Introduction

Thus far, two highly penetrant breast and ovarian cancer susceptibility genes, BRCA1 and BRCA2, have been identified. Germline mutations in these genes are found in the majority of families (75–80%) with both early-onset breast cancer and ovarian cancer, but in a considerably lower fraction of families with site-specific female breast cancer (no ovarian cancer or male breast cancer).1, 2 Recently, the CHEK2 1100delC and Ile157Thr have been identified as new low-risk breast cancer susceptibility alleles.3, 4, 5 It has been suggested that a substantial proportion of the remaining familial aggregation of breast cancer might be attributed to several such low penetrance alleles that act multiplicatively, and thereby form a polygenic basis for breast cancer predisposition.6, 7

One potential candidate gene for breast cancer susceptibility is the BRCA1-associated RING-domain 1, BARD1 (GenBank U76638). The BARD1 protein was originally identified through a yeast two-hybrid screen using a BRCA1 RING domain as bait.8 The two proteins are structurally related: they both have an amino-terminal RING-domain and one (BRCA1) or two (BARD1) carboxy-terminal BRCT domains. The RING domain mediates protein–protein interactions and may also participate in ubiquitination of the proteins, whereas the BRCT domain is found in a large number of proteins that predominantly participate in cell cycle checkpoint or DNA-repair pathways in response to DNA damage. These domains are evolutionarily well conserved, whereas the remaining of the sequences in both genes, with the exception of the ankyrin repeats of BARD1 that also participate in protein–protein interactions, display a markedly lower degree of conservation.

The interaction between BRCA1 and BARD1 is mediated via their RING-finger motifs.8 The interaction is abolished by tumorigenic missense mutations in the RING domain of BRCA1, implying that the formation of a stable complex between these proteins may be an essential aspect of BRCA1-mediated tumor suppression.8 A germline BARD1 mutation has also been reported in a patient diagnosed with breast, ovarian, and endometrial cancer, and somatic BARD1 mutations have been observed in breast and endometrial tumors; this suggests that BARD1 mutations might themselves be contributing factors in the development of some human carcinomas.9 This is further supported by the loss of the wild-type alleles both in the tumor tissue of the aforementioned patient with a germline mutation, as well as in the endometrial tumor with a somatic mutation.9 Functional analysis of the protein with a tumor-associated BARD1 germline mutation, Gln564His, has also shown that it is defective in p53-mediated apoptosis induction and results in reduced binding to polyadenylation factor CstF-50.10, 11 Reduced BARD1 expression has also been observed in some sporadic breast tumors.12

As BARD1 is a plausible breast cancer susceptibility gene, we aimed to evaluate whether there are BARD1 genetic variants that contribute to breast cancer risk by screening the BARD1 gene for germline alterations among 45 Finnish BRCA1/2 negative breast cancer families as well as in seven patients diagnosed with both breast and ovarian cancer. During the course of this study, two of the identified BARD1 variants, Cys557Ser and Val507Met, were suggested to associate with breast cancer risk. Cys557Ser was suggested to increase familial breast cancer risk by four-fold,13 and the Val507Met was associated with a two-fold elevation in postmenopausal breast cancer risk.14 We evaluated these variants in extensive series of unselected and familial breast cancer patients, and healthy controls. We also investigated the possible association of the variants with breast tumor characteristics, age at cancer diagnosis, and survival.

Patients and methods

In the initial mutation analysis, the whole BARD1 coding region was screened in 45 families who have at least three first- or second-degree relatives, including the proband, affected with breast or ovarian cancer. The recruitment of the families, verification of the cancer diagnoses, and exclusion of BRCA1 and BRCA2 mutations have previously been described.4, 15, 16 As a functionally relevant BARD1 germline mutation was found in a patient with three primary tumors (breast, ovarian, and endometrial9), seven more patients with both breast and ovarian cancer were included in the analysis.

The frequency of the BARD1 Cys557Ser was evaluated in two large, independent series of breast cancer patients and healthy population controls originating from Southern and Central Finland. The first series included 888 unselected (described in detail in Syrjakoski et al17 and Kilpivaara et al18) and 946 familial16 breast cancer patients collected at the Helsinki University Central Hospital in Southern Finland. BRCA1 and BRCA2 mutations have been excluded in all the 403 patients with a strong family history of cancer (three or more first- or second-degree relatives with breast or ovarian cancer in the family, including the proband) as well as in 291 cases with one affected relative as previously described2, 15, 17 (also unpublished data). DNA from altogether 736 healthy female subjects collected at the same geographical region of Southern Finland was studied as healthy population controls. The second series where the frequency of the BARD1 Cys557Ser alteration was analyzed consists of 697 unselected breast cancer patients collected in Central Finland at Tampere University Hospital, Tampere, Finland17 (also unpublished data). Of these, 442 patients have no family members affected with breast or ovarian cancer in addition to the proband, 183 have one, and 72 have two or more. BRCA1 and BRCA2 mutations have been excluded in 400 of the patients.17 Also, 358 healthy population controls from the same geographical region were genotyped for the BARD1 Cys557Ser alteration.

