Introduction

Breast and ovarian cancers are a leading cause of morbidity and mortality in women worldwide. The major inherited susceptibilities to these cancers are germline mutations in either BRCA1 or BRCA2 [1]. It has been estimated that BRCA1 mutation carriers have high average lifetime risks of developing breast and ovarian cancers [2]. Notwithstanding, there are substantial differences in the penetrance of disease both within and between families that is best explained by genetic and/or shared environmental factors affecting the risk of breast and ovarian cancers [3].

Both, breast and ovarian cancers are hormonally influenced diseases that point to the importance of mechanisms underlying steroid biosynthesis, metabolism and signaling as well as cell growth. A number of reports have suggested that polymorphisms in steroid receptors and some of their downstream effectors may be associated with an altered risk of disease [4–13]. However, these reports are controversial, since conflicting results have also been published [14–21]. The relationship between aberrant estrogen metabolism and increased levels of DNA damage implicate the various DNA repair pathways that are invoked as a result of an increased level of free-radicals in the target tissue. Excessive levels of estrogen are associated with increased cell division, which requires a large pool of purines and pyrimidines for rapid DNA synthesis. The supply of methionine is crucial for pyrimidine synthesis, as decreased levels are associated with base mis-incorporation, which ultimately results in an increased likelihood of not only genetic errors, but also invocation of DNA repair enzymes.

While a large number of previous studies have reported associations of various polymorphic variants in genes involved in DNA repair, steroid hormone biosynthesis/metabolism/signaling, folate metabolism and cell growth with sporadic breast or ovarian cancer risk [4–33], little data on the potential role of these genetic factors as modifiers of hereditary disease risks have been published [34–42]. Therefore, we investigated whether variants in genes of these four regulatory pathways may act as modifiers of breast and ovarian cancer risk in Polish women carrying one of the three common BRCA1 founder mutations (4153delA, 5328insC and C61G) [43, 44]. Genotyping analysis was performed in 319 breast cancer cases, 145 ovarian cancer cases and 290 controls.

Materials and methods

Study participants

The Pomeranian cancer registry contained clinical and epidemiological data from 1,940 individuals carrying one of the three common Polish BRCA1 founder mutations, 5382insC, 300 T>G and 4153delA. Mutation carriers were selected from families with at least one breast cancer diagnosed before 50 years of age or ovarian cancer at any age or with strong history of breast and/or ovarian cancer. All clinical data have been verified from medical records and were collected from 1997 to 2002. A self-administered questionnaire was used to collect information on potential epidemiological risk factors.

From the registered BRCA1 carriers, 755 female BRCA1 mutation carriers for whom DNA samples were available—319 breast cancer cases, 146 ovarian cancer cases and 290 unaffected controls—were included in this study. All study participants were of Polish origin and resided in Poland. They were considered as breast cancer cases, if they were diagnosed with an invasive primary breast cancer and had not undergone prophylactic mastectomy or adnexectomy prior to the age of breast cancer diagnosis and as ovarian cancer cases, if they were diagnosed with invasive primary ovarian cancer (excluding borderline ovarian carcinoma) and had not undergone prophylactic mastectomy, adnexectomy or tubal ligation prior to the age of ovarian cancer diagnosis. They were considered as controls for breast cancer cases, if they were unaffected by breast cancer, in situ breast carcinoma or any other type of cancer and had not undergone prophylactic mastectomy or adnexectomy and as controls for ovarian cancer cases if they were unaffected by ovarian cancer or any other type of cancer and had not undergone prophylactic mastectomy, adnexectomy or tubal ligation.

The number of subjects in the various subgroups, the median ages of breast and ovarian cancer cases at the time of diagnosis and the median ages of controls at the time of interview are shown in Table 1. The research was approved by the Ethics Committee of Pomeranian Medical University in Szczecin, Poland and all participants gave informed consent prior to enrolling in the study. All women received genetic counseling prior to and at the provision of their test results.

