HSD17B1 Gene Polymorphisms and Risk of Endometrial and Breast Cancer

It is established that endogenous and exogenous estrogens influence endometrial and breast cancer risk (1). The HSD17B1 (17β-hydroxysteroid dehydrogenase type 1) gene produces the enzyme that catalyzes the final step of estradiol biosynthesis, i.e., the conversion of estrone to estradiol. The gene is located on chromosome 17q12-q21, and the protein is expressed in the ovaries, placenta, testis, endometrium, malignant and normal breast epithelium, and prostatic cancer cells (2, 3). Polymorphisms in estrogen-metabolizing genes may potentially cause alterations in their biological function and thus contribute to the susceptibility of an individual to hormone-related cancers. Several single nucleotide polymorphisms (SNPs) in the HSD17B1 gene have been described (2); however, the function of these polymorphisms remains unclear. We hypothesize that these polymorphisms, may increase the activity of HSD17B1 enzyme, increase estradiol level, and, thus, increase breast and endometrial cancer risk. A polymorphism in exon 6 (+1954A/G) that leads to an amino acid change from serine to glycine at position 312 has been studied in relation to breast cancer (3, 4, 5), but no study has been conducted in endometrial cancer.

We validated previously published SNPs in the HSD17B1 gene and inferred haplotypes. We assessed whether polymorphisms in the HSD17B1 gene were associated with endometrial and breast cancer risk, and whether they were associated with plasma steroid hormone levels among postmenopausal women not using hormone replacement therapy (HRT). In addition, we explored gene-environment interaction hypotheses between HSD17B1 genetic variations and a priori risk factors for endometrial and breast cancer in nested case-control studies within the Nurses’ Health Study (NHS).

The NHS began in 1976, when 121,700 female United States registered nurses between the ages of 30 and 55 completed and returned the initial NHS questionnaire. Information regarding endometrial and breast cancer risk factors was obtained from the 1976 baseline questionnaire, subsequent biennial questionnaires, and a questionnaire completed at the time of blood collection. Definition of menopausal status has been described in detail in previous publications (6). Between 1989 and 1990, blood samples were collected from 32,826 women. Approximately 97% of the samples were returned within 26 h of blood draw. These samples were immediately centrifuged; aliquoted into plasma, RBCs, and buffy coat fractions; and stored in liquid nitrogen freezers. Follow-up has been >96% in all of the subsequent questionnaire cycles for this subcohort.

Eligible incident cases for endometrial cancer study consisted of women with pathologically confirmed invasive endometrial cancer from the subcohort who gave a blood specimen. We included both incident (n = 104) and prevalent (n = 118) cases of invasive endometrial carcinoma from the blood subcohort of the NHS. Prevalent cases had pathologically confirmed invasive endometrial cancer diagnosed between 1976 and the date of blood collection, with no previously diagnosed cancer except for nonmelanoma skin cancer. Controls were matched to cases on year of birth (±1 year), menopausal status at blood draw, use of HRT at blood draw, as well as time of day, month, and fasting status at blood draw. For each endometrial cancer case, three controls were randomly selected among participants who gave a blood sample, had not had a hysterectomy, and were free of diagnosed cancer (except nonmelanoma skin cancer) up to and including the interval in which the case was diagnosed. The endometrial nested case-control study consisted of 222 invasive endometrial cancer cases and 666 matched controls.

Eligible incident cases for the breast cancer study consisted of women with pathologically confirmed invasive breast cancer from the subcohort who gave a blood specimen. Breast cancer cases with a diagnosis anytime after blood collection up to June 1, 1998 with no previously diagnosed cancer except for nonmelanoma skin cancer were included. Controls were matched to cases on year of birth (±1 year), menopausal status at blood draw, use of HRT at blood draw, as well as time of day, month, and fasting status at blood draw. For each breast cancer case, one or two controls were randomly selected among women who gave a blood sample and were free of diagnosed cancer (excluding nonmelanoma skin cancer) up to and including the interval in which the case was diagnosed. The nested case-control study of breast cancer consisted of 1007 incident breast cancer cases and 1441 matched controls. The Brigham and Women’s Hospital Committee on Human Subjects approved the protocols for both nested case-control studies.

