Approximately 5–10% of all breast and ovarian cancers are thought to arise from a hereditary predisposition to the disease,¹BRCA1 and BRCA2 being the most important susceptibility genes.^2,3 Genomic alterations in BRCA1 are found in 40–50% of families with a high incidence of breast cancer (six or more cases), and in a majority (75–80%) of the families that display both breast and ovarian cancers.^4,5 However, a significant portion of genetic aberrations predisposing to these cancers, especially in relatively small risk families, still remains unexplained.^6,7 BRCA1 interacts with a variety of proteins and is involved in multiple cellular processes including DNA repair, transcription, and checkpoint control.^8–10 In attempts to identify new breast and ovarian cancer susceptibility genes, much research has focused on BRCA1 associated proteins.

BARD1 was originally identified through its interaction with BRCA1, with which it has a closely related domain structure.¹¹ Both proteins possess an N-terminal RING finger motif and two BRCA1 C-terminal (BRCT) domains present in numerous proteins involved in DNA repair and cell cycle regulation.¹¹ The functionally important BARD1/BRCA1 heterodimer formation is mediated by the RING finger motifs and has also been shown to markedly increase the stability of both proteins.^11–13 The finding of breast cancer associated mutations within the RING finger domain of BRCA1, disrupting BRCA1/BARD1 interaction,^11,14 and the occurrence of BARD1 missense mutations in breast cancer patients,^15–17 implies participation of BARD1 in BRCA1 mediated tumour suppression. BARD1, unlike BRCA1, also contains a centrally located sequence of three ankyrin repeats¹¹ that are found in many proteins involved in transcriptional regulation.¹⁸

Colocalisation of BARD1 with BRCA1 and RAD51 in response to DNA damage indicates a role in DNA repair,^19,20 which is supported by the recent observation of BARD1 participation along with BRCA1 in homology directed repair of chromosome breaks.²¹ Furthermore, interaction between the BARD1/BRCA1 heterodimer and cleavage stimulation factor subunit 1 (CSTF1, also called CstF-50) represses the polyadenylation machinery, presumably to prevent inappropriate mRNA processing at sites of DNA repair.^22,23 The significance of BARD1/BRCA1 collaboration has also been emphasised by studies of its ubiquitin ligase activity that might contribute to tumour suppression and other biological functions of BRCA1.²⁴ In vivo substrates for the ubiquitination are not yet known, but involvement of the RNA polymerase-2 holoenzyme has been proposed.²⁵ BARD1 also regulates the subcellular localisation of BRCA1, both by translocating BRCA1 into the nucleus and by inhibiting its nuclear export.²⁶ The suggested role in TP53 dependent apoptotic signalling²⁷ and interaction with the ankyrin repeats of proto-oncoprotein BCL3, thereby possibly modulating the activity of transcription factor NFKB,²⁸ represent BRCA1 independent functions of BARD1. In addition, the effects of reduced Bard1 expression have been studied in murine mammary epithelial cell cultures, where it was associated with complex cellular changes suggestive of a premalignant phenotype.²⁹ Moreover, a recent study in mice showed that loss of Bard1 results in early embryonic lethality and chromosomal instability, indicating a role of Bard1 in maintenance of genomic integrity. The phenotype of Bard1 null mice was found to be remarkably similar to that of Brca1 nulls, further emphasising the functional relationship between these two proteins.³⁰

Based on these facts, BARD1 seems a plausible target for cancer inducing germline mutations. To explore this possibility, index cases of 126 Finnish breast and/or ovarian cancer families were analysed for mutations located in the protein encoding regions and intron–exon boundaries. Our results suggest a possible role for BARD1 in hereditary susceptibility to breast cancer.

