Abstract
The 17β-hydroxysteroid dehydrogenase 1 gene (HSD17B1) encodes 17HSD1, which catalyzes the final step of estradiol biosynthesis. Despite the important role of HSD17B1 in hormone metabolism, few epidemiologic studies of HSD17B1 and breast cancer have been conducted. This study includes 5,370 breast cancer cases and 7,480 matched controls from five large cohorts in the Breast and Prostate Cancer Cohort Consortium. We characterized variation in HSD17B1 by resequencing and dense genotyping a multiethnic sample and identified haplotype-tagging single nucleotide polymorphisms (htSNP) that capture common variation within a 33.3-kb region around HSD17B1. Four htSNPs, including the previously studied SNP rs605059 (S312G), were genotyped to tag five common haplotypes in all cases and controls. Conditional logistic regression was used to estimate odds ratios (OR) for disease. We found no evidence of association between common HSD17B1 haplotypes or htSNPs and overall risk of breast cancer. The OR for each haplotype relative to the most common haplotype ranged from 0.98 to 1.07 (omnibus test for association: X2 = 3.77, P = 0.58, 5 degrees of freedom). When cases were subdivided by estrogen receptor (ER) status, two common haplotypes were associated with ER-negative tumors (test for trend, Ps = 0.0009 and 0.0076; n = 353 cases). HSD17B1 variants that are common in Caucasians are not associated with overall risk of breast cancer; however, there was an association among the subset of ER-negative tumors. Although the probability that these ER-negative findings are false-positive results is high, these findings were consistent across each cohort examined and warrant further study. (Cancer Res 2006; 66(4): 2468-75)
Introduction
The 17β-hydroxysteroid dehydrogenase 1 gene (HSD17B1 [MIM 109684]) encodes 17HSD1, whose primary function is to catalyze the final step of estradiol biosynthesis, converting estrone to the more biologically active estradiol (1). Its complement, 17HSD2 (encoded by HSD17B2) is the enzyme that predominates in the reverse reaction, the oxidation of estradiol to estrone. The balance of these two enzymes, in part, regulates estrogen concentrations in breast tissue. In normal breast tissue, 17HSD2 activity predominates, whereas 17HSD1 activity predominates in malignant breast tissue (2). HSD17B1 amplification in breast tumors correlates with poorer prognosis, especially among women with estrogen receptor (ER)–positive tumors (2). Furthermore, breast cancer patients with tumors expressing 17HSD1 mRNA or protein had significantly shorter overall (P = 0.001) and disease-free (P = 0.015) survival than other patients, and detection of 17HSD1 mRNA in the tumor was an independent prognostic marker in multivariate analyses (1).
Despite these intriguing data, the relation of germ line variation in HSD17B1 to breast cancer incidence has not been well investigated in epidemiologic studies. To date, only four epidemiologic studies of HSD17B1 variation and breast cancer have been conducted, the largest of which included ∼1,000 cases (3–6). Although several sequence variations in HSD17B1 have been identified (3, 7, 8), three of these studies (4–6) focused only on a single polymorphism in exon 6, designated S312G (rs605059). This single nucleotide polymorphism (SNP) results in an amino acid change from serine (allele A) to glycine (allele G) but does not seem to affect the catalytic or immunologic properties of the enzyme (9). The results of these studies have been inconclusive but have provided some evidence that HSD17B1 may influence risk of breast cancer.
We report here the results from an analysis of HSD17B1 haplotypes and breast cancer risk from a large, collaborative study (The Breast and Prostate Cancer Cohort Consortium, or BPC3), which includes data from five cohorts from the United States and Europe (10). The large size of this study enables us to detect modest genetic effects, explore gene-environment interactions, and examine potentially important subclasses of tumors, such as those defined by stage or hormone receptors.
Materials and Methods
Study population. The BPC3 has been described in detail elsewhere (10). Briefly, the consortium includes large well-established cohorts (or consortia of smaller cohorts) assembled in the United States and Europe that have DNA for genotyping and extensive questionnaire data from cohort members. Written informed consent was obtained from all subjects, and each cohort has been approved by the appropriate institutional review board. This analysis includes 5,370 cases of invasive breast cancer and 7,480 matched controls from five cohorts: the American Cancer Society Cancer Prevention Study II (CPS-II; ref. 11), the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort (12), the Harvard Nurse's Health Study (NHS; ref. 13) and Women's Health Study (WHS; ref. 14), and the Hawaii-Los Angeles Multiethnic Cohort (MEC; ref. 15). With the exception of MEC, most women in these cohorts are Caucasians. The MEC includes U.S. Caucasians, African Americans, Latinos, Japanese, and native Hawaiians (15). Cases were identified in each cohort by self-report, with subsequent confirmation of the diagnosis from medical records or tumor registries and/or linkage with population-based tumor registries. Information on ER and progesterone receptor (PR) status was abstracted from medical records and/or tumor registries and was not available from the EPIC cohort. ER and PR receptor data was reported as positive, negative, borderline, or not available. In all cohorts, questionnaire data was collected prospectively before the diagnosis of cancer. Blood samples were collected before diagnosis in all cohorts except for the MEC and CPS-II cohorts, where most of the blood specimens were collected after diagnosis (10, 11, 16). Cases classified as carcinoma in situ were not included in this analysis. Controls were matched to cases by ethnicity and age, and in some cohorts, additional matching criteria were employed (e.g., EPIC matched on country of residence.) These analyses include a portion of the previously published data from the NHS cohort (3) and MEC (5).
Haplotype discovery and haplotype-tagging SNP selection. The BPC3 adopted a two-stage approach to comprehensively measure genetic variation in and around HSD17B1 among cases and controls. The first stage consists of comprehensive haplotype discovery followed by haplotype-tagging SNP (htSNP) selection. The second stage is genotyping the htSNPs in all the BPC3 cases and controls.
In the first stage, we genotyped a dense set of SNPs spanning the region of interest in five population samples to identify regions of high linkage disequilibrium and low haplotype diversity using the algorithm of Gabriel et al. (17) as implemented in Haploview.25
Novel SNPs were identified by systematically resequencing the HSD17B1 exons in 95 cases of advanced prostate cancer and 95 advanced breast cancer cases from five population groups (equal numbers of U.S. Caucasians, Latinos, Japanese, native Hawaiians, and African Americans) in the MEC (resequencing details at MEC web site).26 Then, SNPs were selected from public databases to cover the introns and flanking regions around HSD17B1.In total, 26 SNPs were selected covering a 42-kb region around HSD17B1 at an average density of one SNP per 1.6 kb. All but one of these SNPs (rs7208557 in the 3′ region) had a minor allele frequency >5% among Caucasians. The SNPs extended in the 5′ direction of HSD17B1 into the adjoining N-acetylglucosaminidase-α (NAGLU) gene and pseudogene for HSD17B1 (HSD17BP1), and in the 3′ direction into the genes for CoA synthase (COASY) and transcription factor-like 4 (TCFL4).