We searched public databases^5,66 and previous articles (2, 3) for SNPs in HSD17B1 gene (also known as EDH17B2). We sequenced regions of exons 1, 4, 5, and 6, and introns 1, 3, 4, and 5 of the HSD17B1 gene in the forward and reverse directions to confirm the presence of 16 SNPs published previously [Table 1; SNP 1–4, 7–9, 11–13, and 15 (2); SNP 14 (3); SNP 5, 6, 10, and 16 (National Center for Biotechnology Information)] and identify new common variants. We chose these SNPs based on their location and/or frequency in a Caucasian population. We obtained DNA samples from 30 Caucasian women within the NHS (17 were controls in nested case-control studies and 13 were women with colon polyps), because 30 was the minimum number required to have >95% power to detect all of the variants with frequencies >5% (7). Genomic DNA was prepared by using a QIAamp 96 spin blood procedure (Qiagen, Chatsworth, CA) and genotyped for the 16 SNPs. Primers for the sequencing reaction were designed to avoid highly homologous areas between a known pseudogene and the HSD17B1 (8), and they are available on request. Briefly, we screened for variants in amplicons of various lengths (200–650 bp) using Big Dye Terminator chemistry (PE Applied Biosystems) on the ABI 377X automated sequencer (PE Applied Biosystems). Base calling of the sample files was performed by using ABI Sequence Analysis software version 3.1. Sequencher; version 3.0 alignment software was used to mark potential heterozygous positions and display them for evaluation. Heterozygotes were called at positions where the secondary peak height was ≥45% of the primary peak height in both forward and reverse sequence reads. Where possible, restriction digests with appropriate enzymes were performed to confirm the polymorphisms.

Haplotype frequencies were computed from phase-unknown genotypes among the 30 subjects using Arlequin 2.0. (L. Excoffier, S. Schneider, and D. Roessli⁷). This program estimates haplotype frequencies by using an expectation maximization algorithm. On the basis of 10 SNPs in the HSD17B1 gene (Table 1), three common (frequency of >5%) haplotypes accounted for 97% of the chromosomes at this locus. Due to the linkage disequilibrium between SNPs, only two SNPs were necessary to reconstruct the three common haplotypes (+1004C/T and +1322C/A; Table 1). Thus, we genotyped +1004C/T, +1322C/A, and an additional previously known nonsynonymous SNP [+1954A/G (Ser312Gly)] in the nested case-control studies of endometrial and breast cancer. Genotyping assays were performed by the 5′ nuclease assay (TaqMan) on the ABI PRISM 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA). TaqMan primers, probes, and conditions for genotyping assays are available on request. Genotyping was performed by laboratory personnel blinded to case-control status, and blinded quality control samples were inserted to validate genotyping procedures; concordance for the blinded samples was 100%. Haplotype frequencies based on the above three SNPs were estimated using PL-EM software (J. Liu, S. Qin, and T. Niu⁸; Ref. 9) and Arlequin 2.0. software for cases and controls.

We measured steroid hormone levels of estradiol, estrone, and estrone sulfate in postmenopausal women (n = 351) who had not used HRT for at least 3 months before the date of blood draw and had no previous diagnosis of cancer (except nonmelanoma skin cancer) in four separate batches. Methods for the plasma hormone assays and information regarding laboratory precision and reproducibility have been published previously (6, 10, 11). Briefly, estradiol and estrone were assayed by RIA preceded by organic extraction and celite chromatography. Estrone sulfate was assayed after extraction of estrone, by RIA (of estrone), after enzyme hydrolysis, organic extraction, and separation by column chromatography. Within-batch laboratory coefficients of variation were ≤13.6%.

