests were performed.
All variables used in the initial analyses were fitted to two multivariate unconditional logistic models after simultaneous adjustment (Table VI). In the first model, which included the age, sex, and polymorphism genotypes for all subjects, the t genotypes (Tt+tt vs. TT) and f genotypes (Ff+ff vs. FF), but not the B genotypes (Bb+BB vs. bb), were associated with a significantly reduced melanoma risk (OR [CI], 0.57 [0.39, 0.84] for TaqI and 1.27 [1.04, 1.55] for FokI). This suggested that the BsmI polymorphism was not an independent melanoma risk factor, consistent with the high linkage disequilibrium between the t and B alleles. Consequently, BsmI was excluded from the second model, and all other selected risk factors were added to the multivariate logistic regression model. The second model included only data from subjects who provided complete questionnaire data (i.e., 712 case subjects and 707 control subjects). Most of the known risk factors were consequently found to be significant independent predictors of melanoma risk, the exceptions being age, sex, skin color, and childhood freckling. Because skin color may be represented by hair or eye color and freckling by the number of sunburns in the same model, and because there was a high correlation between these variables in our study (data not shown), variance in the model was reduced by excluding skin color and freckling from the final model (Table VI). As a result, the VDR TaqI t variant genotypes assessed in the final model were associated with a significantly reduced melanoma risk (OR [CI], 0.68 [0.54, 0.86]), whereas the FokI f variant genotypes were not (1.14 [0.91–1.44]) (Table VI). Further tests for interaction revealed significant associations between the FokI f genotypes and skin color (p = 0.029), moles (p = 0.017), and a family history of cancer (p = 0.013) but not between the TaqI t genotypes and the same variables (data not shown). Since multiple tests were performed, these interactions are only suggestive and require validation in larger future studies.
In this hospital-based case-control study of cutaneous melanoma, we found that TaqI t and BsmI B variant genotypes of the VDR gene (Tt+tt and Bb+BB, respectively) were associated with a reduced risk of melanoma and FokI f variant genotypes (Ff and Ff+ff) with an increased risk when compared with the TT, bb, and FF genotypes, respectively. The tBF and tBf haplotypes were associated with a significantly lower melanoma risk than was the Tbf haplotype. The VDR FokI polymorphisms appeared to interact with other known risk factors to modulate the melanoma risk associated with those factors, while the VDR TaqI polymorphism appeared to exert its protective (i.e., risk-reducing) effect independently of other risk factors.
The VDR protein is expressed in both melanocytes and melanoma cells, and 1,25-[OH]2D3 can apparently inhibit the growth of both normal and malignant melanocytes in vitro 2, 5, 33. However, malignant transformation may inhibit the anticancer actions of 1,252D3 for reasons that include genetic polymorphisms of the VDR gene 34. The VDR gene comprises nine exons harboring several polymorphisms, including a poly-A microsatellite in the 3′ flanking region 35, changes in intron 8 that generate BsmI 36 and ApaI restriction enzyme sites 37, a synonymous change at codon 352 in exon 9 that generates a TaqI restriction enzyme site 38, and a 5′ FokI site in exon 2 8. No apparent association has been found between the TaqI or Bsm1 polymorphisms and altered functional activities. Nevertheless, both of TaqI and BsmI polymorphisms located near to the 3′ end of the gene, thus are thought to affect mRNA stability and VDR gene transcription regulation 39. Among the VDR polymorphisms, the FokI singlenucleotide polymorphism of the translation start site is the only one that results in a VDR protein with a different structure40. This polymorphism is characterized by the presence of either two ATG start codons separated by six nucleotides in the long f-VDR or only one start codon due to a T-to-C substitution in the most 50 ATG codon, resulting in a 3-aa shorter F-VDR protein (424 aa in stead of 427 aa)41.
To date, two other groups have reported their studies on TaqI, BsmI, or FokI polymorphisms in evaluating their association with melanoma risk 19–21 but generated mixed results. In the earliest study of the TaqI polymorphism in 316 melanoma patients and 108 control subjects, neither the Tt nor the combined Tt+tt genotype was associated with altered melanoma risk when compared with the TT genotype 19. However, in the present large study, we found that the t genotypes were in fact associated with a lower melanoma risk. In the Nurses’ Health Study, investigators examined the association between melanoma risk and BsmI polymorphism in 219 melanoma patients and 873 controls and found no association between the two21. However, in our present study, we found an association between Bb+BB genotypes and reduced melanoma risk only in women who carried the FF genotype and not in men (Table V). We found the Ff and Ff+ff genotypes to be associated with increased melanoma risk. Interestingly, even though this finding was consistent with published data from other group 19, it was not consistent with the finding in the Nurses’ Health Study of an association (though not significant) between only the ff genotype and higher melanoma risk21.