To identify regions of high linkage disequilibrium, these 26 SNPs were genotyped in a multiethnic panel of 349 unrelated women from the MEC with no history of cancer. Tagging SNPs were then chosen using the partition-ligation EM algorithm implemented in the program TAGSNPS.27
Selection of htSNPs is based on RH2, a measure of the correlation between observed haplotypes and those predicted based on htSNP genotypes (18). This haplotype-tagging approach is based on the observation that within blocks of high linkage disequilibrium, there is limited haplotype diversity and that common variation in the region is highly correlated with the common haplotype patterns (17). Nineteen SNPs fall into a block of high linkage disequilibrium that could be characterized efficiently with four htSNPs: rs676387, rs598126, rs2010750, and the S312G SNP in exon 6 (rs605059), which we required to be in the htSNP set. These four htSNPs were genotyped in all the BPC3 cases and controls.Genotyping. The 26 SNPs used for haplotype construction were genotyped in a reference panel made up of MEC populations at the Broad Institute using Sequenom and Illumina platforms. Genotyping the four htSNPs in the breast cancer cases and controls was done in four laboratories using a fluorescent 5′ endonuclease assay and the ABI-PRISM 7900 for sequence detection (Taqman). Initial quality control checks of the SNP assays were done at the manufacturer (ABI, Foster City, CA); an additional 500 test reactions were run by the BPC3. Assay characteristics for the four htSNPs for HSD17B1 are available on a public web site.28
Sequence validation for each SNP assay was done, and 100% concordance was observed.29To assess interlaboratory variation, each genotyping center ran assays on a designated set of 94 samples from the Coriell Biorepository (Camden, NJ), showing completion and concordance rates of >99% (19). The internal quality of genotype data at each genotyping center was assessed by typing 5% to 10% blinded samples in duplicate or triplicate (depending on study); resulting concordance was >99%.
Statistical analysis. We used conditional multiple logistic regression to estimate odds ratios (OR) for disease in subjects with a linear (additive) scoring for 0, 1, or 2 copies of the minor allele of each SNP. We also used conditional logistic regression with additive scoring and the most common haplotype as the reference to estimate haplotype-specific ORs using an expectation substitution approach to assign haplotypes based on the unphased genotype data (20). It has been shown (21, 22) that this methods performs well despite the uncertainty in assignment (20, 21). Haplotype frequencies and expected subject-specific haplotype indicators were calculated separately for each cohort (and country within EPIC). To test the global null hypothesis of no association between variation in HSD17B1 haplotypes and htSNPs and risk of breast cancer (or subtypes defined by receptor status), we used a likelihood ratio test comparing a model with additive effects for each common haplotype (treating the most common haplotype as the referent) to the intercept-only model. We combined rare haplotypes (those with estimated individual frequencies <1%) into a single category, which comprised <0.5% of the controls. To test for heterogeneity across cohorts and ethnic groups, we used the Wald χ2 for htSNPs and a likelihood ratio test for the haplotypes.
We considered conditional models both without adjustment and with adjustment for known breast cancer risk factors, including age at menarche, menopausal status, age at menopause, parity, age at first birth, history of benign breast disease, body mass index (BMI, in deciles), first-degree family history of breast cancer, and use of postmenopausal hormones. Because the results remained essentially unchanged regardless of the model used, we present results from the unadjusted conditional model. We evaluated these same covariates for possible interaction effects and also tested whether the association between HSD17B1 and breast cancer differed by stage (localized versus regional or distant metastasis) or hormone receptor (ER and PR) status.
Logistic models to examine associations for specific hormone receptor subtypes (ER positive, ER negative, PR positive, PR negative) included only cases classified as receptor positive or receptor negative, controlled for age and cohort, and were stratified by ethnicity because there was statistical evidence of heterogeneity by ethnicity. Cases without hormone receptor data or with receptor status of “borderline” were not included in the hormone receptor analyses. Controls in these models included all controls from each cohort that provided data on receptor status. Tests for heterogeneity by receptor status were obtained from case-only models comparing receptor-negative cases with receptor-positive cases.
Finally, we established a range of prior probabilities that variation in HSD17B1 is related to breast cancer based on existing epidemiologic and laboratory data to evaluate the false-positive or false-negative report probabilities (23). Based on existing evidence (3–6, 24, 25), we assumed a prior probability of 1%, with a range of 10% to 0.1% for an association between HSD17B1 and overall breast cancer. Although we did not specify prior probabilities a priori for associations with ER-positive and ER-negative tumors, the prior probability for ER-positive tumors should be about the same as for overall breast cancer and perhaps 10-fold lower for ER-negative tumors, where the role of estrogen is less clear. The prior probability for any given haplotype or SNP in the gene is also somewhat less than the prior probability for the gene (23).
Results
The genomic structure of the region around HSD17B1 is shown in Fig. 1. The four htSNPs chosen capture most of the variation, known and unknown, of all common haplotypes in this block (frequencies > 0.05, in at least one ethnic group). In each ethnic group, the four htSNPs predicted common haplotypes with a minimum RH2 above 80%. However, among African Americans, the cumulative frequency of the common haplotypes was only 62% (Supplementary Table 1S). A cumulative frequency of ≥70% among African Americans would be achieved only by genotyping three additional htSNPs, tagging two additional haplotypes, each with a frequency of just under 5%.30
Among the controls in the BPC3 (n = 7,480), the five common htSNP haplotypes account for 99% of all haplotypes. One haplotype (CAAC) is common only among African Americans (frequency = 6.1% among African-American controls). Haplotype frequencies by cohort are shown in Fig. 2. (Haplotype frequencies by population in the MEC are shown in Supplementary Fig. 1S.).SNPs and linkage disequilibrium structure (among Caucasians) across the HSD17B1 locus. Twenty-six SNPs were selected covering a 42-kb region around HSD17B1 at an average density of one SNP per 1.6 kb. Haplotype tagging SNPs (red arrows).
SNPs and linkage disequilibrium structure (among Caucasians) across the HSD17B1 locus. Twenty-six SNPs were selected covering a 42-kb region around HSD17B1 at an average density of one SNP per 1.6 kb. Haplotype tagging SNPs (red arrows).
