Epigenetic silencing and deletion of the BRCA1gene in sporadic breast cancer

Germline mutations in one allele of the BRCA1 or BRCA2 genes significantly increase the risk of developing early-onset breast cancer [1]. Tumour cells from predisposed individuals have consistently lost the wild-type BRCA allele [2, 3]. The most prominent feature of BRCA deficient cells is the inability to repair DNA cross-links and DNA double-strand breaks by error-free homologous recombination, which probably underlies genomic instability and cancer predisposition [4]. Survival of BRCA deficient cells is generally thought to be dependent upon dysfunctional checkpoint mechanisms, in which case tumorigenic potential could be acquired through additional genomic rearrangements and gene mutations. Indeed, familial BRCA1 tumours are associated with mutations in the TP53 checkpoint gene, supporting the notion that genomic instability is an important driving force in early-onset familial BRCA1 tumorigenesis [5].

Although inherited cancer syndromes are rare, the genes accounting for them are generally believed to play an important role in sporadic cancer. It was anticipated, therefore, that somatic BRCA mutations would be found to contribute to sporadic breast carcinogenesis. Surprisingly, somatic BRCA gene mutations have not been found in sporadic breast tumours [6, 7]. On the other hand, allelic imbalance (AI) at the BRCA loci, an indicator for loss of heterozygosity, is know to be a fairly common event in breast cancer [8]. The implications of AI at the BRCA loci are unknown since Knudson's hypothesis predicts an additional inactivating event on top of AI to be required for tumorigenesis to occur [9]. For these reasons, the involvement of the BRCA genes in sporadic breast tumours has been questioned. An alternative mechanism for BRCA1 inactivation has been suggested to be gene silencing by epigenetic mechanisms. Hypermethylation of CpG-island promoters is known to be strongly associated with gene silencing. Once established, methylation is passed on to daughter cells during DNA replication by the activity of DNA methyltransferases, thereby conserving the overall pattern of methylated CpG-islands [10]. The methylation patterns of virtually all types of cancer, including breast carcinoma, have been found to differ extensively from that of the corresponding normal tissue. These alterations are cancer-type specific and include global genomic hypomethylation as well as non-random hypermethylation of normally unmethylated CpG-island promoters [11, 12]. These observations, and others, indicate that epigenetic modifications could be important in cancer etiology [13].

Several studies have reported hypermethylation of the BRCA1 promoter in sporadic breast and ovarian tumours. Furthermore, BRCA1 methylation has only been found in breast and ovarian tumours and has been associated with AI at the BRCA1 locus and reduced BRCA1 gene expression [12, 14]. BRCA2 promoter hypermethylation has not been found in breast tumours, although it has been reported in ovarian tumours [15, 16].

Familial BRCA1 and BRCA2 tumours are associated with young age of onset and are phenotypically distinct from each other as well as from sporadic breast tumours [1, 17‐19]. Conventional histopathological and molecular analyses have demonstrated familial BRCA1 tumours to have a basal-like phenotype and to be significantly associated with certain features, such as AI at the BRCA1 locus, a negative estrogen receptor (ER) and progesterone receptor (PR) status, a medullary tumour histological type, TP53 mutations and, depending on the mutation involved, a high tumour grade [3, 5, 18, 20, 21]. Interestingly, gene expression profiling has revealed similarities between BRCA1 methylated and familial BRCA1 tumours [17, 22]. Similarly, a comparative genomic hybridisation study has reported a specific pattern of genetic alterations to be predictive of familial BRCA1 tumours and BRCA1 methylated tumours [23]. This lends support to the idea that epigenetic silencing of the BRCA1 gene might channel tumour progression, akin to an underlying BRCA1 germline mutation resulting in a BRCA-like phenotype. However, a recent report showing high levels of BRCA1 expression and a low frequency of BRCA1 promoter methylation in basal-like sporadic tumours suggests that this might be more complex [24].

In the present study, we examined the frequency of BRCA1α promoter hypermethylation in 143 unselected primary sporadic breast tumours. All tumours were analysed for AI at the BRCA1 and BRCA2 loci, TP53 mutations, hormonal receptor status and age at diagnosis. Copy number alterations at the BRCA1 locus were further examined by fluorescence in situ hybridisation (FISH) in the BRCA1 methylated tumours which were also analysed for BRCA1 protein expression, histological type and tumour grade. The purpose of the study was to examine whether the BRCA1 gene could be implicated in sporadic breast tumorigenesis through epigenetic modifications.

