Risk of Early-Onset Prostate Cancer in Relation to Germ Line Polymorphisms of the Vitamin D Receptor

There is substantial evidence that the hormonally active form of vitamin D, 1,25-dihydroxyvitamin D (1,25-D), inhibits cancer cell growth, angiogenesis, and metastasis (1-4). Recent research suggests that 1,25-D, which acts as a steroid hormone, may be useful for treatment by blocking integrins important for prostate cancer cell adhesion and migration (5). These reports also suggest that the vitamin D receptor (VDR) and 1,25-D may play an inhibitory role in the development and progression of prostate cancer. The VDR is a ligand-activated transcription factor (6, 7) that mediates the effects of 1,25-D on target gene expression in a variety of tissues, including the normal and malignant prostate cells (1-4). It has been hypothesized that polymorphisms in the VDR may alter 1,25-D function (8, 9) and thereby affect prostate cancer risk (8-23).

Figure 1 shows the locations of five known polymorphisms within or near the human VDR gene. The functions of the variant alleles in several of these polymorphisms have been studied. The FokI (F/f) polymorphism can be detected as a RFLP using the FokI enzyme (14). The variant corresponds to a substitution of C for T, which results in the loss of the first ATG start site that then causes translation to start at the second ATG site located 9 bp downstream. The ATG sequence is designated the f allele (or M1 for the start at the first methionine) and produces a 427-residue VDR protein rather than the shorter 424-residue protein produced by the F allele (or M4, the start at the second methionine; refs. 8, 14). Recent in vitro research in cells (16) and measurement of 1,25-D–induced transactivation of target genes (8, 11, 12) suggest that the long variant or M1 (f form) may have reduced effectiveness compared with the shorter variant or M4 (F form). No functional differences have been associated with the BsmI and ApaI restriction site polymorphisms (both occurring in the intron separating exons 8 and 9) or with the TaqI site (a silent substitution in exon 9; refs. 13, 15). Conversely, Jurutka et al. (11) and Whitfield et al. (12) observed the long (L) allele (≥17 A's) of the polyadenylic acid [Poly(A)] microsatellite repeat (in the 3′ untranslated region) and the F (M4) allele to be more functionally active than the short (S) allele (<17 A's) and the f (MI) allele.

As suggested by the locations of the five sites (Fig. 1), the extent of linkage disequilibrium between pairs of sites is strong for the BsmI, ApaI, and TaqI sites. The BsmI and TaqI sites are very tightly linked in Whites, with Bt and bT haplotypes accounting for virtually all haplotypes observed (9, 24). These two sites are somewhat less tightly linked in African Americans (25). The intermediate ApaI site is less tightly linked, with A and a alleles noted on both Bt and bT haplotypes, suggesting that the mutation responsible for this polymorphism occurred more recently than did those causing the BsmI and TaqI variation. The PolyA microsatellite shows linkage disequilibrium with the BsmI and TaqI polymorphisms (25). In contrast, linkage disequilibrium is not expected between alleles at any of these four sites and those at the FokI site, which is located at least 30 kb from the others.

Some case-control studies, conducted largely among Whites, indicate positive associations between prostate cancer risk and VDR alleles at these sites, which are often stronger among men with advanced disease (17-23). However, other studies do not show such associations (26-31). The most striking findings in Whites were a 3-fold increased risk of prostate cancer associated with carrying at least one T allele among 108 cases compared with 170 urology clinic controls (17) and a 4-fold increased risk associated with carrying at least one L allele among 57 cases compared with 169 controls (18). In an analysis of 49 advanced cases and 174 disease-free African Americans, a 2-fold increased risk of advanced prostate cancer was associated with the BL haplotype (19). In contrast, a 3-fold increased risk of prostate cancer has been associated with the bb genotype among 222 Japanese with prostate cancer compared with 209 men with benign prostatic hyperplasia and 128 disease-free men (21).

It is important to confirm or refute these findings because polymorphisms associated with increased prostate cancer risk could be useful in decisions concerning screening and treatment. Because the prevalences of VDR variants in families at high risk for prostate cancer may differ from prevalences among men in the general population with and without the disease, it also is useful to compare prevalences and measures of association based on multiple-case families with those based on population-based samples of cases and controls. Here, we report prevalences of genotypes at five polymorphic sites of the VDR gene among men in 98 predominantly White families with three or more verified cases of prostate cancer within first-degree or second-degree relatives and in population-based samples of 242 African American and 414 White men with and without prostate cancer at age <65 years.