The genotyping success rate was ≥94% for each of the four htSNPS at each genotyping center. No deviation from Hardy-Weinberg equilibrium was observed among the controls in each cohort (at the P < 0.01 level) or across more than one cohort for any given assay.
Study characteristics of each cohort are provided in Table 1. Case and control characteristics were comparable across cohorts. The majority of cases were postmenopausal. The mean age of diagnosis ranged from 57.8 in EPIC to 70.2 in CPS-II, reflecting differences in the age and length of follow-up in these cohorts. The percentage of women who reported age at menarche over 14 years was higher among European women in EPIC (18% of controls) than in the North American cohorts (6-12% of controls). European women also reported less hormone replacement therapy use than U.S. women.
Descriptive characteristics of invasive breast cancer cases (n = 5,370) and controls (n = 7,480) by cohort in the BPC3
. | CPS-II . | . | EPIC . | . | MEC . | . | NHS . | . | WHS . | . | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Cases . | Controls . | Cases . | Controls . | Cases . | Controls . | Cases . | Controls . | Cases . | Controls . | ||||||||||
Number | 395 | 505 | 1,610 | 2,844 | 1,601 | 1,962 | 1,059 | 1,464 | 705 | 705 | ||||||||||
Ethnicity | ||||||||||||||||||||
White | 98 | 99 | 100 | 100 | 25 | 22 | 94 | 94 | 96 | 96 | ||||||||||
Hispanic | 1 | 0 | 21 | 20 | 0 | 0 | 1 | 1 | ||||||||||||
African American | 1 | 1 | 21 | 22 | 1 | 1 | 1 | 1 | ||||||||||||
Asian | 0 | 0 | 26 | 21 | 0 | 0 | 1 | 1 | ||||||||||||
Hawaiian | 0 | 0 | 7 | 15 | 0 | 0 | 0 | 0 | ||||||||||||
Other/missing | 0 | 0 | 0 | 0 | 5 | 5 | 2 | 2 | ||||||||||||
Age at diagnosis (mean) | 70.2 | 57.8 | 65.1 | 63.2 | 60.3 | |||||||||||||||
Menopausal status (%) | ||||||||||||||||||||
Premenopausal | 23 | 28 | 11 | 16 | 19 | 17 | 22 | 21 | ||||||||||||
Postmenopausal | 100 | 100 | 68 | 63 | 87 | 82 | 71 | 75 | 63 | 60 | ||||||||||
Unknown/missing | 9 | 9 | 2 | 2 | 9 | 8 | 15 | 19 | ||||||||||||
Age at menarche (%) | ||||||||||||||||||||
≤12 | 45 | 45 | 34 | 35 | 53 | 49 | 52 | 48 | 57 | 52 | ||||||||||
13-14 | 44 | 46 | 46 | 44 | 35 | 38 | 39 | 43 | 37 | 43 | ||||||||||
≥15 | 9 | 8 | 15 | 18 | 10 | 12 | 8 | 8 | 6 | 6 | ||||||||||
Unknown/missing | 2 | 1 | 4 | 3 | 2 | 1 | 1 | 1 | 0 | 0 | ||||||||||
Age at menopause* | ||||||||||||||||||||
≤44 | 20 | 22 | 11 | 14 | 30 | 35 | 22 | 22 | 16 | 22 | ||||||||||
45-49 | 20 | 28 | 22 | 27 | 26 | 27 | 28 | 28 | 29 | 30 | ||||||||||
50-54 | 44 | 39 | 37 | 37 | 32 | 28 | 44 | 44 | 41 | 35 | ||||||||||
≥55 | 14 | 10 | 8 | 8 | 9 | 7 | 6 | 6 | 7 | 8 | ||||||||||
Unknown/missing | 3 | 2 | 22 | 15 | 3 | 3 | 0 | 0 | 6 | 6 | ||||||||||
Parity (%) | ||||||||||||||||||||
Nulliparous | 8 | 9 | 13 | 13 | 14 | 11 | 8 | 7 | 15 | 14 | ||||||||||
≤2 children | 36 | 27 | 55 | 52 | 36 | 35 | 33 | 31 | 40 | 34 | ||||||||||
≥3 children | 53 | 62 | 26 | 30 | 48 | 53 | 59 | 62 | 45 | 52 | ||||||||||
Unknown/missing | 3 | 2 | 6 | 5 | 1 | 1 | 1 | 1 | 0 | 0 | ||||||||||
First-degree family history (%) | ||||||||||||||||||||
Yes | 20 | 15 | 17 | 11 | 19 | 14 | 20 | 16 | ||||||||||||
No | 78 | 81 | 83 | 89 | 81 | 86 | 79 | 83 | ||||||||||||
Unknown | 2 | 4 | 100 | 100 | 0 | 0 | 0 | 0 | 2 | 1 | ||||||||||
Hormone replacement therapy use (%)* | ||||||||||||||||||||
Never | 34 | 41 | 54 | 64 | 36 | 41 | 23 | 28 | 30 | 38 | ||||||||||
Ever | 66 | 59 | 42 | 33 | 64 | 58 | 77 | 72 | 65 | 57 | ||||||||||
Unknown/missing | 0 | 0 | 4 | 3 | 0 | 1 | 0 | 0 | 5 | 5 | ||||||||||
ER (%) | ||||||||||||||||||||
Positive | 62 | 62 | 68 | 80 | ||||||||||||||||
Negative | 8 | 16 | 18 | 14 | ||||||||||||||||
Borderline | 1 | 1 | 1 | 0 | ||||||||||||||||
Not available | 29 | 100 | 21 | 13 | 5 | |||||||||||||||
PR (%) | ||||||||||||||||||||
Positive | 53 | 51 | 58 | 72 | ||||||||||||||||
Negative | 15 | 23 | 26 | 21 | ||||||||||||||||
Borderline | 1 | 1 | 1 | 1 | ||||||||||||||||
Not available | 31 | 100 | 25 | 15 | 6 |
