CYP17 Genotypes Differ in Salivary 17-β Estradiol Levels: A Study Based on Hormonal Profiles from Entire Menstrual Cycles

Sex-steroid hormones are implicated in the development and growth of breast cancer and most of the risk factors for that disease exert their effect by influencing levels of sex-steroid hormones, especially estrogens (1-4). Levels of sex-steroids are also important for fecundity, risks of osteoporosis, cardiovascular health, and psychological well-being (5-9).

Considerable variation has been documented in the levels of sex-steroid hormones among populations and among healthy, premenopausal women within a population (10). Many sources of this variation have been identified. Levels of ovarian hormones are sensitive to factors related to fetal development (11, 12), childhood growth (13), and adult life (14-22). It is also likely that variation in sex-steroid levels results partly from genetic variation (i.e., polymorphism of genes that control steroid hormone biosynthesis; refs. 23-28).

The gene CYP17 encodes cytochrome P450c17α, which mediates the activity of 17α-hydroxylase and 17,20-lyase, both involved in the biosynthesis of estradiol (E₂; ref. 29). In women, CYP17 is expressed in the ovarian theca cells, the corpus luteum, adrenals, and adipose tissue (30-32). A single nucleotide polymorphism in the 5′-untranslated region of CYP17 is relatively common and the presence of the A2 allele is thought to increase transcription rates (26), although other studies suggest that no additional transcription factor binding activity is associated with the presence of the A2 allele (33, 34).

Studies investigating the relationship between CYP17 polymorphism and levels of E₂ in premenopausal women (24, 25, 27, 35-37) have yielded inconsistent results. E₂ levels measured around day 11 of the menstrual cycle have been reported to be 11% and 57% higher among women with genotypes A1/A2 and A2/A2, respectively, compared with A1/A1 women (24). In the same study, E₂ levels during the luteal phase, around day 22 of the cycle, were reported to be 7% and 28% higher for women with A1/A2 and A2/A2, respectively. Another study found that women with the A2/A2 genotype had 42% and heterozygotes 19% higher E₂ than the A1/A1 genotype but only among women with body mass index values 25 kg/m² and under (25). Among women with higher body mass indexes, the genotypes did not differ in E₂ levels. Both of these studies are based on only one or two E₂ values per woman.

A third study of 173 premenopausal women did not find any differences in E₂ levels among CYP17 genotypes but documented significant differences in the levels of steroid hormone dehydroepiandrosterone, a precursor for E₂ synthesis (37). E₂ levels were measured in a single blood sample presumably collected in the luteal phase of the cycle, between day 20 and 24 from the beginning of the cycle. Two other studies that did not find statistically significant differences in E₂ levels among CYP17 genotypes also used a single blood sample for hormonal measurements (35, 36). Only one study attempted to control the within-cycle variability in E₂ levels by sampling, on the average, 4.4 days per woman over a 2-year period, but did not find significant differences in relation to the CYP17 genotype (27).

We document, for the first time, a relationship between polymorphism in CYP17 and full cycle profiles of 17-β E₂ among healthy, regularly menstruating women in midreproductive years. In contrast to the earlier studies that relied on one or a few samples for assessing the E₂ levels of an individual woman, our study is based on daily collected saliva samples for one entire menstrual cycle.

The subjects were 60 urban (n = 22) and rural (n = 38) women from Poland recruited for the study by advertisements. Women were selected for participation if they met the following criteria: age between 24 and 36 years, regular menstrual cycles and no fertility problems, no gynecologic and/or chronic disorders (i.e., diabetes and hypothyroidism/hyperthyroidism), not taking any hormonal medication or using hormonal contraception during the 6 months before recruitment, and not being pregnant or lactating during the 6 months before recruitment. The recruited women signed a consent form after being informed about the aims and requirements of the study, which had been approved by the Jagiellonian University Research Ethics Committee.

Subjects' body weight, height, and percentage body fat (by bioimpedance) were measured by a trained anthropologist. A general questionnaire (one part completed by interview and one part self-administered) was used to collect information on education, reproductive history, and past use of hormonal medication, tobacco, and alcohol. Data on birth weight and birth length were recorded at birth by a nurse and obtained from subjects' personal “health books” issued by hospitals following birth. Detailed description of the methods was published elsewhere (11, 17).

Women collected daily morning saliva samples for one entire menstrual cycle. Saliva samples from 20 days (reverse cycle days −5 to −24, where the last day of each cycle was marked as day −1) of each cycle were analyzed for the concentration of E₂ using an I-125 based RIA kit (Diagnostic Systems Laboratories, Webster, TX) with published (16) modifications to the manufacturer's protocol. The sensitivity of the E₂ assay is 4 pmol/L. Average intra-assay variability was 9%, and inter-assay variability ranged from 23% for lower (15 pmol/L) to 13% for higher (50 pmol/L) values. Before other statistical analyses, cycles were aligned based on identification of the day of the midcycle E₂ drop (day 0; Fig. 1), which provides a reasonable estimate of the day of ovulation (5). E₂ values from 18 consecutive days of each cycle aligned on day 0 were used in analyses. Reliable identification of the day of the midcycle E₂ drop could not be made for two subjects. Therefore, we used data from 58 women in statistical analyses.

