The Relationship between a Polymorphism in CYP17 with Plasma Hormone Levels and Prostate Cancer¹

Evidence from studies in both animals and humans support a role for androgens in prostate carcinogenesis (1, 2). Variation in lifetime exposure to endogenous androgens may underlie interindividual differences in prostate cancer risk and racial/ethnic differences in prostate cancer incidence. Polymorphisms in genes involved in androgen biosynthesis, transport, and metabolism (SRD5A2³ and CYP17) and the activation of androgen-responsive genes in prostate cells (androgen receptor and amplified in breast-1) are reasonable candidates to evaluate as surrogate markers of altered long-term androgen exposure and biomarkers of prostate cancer susceptibility (Refs. 3, 4, 5, 6, 7 and reviewed in 8).

CYP17 encodes a cytochrome P450 enzyme (P450c17α) that is involved in the biosynthesis of androgens. A single-bp change (T-C) in the 5′-transcribed but untranslated region of CYP17 has been positively associated with familial polycystic ovarian syndrome and male pattern baldness (9), disorders associated with excess androgen production. This bp substitution in CYP17 was originally hypothesized to create an additional binding site for the transcription factor Sp-1, that may lead to increased transcription of the enzyme and enhanced steroid hormone production. However, in an in-vitro assay, Kristensen et al.(10) did not observe Sp-1 binding at this polymorphic site or within the promoter sequence of CYP17; whether other regulatory factors differentially bind to this polymorphic motif remains unknown. The A2 allele (C nucleotide) of CYP17 has been studied extensively in relation to hormonal-related cancers among men and women (Refs. 5, 6, 11, 12, 13 and reviewed in 14). Among healthy pre- and postmenopausal women, the A2/A2 CYP17 genotype has been associated with modestly higher levels of circulating estrogens (12, 15). In a large study among Caucasian men (n = 621; Ref. 16), CYP17 genotype was not associated with circulating androgen levels, whereas results from case-control studies evaluating CYP17 genotype and prostate cancer risk have been contradictory. In a study among Caucasians from the United States (n = 96 cases and n = 159 controls), Lunn et al.(5) observed the A2 allele to be overrepresented among prostate cancer cases (A1/A2 and A2/A2 genotypes: OR, 1.7; 95% CI: 1.0–3.1). In a smaller study (n = 63 cases and n = 126 controls with benign prostatic hyperplasia), Gsur et al.(17) observed a positive association limited to men with the A2/A2 genotype (versus A1/A1 genotype: OR, 2.80; 95% CI: 1.02–7.76). In contrast, in a study comprised of Swedish Caucasians (n = 178 cases and n = 160 controls), Wadelius et al.(6) reported an increased risk with the A1/A1 genotype (versus A1/A2 and A2/A2 genotypes: OR, 1.61; 95% CI; 1.02–2.53). A more recent study among Japanese in Japan (n = 252 cases and n = 131 controls) also reported men with the A1/A1 genotype to have increased risk of prostate cancer (versus A2/A2 genotype: OR, 2.57; 95% CI; 1.39–4.78; Ref. 18).

In the prostate, conversion of testosterone to the more active androgen, DHT, is catalyzed by SRD5A2. The V89L polymorphism (leucine allele) of SRD5A2 has been associated with lower circulating AAG levels, a surrogate marker of SRD5A2 activity, among Asian and Caucasian men (4, 16, 19). Racial/ethnic differences in V89L allele frequencies have also been hypothesized to explain a proportion of racial/ethnic variation in prostate cancer incidence (19). In a previous study within the Physician’s Health Study (4), no significant association was observed between the V89L alleles and prostate cancer risk. An additional study among Caucasians also does not support an association between this genetic variant and prostate cancer risk (5).

To evaluate further whether CYP17 genotype is associated with prostate cancer risk among Caucasian men, we investigated the relationship in a large nested case-control study within the Physicians’ Health Study. We also examined the relationship between CYP17 genotype and endogenous plasma steroid hormone levels to assess the biological relevance of this polymorphism among men. We also studied the potential interaction between CYP17 and SRD5A2 V89L polymorphisms in relationship with prostate cancer risk as well as the combined influence of these genetic variants on circulating androgen levels.

The source population for this study was the Physicians’ Health Study cohort, initiated in 1982 as a randomized, double-blinded, placebo-controlled trial of aspirin and β-carotene among 22,071 healthy and predominantly Caucasian (93%) United States male physicians, aged 40–84 years. Men were excluded at baseline if they had a history of myocardial infarction, stroke, or transient ischemic attack, unstable angina, cancer (except nonmelanoma skin cancer), current renal or liver disease, peptic ulcer, gout, contraindications to the use of aspirin, current use of aspirin or other platelet-active agents, or current use of vitamin A or β-carotene supplements.