METHODS

The index cases of 126 breast and/or ovarian cancer families originating from northern Finland were selected for the screening of BARD1 germline mutations. Of the studied families, 94 were associated with breast, 29 with breast-ovarian and three with ovarian cancer. All index cases had been diagnosed with either breast or ovarian cancer. Inclusion criteria for the 75 high risk families were the following: (1) three or more cases of breast and/or ovarian cancer in first or second degree relatives, or (2) two cases of breast and/or ovarian cancer in first or second degree relatives, of which at least one had early disease onset (⩽35 years), bilateral disease, or multiple primary tumours. Most of the high risk families displayed three or more cancer cases. The remaining 51 families with moderate disease susceptibility displayed two cases of breast and/or ovarian cancer in first or second degree relatives. All of the high risk families had previously been screened for germline mutations in BRCA1, BRCA2, CHK2, and TP53,^31–33 and ten families showed alterations in BRCA1 or BRCA2. The occurrence of the BARD1 Cys557Ser alteration was also tested in 188 breast cancer patients without known family history of the disease, and DNA samples from 1018 anonymous cancer free blood donors served as controls. Both reference groups originated from the same geographical region as the studied cancer families. All patients had given their informed consent for obtaining pedigree data and blood specimens for a study on cancer susceptibility gene mutations. Approval to perform the study was obtained from the Ethical Board of the Northern Ostrobotnia Health Care District and the Finnish Ministry of Social Affairs and Health.

DNA was extracted from blood lymphocytes using either the standard phenol-chloroform protocol or the Puregene D-50K purification kit (Gentra, Minneapolis, MN, USA). The screening of protein encoding and exon–intron boundary regions of BARD1 was done by conformation sensitive gel electrophoresis (CSGE),³⁴ and samples with a band shift were reamplified and purified with the QIAquick PCR Purification Kit (Qiagen, Hilden, Germany). Sequencing analysis was performed using the Li-Cor IR² 4200-S DNA Analysis system (Li-Cor, Lincoln, NS, USA) and the SequiTherm EXCEL II DNA Sequencing Kit-LC (Epicentre Technologies, Madison, WI, USA). Oligonucleotides for CSGE and sequencing were designed by using Primer3 software, utilising sequence information obtained from public databases. Oligonucleotide sequences and PCR conditions for CSGE and sequencing are available upon request.

The observed differences in mutation frequencies between the hereditary or sporadic group of breast cancer patients and control samples were analysed by Fisher’s exact test. Mann-Whitney’s U test was used to compare mean ages of disease onset between alteration carriers and non-carriers. All p values were two sided. Statistical analyses were carried out with SPSS version 12.0 for Windows (SPSS, Chicago, IL, USA).

RESULTS AND DISCUSSION

The screening of index cases of breast and/or ovarian cancer families for BARD1 germline mutations revealed four missense and three synonymous changes within the protein encoding region (table 1). Of these, the previously observed Cys557Ser missense alteration occurred in multiple families and appeared to be associated with cancer predisposition. The other six coding region variants were considered neutral polymorphisms, as most of them also displayed similar allele frequencies in controls, occurred in the SNP database, or have been reported in other studies,^15,17 and none of them appeared to have a significant effect on protein function. In addition, a large number of apparently harmless intronic changes were found (data not shown).

Cys557Ser was the only possible disease related alteration in BARD1 observed in the present study. Of the index cases from the studied breast and/or ovarian cancer families, 5.6% (7/126) were found heterozygous for the Cys557Ser allele. The same alteration was also seen in 1.4% of the cancer free controls (n = 1018), but at significantly lower incidence (p = 0.005; OR 4.2; 95% CI 1.7 to 10.7). Interestingly, in the hereditary material all seven index cases with Cys557Ser belong to a subgroup of 94 families displaying breast but not ovarian cancer, giving a prevalence of 7.4% (p = 0.001; OR 5.8; 95% CI 2.3 to 14.7). All index patients exhibiting the Cys557Ser allele were negative for BRCA1 and BRCA2 germline mutations. Cys557Ser was also observed in five of 188 (2.7%) breast cancer cases without known family history of the disease, the frequency being slightly, but not significantly, higher than among control subjects (p = 0.2; OR = 2.0; 95% CI 0.7 to 5.5). Altogether, the obtained results indicate that Cys557Ser may be a low penetrance breast cancer susceptibility allele. Interestingly, BARD1 maps to 2q34-q35, which is relative close to the 2q32 chromosome region indicated by a genome-wide linkage analysis to segregate with breast cancer predisposition in some Finnish breast cancer families.³⁵