The frequency of the BARD1 variant Val507Met was studied in 888 unselected and in 710 familial breast cancer patients as well as in 736 population controls from the Helsinki series. As the variant has been associated with an elevation in postmenopausal breast cancer risk, the patients as well as the population controls were subgrouped according to their menopausal status (age 50 years was chosen as a surrogate for menopause: patients diagnosed <50 years were considered as premenopausal, and patients diagnosed ≥50 years as postmenopausal).

Information on tumor characteristics (histology, grade, and size of the tumor; nodal status and distant metastasis; estrogen and progesterone receptor status) and age at breast cancer diagnosis was available from 1296 unselected breast cancer patients (altogether 1340 tumors).17, 18 Written informed consent was obtained from each participating patient. This study has been performed with appropriate permissions from The Ministry of Social Affairs and Health in Finland as well as The Ethics Committees of the Departments of Obstetrics and Gynecology, and Oncology, in Helsinki University Central Hospital, Helsinki, and in the Tampere University Hospital, Tampere, Finland.

Whole-gene mutation analysis and genotyping

The mutation screening of the whole BARD1 coding region and exon–intron boundaries was carried out by conformation-sensitive gel electrophoresis (CSGE). The detection was based either on autoradiography, where one of the PCR primers was labeled with [γ-33P]dATP (Amersham Biosciences, Uppsala, Sweden), or by staining the CSGE gels with silver nitrate (MERCK, Darmstadt, Germany). Primer sequences and PCR conditions used in CSGE analysis are available upon request. Samples with a band shift were reamplified, and the nucleotide changes were verified by direct sequencing using BigDye Terminator v3.1 Cycle Sequencing Kit and ABI 310 Sequencer (Applied Biosystems, Foster City, CA, USA).

The genotyping of DNA samples for the nt 1743 G → C (Cys557Ser) and nt 1592 G → A (Val507Met) variants in the Helsinki patient series and population controls was performed using Amplifluor™ fluorescent genotyping (K-Biosciences, Cambridge, UK). The genotyping for codon 557 was successful in 868/888 (98%), 926/946 (98%), and 725/736 (99%) of the unselected and familial breast cancer patient and control samples, respectively, and 867/888 (98%), 663/710 (93%), and 718/736 (98%) of the studied samples were successfully analyzed for codon 507, respectively. The genotypes of Tampere patients and healthy controls were determined by minisequencing.19

Statistical analyses

Associations between the BARD1 genetic alterations and breast cancer risk, as well as the tumor characteristics of the breast tumors, were tested by the χ2 test. The statistical analyses were carried out in the SPSS software (version 12.0 for Windows, SPSS, Chicago, IL, USA). All P-values were two-sided, and P-value <0.01 was considered statistically significant as suggested for the association studies for potential low-penetrance alleles.20

Results and discussion

Screening of the BARD1 gene for germline variants revealed seven coding sequence alterations (Table 1). The rare variants 1145del21 and Arg658Cys found in one family each (1.9%) were also detected in 0.8% (8/974) and 1.9% (18/975) of healthy population controls, respectively (data not shown). The heterozygote carrier frequency of Arg378Ser was also similar in familial breast cancer patients (40.4%, 21/52) and healthy population controls (39.8%, 97/244). These, as well as the silent variants Thr351Thr and His506His, have been described previously, and classified as polymorphisms with likely having no effect on cancer risk.9, 13, 14 Comprehensive analyses were carried out on the identified variants Cys557Ser and Val507Met that were recently suggested to associate with an increased breast cancer risk. These are described in more detail in the following.

Table 1 Observed coding region variants of the BARD1 gene
Full size table

The BARD1 cysteine 557 is a conserved amino acid that resides between the ankyrin repeats and the BRCT domain of a protein. This region has been shown to be essential for apoptosis induction and the interaction of BARD1 with a proto-oncogene Bcl-3.21, 22 The Cys557Ser alteration was recently suggested as a familial breast cancer allele in breast cancer families from Northern Finland.13 We found the variant here in one of the familial patients screened in the initial analysis, and further studied it in two independent series of breast cancer patients and population controls. In the first study series, the Cys557Ser variant was observed with a low frequency both among unselected (2.0%, 17/868) and familial (1.3%, 12/926) breast cancer patients. It was also present among healthy population controls with a frequency of 2.5% (19/725) (Table 2). In the second independent series, the results were similar, with a frequencies of 2.4% (13/442) among unselected and 1.6% (4/255) among familial breast cancer patients, and 2.2% (8/358) among healthy controls (Table 2). Breast cancer patients were further divided into subgroups based on the number of affected relatives and whether or not there were ovarian cancer patients in the family (Table 2). No evidence for elevated breast cancer risk of the BARD1 Cys557Ser variant was observed in any of the study cohorts or subgroups; indeed, the variant frequency was slightly lower among the familial patient series than among population controls. No association between the variant and any of the tumor characteristics (tumor histology, grade, and size; nodal status and distant metastasis; estrogen and progesterone receptor status) was seen either (data not shown), and the mean age at breast cancer diagnosis was similar between the carriers and noncarriers of the variant (57.3 and 58.2 years, respectively (P=0.723)).