Study participants completed two mailed questionnaires before randomization, and additional questionnaires at 6 and 12 months and annually thereafter. WHRs were calculated from torso and hip measurements reported on the 1989 questionnaire. Men reported their hair balding pattern at age 45 on the 1991 questionnaire; there were five choices ranging from full hair to almost no hair. A blood specimen was collected from two-thirds of the men (n = 14,916) before randomization. Prostate cancer cases and matched controls were drawn from among the participants who supplied blood specimens. For men reporting prostate cancer diagnosis on the follow-up questionnaire, medical records and pathology reports were obtained and reviewed by study physicians to confirm the diagnosis and to determine stage at diagnosis (modified Whitmore-Jewett classification scheme; Ref. 20), tumor grade, and Gleason score. If pathological staging was not available, the case was considered to be of indeterminate stage unless metastasis was clinically evident. We categorized cases as high stage/grade if diagnosed at stages C or D and/or had a Gleason score of ≥7 or poor histological differentiation.

For each case, one or two controls were randomly selected from among the men who returned a blood specimen and who had not been diagnosed with prostate cancer by the date of case diagnosis. Cases and controls were matched on age, within 1 year (± 2 years for elderly cases) and smoking status (current, former, or never) at baseline. The nested case-control study consists of 600 incident prostate cancer cases and 804 controls; 590 cases and 782 controls were successfully genotyped for CYP17. The study sample for the genotype-steroid hormone analysis is composed of 377 controls that were included in a previous study investigating the relationship between sex steroid hormones levels and prostate cancer risk in the Physicians’ Health Study (2). The protocol was approved by the Institutional Review Board at Brigham and Women’s Hospital, Boston, MA.

CYP17 genotyping analysis was performed by the Taqman allelic discrimination method (21) using the ABI 7700 Sequence Detection System (Applied Biosystems, Foster City, CA). This assay measures fluorescent intensity released from allele-specific fluorogenic probes and allows for high-throughput genotyping without post-PCR processing. A 97-bp fragment that included the T-C polymorphism was amplified in a 96-plate format using the following primers: 5′-AGGCCTCCTTGTGCCCTAGA-3′ and 5′-GAGCCACGAGCTCCCACAT-3′. The fluorescently labeled, allele-specific probes were: FAM-CTTCTACTCCACTGCTGTCTTGCCTG-TAMRA and VIC-CTTCTACTCCACCGCTGTCTTGCCT-TAMRA. Primers and probes were designed using Primer Express software (Applied Biosystems). Genomic DNA was used (40 ng) per 25 μl reaction, with 900 nm of each primer, 100 nm of the FAM probe, 200 nm of the VIC probe, and 1× Taqman Universal PCR Master Mix (Applied Biosystems). Amplification conditions were 2 min at 50°C, 10 min at 95°C, and 40 cycles of 15 s at 95°C and 1 min at 63.5°C. The MspAI restriction enzyme was used to identify homozygous (CC and TT) controls required for this genotyping assay. Sequence Detection System software (Applied Biosystems) was used to determine CYP17 genotype, and all genotyping was performed by laboratory personnel unaware of case-control status. To assess genotyping reproducibility, a random 10% selection of the case-control sample was regenotyped; all genotypes matched initial designated genotypes. The genotyping methods for the V89L polymorphism of SRD5A2 have been previously described (4).

Plasma steroid hormone fractions of estradiol, testosterone, SHBG, DHT, and AAG were measured as previously described (2). Intra-assay coefficients of variation for each hormone fraction were < 9%.

The χ² test was used to compare allele and genotype frequencies between cases and controls. ORs and 95% CIs were calculated using unconditional logistic regression controlling for the matching variables at baseline: age (5-year intervals) and smoking status (never, former, and current). Indicator variables for all three genotypes were created using the A1/A1 genotype as the reference category in the multivariate models. Because a gene-dosage effect on prostate cancer risk was not apparent, genotype was also evaluated as a dichotomous variable with A1/A2 and A2/A2 subjects combined. We used linear regression to evaluate the potential relationships between the physiological conditions, male pattern baldness, and WHR, which have been associated with abnormal circulating androgen levels, by calculating age-adjusted least-squared means for each genotype. We also performed case-only analyses to evaluate associations between CYP17 genotype and stage/grade of prostate cancer among all cases and in strata of age (dichotomized at the median age at blood draw: ≤ 61 years and > 61 years). The combined effect of CYP17 and SRD5A2 V89L genotypes on prostate cancer risk was evaluated with the CYP17, A1/A1, and SRD5A2, V/V group serving as the reference category. The interaction between CYP17 and SRD5A2 genotypes was evaluated by including the gene-gene interaction term into the unconditional multivariate logistic regression models. The Wald test was used to assess the statistical significance of the interaction.