Prostate Cancer Cases and Controls. We obtained epidemiologic data and blood samples from a population-based series of incident African American and White prostate cancer cases diagnosed in the San Francisco-Oakland Bay Area at ages ≤65 years. Potentially eligible cases were those listed in the Greater Bay Area Cancer Registry with histologically confirmed prostate cancer in one of the nine Bay Area counties (Alameda, Contra Costa, Marin, Monterey, San Benito, San Francisco, San Mateo, Santa Clara, and Santa Cruz) during the period January 1, 1993 to December 31, 1995. Under statewide cancer reporting legislation, the Greater Bay Area Cancer Registry collects and validates information on persons newly diagnosed with cancer as reported by hospitals, physicians, and other cancer treatment facilities. Cases were selected using a stratified sampling design, with higher sampling fractions among African Americans and among patients diagnosed with aggressive disease. Aggressive prostate cancers were those diagnosed in regional or distant stages or with poorly differentiated or undifferentiated grades. We did not sample cases with both unknown stage and grade, those with unknown grade and local stage, or those with unknown stage and well or moderately differentiated grade. A total of 113 (58%) of the 194 sampled African American patients participated by donating a blood sample. Among the nonparticipants, 22 (11%) were deceased, 37 (19%) refused to participate, and 20 (10%) could not be located. Among the 338 sampled White patients, 232 (69%) donated a blood sample. Among nonparticipants, 16 (5%) were deceased, 82 (24%) refused, and 8 (2%) could not be located.

Genotypes of prostate cancer cases were compared with those of 121 African American controls (residing in Los Angeles or San Francisco) and 171 White controls (from Hawaii, Los Angeles, or San Francisco) who did not have clinically diagnosed prostate cancer and were ages ≤70 years at blood draw. From 1989 to 1991, these men were identified by random digit dial as part of a population-based case-control study of prostate cancer conducted in Hawaii, Los Angeles, the San Francisco-Oakland Bay Area, and Vancouver, British Columbia. Controls were frequency matched to case patients by ethnicity, age (5-year intervals), and region of residence. Among eligible White control subjects, 63% provided an interview, and of these, 89% provided a blood sample. Corresponding response rates for African Americans were 45% and 68%, respectively. Additional details of the study protocol can be found in Whittemore et al. (32).

Prostate Cancer Families. We analyzed the genotypes of members of 98 unrelated families containing three or more medically verified diagnoses of prostate cancer among first-degree or second-degree relatives. The families were identified from a multiethnic case-control study, from screens of the British Columbia and the San Francisco-Oakland cancer registries, and from publicity in the San Jose Mercury News. Eighty-two of these families fulfilled one or more of the proposed criteria for families whose prostate cancer is likely to be hereditary (i.e., three or more affected individuals within one nuclear family; affected individuals in three generations; and/or two or more individuals affected before age 55 years). Eighty-three families were White, seven were African American, five were Japanese American, and three were Chinese American. The mean number per family of affected and genotyped individuals was 2.6 (range 2 to 5). Further details about these families can be found in Hsieh et al. (33).

DNA was extracted from buffy coats, and the genotypes of the F, B, A, and T sites were determined using an aliquot of 1 to 2 μL DNA solution as template for amplification of regions of the VDR gene (Qiagen PCR Core Kit, Qiagen, Valencia, CA; refs. 14, 34). The reaction mixture of 25 μL contained buffer, deoxynucleotide triphosphates, solution Q, Taq DNA polymerase, template, and appropriate primers at concentrations up to 1 μmol/L. Amplifications were done in a MJ Research thermocycler (MJ Research, Waltham, MA) with a heated lid. The cycle conditions and primers were described previously (14, 34, 35). Digestion products were electrophoresed on a 3% agarose gel containing TAE buffer and ethidium bromide (0.5 μg/mL). For each of the four genotype assays, we included control samples whose sequence-determined genotype was unknown to the technician.