Genotyping was done according to published methods (38). Genomic DNA was extracted from 0.2 mL peripheral blood using extraction kit (Qiagen GmbH, Hilden, Germany). PCR was done using primers that amplify restriction sites for Msp A1 I (A2 polymorphism-CYP17): 5′-CAAGGTGAAGATCAGGGTAG-3′ (forward) and 5′-GCTAGGGTAAGCAGCAAGAG-3′ (reverse). A 145-bp fragment encompassing biallelic single gene polymorphism (T→C) in the untranslated 5′ region of CYP17 was amplified. For the PCR, 25 μL final volume of reaction mixture contained the following: 200 pg of genomic DNA, 0.45 μmol/L of each primer, 2 mmol/L MgSO₄, 200 μmol/L of each deoxynucleotide triphosphate, and 1 unit plaque-forming unit polymerase. PCR cycling conditions were 94°C for 45 seconds, 57°C for 60 seconds, and 72°C for 60 seconds for a total of 35 cycles. PCR product was digested by restriction enzyme Msp A1 I for 3 hours at 37°C, separated on 4% agarose, and visualized by ethidium bromide fluorescence under UV light. The polymorphism was identified by digestion of the PCR fragment, resulting in 70- and 75-bp DNA fragments. Genotypes identified by gel electrophoresis were assigned as homozygous wild-type (A1/A1) 145-bp band, heterozygous variant (A1/A2) 145-, 75-, and 70-bp bands, and homozygous variant (A2/A2) 75- and 70-bp bands.

The protocol was run twice for each DNA sample. Additionally, 20% of templates were rerun according to an alternative method (39). A control sample which did not contain DNA was run in every reaction to confirm the absence of contamination. The quality of the PCR product was checked by running 4 mL of each sample on 1% agarose gel.

Differences among the three genotypes in mean E₂ were tested in two-way repeated measure univariate and multivariate ANOVAs. Mean values of E₂ were calculated for each woman for each 3 consecutive days of the menstrual cycle (i.e., the first arithmetic mean was calculated from untransformed values for cycle days −9, −8, and −7; the second mean from cycle days −6, −4, and −5, etc.), obtaining a six-level hormonal profile for each woman. The profile was analyzed as the repeated measure variable and the genotypes were the between-group factor (A1/A1, A1/A2, and A2/A2). The ANOVA was followed by contrast analyses; an α level of 0.0167 (with the Bonferroni correction) was used to indicate statistical significance. Differences among genotypes in log-transformed E₂ levels during the cycle day −1 (preovulatory day; Fig. 1) were tested in a one-way ANOVA.

Differences among the genotypes in age, reproductive characteristics, length of menstrual cycle, size at birth, anthropometrics and body composition, tobacco smoking, alcohol consumption, and mean 24-hour physical activity were tested in factorial, fixed-model, one-way ANOVA analyses followed by Tukey-Kramer post-hoc tests.

Table 1 shows the characteristics of the study group stratified by genotype. Genotypes did not show statistically significant differences in age, reproductive characteristics, size at birth, anthropometrics and body composition, tobacco smoking, alcohol consumption, and physical activity. The length of the menstrual cycle during which samples were collected did not differ among genotypes (P = 0.8).

There was significant variation among genotypes in mean salivary E₂ levels (F_2,54 = 3.535; P = 0.036; see Table 2). The overall interaction between the genotypes and the six-level hormonal profile was significant in the univariate repeated measure ANOVA (F_10,270 = 3.391; P = 0.0014 with Greenhouse-Geisser or P = 0.0008 with Hyun-Feldt adjustments) and marginally significant in the multivariate test [Wilks' λ = 0.718; degrees of freedom (df) = 10,100; P = 0.071]. Contrasts showed that the A2/A2 genotype had significantly different six-level E₂ profile from the A1/A1 genotype (F = 6.451; df = 5; P = 0.0001) and from the A1/A2 heterozygote (F = 4.347; df = 5; P = 0.0008). The heterozygote A1/A2 did not differ significantly from A1/A1 homozygote in mean E₂ profiles (F = 0.990; df = 5; P = 0.424). However, because in the E₂ profile the heterozygote means are numerically higher than the A1/A1 homozygote means on 17 of 18 days, we did a nonparametric Wilcoxon signed-rank test to obtain an approximate statistical assessment of this difference. In this test, the profiles of A1/A1 and A1/A2 genotypes are significantly different at P = 0.0007 level (T = 8; N = 18), but larger sample sizes would be required to confirm this result in parametric tests.

