Constitutional Methylation of the BRCA1 Promoter Is Specifically Associated with BRCA1 Mutation-Associated Pathology in Early-Onset Breast Cancer

Read the Commentary on this article by Issa et al., p. 6

Inactivating germline mutations in BRCA1 act as an initial driver of breast carcinogenesis (1). Carriers of BRCA1 germline mutations typically develop tumors that have characteristic morphologic features that include a high mitotic index, a trabecular growth pattern, little or no tubule formation, malignant nuclear grade, necrosis, nuclear pleomorphism, circumscribed growth pattern, pushing margins, and a moderate to intense lymphocytic infiltrate (refs. 2–3; Southey et al., submitted manuscript).

Germline mutations in BRCA1 currently explain less than 5% of all breast cancers, even for those women diagnosed before the age of 40 years (4–7). However, a substantial proportion of breast cancers diagnosed in women who have been found not to carry BRCA1 mutations have the characteristic morphologic features of the tumors arising in women with germline BRCA1 mutations. Indeed, only 25% of breast cancers diagnosed before age of 40 years with a BRCA1 mutation-associated morphology—defined by 5 or more of the 9 features described earlier—occur in women with a detectable BRCA1 germline mutation (Southey et al., submitted manuscript).

Consequently, alternative mechanisms that could inactivate BRCA1 are of interest. Constitutional methylation (methylation of specific genes detectable in normal tissues either in complete or in mosaic form) is increasingly being considered as a mechanism for cancer predisposition (8). Notably, constitutional methylation of the MLH1 gene has been reported in a proportion of colorectal cancer patients whose tumors phenocopy those arising in individuals with germline mutations of the MLH1 gene (9–11). Peripheral blood is a convenient tissue to assay for constitutional methylation as its collection is considered noninvasive.

It is likely that constitutional methylation of BRCA1 could also be associated with increased risk of breast carcinogenesis. We previously published a pilot study of familial breast cancer patients in which low to moderate levels of BRCA1 methylation were detected in the peripheral blood-derived DNA of 3 of 7 mutation-negative women who had tumors with typical BRCA1-type morphology (12). Given that documented cases of transmission of germline transmission of altered methylation patterns are rare (13), we considered it desirable to analyze constitutional methylation of BRCA1 independently of family history.

In this study, we have used a large population-based study of early-onset breast cancer in which the tumors have been carefully analyzed with regard to their pathology, to determine the relationship of BRCA1 promoter methylation in readily accessible normal tissue (peripheral blood) to breast cancer susceptibility and tumor morphology.

Investigations were made after approval by Human Research Ethics Committees of the Peter MacCallum Cancer Institute and The University of Melbourne. Individuals in the study were enrolled in the Australian site (ABCFR) of the Breast Cancer Family Registry (BCFR). The ABCFR includes a population-based case–control family study of early-onset breast cancer diagnosed before the age of 40 years (14–16). Written informed consent was obtained from each subject. DNA was extracted from stored buffy coats as previously described (6) for those recruited before 1995, and from whole peripheral blood by the use of spin columns (Mini blood spin columns; Qiagen) for those recruited from 1995 onward. BRCA1 mutation screening of these subjects included a combination of protein truncation testing, Sanger sequencing of exonic and flanking intronic regions, screening for large genomic alterations, specific testing for founder mutations, 2-dimensional gel electrophoresis, and via BRCAnalysis at Myriad Genetics (6, 14–15, 17–19).

The first primary invasive breast tumor for each woman was retrieved for 457 of 856 tumors (54%) arising and reviewed as of December 2007. The reasons for not being able to retrieve tissue included lack of participant consent, loss of material at the diagnostic center, the diagnostic laboratory not being willing or able to provide the material, and inability to locate material in centers that had moved or had undergone amalgamation. These are unlikely to be related to age at diagnosis, family history or other factors potentially confounding the analyses of this study.

Retrieved cases were systematically reviewed and scored for morphology features. The pathology scoring system used defines BRCA1 mutation-associated morphology when the tumor has at least 5 of the following 9 features: mitotic index of 50 per 10 high power fields or greater, malignant nuclear grade, little or no (<10%) tubule formation, a trabecular growth pattern (primary or secondary), a syncytial growth pattern, pushing margins (>50%), circumscribed, necrosis, moderate, or intense lymphocytic infiltrate (Southey et al., submitted manuscript). Estrogen receptor (ER) and progesterone receptor (PR) status was collected for cancers from the diagnostic centers or obtained via standard immunohistochemistry staining as previously described (20).