Linear regression models were used to evaluate associations between CYP17 genotype and circulating steroid hormone levels among controls, controlling for the matching variables. Least-squared geometric mean hormone levels were estimated, and differences in hormone levels between genotypes were evaluated with the A1/A1 group as the reference category. We adjusted for SHBG levels in analyses evaluating the association between CYP17 genotype and testosterone levels, because we have previously observed that plasma testosterone adjusted for SHBG is a stronger predictor of prostate cancer risk than total testosterone levels (2). The natural logarithm of the plasma hormone values were used in the analysis to reduce the skewness of the regression residuals. Hormone values >3 interquartile ranges from the 75th percentile were treated as outliers and were excluded (estradiol, n = 1; testosterone, n = 1; DHT, n = 2; AAG, n = 1). We also evaluated associations between the combination of CYP17 and SRD5A2 V89L genotypes and plasma androgen levels with the A1/A1, V/V group as the reference category. The interaction between the CYP17-SRD5A2 genotypes was assessed by including the gene-gene interaction term in the linear regression models. We used the SAS statistical package for all analyses (SAS Institute, Inc., Cary, NC).

The distributions of CYP17 genotypes were in accordance with Hardy-Weinberg equilibrium among both cases (χ² = 0.83; df, 1; P = 0.36) and controls (χ² = 2.62; df, 1; P = 0.11). Among controls, the frequency of the A2 allele was similar to that reported among Caucasians (5, 6, 13). The A2 allele was not significantly overrepresented among cases (cases, 41% versus controls, 39%; P = 0.17). There was also no significant difference in genotype frequencies between cases and controls (χ² = 3.49; df, 2; P = 0.18). Compared with the A1/A1 genotype, the adjusted ORs for A1/A2 and A2/A2 genotypes were 1.26 (95% CI, 0.99–1.59) and 1.17 (95% CI, 0.85–1.61), respectively (Table 1). The adjusted OR for men with at least one A2 allele was 1.23 (95% CI, 0.99–1.54). Results did not change when excluding nonwhite racial/ethnic groups (∼5% of cases and controls).

Compared with controls, the A2 allele was not overrepresented among prostate cancer cases with high stage/grade disease (42% versus 39%; P = 0.14). In case-only analyses, the A2 allele was not associated with stage/grade of disease (high stage/grade, 42% versus low stage/grade, 40%; P = 0.43). We also observed no material difference in genotype frequencies between high and low stage/grade disease (χ² = 1.69; df, 2; P = 0.43). In case-only analyses, the frequency of the A2 allele was similar among men with high and low stage/grade disease in both older (>61 years: 40% versus 40%; P = 0.97) and younger cases (≤ 61 years: 44% versus 40%; P = 0.32).

We also evaluated associations between CYP17 genotype and phenotypes associated with an abnormal hormonal milieu. Among controls, WHRs were nonsignificantly elevated among men with the A2/A2 genotype [A1/A1 genotype (reference; n = 269), WHR (mean) = 0.94, A1/A2 (n = 319), WHR = 0.94, P = 0.94; A2/A2 (n = 109), WHR = 0.95, P = 0.16]. Similar associations were observed among cases and when cases and controls were analyzed together [cases and controls: A1/A1 genotype (reference; n = 447), WHR = 0.94; A1/A2 (n = 563), WHR = 0.95, P = 0.64; A2/A2 (n = 192), WHR = 0.96, P = 0.02]. We also evaluated the association between CYP17 genotype and hair balding patterns. No correlation was observed between the A2 allele or A2/A2 genotype and baldness pattern among young (≤ 61 years) or old men (>61; data not shown).

In CYP17 genotype-plasma steroid hormone analyses, we calculated least-squared geometric mean hormone levels among controls for each genotype (Table 2). Compared with A1 homozygotes, carriers of one or more A2 alleles did not have significantly elevated levels of any hormone fraction.

We next examined the interaction between CYP17 and SRD5A2 V89L polymorphisms and prostate cancer risk. An earlier study within this cohort found no significant relationship between the SRD5A2 V89L polymorphism and prostate cancer risk (4). In the present study, compared with men homozygous at both loci (A1/A1 and V/V genotypes), we did not observe consistent associations between carriers of additional hypothesized high- or low-risk CYP17 or SRD5A2 V89L alleles and risk of prostate cancer (Table 3). We did observe a borderline significant association among men with the A1/A2 and V/V genotypes (OR, 1.68; 95% CI, 1.20–2.35) and a significant trend for the V89L genotype limited to men with the A1/A2 genotype (P = 0.04); however, the overall test for interaction between these polymorphisms was not statistically significant (P = 0.60).