. | CPS-II . | . | EPIC . | . | MEC . | . | NHS . | . | WHS . | . | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Cases . | Controls . | Cases . | Controls . | Cases . | Controls . | Cases . | Controls . | Cases . | Controls . | ||||||||||
Number | 395 | 505 | 1,610 | 2,844 | 1,601 | 1,962 | 1,059 | 1,464 | 705 | 705 | ||||||||||
Ethnicity | ||||||||||||||||||||
White | 98 | 99 | 100 | 100 | 25 | 22 | 94 | 94 | 96 | 96 | ||||||||||
Hispanic | 1 | 0 | 21 | 20 | 0 | 0 | 1 | 1 | ||||||||||||
African American | 1 | 1 | 21 | 22 | 1 | 1 | 1 | 1 | ||||||||||||
Asian | 0 | 0 | 26 | 21 | 0 | 0 | 1 | 1 | ||||||||||||
Hawaiian | 0 | 0 | 7 | 15 | 0 | 0 | 0 | 0 | ||||||||||||
Other/missing | 0 | 0 | 0 | 0 | 5 | 5 | 2 | 2 | ||||||||||||
Age at diagnosis (mean) | 70.2 | 57.8 | 65.1 | 63.2 | 60.3 | |||||||||||||||
Menopausal status (%) | ||||||||||||||||||||
Premenopausal | 23 | 28 | 11 | 16 | 19 | 17 | 22 | 21 | ||||||||||||
Postmenopausal | 100 | 100 | 68 | 63 | 87 | 82 | 71 | 75 | 63 | 60 | ||||||||||
Unknown/missing | 9 | 9 | 2 | 2 | 9 | 8 | 15 | 19 | ||||||||||||
Age at menarche (%) | ||||||||||||||||||||
≤12 | 45 | 45 | 34 | 35 | 53 | 49 | 52 | 48 | 57 | 52 | ||||||||||
13-14 | 44 | 46 | 46 | 44 | 35 | 38 | 39 | 43 | 37 | 43 | ||||||||||
≥15 | 9 | 8 | 15 | 18 | 10 | 12 | 8 | 8 | 6 | 6 | ||||||||||
Unknown/missing | 2 | 1 | 4 | 3 | 2 | 1 | 1 | 1 | 0 | 0 | ||||||||||
Age at menopause* | ||||||||||||||||||||
≤44 | 20 | 22 | 11 | 14 | 30 | 35 | 22 | 22 | 16 | 22 | ||||||||||
45-49 | 20 | 28 | 22 | 27 | 26 | 27 | 28 | 28 | 29 | 30 | ||||||||||
50-54 | 44 | 39 | 37 | 37 | 32 | 28 | 44 | 44 | 41 | 35 | ||||||||||
≥55 | 14 | 10 | 8 | 8 | 9 | 7 | 6 | 6 | 7 | 8 | ||||||||||
Unknown/missing | 3 | 2 | 22 | 15 | 3 | 3 | 0 | 0 | 6 | 6 | ||||||||||
Parity (%) | ||||||||||||||||||||
Nulliparous | 8 | 9 | 13 | 13 | 14 | 11 | 8 | 7 | 15 | 14 | ||||||||||
≤2 children | 36 | 27 | 55 | 52 | 36 | 35 | 33 | 31 | 40 | 34 | ||||||||||
≥3 children | 53 | 62 | 26 | 30 | 48 | 53 | 59 | 62 | 45 | 52 | ||||||||||
Unknown/missing | 3 | 2 | 6 | 5 | 1 | 1 | 1 | 1 | 0 | 0 | ||||||||||
First-degree family history (%) | ||||||||||||||||||||
Yes | 20 | 15 | 17 | 11 | 19 | 14 | 20 | 16 | ||||||||||||
No | 78 | 81 | 83 | 89 | 81 | 86 | 79 | 83 | ||||||||||||
Unknown | 2 | 4 | 100 | 100 | 0 | 0 | 0 | 0 | 2 | 1 | ||||||||||
Hormone replacement therapy use (%)* | ||||||||||||||||||||
Never | 34 | 41 | 54 | 64 | 36 | 41 | 23 | 28 | 30 | 38 | ||||||||||
Ever | 66 | 59 | 42 | 33 | 64 | 58 | 77 | 72 | 65 | 57 | ||||||||||
Unknown/missing | 0 | 0 | 4 | 3 | 0 | 1 | 0 | 0 | 5 | 5 | ||||||||||
ER (%) | ||||||||||||||||||||
Positive | 62 | 62 | 68 | 80 | ||||||||||||||||
Negative | 8 | 16 | 18 | 14 | ||||||||||||||||
Borderline | 1 | 1 | 1 | 0 | ||||||||||||||||
Not available | 29 | 100 | 21 | 13 | 5 | |||||||||||||||
PR (%) | ||||||||||||||||||||
Positive | 53 | 51 | 58 | 72 | ||||||||||||||||
Negative | 15 | 23 | 26 | 21 | ||||||||||||||||
Borderline | 1 | 1 | 1 | 1 | ||||||||||||||||
Not available | 31 | 100 | 25 | 15 | 6 |
Percentages for age at menopause and hormone replacement therapy use include postmenopausal women only.
As there was no heterogeneity in results for any of the main effects by cohort, we report results from only pooled analyses here; cohort-specific results are available in supplemental tables. Table 2 presents the pooled haplotype and SNP associations for HSD17B1 and breast cancer. There were no significant differences in the haplotype frequencies between cases and controls (P of the omnibus test for association with breast cancer: X2 = 3.77 with 5 degrees of freedom, P = 0.58). None of the individual SNPs, including the S312G SNP (rs605059), were associated with breast cancer. There was also no evidence that any of the individual rare haplotypes (defined as any haplotype with a frequency of <5%) were over represented in the cases (data not shown). Cohort-specific and population-specific results are shown in Supplementary Tables 2S-3S.