To determine length of the poly(A) microsatellite, we amplified template DNA as described above but with primers 5′-GACAGAGGAGGGCGTGACTC-3′ and 5′-GTGTAGTGAAAAGGACACCGGA-3′. We used the restriction enzymes HaeIII and DdeI for 3 hours at 37°C to digest the PCR products. The digestion products were electrophoresed on a 6% polyacrylamide gel containing 1% Spreadex (Amresco, Inc., Solon, OH). The gel was stained in water containing ethidium bromide (50 μg/μL) for 15 minutes. The PCR products from some subjects were cloned and sequenced for use as standards and included in all assays. Poly(A) alleles were designated as short (S ≤17 A's) and long (L >17 A's; ref. 18).

For the case-control comparisons, we used race-specific unconditional logistic regression to evaluate associations between VDR genotypes and prostate cancer risk. Associations are reported as odds ratios (OR) relating risk in heterozygotes and risk in homozygotes for the “variant” allele (F, B, A, T, or S) to that of homozygotes for the f, b, a, t, or L variant. We also computed a test for trend in risk with the number of variant alleles in the genotype. All analyses were adjusted for age at diagnosis (cases) or at interview (controls) by including age categories (<50, 50 to 54, 55 to 59, or 60+ years) as dummy variables. We estimated haplotype frequencies by the method of maximum likelihood as described by Excoffier and Slatkin (36) as implemented in the software SNPHAP (37). We used likelihood ratio statistics to test the null hypothesis of equal haplotype frequencies among cases and controls, specific for race. These statistics allow for phase uncertainty in the estimated haplotypes. When using the estimated haplotypes, each man was counted twice, once for each imputed chromosomal haplotype.

We used a family-based association statistic to test for association between prostate cancer risk and VDR genotypes in the multiple-case families (38-40). This test statistic evaluates departures from Mendelian transmission of variant alleles from parents to offspring; it generalizes the transmission disequilibrium test to include censored age at onset data and families with untyped parents. The statistic has a standard Gaussian distribution under the null hypothesis of no association between prostate cancer risk and the genotype(s) of interest and under the null hypothesis of Mendelian allele transmission from parent to child. We tested the null hypothesis that prostate cancer risk was unrelated to (1) carrier versus noncarrier status of the variant allele (corresponding to a dominant alternative), (2) homozygous carrier status of the variant allele versus other genotypes (a recessive alternative), and (3) number of variant alleles (0, 1, or 2) carried (an additive alternative). Because the software for computing this statistic applies only to nuclear families, we decomposed the 98 families into 127 nuclear families (109 White, 10 African American, and 8 Asian American).

Table 1 summarizes characteristics of the population-based cases and controls. In both races, controls were older than cases. Among cases, African Americans were less likely than Whites to report a family history of prostate cancer.

Table 2 shows the genotype distribution for the FokI, BsmI, ApaI, TaqI, and Poly(A) sites among the prostate cancer cases and controls by race. Among controls, we found no departures from Hardy-Weinberg genotype frequencies for the four RFLP sites. However, the prevalence of SL heterozygotes among African American controls was lower than Hardy-Weinberg expectation (P = 0.02).

As seen in Table 2, none of the five polymorphisms were associated with risk among Whites. However, the table shows a marginally significant trend of increased risk with increasing number of F alleles at the FokI site among African Americans (P = 0.059). The association in Whites, although it did not achieve statistical significance, was in the same direction. Few African American men were homozygous for the f allele; when the combined group of all carriers of the f allele were compared with FF homozygotes, we found elevated risk in the FF homozygotes (OR 1.9, 95% CI 1.0-3.3; online supplementary data). This association also was seen when analysis was restricted to cases with high grade/advanced disease (OR 2.1, 95% CI 1.1-4.1). However, the association was weaker when analysis was restricted to cases with localized disease (OR 1.4, 95% CI 0.65-3.2). Polymorphisms at the other four sites were unassociated with risk among African Americans.

Table 3 shows estimated haplotype frequencies among case and control chromosomes by race. Whereas as many as 2⁵ = 32 haplotypes might have been observed, we actually found <20 haplotypes with nonzero estimated frequencies (13 haplotypes among Whites and 19 haplotypes among African Americans). In Table 3, haplotypes having estimated frequency of <5% among White controls were grouped into a single category labeled “other” haplotypes. The estimated measures of linkage disequilibrium between pairs of polymorphisms were consistent with those observed by other authors.