In analyses of the relationship between CYP17 and SRD5A2 genotypes and steroid hormone levels among controls, we calculated geometric mean plasma androgen levels for each genotype combination. Compared with men homozygous for both the A1 and V alleles, men with other genotype combinations did not have significantly increased or decreased mean levels of testosterone or AAG (Table 4). We did observe nonsignificant trends for elevated testosterone levels across SRD5A2 genotypes within strata of CYP17 genotype, however, absolute levels did not depend on CYP17 genotype. The overall test for this interaction was nonsignificant (P = 0.46). For AAG, a trend was observed for decreasing levels across SRD5A2 genotypes among men with A1/A1 and A1/A2 genotypes; this test for interaction was also not significant (P = 0.82).

In this prospective case-control study among Caucasian men, we observed a borderline significant association between carriage of the CYP17 A2 allele and prostate cancer risk. However, we did not observe a gene-dosage effect; the borderline positive association was stronger in those heterozygous for the A2 allele. We did not find the A2 allele to be overrepresented among men with aggressive prostate cancer regardless of age at diagnosis. In addition, we did not detect mean steroid hormone levels to be significantly elevated among carriers of the A2 allele or evidence of a gene-gene interaction between CYP17 and SRD5A2 V89L polymorphisms on prostate cancer risk or endogenous steroid hormone levels.

A positive association first reported by Carey et al.(9) between the A2 allele of CYP17 and hyperandrogenic diseases, polycystic ovarian syndrome, and male pattern baldness, led to the selection of CYP17 as a candidate gene for study in relation to hormonal-related cancers. We and others have provided preliminary results to support the proposed hypothesis that this genetic variant may have a modest effect on estrogen biosynthesis among women (12, 15). However, data from multiple studies suggests that this polymorphism in CYP17 does not independently predict increased risks of breast cancer or advanced breast cancer among Caucasian postmenopausal women (13, 14).

This CYP17 polymorphism has also been evaluated among men as a marker of lifetime steroid hormone exposure and cancer risk. Among Caucasian men of Scottish ancestry, the A2 allele was positively associated with increased risk of male breast cancer (A2 allele: OR, 2.10; 95% CI, 1.04–4.27; Ref. 22). However, results have been inconsistent in the published studies that have examined the association between CYP17 genotype and prostate cancer risk. In a North Carolinian Caucasian population, Lunn et al.(5) observed a borderline significant association between the A1/A2 genotype and prostate cancer risk (versus A1/A1: A1/A2; OR, 1.7; 95% CI, 1.0–3.2; A2/A2; OR, 1.7; 95% CI, 0.7–4.2). In a smaller study, Gsur et al.(17) also reported increased risk for Caucasian men homozygous for the A2 allele. In contrast, among Japanese men (n = 252 cases and n = 131 controls), Habuchi et al.(18) observed men with the A1/A1 genotype to have increased risk of prostate cancer (versus A2/A2 genotype: OR, 2.57; 95% CI, 1.39–4.78) and benign prostatic hyperplasia (OR, 2.44; 95% CI, 1.26–4.72). Similarly, in a Swedish study, (cases, n = 178; controls, n = 160), Wadelius et al.(6) also reported a significant elevation in risk for the A1/A1 genotype (versus A2 carriers: OR, 1.61; 95% CI, 1.02–2.53). Our findings among Caucasians in the United States are more similar to those of Lunn et al.(5), as we also observed a marginally significant relation between carrying the A2 allele and prostate cancer risk, and a significant positive association for men with the A1/A2 genotype. The opposing findings of the previous studies may be attributable to inaccurate genotype frequency estimates resulting from small sample sizes (all studies have less than ∼250 cases and ∼160 controls). Linkage disequilibrium between the A1 and A2 alleles of CYP17 with different functionally relevant polymorphisms in specific ethnic groups may explain the contradictory findings observed in studies conducted in different white or nonwhite ethnic populations.

In a large study of the relationship between CYP17 genotype and circulating steroid hormone levels among men (n = 621), Allen et al.(16) did not observe an association between the A2 allele and endogenous androgen levels. Our results support the previous observation that men with the A2 allele do not have substantially higher circulating androgen levels. Furthermore, we also did not observe a combined influence of CYP17 and SRD5A2 V89L polymorphisms on AAG or testosterone levels or an interaction between these loci on prostate cancer risk.

The relatively low prevalence of the variant alleles (A2 and L), and a moderate sample size limited our ability to thoroughly evaluate differences in mean hormone levels and relative risks for all CYP17 and SRD5A2 V89L genotype combinations. In the future, larger studies will be needed to examine gene-gene interactions and to detect a significant positive association with the A2/A2 genotype, if a weak relation exists. In this relatively large, prospective case-control study among Caucasian men, we observed evidence suggesting a possible weak association between the A2 allele of CYP17 and prostate cancer risk. However, our data does not suggest a role for CYP17 genotype as a modifier of androgen metabolism in men.

We are indebted to the participants of the Physicians’ Health Study for their dedication and cooperation.