To further evaluate the association between Cys557Ser and cancer susceptibility, the mutation status of other available affected or unaffected family members was analysed. In the pedigree shown in fig 1, the Cys557Ser allele was seen in a woman with breast cancer at age 75, and in her two daughters diagnosed with breast cancer at ages 49 and 53. The third daughter did not show the missense alteration and was cancer free at age 63. In addition, the sister of the proband developed uterine cancer at the age of 40, but no DNA was available for the testing of mutation status. Unfortunately, in the remaining six families displaying the Cys557Ser alteration, only a small number of additional DNA samples were available for analysis: one individual with breast cancer at age 80, another with bilateral breast cancer at ages 40 and 53, and two individuals with uterine cancer did not carry the alteration. Thus, despite of a history of breast, and in some members of the families, other cancers as well, the segregation of Cys557Ser with the disease could not be conclusively determined. In the studied families, the mean age at breast cancer diagnosis in Cys557Ser carriers was 56.8 years (ranging from 42 to 75 years), and thus somewhat higher compared to 51.9 years in index cases without BARD1 alteration, and 48.2 years in patients belonging to BRCA1 and BRCA2 families. However, the difference between ages of onset was not statistically significant.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

 Pedigree of a Finnish cancer family showing the occurrence of the BARD1 Cys557Ser allele. Filled symbols indicate individuals with cancer, and the age at diagnosis is marked after the malignancy. For the disease free individual homozygous for the wildtype allele, the age at monitoring is shown. The arrow shows the index case.

In addition to the current study, involvement of BARD1 alterations in breast and ovarian cancer development has been suggested in three previous studies, two of which also reported the Cys557Ser variant.^15–17 Cys557Ser was initially seen in sporadic Caucasian breast/ovarian/uterine tumour material with a frequency of 2%, and was classified as a rare polymorphism.¹⁵ Later, in an Italian study of 40 breast and/or ovarian cancer families, the alteration was observed in four women belonging to the same family.¹⁶ Three of the women had breast or breast/ovarian cancer and the fourth was unaffected by the age of 39. As Cys557Ser was not present among any of the studied 60 controls or 20 sporadic cases, it was suggested to be a possible cancer susceptibility allele. The prevalence of 5.6% for BARD1 Cys557Ser in the Finnish cancer family material is somewhat higher than the 2.5% reported in the Italian study,¹⁶ and as the alteration was not observed in Japanese families,¹⁷ it may indicate that it is of European/Caucasian origin. In contrast to previous studies, we observed that the occurrence of the Cys557Ser allele in the hereditary material was restricted to patients whose family history did not include ovarian cancer. However, due to the small number of ovarian and breast/ovarian cancer families analysed in the current study, a possible association of Cys557Ser with ovarian cancer cannot be excluded. Besides Cys557Ser, the following BARD1 alterations, Asn295Ser, Lys312Asn, Asn470Ser, Gln564His, and 1144del21bp, have also been proposed to be related to cancer predisposition,^15–17 but none of these changes were seen in our study.