We used the χ² test to assess whether the HSD17B1 genotypes were in Hardy-Weinberg equilibrium and used Fisher’s Exact test to determine Ps for differences in haplotype frequencies between cases and controls. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated by using conditional and, for the analyses stratified by menopausal status, by unconditional logistic regression. For the endometrial cancer analysis, in addition to the matching variables, we included the following risk factors in the regression models, body mass index (BMI) at age 18 (kg/m²); weight gain since age 18 (<5, 5–19.9, and ≥20 kg); HRT use at diagnosis (nonusers versus current users); age at menarche (<12, 12, 13, >13 years); age at menopause (<48, 48–<50, 50–<53, ≥53 years); parity/age at first birth (nulliparous, ≥1 child and ≤24 years, ≥1 child and >24 years); pack-years of smoking (never, <30, ≥30); first-degree family history of endometrial cancer (yes/no) and colorectal cancer (yes/no). For the breast cancer analyses, the following risk factors were adjusted for in addition to matching factors, duration of HRT use (past users <5 years, past users ≥5 years, current users <5 years, current users ≥5 years); age at menarche (<12, 12, 13, >13 years); age at menopause (<48, 48–<50, 50–<53, ≥53 years); BMI at age 18 (kg/m²); weight gain since age 18 (<5, 5–19.9, and ≥20 kg); parity/age at first birth (nulliparous, 1–2 children and ≤24 years, 1–2 children and >24 years, >2 children and ≤24 years, >2 children and >24 years); family history of breast cancer (yes/no); and personal history of benign breast disease (yes/no). For consistency with previous publications (4, 5), the GG genotype of +1954A/G was treated as the reference category in the regression model, and when evaluating gene-environment interactions, AG and GG of +1954A/G were grouped together as the reference category. For the other two SNPs, homozygosity for the more common allele (CC for both +1004C/T and +1322C/A) was used as the reference category.

Because of the low prevalence of the homozygote variants, we combined heterozygotes and homozygotes in the interaction analyses. To test statistical significance of interactions between environmental exposures and HSD17B1 genotypes, we used a likelihood-ratio test to compare nested models that included terms for all combinations of the HSD17B1 genotype and levels of environmental exposure to the models with indicator variables for the main effects only (nominal likelihood-ratio test).

Multiple linear regression models adjusting for BMI, age, time and date of blood draw, fasting status at blood draw, and laboratory batch were used to evaluate the association between genotypes and plasma steroid hormone levels in postmenopausal women not using HRT. The natural logarithm of the plasma hormone values was used in the analyses to reduce the skewness of the regression residuals. We used the SAS (SAS Institute, Cary, NC) statistical package for all of the analyses (SAS; version 8.2). All Ps were two-sided.

Six of 16 previously described SNPs were not detected in the 30 DNA samples from the NHS (Table 1). The distribution of all genotypes was in accordance with Hardy Weinberg equilibrium, and we confirmed that 7 SNPs (SNP ID 1, 2, 3, 7, 8, 9, and 15) were in complete linkage disequilibrium, as reported previously (2). We did not identify any additional new variants during the SNP validation procedure. We estimated four haplotypes of a theoretically possible 1024 (2¹⁰). Three common (frequency of >5%) haplotypes accounted for ∼97% of the chromosomes for the 30 NHS samples. Due to the complete linkage disequilibrium between SNPs, only two SNPs were necessary to distinguish the three common haplotypes (+1004C/T and +1322C/A). In addition, we genotyped +1954A/G (Ser312Gly) SNP because it alters an amino acid and has been studied previously in relation to breast cancer.

The characteristics of endometrial cancer cases and controls have been described previously (12). Briefly, endometrial cancer cases were more likely to be current HRT users, have higher BMI at diagnosis, gain more weight since age 18, smoke less, and have a family history of endometrial and colorectal cancer than controls. Breast cancer cases were more likely to be current HRT users, have a family history of breast cancer and have a personal history of benign breast disease than controls (13).

The haplotypes and their frequencies generated by PL-EM and ARLEQUIN 2.0. software were similar (only results from PL-EM are shown). In our study of mostly Caucasian women, we estimated six haplotypes of a theoretically possible eight (2³) in both endometrial and breast cancer nested case-control studies (Table 2). By using the Fisher’s Exact test, we observed no significant difference in frequency distribution in cancer cases and controls.