Table 1 Study groups, median ages of cases at diagnosis of breast and ovarian cancers and of controls at the time of interview
Full size table

Genetic analyses

Twenty-nine polymorphisms in 20 genes involved in DNA repair, steroid hormone biosynthesis/metabolism/signaling, folate metabolism and cell growth were genotyped. The genes, variants under investigation and genotyping methods are given in Supplementary Tables 1 and 2.

Genomic DNA was isolated from peripheral blood leukocytes according to Lahiri and Schnabel [45]. Seventeen restriction fragment lengths polymorphisms (RFLPs), six insertion/deletion polymorphisms and six repeat polymorphisms were genotyped by PCR-based fragment analyses using the CEQ™ 8000 Genetic DNA Analysis System (Beckman Coulter, Krefeld, Germany) and/or ethidium bromide-stained agarose gels (Sigma–Aldrich, Steinheim, Germany), which were scored by UV visualization. Primer sequences, PCR and genotyping conditions are presented in Supplementary Table 2. CYP17A1 -34T>C, CYP19A1 15210(TTTA)n, CYP19A1 15262delCTT and COMT 1222G>A were genotyped as described earlier [46].

Genotyping was performed by laboratory personnel blinded to the case–control status. The reproducibility of the genotyping data was assessed by repeated analysis of 5% of randomly selected DNA samples. Genotyping accuracy was assessed by sequencing of several samples with detected polymorphic variants. Call rates for individual polymorphisms were greater than 96% and concordance of duplicates was 100%.

Statistical analysis

Risk estimates were calculated as odds ratios (OR) with 95% confidence intervals (CI) using both univariate and multivariate unconditional logistic regression. Crude ORs (ORcrude) were calculated for 319 breast cancer cases and 290 controls and 146 ovarian cancer cases and 280 controls. For 232 breast cancer cases and 225 controls and 85 ovarian cancer cases and 215 controls, we adjusted for potential breast/ovarian cancer risk factors, referred to as adjusted ORs (ORadj), by including age at first live birth (0, ≤24, >24 years), parity, lifetime cumulative months of breastfeeding (≤12, >12 months), age at menarche, oral contraceptive (OC) use (<5, ≥5 years), smoking (0, <4, ≥4 pack-years), body mass index (BMI) (at age of breast cancer diagnosis for cases and at time of interview for controls), year of birth and BRCA1 mutation in the multivariate logistic regression model.

The statistical analyses were performed using the SAS/STAT(r) software (Version 9.1 of the SAS System for Windows, Cary, NC, USA, SAS Institute Inc. 2003), with the procedure LOGISTIC.

Repeat polymorphisms were treated as a two- or three-level categorized variable defined by a certain cutpoint. The cutpoints used were described earlier and provided the categories <28/<28+<28/≥28 and ≥28/≥28 for AIB1 28824(CAG/CAA)n [34–37], ≥29/≥29+≥29/<29 and <29/<29 for AR 171(CAG)n [38, 39], <10/<10, <10/≥10 and ≥10/≥10 for CYP19A1 15210(TTTA)n [9], 3/3, 2/2 and 2/3+2/4 for TYMS -97(28)n [11] and 6/6+6/9 and 9/9+9/10 for TGFBR1 52(GCG)n [32]. For XRCC3 204(GT)n, cutpoint 14 was selected using maximal selected rank statistics [47], yielding the categories ≤14/≤14, ≤14/>14 and >14/>14. For CYP11A1 (AAAAT)n, the categories were 4/4, 4/6, 6/6, 8/any and others [26]. Allele frequencies of repeat polymorphisms between cases and controls were compared using Pearson’s chi square test. For each SNP, statistical modeling was performed using unconditional logistic regression to assess the predicted power of the study. Bonferroni correction was used to account for multiple testing and a two-tailed P value < 0.0017 was considered statistically significant.