HSD17B1 haplotype and SNP associations with invasive breast cancer among U.S. and European women in BPC3
. | Cases (%*), n = 5,370 . | Controls (%*), n = 7,480 . | OR† (95% confidence interval) . | |||
---|---|---|---|---|---|---|
Haplotypes‡ | ||||||
CGAT (reference) | 4,228.5 (39) | 5,919.5 (40) | 1.00 | |||
AAGC | 3,048.2 (28) | 4,242.9 (28) | 1.03 (0.96-1.09) | |||
CAGC | 2,479.0 (23) | 3,535.8 (24) | 0.98 (0.92-1.05) | |||
CGAC | 793.5 (7) | 1,052.9 (7) | 1.05 (0.95-1.17) | |||
CAAC§ | 50.9 (1) | 59.9 (0) | 1.07 (0.71-1.61) | |||
All rare | 59.9 (1) | 62.9 (0) | 1.26 (0.88-1.80) | |||
rs676387∥ | ||||||
C/C | 2,666 (50) | 3,740 (50) | 1.00 | |||
C/A | 2,088 (39) | 2,931 (39) | 1.01 (0.94-1.10) | |||
A/A | 457 (9) | 618 (8) | 1.08 (0.95-1.24) | |||
S312G (rs605059) | ||||||
A/A | 1,406 (26) | 2,011 (27) | 1.00 | |||
A/G | 2,588 (48) | 3,610 (48) | 1.02 (0.93-1.11) | |||
G/G | 1,139 (21) | 1,580 (21) | 1.01 (0.91-1.12) | |||
rs598126 | ||||||
G/G | 1,399 (26) | 1,968 (28) | 1.00 | |||
G/A | 2,600 (48) | 363 (49) | 1.00 (0.92-1.09) | |||
A/A | 1,182 (22) | 1,642 (22) | 0.98 (0.89-1.09) | |||
rs2010750 | ||||||
C/C | 1,854 (35) | 2,593 (35) | 1.00 | |||
C/T | 2,453 (46) | 3,475 (46) | 0.98 (0.90-1.06) | |||
T/T | 834 (15) | 1,169 (16) | 0.98 (0.88-1.09) |
. | Cases (%*), n = 5,370 . | Controls (%*), n = 7,480 . | OR† (95% confidence interval) . | |||
---|---|---|---|---|---|---|
Haplotypes‡ | ||||||
CGAT (reference) | 4,228.5 (39) | 5,919.5 (40) | 1.00 | |||
AAGC | 3,048.2 (28) | 4,242.9 (28) | 1.03 (0.96-1.09) | |||
CAGC | 2,479.0 (23) | 3,535.8 (24) | 0.98 (0.92-1.05) | |||
CGAC | 793.5 (7) | 1,052.9 (7) | 1.05 (0.95-1.17) | |||
CAAC§ | 50.9 (1) | 59.9 (0) | 1.07 (0.71-1.61) | |||
All rare | 59.9 (1) | 62.9 (0) | 1.26 (0.88-1.80) | |||
rs676387∥ | ||||||
C/C | 2,666 (50) | 3,740 (50) | 1.00 | |||
C/A | 2,088 (39) | 2,931 (39) | 1.01 (0.94-1.10) | |||
A/A | 457 (9) | 618 (8) | 1.08 (0.95-1.24) | |||
S312G (rs605059) | ||||||
A/A | 1,406 (26) | 2,011 (27) | 1.00 | |||
A/G | 2,588 (48) | 3,610 (48) | 1.02 (0.93-1.11) | |||
G/G | 1,139 (21) | 1,580 (21) | 1.01 (0.91-1.12) | |||
rs598126 | ||||||
G/G | 1,399 (26) | 1,968 (28) | 1.00 | |||
G/A | 2,600 (48) | 363 (49) | 1.00 (0.92-1.09) | |||
A/A | 1,182 (22) | 1,642 (22) | 0.98 (0.89-1.09) | |||
rs2010750 | ||||||
C/C | 1,854 (35) | 2,593 (35) | 1.00 | |||
C/T | 2,453 (46) | 3,475 (46) | 0.98 (0.90-1.06) | |||
T/T | 834 (15) | 1,169 (16) | 0.98 (0.88-1.09) |
Percentages do not sum to 100% because of missing genotype data.
Matched for age, cohort, and ethnicity.
Alleles listed in 5′ to 3′ order; global test for haplotype association with breast cancer: X2 = 3.77, with 5 degrees of freedom for P = 0.58.
Only common among African Americans (frequency of 6% among African-American controls).
htSNPs listed in 5′ to 3′ order.
We tested for statistical interaction with HSD17B1 and the following breast cancer risk factors: age at menarche, menopausal status, age at menopause, parity, age at first birth, history of benign breast disease, BMI (in deciles), first-degree family history of breast cancer, and use of postmenopausal hormones (estrogen-only therapy and combination estrogen-progesterone therapy). We did not find any consistent evidence of effect modification with any of the examined risk factors for any of the SNPs or common haplotypes, nor did the association between HSD17B1 variation and breast cancer differ by tumor stage at diagnosis (data not shown).
Data on receptor status were available from four of the five participating cohorts, including 353 cases of ER-negative tumors, 1,723 cases of ER-positive tumors, and 352 unclassified tumors among U.S. Caucasian women (Table 1). Missing data varied by cohort. The WHS had ER receptor data available on all but 5% of cases, whereas CPS-II had missing ER data for 29% of cases; no receptor data was available from the EPIC cohort. When the data were stratified based on ER status of the tumors, we found statistical evidence of heterogeneity for two htSNPs and one haplotype (at P < 0.05). As shown in Table 3, each of the four htSNPs is statistically significantly associated with ER-negative tumors but not with ER-positive tumors. Each of the corresponding haplotypes that carry all high-risk alleles (CGAT) or all low-risk alleles (AAGC) for each htSNP is also associated with ER-negative tumors (Ptrend = 0.0076 and 0.0009, respectively). All these htSNP and haplotype associations show evidence of a dose-response relationship for ER-negative tumors, with stronger associations for homozygotes than heterozygotes. Furthermore, analysis of haplotype combinations (Table 4) shows that every common haplotype combination that includes AAGC is associated with a reduced risk of ER-negative breast cancer. This association with the AAGC haplotype and ER-negative tumors is present among U.S. Caucasians in each of the four cohorts that contributed information on receptor status. (Cohort specific associations by ER status and specific numbers of cases and controls from each cohort are shown in Supplementary Tables 4S-5S).