The distributions of HSD17B1 genotypes were in accordance with Hardy-Weinberg equilibrium, and were similar between incident and prevalent cases, so they were combined for all of the statistical analysis. The prevalence of variant carriers was (cases versus controls) 32% versus 27% for +1954A/G, 45% versus 45% for +1004C/T, and 50% versus 43% for +1322C/A(Table 3). After adjusting for potential confounders, we observed no statistically significant associations between +1954A/G (OR, 1.20; 95% CI, 0.83–1.74), +1004C/T (OR, 1.04; 95% CI, 0.74–1.46) and +1322C/A (OR, 1.37; 95% CI, 0.96–1.95) and endometrial cancer. There were too few endometrial cancer cases to allow additional stratification by menopausal status.

The distributions of HSD17B1 genotypes in cases and controls were in accordance with Hardy-Weinberg equilibrium and the prevalence of variant carriers was (cases versus controls) 27% versus 28% for +1954A/G, 48% versus 47% for +1004C/T, and 44% versus 46% for +1322C/A for all women (Table 4). After adjusting for potential confounders, no statistically significant association was observed between +1954A/G (OR, 0.94; 95% CI, 0.74–1.19), +1004C/T (OR, 0.96; 95% CI, 0.77–1.18), or +1322C/A (OR, 1.01; 95% CI, 0.83–1.24) and breast cancer among all women, or after stratification by menopausal status at diagnosis.

Because HSD17B1 is involved in estrogen metabolism, we chose BMI and HRT use among postmenopausal women as potential effect modifiers based on biological plausibility and their potential influence on estrogen levels. No statistically significant interaction was observed between BMI and HSD17B1, and endometrial cancer (data not shown). Because postmenopausal obesity has been associated with an increase in breast cancer risk, we limited our analysis to postmenopausal women only. In postmenopausal breast cancer, we observed potential interactions between BMI and both the +1954A/G genotype (nominal likelihood-ratio test P = 0.05) and the +1322C/A genotype (nominal likelihood-ratio test P = 0.01). Among women with a BMI >30 kg/m², the AA genotype of +1954A/G was associated with an increased breast cancer risk (OR, 1.77; 95% CI, 0.99–3.17) when compared with the AG and the GG genotype. Among lean women (BMI <25 kg/m²), we observed an inverse association between the CA and the AA genotype of +1322C/A (OR, 0.74; 95% CI, 0.57–0.97) when compared with the CC genotype, which was not apparent in overweight women. Although HRT use is another well-established risk factor for both endometrial and breast cancer, we observed no significant interaction between HSD17B1 polymorphisms, with never, past, or current HRT use in both cancers (data not shown).

Among lean women (BMI <25 kg/m²), the AA genotype of the +1954A/G (Ser312Gly) polymorphism was associated with a significant increase in the level of estradiol (P = 0.01) compared with the AG and GG genotype (Table 5). The ratio of estrone to estradiol, which may be a better measure of enzyme activity, was also significantly different (P = 0.03) in lean women with the AA genotype of the +1954A/G (Ser312Gly) polymorphism compared with the AG and GG genotype. No significant difference was observed between each genotype of HSD17B1 and other plasma steroid hormone levels (estrone sulfate and estrone) among postmenopausal women not using HRT.

To our knowledge, our study is the first population-based study that describes polymorphisms in the HSD17B1 gene and their potential relationship to endometrial cancer risk. Three population-based studies have evaluated the HSD17B1 polymorphisms +1954A/G (Ser312Gly) in relation to breast cancer, but none has reported data on the other two common polymorphisms in the HSD17B1 gene (+1004C/T and +1322C/A) in breast cancer risk and how they relate to plasma steroid hormone levels. In addition, our study is the first to infer haplotypes in the HSD17B1 gene, and compare their frequency distributions in cancer cases and controls.

We used SNP data from Normand et al. (2) and Mannermaa et al. (3), and supplemented them with SNPs from the public databases for the HSD17B1 gene. We validated the SNPs in a subset of the population and confirmed Normand et al. (2) findings in their study of 29 Caucasians and 1 Asian that seven SNPs are in complete linkage disequilibrium. We validated the presence of three common SNPs [+1954A/G (Ser312Gly), +1004C/T, and +1322C/A] in our study, and their frequencies are consistent with a previous report (2); we did not observe any additional novel variants in the amplicons sequenced to validate these SNPs.