Critical mutations in either BRCA1 or BARD1 have been shown to inhibit many of the essential functions of the BARD1 protein. For instance, the cancer associated BARD1 Gln564His change, which affects a residue that is located only eight amino acids from Cys557Ser (fig 2), results in reduced binding of CSTF1. This was found to abolish the inhibition of polyadenylation in response to DNA damage, indicating a link between mRNA processing and tumour suppression.²³ The Gln564His allele is also defective in apoptotic activity, suggesting a role of BARD1 in cancer prevention by mediating the signalling between genotoxic stress and induction of TP53 dependent apoptosis.²⁷ Recently, it was discovered that the minimal region in mouse Bard1 required for apoptosis induction maps between the ankyrin repeats and the BRCT domains, encompassing amino acids 510 through 604.³⁶ Furthermore, a BARD1 fragment composed of half of the ankyrin region through the BRCT domains (residues 464–777) binds in vitro to the ankyrin repeats of BCL3.²⁸ The minimal segments crucial for the binding to BCL3 or CSTF1 have not yet been determined. In this context, it would be important to examine whether Cys557Ser also has an effect on protein function. Cysteine residues are often crucial for the formation of disulfide bridges within and between proteins, and the substitution can therefore have profound effects on protein stability and folding. Cysteine 557 is conserved between several species (Homo sapiens, Mus musculus, Rattus norvegicus, Xenopus laevis; fig 2), as is also the case for Gln564, even if neither of them falls within the putative functional domains of the protein. Recently, BRCA1/BARD1 orthologs were also discovered in Caenorhabditis elegans,³⁷ but neither Cys557 nor Gln564 was found conserved in the worm Ce-BRD-1 protein. Currently, corresponding sequence data from other more diverged species, such as yeast or flies, are not available.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

 Schematic structure of the human BARD1 protein with highlighted window showing the amino acid (aa) sequence for residues 510–604, a region crucial for induction of apoptosis,³⁶ and a comparison of homology between H. sapiens (Hs), M. musculus (Mm), R. norvegicus (Rn), and X. laevis (Xl). The RING domain (blue), three ankyrin repeats (orange) and two BRCT regions (purple), and their aa positions are displayed. The location of the possible breast cancer related Cys557Ser and other non-synonymous exon region variants seen in the present study are shown above the diagram. In the sequence alignment, residues identical in all four species are coloured red and biochemically similar residues are in green. The positions of Cys557Ser and the previously observed breast/ovarian/uterine cancer associated Gln564His variant are boxed.

The two main susceptibility genes BRCA1 and BRCA2 explain only part of the hereditary predisposition to breast and/or ovarian cancer, the mutation incidence being particularly low in families with four or less cases of breast cancer and typically without ovarian cancer. As it currently appears unlikely that mutations in any single novel susceptibility gene could account for a significant fraction of the remaining families, a polygenic model has been proposed.^38,39 This model suggests that in the absence of BRCA1 and BRCA2 alterations, a majority of the cancer cases clustering within a family could be due to the multiplicative effect of various low to moderate penetrance genes, perhaps in combination with environmental factors. Thus, the finding that BARD1 Cys557Ser associates with breast cancer predisposition in families negative for mutations in BRCA1 and BRCA2, suggests that the effect on its own may be sufficient. On the other hand, it is possible that the Cys557Ser alteration may cosegregate with other currently unknown disease related alleles, and that the variation in age at disease onset could be a consequence of this phenomenon. The slight, but not statistically significant increase in the frequency of the Cys557Ser allele also among breast cancer patients without a known family history of the disease (2.7%) may be a reflection of its incomplete phenotypic expression. Interestingly, a large previous investigation has shown that the cancer associated CHK2 1100delC allele also displays similar frequencies in BRCA1 and BRCA2 mutation negative breast cancer families, unselected breast cancer cases, and healthy controls,⁴⁰ as was seen in the current study for BARD1 Cys557Ser. The effect of these two alterations on cancer risk may also be comparable. Altogether, these findings indicate there are common sequence variants which slightly, but notably, enhance the risk of cancer.⁴⁰

Based on published data, the role of BARD1 in predisposition to breast cancer remains equivocal and unquestionably warrants further studies. According to the results obtained in the current study, individuals heterozygous for the Cys557Ser allele appear to be at increased risk of breast cancer. The observed incidence of 7.4% in the studied breast cancer families compared to 1.4% in healthy controls suggests that it may be a relatively common susceptibility allele associated with incomplete disease penetrance. However, in the absence of direct functional assays, the significance of Cys557Ser on breast cancer development cannot be properly assessed. Therefore, despite interesting results, more extensive studies will still be needed to address the role of BARD1 in breast cancer predisposition.