Association between HSD17B1 and invasive breast cancer among U.S. Caucasian women by ER status
. | ER-positive* . | . | . | . | ER-negative* . | . | . | . | Phet§ . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Cases (%)†, n = 1,737 . | Controls (%)†, n = 2,982 . | OR (95% confidence interval) . | Ptrend‡ . | Cases (%)†, n = 354 . | Controls (%)†, (n = 2,982) . | OR (95% confidence interval) . | Ptrend‡ . | . | |||||||||
rs676387∥ | ||||||||||||||||||
C/C | 909 (52) | 1,529 (51) | 1.00 | 217 (61) | 1,529 (51) | 1.00 | 0.018 | |||||||||||
C/A | 652 (38) | 1,143 (38) | 0.97 (0.85 -1.10) | 111 (31) | 1,143 (38) | 0.67 (0.53-0.86) | ||||||||||||
A/A | 122 (7) | 210 (7) | 0.99 (0.78-1.26) | 0.73 | 18 (5) | 210 (7) | 0.61 (0.37-1.01) | 0.0009 | ||||||||||
S312G (rs605059) | ||||||||||||||||||
A/A | 485 (28) | 819 (27) | 1.00 | 77 (22) | 819 (27) | 1.00 | 0.032 | |||||||||||
A/G | 802 (46) | 1,431 (48) | 0.94 (0.82-1.09) | 171 (48) | 1,431 (48) | 1.28 (0.97-1.71) | ||||||||||||
G/G | 357 (21) | 573 (19) | 1.04 (0.87-1.24) | 0.78 | 91 (26) | 573 (19) | 1.71 (1.24-2.37) | 0.0013 | ||||||||||
rs598126 | ||||||||||||||||||
G/G | 482 (28) | 802 (27) | 1.00 | 78 (22) | 802 (27) | 1.00 | 0.059 | |||||||||||
G/A | 822 (47) | 1,452 (49) | 0.94 (0.82-1.09) | 170 (48) | 1,452 (49) | 1.22 (0.92-1.62) | ||||||||||||
A/A | 372 (21) | 599 (20) | 1.03 (0.86-1.22) | 0.86 | 93 (26) | 599 (20) | 1.63 (1.18-2.25) | 0.0028 | ||||||||||
rs2010750 | ||||||||||||||||||
C/C | 593 (34) | 1,000 (34) | 1.00 | 102 (29) | 1,000 (34) | 1.00 | 0.12 | |||||||||||
C/T | 792 (46) | 1,408 (47) | 0.94 (0.82-1.08) | 170 (48) | 1,408 (47) | 1.18 (0.91-1.53) | ||||||||||||
T/T | 264 (15) | 445 (15) | 1.00 (0.83-1.20) | 0.75 | 67 (19) | 445 (15) | 1.50 (1.07-2.08) | 0.020 | ||||||||||
Haplotype hCGAT | ||||||||||||||||||
None | 633.0 (36) | 1,063.0 (36) | 1.00 | 106.3 (30) | 1,063.0 (36) | 1.00 | 0.063 | |||||||||||
One | 821.2 (47) | 1,449.4 (49) | 0.94 (0.83-1.08) | 177.5 (50) | 1,449.4 (49) | 1.22 (0.94-1.58) | ||||||||||||
Two | 268.8 (15) | 451.6 (15) | 1.00 (0.83-1.20) | 0.77 | 69.2 (20) | 451.6 (15) | 1.57 (1.13-2.17) | 0.0076 | ||||||||||
Haplotype hAAGC | ||||||||||||||||||
None | 936.2 (54) | 1,579.2 (53) | 1.00 | 220.8 (62) | 1,579.2 (53) | 1.00 | 0.024 | |||||||||||
One | 665.4 (38) | 1,169.4 (39) | 0.97 (0.85-1.10) | 113.9 (32) | 1,169.4 (39) | 0.68 (0.53-0.87) | ||||||||||||
Two | 121.4 (7) | 215.4 (7) | 0.96 (0.75-1.22) | 0.60 | 18.3 (5) | 215.4 (7) | 0.60 (0.36-1.00) | 0.0009 | ||||||||||
Haplotype hCAGC | ||||||||||||||||||
None | 917.4 (53) | 1,584.5 (53) | 1.00 | 187.4 (53) | 1,584.5 (53) | 1.00 | 0.91 | |||||||||||
One | 677.6 (39) | 1,171.9 (39) | 1.00 (0.88-1.14) | 142.6 (40) | 1,171.9 (39) | 1.03 (0.81-1.30) | ||||||||||||
Two | 128.0 (7) | 207.6 (7) | 1.06 (0.83-1.35) | 0.75 | 23.0 (7) | 207.6 (7) | 0.93 (0.58-1.49) | 0.95 | ||||||||||
Haplotype hCGAC | ||||||||||||||||||
None | 1,501.1 (86) | 2,616.5 (88) | 1.00 | 306.2 (87) | 2,616.5 (88) | 1.00 | 0.57 | |||||||||||
One | 216.7 (12) | 339.1 (11) | 1.11 (0.92-1.34) | 44.6 (13) | 339.1 (11) | 1.15 (0.81-1.62) | ||||||||||||
Two | 5.2 (<1) | 8.5 (<1) | 1.03 (0.33-3.19) | 0.30 | 2.1 (1) | 8.5 (<1) | 2.15 (0.46-10.09) | 0.29 |
. | ER-positive* . | . | . | . | ER-negative* . | . | . | . | Phet§ . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Cases (%)†, n = 1,737 . | Controls (%)†, n = 2,982 . | OR (95% confidence interval) . | Ptrend‡ . | Cases (%)†, n = 354 . | Controls (%)†, (n = 2,982) . | OR (95% confidence interval) . | Ptrend‡ . | . | |||||||||
rs676387∥ | ||||||||||||||||||
C/C | 909 (52) | 1,529 (51) | 1.00 | 217 (61) | 1,529 (51) | 1.00 | 0.018 | |||||||||||
C/A | 652 (38) | 1,143 (38) | 0.97 (0.85 -1.10) | 111 (31) | 1,143 (38) | 0.67 (0.53-0.86) | ||||||||||||
A/A | 122 (7) | 210 (7) | 0.99 (0.78-1.26) | 0.73 | 18 (5) | 210 (7) | 0.61 (0.37-1.01) | 0.0009 | ||||||||||
S312G (rs605059) | ||||||||||||||||||
A/A | 485 (28) | 819 (27) | 1.00 | 77 (22) | 819 (27) | 1.00 | 0.032 | |||||||||||
A/G | 802 (46) | 1,431 (48) | 0.94 (0.82-1.09) | 171 (48) | 1,431 (48) | 1.28 (0.97-1.71) | ||||||||||||
G/G | 357 (21) | 573 (19) | 1.04 (0.87-1.24) | 0.78 | 91 (26) | 573 (19) | 1.71 (1.24-2.37) | 0.0013 | ||||||||||
rs598126 | ||||||||||||||||||
G/G | 482 (28) | 802 (27) | 1.00 | 78 (22) | 802 (27) | 1.00 | 0.059 | |||||||||||
G/A | 822 (47) | 1,452 (49) | 0.94 (0.82-1.09) | 170 (48) | 1,452 (49) | 1.22 (0.92-1.62) | ||||||||||||
A/A | 372 (21) | 599 (20) | 1.03 (0.86-1.22) | 0.86 | 93 (26) | 599 (20) | 1.63 (1.18-2.25) | 0.0028 | ||||||||||
rs2010750 | ||||||||||||||||||
C/C | 593 (34) | 1,000 (34) | 1.00 | 102 (29) | 1,000 (34) | 1.00 | 0.12 | |||||||||||
C/T | 792 (46) | 1,408 (47) | 0.94 (0.82-1.08) | 170 (48) | 1,408 (47) | 1.18 (0.91-1.53) | ||||||||||||
T/T | 264 (15) | 445 (15) | 1.00 (0.83-1.20) | 0.75 | 67 (19) | 445 (15) | 1.50 (1.07-2.08) | 0.020 | ||||||||||
Haplotype hCGAT | ||||||||||||||||||