The estimated haplotype frequencies in Table 3 show greater diversity among African Americans compared with Whites: some 20% of African American control chromosomes and 33% of African American case chromosomes carry one of the infrequent haplotypes compared with 6% of the White control and 3% of the White case chromosomes. In addition, the estimated haplotype frequencies among African American case chromosomes differ significantly from those of the African American control chromosomes (χ₅² = 18.1; P < 0.01). We observed no significant differences between cases and controls of either race in estimated frequencies of the four BL, bL, BS, and bS haplotypes. The estimated frequencies were 0.10, 0.59, 0.21, and 0.10 in Black cases; 0.07, 0.62, 0.24, and 0.07 in Black controls; 0.02, 0.56, 0.42, and 0.01 in White cases; and 0.01, 0.58, 0.41, and 0.00 in White controls, respectively. Frequencies among cases with aggressive disease were similar.

The family-based test statistics provided little evidence for association between prostate cancer risk and any of the five polymorphisms regardless of whether we included all 98 families or restricted analysis to the 83 White families. There were too few African American (n = 10) or Asian American (n = 8) families for separate analyses in these races. For example, FF homozygosity was present in 12 of 18 (67%) affected sons and 6 of 7 unaffected sons in the African American families (none of the members of these families carried the ff genotype). The estimated genotype frequencies among parents in the nuclear families were similar to those among the White cases and controls.

The present analysis of family and case-control data fails to support associations between prostate cancer risk and alleles at the FokI, BsmI, ApaI, TaqI, or Poly(A) polymorphisms in Whites. Among African American men, we observed an increased risk associated with homozygosity for the F allele at the FokI site that was stronger in cases with advanced disease than in those with localized disease. We also found a significant difference in estimated five-locus haplotype frequencies between African American cases and controls.

Several limitations of the study suggest caution in interpreting these findings. These include our inability to genotype all identified cases and controls (with a lower response rate in African Americans than Whites), the lack of consistency in findings across the two races, and the lack of a dose-response trend in risk with number of F alleles in the genotypes, so that excess risk was found only among African Americans homozygous for the FF genotype. Thus, it is important that the findings be replicated in other studies.

Xu et al. (41) observed that the cancers in White prostate cancer cases with one or more F alleles were more likely to be aggressive than those of cases with genotype ff. This observation contrasts with those of other studies in Whites that suggest no association between prostate cancer risk and alleles at the FokI site (27). The current results and some prior studies in Whites also suggest no association between prostate cancer risk and alleles at the other four polymorphsims (20, 22, 28, 31). However, the results presented here conflict with other studies that observed an association between alleles at the TaqI (17, 23, 27) and Poly(A) (18, 27) sites and prostate cancer risk in Whites.

In African Americans, there is a dearth of information regarding associations between prostate cancer risk and VDR polymorphisms. In the only other study to include a substantial number of African Americans, Ingles et al. (19) found that advanced prostate cancer was positively associated with the BL haplotype and negatively associated with the bL haplotype. The present data do not support these findings. If the BL haplotype were associated with poor survival, the lack of association seen in the present data could reflect bias, because many cases with advanced disease died prior to interview.

We observed that the distribution of VDR haplotypes varied by race as cited previously (35, 42, 43) and by disease status among African Americans (35). This may explain some of the apparent inconsistent findings as the same set of genotypes or haplotypes may confer different cancer risks in different races (35). Such differences by race have been observed with the chemokine receptor CCR5 haplotypes and the rate of HIV-1 disease progression in Whites and Blacks (44). For prostate cancer, risk may be related to the interaction of several alleles or haplotypes that are distributed differently by race due to different evolutionary pressures. If, among African Americans, the F allele is also associated with more aggressive disease, as observed among Whites by Xu et al. (41), the low prevalence of the ff genotype may account for not only the increased incidence but also the poorer survival in African Americans. In fact, we may have underestimated the risk associated with the F allele in African American men if poorer survival in F allele carriers contributed disproportionately to the deaths in African American cases (11%).

We did not find evidence of an association among BsmI, ApaI, or TaqI RFLPs and the Poly(A) microsatellite with prostate cancer risk in the family or case-control data. If the present finding among African Americans at the FokI site can be confirmed, the race-specific significance of the FokI site in prostate cancer etiology will be an important area for additional research and may result in screening and treatments that reduce the disease impact on African Americans.

We thank Rong Xu for assistance with genotyping.