Results

The distribution of selected cancer risk factors including year of birth, age at menarche, age at first live birth, parity, duration of breastfeeding, smoking, BMI and BRCA1 mutation was similar between breast and ovarian cancer cases and their corresponding controls (Table 2).

Table 2 Characteristics of breast and ovarian cancer cases and their corresponding controls
Full size table

In total, 755 study participants were genotyped at 29 polymorphisms in 20 genes, which included eleven variants in five DNA repair genes (ERCC2, XRCC1, XRCC2, XRCC3 and Lig4) and the cell cycle check point control gene (TP53), five variants in five growth factor and receptor genes (HER2, IL6, LRP1, TGFB1 and TGFBR1), two variants in the thymidylate synthetase gene (TYMS) and eleven variants in eight genes associated with steroid hormone biosynthesis, metabolism and signaling (AIB1, AR, COMT, CYP11A1, CYP17A1, CYP19A1, HSD17B1 and PGR).

The distribution of the genotype frequencies among breast and ovarian cancer cases and their respective controls is shown in Table 3. Three polymorphisms, XRCC3 204(GT)n, COMT 1222G>A and CYP11A1 (AAAAT)n appeared to be related to disease expression in ovarian cancer (ORcrude 2.24, 95% CI 1.02–4.92 ORcrude 1.85, 95% CI 1.01–3.48 and ORcrude 0.54, 95% CI 0.31–0.95, respectively). Two other polymorphisms, TP53 441G>C and TGFB1 -1349C>T, were associated with an increased breast cancer risk after adjustment for potential risk factors (ORadj 1.56, 95% CI 1.04–2.34 and ORadj 1.91, 95% CI 1.01–3.60, respectively). However, upon correction for multiple testing the significance of these effects vanished.

Table 3 Genotype frequencies in the breast and ovarian cancer cases and their corresponding controls
Full size table

There were no significant differences in the allele frequencies of the AIB1 28824(CAG/CAA)n, AR 171(CAG)n, CYP11A1 (AAAAT)n, CYP19A1 15210(TTTA)n, TYMS -97(28)n, TGFBR1 52(GCG)n and XRCC3 204(GT)n repeat polymorphisms between breast and ovarian cancer cases and their corresponding controls (Supplementary Table 3).

Discussion

We performed association studies at 29 polymorphic loci in 20 genes encoding proteins involved in DNA repair, folate metabolism, steroid hormone biosynthesis/metabolism/signaling and cell growth to identify genetic variants that may modify BRCA1-associated breast and ovarian cancer risk. Numerous studies investigated polymorphisms in sporadic breast and ovarian cancer populations often yielding controversial results [4–31]; however, little is known about the effects of these polymorphisms on disease risk in women with an inherited predisposition.

A few of the polymorphisms investigated in this study have been previously analyzed for their potential risk modifying effect in BRCA1/2 mutation carriers. An association of long AIB1 28824(CAG/CAA)n alleles (≥28 and ≥29 repeats) with an increased breast cancer risk was reported in two studies among BRCA1/2 mutation carriers [34, 35], which, however, was not confirmed in two other studies [36, 37] or in the present study. Similarly, conflicting results were published for AR 171(CAG)n. One study reported an association of long AR alleles (≥28, ≥29 and ≥30 repeats) with an increased breast cancer risk among BRCA1/2 mutation carriers [38], whereas in three other studies, including the present study, this observation was not confirmed [39, 40]. The inconsistent results may be explained by differences between these studies including size, inclusion of BRCA1, BRCA2 or both BRCA1 and BRCA2 carriers, type of BRCA1/2 mutations or ethnicity. The latter explanation is supported by data from Kadouri et al. of significantly shorter allele lengths, on average, in 256 BRCA1/2 mutation carriers of Ashkenazi origin compared with 55 British non-Ashkenazi carriers (P trend 0.0007) [35]. Similar findings were observed for AR 171 (CAG)n, where median CAG repeat length was significantly longer in the Ashkenazi carriers compared with non-Ashkenazi carriers (23 and 21, respectively, two-sided P = 0.01) and at least one allele with repeat length ≥ 28 was found in 17/154 (11%) of the Ashkenazi carriers, but in none of the 30 non-Ashkenazi carriers [39]. These observations suggest that differences between studied populations are an important factor in association studies on modifiers of BRCA1/2-associated breast or ovarian cancer risk.