None | 633.0 (36) | 1,063.0 (36) | 1.00 | 106.3 (30) | 1,063.0 (36) | 1.00 | 0.063 | |||||||||||
One | 821.2 (47) | 1,449.4 (49) | 0.94 (0.83-1.08) | 177.5 (50) | 1,449.4 (49) | 1.22 (0.94-1.58) | ||||||||||||
Two | 268.8 (15) | 451.6 (15) | 1.00 (0.83-1.20) | 0.77 | 69.2 (20) | 451.6 (15) | 1.57 (1.13-2.17) | 0.0076 | ||||||||||
Haplotype hAAGC | ||||||||||||||||||
None | 936.2 (54) | 1,579.2 (53) | 1.00 | 220.8 (62) | 1,579.2 (53) | 1.00 | 0.024 | |||||||||||
One | 665.4 (38) | 1,169.4 (39) | 0.97 (0.85-1.10) | 113.9 (32) | 1,169.4 (39) | 0.68 (0.53-0.87) | ||||||||||||
Two | 121.4 (7) | 215.4 (7) | 0.96 (0.75-1.22) | 0.60 | 18.3 (5) | 215.4 (7) | 0.60 (0.36-1.00) | 0.0009 | ||||||||||
Haplotype hCAGC | ||||||||||||||||||
None | 917.4 (53) | 1,584.5 (53) | 1.00 | 187.4 (53) | 1,584.5 (53) | 1.00 | 0.91 | |||||||||||
One | 677.6 (39) | 1,171.9 (39) | 1.00 (0.88-1.14) | 142.6 (40) | 1,171.9 (39) | 1.03 (0.81-1.30) | ||||||||||||
Two | 128.0 (7) | 207.6 (7) | 1.06 (0.83-1.35) | 0.75 | 23.0 (7) | 207.6 (7) | 0.93 (0.58-1.49) | 0.95 | ||||||||||
Haplotype hCGAC | ||||||||||||||||||
None | 1,501.1 (86) | 2,616.5 (88) | 1.00 | 306.2 (87) | 2,616.5 (88) | 1.00 | 0.57 | |||||||||||
One | 216.7 (12) | 339.1 (11) | 1.11 (0.92-1.34) | 44.6 (13) | 339.1 (11) | 1.15 (0.81-1.62) | ||||||||||||
Two | 5.2 (<1) | 8.5 (<1) | 1.03 (0.33-3.19) | 0.30 | 2.1 (1) | 8.5 (<1) | 2.15 (0.46-10.09) | 0.29 |
NOTE: Data on receptor status was available from the following cohorts: CPS-II, NHS, WHS, MEC.
Global test of haplotype association with ER-positive breast cancer: X2 = 1.39 with 5 degrees of freedom for P = 0.925; global test of haplotype association with ER-negative breast cancer: X2 = 14.61 with 5 degrees of freedom for P = 0.012.
Percentages do not add to 100% because of missing genotype data; same controls were used for both ER-positive and ER-negative models.
Ps for test of linear trend.
Ps for etiologic heterogeneity by ER status were obtained from case-only models comparing ER-negative cases to ER-positive cases.
htSNPs listed in 5′ to 3′ order.
Association between HSD17B1 haplotype combinations and invasive breast cancer among U.S. Caucasian women by ER status
Observed combination* . | ER-positive† . | . | . | ER-negative† . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
. | Cases (%), n = 1,737 . | Controls (%), n = 2,982 . | OR (95% confidence interval) . | Cases (%), n = 354 . | Controls (%), n = 2,982 . | OR (95% confidence interval) . | ||||
CGAT-CGAT‡ | 270 (16) | 451 (15) | 1.00 | 69 (19) | 451 (15) | 1.00 | ||||
CGAT-AAGC | 344 (20) | 639 (22) | 0.90 (0.74-1.10) | 64 (18) | 639 (22) | 0.64 (0.44-0.92) | ||||
CAGC-AAGC | 257 (15) | 431 (14) | 1.00 (0.81-1.25) | 41 (12) | 431 (14) | 0.59 (0.39-0.90) | ||||
CGAC-AAGC | 59 (3) | 87 (3) | 1.14 (0.79-1.65) | 7 (2) | 87 (3) | 0.50 (0.22-1.14) | ||||
AAGC-AAGC | 118 (7) | 207 (7) | 0.96 (0.73-1.26) | 18 (5) | 207 (7) | 0.56 (0.33-0.97) | ||||
CGAT-CAGC | 373 (21) | 645 (22) | 0.96 (0.79-1.17) | 90 (25) | 645 (22) | 0.89 (0.63-1.25) | ||||
CGAT-CGAC | 97 (6) | >142 (5) | 1.10 (0.82-1.49) | 22 (6) | >142 (5) | 0.96 (0.57-1.62) | ||||
CAGC-CAGC | 124 (7) | 198 (7) | 1.03 (0.79-1.36) | 23 (7) | 198 (7) | 0.75 (0.45-1.24) | ||||
CAGC-CGAC | 50 (3) | 97 (3) | 0.88 (0.60-1.28) | 13 (4) | 97 (3) | 0.95 (0.50-1.80) | ||||
CGAC-CGAC | 5 (<1) | 8 (<1) | 1.00 (0.32-3.12) | 2 (1) | 8 (<1) | 1.55 (0.32-7.53) | ||||
Other combinations | 15 (1) | 28 (1) | 0.87 (0.45-1.67) | 2 (1) | 28 (1) | 0.48 (0.11-2.06) |
Observed combination* . | ER-positive† . | . | . | ER-negative† . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
. | Cases (%), n = 1,737 . | Controls (%), n = 2,982 . | OR (95% confidence interval) . | Cases (%), n = 354 . | Controls (%), n = 2,982 . | OR (95% confidence interval) . | ||||
CGAT-CGAT‡ | 270 (16) | 451 (15) | 1.00 | 69 (19) | 451 (15) | 1.00 | ||||
CGAT-AAGC | 344 (20) | 639 (22) | 0.90 (0.74-1.10) | 64 (18) | 639 (22) | 0.64 (0.44-0.92) | ||||
CAGC-AAGC | 257 (15) | 431 (14) | 1.00 (0.81-1.25) | 41 (12) | 431 (14) | 0.59 (0.39-0.90) | ||||
CGAC-AAGC | 59 (3) | 87 (3) | 1.14 (0.79-1.65) | 7 (2) | 87 (3) | 0.50 (0.22-1.14) | ||||
AAGC-AAGC | 118 (7) | 207 (7) | 0.96 (0.73-1.26) | 18 (5) | 207 (7) | 0.56 (0.33-0.97) | ||||
CGAT-CAGC | 373 (21) | 645 (22) | 0.96 (0.79-1.17) | 90 (25) | 645 (22) | 0.89 (0.63-1.25) | ||||
CGAT-CGAC | 97 (6) | >142 (5) | 1.10 (0.82-1.49) | 22 (6) | >142 (5) | 0.96 (0.57-1.62) | ||||
CAGC-CAGC | 124 (7) | 198 (7) | 1.03 (0.79-1.36) | 23 (7) | 198 (7) | 0.75 (0.45-1.24) | ||||
CAGC-CGAC | 50 (3) | 97 (3) | 0.88 (0.60-1.28) | 13 (4) | 97 (3) | 0.95 (0.50-1.80) | ||||
CGAC-CGAC | 5 (<1) | 8 (<1) | 1.00 (0.32-3.12) | 2 (1) | 8 (<1) | 1.55 (0.32-7.53) | ||||
Other combinations | 15 (1) | 28 (1) | 0.87 (0.45-1.67) | 2 (1) | 28 (1) | 0.48 (0.11-2.06) |
NOTE: Data on receptor status was available from the following cohorts: CPS-II, NHS, WHS, MEC.