Table 2 BARD1 Cys557Ser frequencies among two independent breast cancer study cohorts and healthy controls
Full size table

Previously, the Cys557Ser variant has been observed in two sporadic Caucasian breast/uterine/ovarian tumors from the US population,9 and in one Italian breast/ovarian cancer family.23 In the former study, the variant was classified as a polymorphism and in the latter, possibly cancer related. Association analysis to evaluate breast cancer risk was not possible in either study. In another recent study, the variant was found in seven out of 126 (5.6%) breast cancer families from Northern Finland and in 14 out of 1018 (1.4%) healthy population controls, suggesting that the carriers would have a four-fold increased risk for familial breast cancer.13 The prevalence was highest among patients whose family history included only breast and not ovarian cancer (7/94, 7.4% vs 14/1018, 1.4%; OR 5.8; 95% CI 2.3–14.7, P=0.001). Our results with two extensive and independent series of breast cancer patients do not support association of Cys557Ser with breast cancer risk either among familial or unselected breast cancer patients. We found the variant altogether in 1.4% of 1181 familial breast cancer patients, in 2.2% of 1565 unselected breast cancer patients, and in 2.5% of 1083 healthy controls, respectively, and there was no difference in the frequencies by different family history. The discrepancy between the present study and the study by Karppinen et al13 (2004) could be due to the small sample size in the latter, which may have resulted in a positive association by chance. Differences in the population stratification of the breast cancer families and population controls used could also be a possible cause for a chance association. However, the possibility that the Cys557Ser variant interacts with another, unknown risk allele in the breast cancer families from Northern Finland, but not in Central or Southern Finland, cannot be excluded.

The Met-allele of the common BARD1 missense substitution Val507Met has been suggested to associate with an increased breast cancer risk in postmenopausal women in a study including 73 pre- and 70 postmenopausal breast cancer patients and 152 healthy controls (adjusted OR 2.05; 95% CI 1.01–4.16).14 The residue resides within the conserved ankyrin repeat domain, which is responsible for the protein–protein interactions. The particular codon is, however, not fully conserved.

The BARD1 Val507Met was found to be very common also in our study population. The allele frequencies for G (coding for Val) and A (coding for Met) were 0.54 and 0.46 among the unselected breast cancer patients, 0.51 and 0.49 among the familial breast cancer patients, and 0.50 and 0.50 among the healthy population controls, respectively. The Val-allele (either in homo- or heterozygous form) was slightly more common among unselected breast cancer patients than among healthy controls (OR 1.29; 95% CI 1.02–1.63; P=0.037; Table 3). The difference in frequency was highest for the Val-homozygotes, although the statistical significance was not reached (Table 2). When the patients were subgrouped according to their menopausal status, the higher Val-allele carrier frequency was found in the pre- (OR 1.47; 95% CI 1.018–2.12; P=0.039) but not in the postmenopausal women (OR 1.17; 95% CI 0.82–1.66; P=0.392), when compared to Met-homozygotes. Similar to the whole cohort, the difference in frequency was highest for the Val-homozygotes (OR 1.61; 95% CI 1.05–2.47; P=0.028). Results from the three-value trend test suggested a slightly elevated cancer risk for the Val-allele carriers in the premonopausal patient group but not in the whole cohort of unselected patients (data not shown). The age at breast cancer diagnosis did not differ between the different genotype carriers (55.8 years for the Met-homozygotes, 56.2 years for the heterozygotes, and 55.9 years for the Val-homozygotes). No association between the Val507Met and any of the tumor characteristics or cumulative 5-year survival among the unselected patients was observed (data not shown). Also, no increase in breast cancer risk was observed in familial breast cancer patients. Altogether, our results suggest that this variant does not have an effect on familial breast cancer susceptibility, whereas some evidence for a modest elevation in breast cancer risk among unselected breast cancer patients, especially among those below 50 years, was obtained.

Table 3 Genotypes of the BARD1 Val507Met in 867 unselected and 663 familial breast cancer patients, and in 718 healthy controls
Full size table

In summary, rarity or absence of clearly pathologic BARD1 mutations in this study and previous studies suggests that BARD1 is not a major gene underlying familial breast cancer. Our results also indicate that neither the BARD1 Cys557Ser nor Val507Met variants predispose to familial breast cancer.