Mean E₂ on the preovulatory day −1 varied among genotypes (F_2,53 = 3.56; P = 0.03). The A2/A2 genotype had significantly higher levels of E₂ than the A1/A1 genotype (P = 0.01) and than the A1/A2 heterozygote (P = 0.03), whereas heterozygote A1/A2 did not differ significantly from the A1/A1 homozygote (P = 0.41).

Our results show that variation in the levels of E₂ produced during menstrual cycles can be partially explained by polymorphism at the CYP17 locus. Women with A2/A2 genotypes had 54% higher mean E₂ levels than women with A1/A1 genotypes and 37% higher mean E₂ levels than women who had only one A2 allele. Heterozygous A1/A2 women had 13 % higher E₂ levels than homozygous A1/A1 women. Levels of E₂ during preovulatory day were 72% higher in A2/A2 compared with A1/A1 and 52% higher compared with A1/A2. Polymorphism of CYP17 gene investigated by us involves a single bp change T→C in the 5′-untranslated region at 27 bp upstream of the start of transcription. It is still controversial if this substitution influences the transcription of CYP17 or enzyme expression (33, 34, 40-44). Apparently, transcriptional efficacy is affected by other, still unknown factor(s).

Our results should be treated as preliminary due to relatively small sample size, especially for the A2/A2 genotype. The frequency of the A2/A2 genotype (13%) is similar to the frequency of that genotype (14%) in the pooled sample of populations of European descent calculated from the published studies (26).

Our study is the first which measured E₂ levels in daily collected saliva samples for the entire menstrual cycle. Previous studies addressed the question of variation in E₂ in menstrual cycles by measuring E₂ levels in only one or two serum or urine samples collected per woman (24, 25, 27, 35-37) and one study had, on the average, 4.4 samples per woman (27). Due to substantial intracycle variation in E₂ levels, such sampling is vastly insufficient and can lead to errors in estimating mean E₂ levels for individual women.⁷

In our study, E₂ was measured in saliva reflecting free, unbound, biologically active fraction of this hormone (45). By contrast, E₂ concentrations measured in blood samples are influenced by levels of sex hormone-binding globulin. Sex hormone-binding globulin binds estrogens and regulates the bioavailability of sex steroids. Sex hormone-binding globulin has substantial interindividual variation, with additive genetic factors accounting for 68% of the total phenotypic variation (46). Urinary measurements, used in some studies, allow only for the assessment of E₂ metabolites, levels of which may be influenced by the rate of metabolic clearance, as well as genetic polymorphism in CYP1B1 or COMT involved in the 4-hydroxylation and the O-methylation of E₂ and estrone (35). Conflicting results of the previous studies may therefore result from methodologic difficulties encountered by researchers in reliable characterization of individual hormone levels.

We accounted for a potential influence of lifestyle variation on E₂ levels by using selection criteria for participation and by controlling for factors known to influence levels of ovarian steroid hormones. Women in our study were in peak of reproductive years and in the age range (24-36 years) when relatively little age-related variation in levels of E₂ is expected (47). They did not use steroid contraception or other steroid medication for at least 6 months before the study. Because lactation and postpartum period are associated with changes in sex steroids, women were not recruited unless at least 6 months elapsed since these reproductive events. All women had maintained regular menstrual cycles for at least 6 months before the study.

Levels of ovarian steroid hormones produced in menstrual cycles are both important determinants of fecundity (5, 48) and are related to risk of several diseases, especially in postmenopausal years (8). Cumulative lifetime levels of sex-steroid hormones are implicated in development of hormone-dependent cancers (1-4). Due to individual differences in exposure to factors modifying genetic levels of E₂, however, a relationship between CYP17 genetic polymorphism and risk of breast cancer may not be detectable. Several studies tested the relationship between CYP17 polymorphism and risk of breast cancer (23, 37, 42, 49-59), but results are conflicting.

The existence of genetic variation underling variation in levels of E₂ does not preclude the importance of other factors. Well documented is the influence of lifestyle on sex-steroid levels, especially factors related to availability of metabolic energy during fetal and adult life (11, 12, 14-22, 60-62). Future studies should investigate a relationship between environmental effects and genetic polymorphism in CYP17 from the perspective of genotype-environment interactions. In addition, interactive effects of other genes involved in estrogen metabolic pathways, such as CYP19, COMT, CYP1A1, SRD5A2, and HSD17B, should be studied to account for potential epistatic interactions among genetic loci.

We thank the women who participated in this study, Dr. Susan F. Lipson who oversaw the steroid measurements, and students of the Faculty of Public Health, Jagiellonian University who worked as research assistants.