Calculated by rounding the estimated haplotype frequencies to whole numbers. Observations with rounded haplotypes that did not total two copies were excluded, as well as people with unknown genotype for three of the four SNPs.
Global test of diplotype association with ER-positive breast cancer: X2 = 4.20 with 10 degrees of freedom for P = 0.937; global test of diplotype association with ER-negative breast cancer: X2 = 15.47 with 10 degrees of freedom for a P = 0.116.
The reference group is made up of observations with two copies of the most common haplotype (CGAT).
Similar but nonsignificant associations were observed among U.S. Caucasian women with PR-negative tumors (Supplementary Table 6S). Statistical testing indicated heterogeneity by ethnicity (Supplementary Table 7S), but no single population group other than Caucasians was of sufficient size to examine the association by ER status.
The false-positive report probability (FPRP) values for the ER-negative tumors for prior probabilities of 0.01, 0.001, and 0.0001 are 0.23, 0.75, and 0.97, respectively, with statistical power near 1.0 for a trend test (26), assuming that carriers of the second most common haplotype had risk of 1.2 for each copy of the haplotype and the risks associated with other haplotypes were all equal to each other. Similarly, the FPRP values for the htSNPs that were significant at 0.05 level in ER-negative tumors were below 0.2 for priors of 0.01 but not for priors of 0.001 or below (data not shown).
Discussion
This is the first study of HSD17B1 and breast cancer that comprehensively characterizes the variation in and around HSD17B1. Despite the large sample size and comprehensive approach for identifying common SNPs and haplotypes, we found no evidence of association overall between breast cancer and variants of HSD17B1 that are common in Caucasians. However, among the subset of U.S. Caucasian cases with ER-negative tumors, we found evidence of an association with each of the four htSNPs, and with the corresponding haplotypes that carry all high-risk alleles (CGAT) or all low-risk alleles (AAGC). Analysis of haplotype combinations showed that every common haplotype combination that included AAGC was associated with a reduced risk of ER-negative breast cancer. However, the FPRP calculations suggest that these associations may be due to chance and thus should be considered preliminary until they can be confirmed in additional studies that have a large number of ER-negative tumors.
Results from previous epidemiologic studies (3–6) have found no overall association with HSD17B1 and breast cancer but have suggested that an association may exist in specific subgroups defined by BMI (3, 5), advanced stage (5), menopausal status (6), or parity (6). We did not find evidence of any association in these subgroups, despite the large size of our study and comprehensive evaluation of the gene.
No previous study has reported an association with HSD17B1 in a subgroup of ER-negative tumors, but it would be difficult to detect in a smaller study because ER-negative tumors typically make up <25% of all breast cancers in Western countries (27, 28). Given our a priori knowledge about the activity of this gene in breast tissue (2, 29), this finding was unexpected. However, etiologic differences between ER-positive and ER-negative tumors is a topic of considerable debate and active research (27, 30). Clinical, epidemiologic, and laboratory data show that ER-positive and ER-negative breast tumors have important differences (31). Epidemiologic studies suggest that risk factors differ by receptor status (30, 32, 33), and ER-positive and ER-negative tumors display different gene expression profiles (34, 35). Our data suggest that germ line variation may also influence ER status.
Further investigation is needed to confirm this association with ER-negative breast cancer, and, if confirmed, isolate the causal variant responsible for this association with HSD17B1-containing haplotypes that define a high linkage disequilibrium block spanning 33 kb, including HSD17B1 and its 5′ and 3′ regions. There is evidence that several upstream regions that lie well within this block participate in the regulation of HSD17B1 expression (8, 29). This region also includes two other genes: NAGLU (5′) and TCFL4 (3′) (Fig. 1). Although there is no a priori reason to suspect that these other genes are associated with breast cancer, they cannot be definitively excluded as possible candidates until further characterization of this region is complete.
The strengths of the BPC3 include its unprecedented sample size and comprehensive characterization of variation around the HSD17B1 locus. Our analysis provides powerful null evidence against a main effect association between the overall risk of breast cancer and variants in HSD17B1 that are common among Caucasians and in subgroups defined by common breast cancer risk factors. The subgroup association that we did observe among ER-negative tumors should be viewed as preliminary and evaluated in future studies.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
H.S. Feigelson, D.G. Cox, H.M. Cann, S. Wacholder, R. Kaaks, and B.E. Henderson were the writing committee for this article.
Acknowledgments
Grant support: National Cancer Institute cooperative agreements UO1-CA98233, UO1-CA98710, UO1-CA98216, and UO1-CA98758 and Intramural Research Program of the NIH, National Cancer Institute, Division of Cancer Epidemiology and Genetics.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank the participants in the component cohort studies and the expert contributions of Hardeep Ranu, Craig Labadie, Lisa Cardinale, Shamika Ketkar (Harvard University), Merideth Yeager, Robert Welch, Cynthia Glaser, Laurie Burdett (National Cancer Institute), Loreall Pooler (University of Southern Califonia), Antonia Trichopoulou, H. Bas Bueno de Mesquita, Heiner Boeing, Domenico Palli, Salvatore Panico, Rosario Tumino, Paolo Vineis, Carlos A. Gonzalez, Carmen Martinez-Garcia, Miren Dorronsoro, and Goran Hallmans (EPIC).