In both prospective nested case-control studies, we observed no significant difference in haplotype distribution in cancer cases and controls, and we also did not observe any statistically significant association between the HSD17B1 polymorphisms +1954A/G, +1004C/T, or +1322C/A and the risk of endometrial or breast cancer. The +1954A/G is the only common polymorphism in the coding region (exon 6) that results in an amino acid change, serine to glycine at position 312 in the HSD17B1 protein. Site-directed mutagenesis experiments have failed to demonstrate any changes in the catalytic or immunological properties of the HSD17B1 enzyme resulting from this amino acid change (14). The function of +1004C/T and +1322C/A has not been studied, but they are not expected to influence RNA splicing, because they are not at the exon-intron boundary.

The earliest study of the HSD17B1 +1954A/G polymorphism and breast cancer suggested a difference in the genotype frequencies between cases and controls; however, the genotype frequency in the control group was not in accordance with Hardy-Weinberg equilibrium (3). This study included breast cancer cases from Finland (n = 149) and the United Kingdom (familial breast cancer, n = 41), and controls (161 Finnish and 29 United Kingdom), and no relevant risk factor information was available for statistical adjustment. The second case-control study of the HSD17B1 +1954A/G polymorphism and breast cancer was nested within the Multiethnic Cohort Study, which included African-Americans, Japanese, Native Hawaiians, non-Latino Caucasians, and Latino women (cases, n = 615; Ref. 4). After adjusting for a CYP17 polymorphism, age, weight, and ethnicity, the authors found no statistically significant association between the HSD17B1 +1954A/G polymorphism and stage 1 breast cancer (AA versus GG genotype: OR, 0.99; 95% CI, 0.76–1.29) or with advanced cases (AA versus GG genotype: OR, 1.28; 95% CI, 0.85–1.93). A study of Chinese women in Singapore by Wu et al. (5) showed no association with overall breast cancer risk (AA versus AG/GG genotype: OR, 1.37; 95% CI, 0.90–2.07). After stratification by menopausal status, the authors observed a statistically significant association in postmenopausal women (OR, 1.86; 95% CI, 1.14–3.03). Our results did not indicate an association between +1954A/G HSD17B1 and breast cancer among postmenopausal women (OR, 0.95; 95% CI, 0.76–1.19). It should be noted that Wu et al. (5) considered their findings preliminary, as their number of cases was small (124 postmenopausal breast cancer cases).

Because the HSD17B1 enzyme catalyzes the conversion of estrone to estradiol, we examined whether HSD17B1 polymorphisms influence the plasma levels of estrone, estrone sulfate, and estradiol in postmenopausal women who were not using HRT. Our data suggest that in lean women there is a significant increase in the level of estradiol and the ratio of estradiol to estrone associated with the AA genotype of the +1954A/G (Ser312Gly) polymorphism. It has been hypothesized that the effect of polymorphisms in estrogen metabolizing genes on the level of circulating estrogens would be most pronounced in women without other sources of estrogens, i.e., lean women in whom peripheral conversion of androgens in the adipose tissue would be minimal and women who are not currently using HRT (4). This elevation, however, did not translate into an increased risk of breast cancer among lean women; indeed, it was among obese women that we observed a marginally significant increase in risk associated with the AA genotype. It is possible that the AA genotype gives information about increase of estradiol exposure earlier in life, whereas among obese postmenopausal women the elevation in estradiol due to the AA genotype is masked by the strong effect of obesity on estradiol levels, although a possibility of chance finding cannot be excluded.

The relatively large sample size, prospective design, and extensive relevant lifestyle information are among the strengths of this study. In summary, we found that three common SNPs within the HSD17B1 gene were not associated with endometrial or breast cancer risk; however, we observed an association between HSD17B1 and circulating estradiol levels in lean postmenopausal women and a potential interaction with BMI in postmenopausal breast cancer that requires confirmation.

We thank Rong Chen, Hardeep Ranu, and Monica McGrath for their technical assistance. We are also indebted to the participants in the NHS for their dedication and commitment.