There are several possible reasons for the apparent discrepancy between our results and those reported by others. One is the relatively larger size of our control population, and another is the potential for selection bias in our control population. Our present study, with its 805 melanoma cases and 841 control subjects, is the largest study so far to have addressed the possible association between VDR polymorphism and melanoma risk; in addition, the VDR t, B, and f allele frequencies we report here are similar to those reported previously in a large meta-analysis of studies in whites 14. Therefore, we believe it unlikely that the association between the VDR polymorphisms and melanoma risk demonstrated in the present study was biased by our selection of controls. A third possible reason for the discrepancy between our findings and those previously reported is variations in the serum vitamin D levels of study subjects between studies. Indeed, the serum vitamin D level may have affected our results. Unfortunately, none of the studies published so far gathered data on serum vitamin D levels in their subjects. A fourth possible reason is recall biases in exposure data, to which a retrospective study might be prone. Thus, larger, population-based studies are needed to verify our findings.
The functional significance of the VDR TaqI polymorphisms is unknown. As a synonymous change in exon 9, the TaqI polymorphism does not cause the amio acid substitution (http://egp.gs.washington.edu/data/vdr/vdrxx.csnps.txt; http://snp500cancer.nci.nih.gov/snplist.cfm). In addition, no apparent association has been found between the BsmI polymorphism and altered functional activities42. Nevertheless, these polymorphisms might be functional themselves or in linkage disequilibrium with other functional SNPs and associated with melanoma risk. Indeed, previous in vitro functional studies have revealed the baT haplotype (haplotype of Bsm1/ApaI/TaqI) inserted in transfection constructs resulted in lower reporter gene activity compared with BAt 43 and associated with low VDR mRNA expression 44, which is in agreement with our findings that both of the BsmI B allele and TaqI t allele are protective against melanoma in the present study. VDR FokI is the only polymorphism that is not linked to any of the other VDR polymorphisms41. A study recently provided evidence that the VDR FokI polymorphism affects immune cell behavior, with a more active immune system for the short F-VDR45. Consistent with this functional study, we found Ff+ff genotype associated with significantly increased melanoma risk, which might due to the f allele-related reduced anti-tumor immune activity. However, since some in vitro data may not accurately reflect the biologic environment, in which a marker may be acting in humans, our consideration regarding the putative function of VDR polymorphisms should be adequately validated in further functional studies.
As some epidemiologic studies have suggested, adequate vitamin D levels (including sunlight induced) may provide very important protection against colon, breast, and prostate cancers 46–48. However, its protection against skin cancer is a more complex issue. One potential complication is that ultraviolet light exposure not only promotes vitamin D-3 (cholecalciferol) synthesis in the skin but also increases the risk of skin cancer by inducing DNA damage. Therefore, it is very important to consider gene-environment interactions as well as locus-locus interactions when studying associations between VDR polymorphisms and melanoma risk. For example, one recent case-only analysis study revealed an association between the TaqI tt genotype and reduced prostate cancer risk, but only in association with high sun exposure 16. However, in the present study, we found that the TaqI t variant genotypes exerted their protective effects independently of other genotypes and known risk factors. Meanwhile, the effect of FokI genotypes on melanoma risk appeared to be independent of the TaqI polymorphism but dependent on other known risk factors, suggesting that some environmental modification of the VDR gene may have occurred. Indeed, we found that the FokI polymorphism interacted with the known melanoma risk factors of skin color, moles, and family history of cancer. However, because of our study’s limited size and current uncertainty about the biological mechanisms underlying such interactions, these findings are only suggestive. Again, larger, population-based, and functional studies are necessary to validate these interactions.
The present study has several limitations. First, it was a hospital-based case-control study in which selection of the unrepresentive population and retrospective collection of exposure data may have led to uncontrolled biases. Second, despite being the largest study of its kind ever published, the present study was still too underpowered to detect gene-gene or gene-environment interactions. Third, the self-reporting of skin conditions by both case and control subjects created an additional source of potential bias. Finally, like most previous studies on the subject, ours could not account for serum vitamin D and thus did not allow for genotype-phenotype correlation analysis. These limitations can only be overcome in large, well-designed prospective studies that gather data on both genotypes and phenotypes of vitamin D metabolism.
In summary, the VDR TaqI, BsmI, and FokI polymorphisms and their combined variant genotypes do affect melanoma risk. The VDR TaqI polymorphism alters risk independently of BsmI, FokI, and other known melanoma risk factors, while the VDR FokI polymorphism may modify it through interaction with sun exposure-related melanoma risk factors. Larger, population-based studies are needed to confirm these findings.
We thank Margaret Lung, Cesar A. Maldonado, and Amanda Francofor assistance in recruiting subjects; Zhaozheng Guo, Yawei Qiao, Jianzhong He, and Kejing Xu for laboratory assistance; Monica Domingue for assistance in preparing the manuscript; and Jude Richard, ELS, for editing the manuscript.
Grant sponsors: National Cancer Institute grants R01 CA 100264 (QW) and P50 CA 093459 (EAG) and National Institute of Environmental Health Sciences grants R01 ES11740 (QW) and P30 CA16672 (M. D. Anderson Cancer Center).