We report here that hypermethylation of the BRCA1 gene promoter is found in a considerable proportion of primary sporadic breast carcinomas, that is, 13 of 143 (9.1%), which is in the lower end of previously reported frequencies for this alteration in sporadic breast tumours [14, 33, 34].

Absent or markedly reduced BRCA1 protein expression was evident in the majority of the BRCA1 methylated tumours (9 of 13), suggesting transcriptional silencing in these tumours by epigenetic modifications. A trend for AI at the BRCA1 locus was observed in the subset of BRCA1 methylated tumours (P = 0.0731). All the BRCA1 methylated tumours that had AI at the BRCA1 locus and were informative for AI at the exogenic and an intragenic marker displayed AI at both regions, indicating a rather large deletion at chromosome 17. This is supported by the FISH analysis, which revealed deletion of chromosome 17 in most of the BRCA1 methylated tumours that had a detectable BRCA1 deletion. Importantly, the FISH analysis revealed substantial heterogeneity in BRCA1 gene copy numbers between individual cells in the BRCA1 methylated tumours, demonstrating that AI as detected by polymorphic microsatellite PCR analysis does not infer a simple loss of one BRCA1 allele but, rather, it appears to reflect complex genetic rearrangements.

AI at the BRCA1 and BRCA2 loci are know to be relatively common in breast tumours [8]. The implications of AI at the BRCA1 and/or BRCA2 loci for sporadic breast tumorigenesis remain unknown since Knudson's hypothesis predicts that two 'hits' are required for tumorigenesis to occur [9]. Our results confirm that AI at the BRCA1 and BRCA2 loci are common events in sporadic breast tumours, present in 37.4% (49/131) and 31.1% (42/135) of primary sporadic breast tumours, respectively. A significant association was found between AI at the BRCA1 and BRCA2 loci (P < 0.0001). Importantly, we found an indication for AI at the BRCA1 locus to be associated with BRCA1 methylation (P = 0.0731) whereas AI at the BRCA2 locus was not found to be associated with BRCA1 methylation (P = 0.5420). This has not been shown previously and suggests that AI at the BRCA1 locus is specifically associated with BRCA1 methylation. Thus, copy number alterations and epigenetic silencing of the BRCA1 gene in sporadic breast cancer could serve as Knudson's 'hits', which has previously been proposed by Esteller and colleagues [35]. Indeed, all but one of the BRCA1 methylated tumours that had absent/markedly reduced BRCA1 protein expression (8 of 9) also had a detectable deletion of the BRCA1 gene. Collectively, these results suggest that the BRCA1 gene is implicated in sporadic breast tumorigenesis through epigenetic silencing and deletion of the BRCA1 gene. Indications that BRCA1 methylation is important in hereditary breast cancer have been reported [35].

The failure to detect a BRCA1 deletion in one of the tumours that exhibited absent or markedly reduced BRCA1 expression could mean that promoter hypermethylation is present on both alleles, thereby alleviating any selection pressure for deletion at the BRCA1 locus. Alternatively, the level of detection in the FISH analysis could be limited by the small proportion of tumour cells present in each tumour section analysed. This might also apply for those tumours in which AI was present without a detectable deletion by FISH. Conversely, the detection level of the AI analysis was limited by the fact that none of the tumours were micro/macrodissected prior to DNA isolation, which also means that unmethylated BRCA1 alelles are always detected in the tumour samples due to the presence of normal DNA.

The four BRCA1 methylated tumours that did not exhibit significantly reduced BRCA1 expression could possibly be heterogenous with respect to this alteration. None of the four tumours exhibited BRCA1 deletion by FISH and only one displayed AI at the BRCA1 locus. Alternatively, DNA methylation might not bring about transcriptional silencing in all instances.