In this study, TP53 441G>C and TGFB1 -1349C>T were associated with an increased breast cancer risk in analyses adjusted for potential risk factors. However, these effects were not stable and vanished upon correction for multiple testing. We also found no association of the HER2 23097A>G polymorphism with BRCA1-associated breast cancer risk. This contrasts with findings from another study among Ashkenazi carriers of the three common BRCA1/2 founder mutations, which reported an association of this SNP with an increased breast cancer risk. This study however was based on a small number of 120 BRCA1/2 carriers [41].

Three polymorphisms, XRCC3 204(GT)n, COMT 1222G>A and CYP11A1 (AAAAT)n, were associated with ovarian cancer risk in the present study. These results however were also not stable and vanished once corrected for multiple testing. We also observed no association of the TP53 237ins16 and 1798G>A polymorphisms with disease risk. The findings of the TP53 237ins16 are in line with those reported in a previous study [42].

In this study, patients affected by breast or ovarian cancer as well as healthy individuals were of Polish origin and all carried one of the three Polish BRCA1 founder mutations. Because of the strong founder effect and the Polish population being relatively stable and ethnically homogeneous, it is ideal for association studies of risk modifying genes not influenced by BRCA1 allelic or ethnic variation.

Based on statistical modeling, our study on 319 breast cancer cases, 146 ovarian cancer cases and 290 unaffected controls was sufficiently large to identify any potential association. For the breast cancer population, all comparisons of the reference genotype with the two genotypes had a power of at least 90% to detect an OR of at least 2.0 at the significance level of 5%, except for TGFBR1 52(GCG)n and AR 171(CAG)n. Those had a power larger than 80 and 60%, respectively, when compared with the heterozygous genotype, and larger than 80% when compared with the homozygous variant genotype, but then only for an OR of 2.5. For the ovarian cancer population, all comparisons of the reference genotype with the two genotypes had a power of at least 80% to detect an OR of at least 2 (in about half of the polymorphisms larger than 90%) at the significance level of 5%, except for AR 171(CAG)n, HSD17B1 1995A>G, XRCC2 27158G>A, AIB1 28824(CAG/CAA)n, TGFBR1 52(GCG)n and PGR -413G>A (power of 20, 65, 66, 74, 78 and 78%, respectively) when comparing with the heterozygous genotype, and except for TP53 1798G>A, CYP11A1 (AAAAT)n, TP53 441G>C and CYP19A1 15162delCTT (power of 44, 66, 78 and 78%, respectively) when comparing with the homozygous variant genotype.

Our study has some limitations. A general selection bias may have occurred due to the use of study participants from a registry. Furthermore, due to the inclusion of incident and prevalent cases in the Polish registry, the presence of prevalent cases among our study participants may have led to a survival-bias, which is a general limitation of this type of retrospective study. However, because the frequency of BRCA1 mutation carriers is low, in the Polish population nearly 0.5% [48], it makes a recruitment of study participants from the population unrealistic. Therefore, the use of registry participants seems to be the only acceptable approach to perform this kind of study.

In summary, no association of the 29 polymorphisms in genes involved in DNA repair, folate metabolism, steroid hormone biosynthesis/metabolism/signaling and cell growth with BRCA1-associated breast and ovarian cancer risks was observed. These findings indicate that these genetic variants are likely to play no or a minor role as modifiers of breast and ovarian cancer risks in Polish women carrying one of the three common BRCA1 founder mutations.