Although the etiology of cancer predisposition in individuals carrying a germline BRCA1 mutation is not clear, increased genomic instability in BRCA1 deficient cells is undoubtedly of importance since it is predicted to result in increased probability of further genetic alterations and gene mutations, which might result in functional consequences by which tumorigenic potential could be acquired. Genomic instability, however, is a potent inducer of apoptosis where cell survival is dependent upon dysfunctional checkpoint mechanisms [4]. Indeed, familial BRCA1 tumours are associated with mutations in the TP53 checkpoint gene, supporting the notion that genomic instability is an important driving force in early-onset familial BRCA1 tumorigenesis [5]. Association of BRCA1 methylation with TP53 mutations has not been shown previously. Our results show a higher frequency of TP53 mutations among the BRCA1 methylated tumours compared with the non-methylated tumours or 38.5% (5 of 13) and 17.2% (22 of 128), respectively (P = 0.1299, OR = 3.0, 95%CI = 0.9–10.1). This association was not statistically significant, although the TP53 mutations were found to be entirely limited to those BRCA1 methylated tumours that exhibited absent or markedly reduced BRCA1 expression, in which case the frequency of TP53 mutations becomes 55.5% (5 of 9) and the association statistically significant (P = 0.01317, OR = 6.13, 95%CI = 1.21–33.51). Reinforcing this idea is the observation that all the five BRCA1 methylated tumours with a TP53 mutation had a detectable BRCA1 copy number reduction and the majority of these tumours had a relatively high DNA index, suggesting genomic instability (Table 2).

It has previously been suggested that BRCA1 methylated tumours might phenocopy familial BRCA1 tumours [36]. In support of this notion, we observed ER negativity to be significantly associated with BRCA1 methylation (P = 0.0475), a well established characteristic of familial BRCA1 tumours previously reported by Catteau and colleagues [37] and others. However, Matros and colleagues [24], looking at gene expression profiles, found a high frequency of BRCA1 promoter methylation among high-grade ER positive tumours, suggesting a more complex phenotype association. We found an indication for BRCA1 methylation to be specifically associated with AI at the BRCA1 locus and an elevated frequency of TP53 mutations, which has not been reported previously. In addition, we found a considerable proportion of the BRCA1 methylated tumours (7 of 13) to be of grade 3, with only one tumour of grade 1, as well as an indication of an association between BRCA1 methylation and an early age of onset (P = 0.0898) as previously reported by Wei and colleagues [34].

It has been suggested that breast cancers arising in individuals carrying a germline mutation in the BRCA genes could benefit from therapeutic agents that lead to DNA cross-links or double-strand breaks at replication forks, for example, mitomycin C, cisplatin, diepoxybutane and, more recently, poly(ADP-ribose) polymerase (PARP) inhibitors [38]. These therapeutic agents could also be effective for sporadic breast cancers with abnormalities in the BRCA genes, which is, as shown here, a considerably larger proportion of all breast cancer patients than germline BRCA1 or BRCA2 mutation carriers. In addition, abnormalities in other genes regulating homologous recombination could also be of relevance. This emphasizes the importance of developing methods for identifying BRCA-like cancers, regardless of the underlying alterations [36].

Our results show promoter hypermethylation of the BRCA1 gene in a considerable proportion of all primary sporadic breast tumours. The majority of the BRCA1 methylated tumours were found to have absent or markedly reduced BRCA1 expression, suggesting transcriptional silencing by epigenetic modifications. In addition, we found an indication for AI at the BRCA1 locus to be associated with BRCA1 methylation whereas AI at the BRCA2 locus was not associated with BRCA1 methylation. This indicates that AI at the BRCA1 locus is specifically associated with BRCA1 methylation. The genetic alterations at the BRCA1 locus were further examined by FISH, which revealed chromosome 17 deletions and heterogeneity with respect to chromosomal abnormalities. These results imply that methylation of the BRCA1 gene is accompanied by genomic rearrangements at the BRCA1 locus, resulting in loss of genetic material containing non-methylated BRCA1 alleles and retention of methylated BRCA1 alleles. We also found a substantially elevated frequency of TP53 mutations in the subset of BRCA1 methylated tumours, which has not been reported previously, suggesting that BRCA1 methylation might lead to alterations in the same molecular pathways as those known to be commonly altered in familial BRCA1 tumours. Collectively, these results implicate epigenetic silencing of the BRCA1 gene in sporadic breast tumorigenesis.

Medullary histological type was not found in the BRCA1 methylated tumours. However, we observed ER negativity to be significantly associated with BRCA1 methylation. We also found a substantial proportion of the BRCA1 methylated tumours to be of grade 3 and an indication for an association between BRCA1 methylation and early age of onset. Thus, our results indicate phenotypic similarities between BRCA1 methylated and familial BRCA1 breast tumours.