Of the cases without a germline BRCA1 mutation (or mutations in the breast cancer susceptibility genes BRCA2, TP53, CHEK2, ATM, and PALB2), 3 groups were selected on the basis of having high, intermediate, and low numbers of the 9 BRCA1 mutation-associated morphologic features. Group 1 consisted of cases who were considered to have BRCA1 mutation-associated pathology as their tumors had 5 or more features (n = 52), group 2 cases had 4 features (n = 39) and group 3 cases had 3 or less features (n = 164). Controls were women (n = 169) who were unaffected at the time of recruitment into the ABCFR using the electoral rolls and frequency-matched for age.

Macrodissection to enrich for tumor tissue was done using a 21-gauge syringe needle following pathologic identification of the tumor rich areas. DNA was prepared using the DNeasy Tissue Kit (Qiagen) as per the manufacturer's protocol, except that proteinase K incubation at 56°C was carried out for 72 hours with supplementation of the proteinase K at 24 and 48 hours.

Where 1 μg DNA was available, DNA was sodium bisulfite modified using the MethylEasy DNA Bisulfite Modification Kit (Human Genetic Signatures) as per the manufacturer's protocol. Where less than 1 μg DNA was obtained (predominantly when using formalin-fixed paraffin-embedded samples), DNA was sodium bisulfite modified using the Epitect Bisulfite Kit (Qiagen). CpGenome Universal Methylated DNA (Chemicon) was used as fully methylated control DNA (100%). For each modification experiment, similar amounts of fully methylated DNA and DNA extracted from a panel of normal individuals (unmethylated DNA) were also modified.

All bisulfite-modified DNA samples were assayed for relative amounts of DNA using MethyLight analysis of a region free of CpG dinucleotides within the hydroxymethylbilane synthase gene (HMBS; Genbank: NC_000011.8) on the Rotor-Gene 3000 (Corbett Life Sciences). The primers used were 5′-ggTTtgaTTTtTtgTttTagggttatT and 5′-tAccaccaAtcaacactcctcaAA with the dual-labeled probe (5′-TtgTTTtaggTtTTaTTaTtgaagtagaggTagggg). Capital letters denote bases corresponding to converted cytosines. A combined pool of bisulfite-modified DNA from normal individuals was diluted with PCR-grade water to obtain a standard curve of relative DNA amounts of 100%, 50%, 10%, 5%, 1%, and 0%.

The MethyLight primers encompassed CpG dinucleotides at −37 and −29 (forward primer: 5′-ggtagTTTTttggtttTcgtggTaac) and +27 and +44 (reverse primer: 5′-cccgtccaAAaaAtctcaAcg) with the MGB-labeled probe (ActcacgccgcgcaA) spanning CpG dinucleotides at +14, +16, and +19. The relative methylation level (percentage methylated reference) was determined using Ct values relative to the HMBS assay. A standard curve including fully methylated DNA and unmethylated DNA samples was included in each experiment. Amplification efficiency (E) and take-off (T) for each sample were obtained using the RotorGene software (21).

Methylation-sensitive high resolution melting (MS-HRM) primers assessed either 9 CpG dinucleotides (−37, −29, −21, −19, +8, +14, +16, +19, +27) in a 122 base pair (bp) amplicon for the peripheral blood analysis (forward primer: 5′-TtgTtgTttagcggtagTTTTttggtt; reverse primer: 5′-aAcctAtcccccgtccaAAaa) or 4 CpG dinucleotides (−37, −29, −21, and −19) in a 79 bp amplicon for the formalin-fixed (FFPE) tumor analysis (forward primer as above; reverse primer: 5′-caAtcgcaAttttaatttatctAtaattccc).

Bisulfite-modified fully methylated DNA was diluted in pooled bisulfite-modified DNA from normal individuals to obtain a standard curve comprising 100%, 10%, 5%, 1%, 0.1%, and 0% methylated alleles. Each bisulfite-modified peripheral blood DNA sample was run in triplicate on the RotorGene 6000 (Corbett, now available as the RotorGeneQ, Qiagen). When assaying bisulfite-modified tumor-enriched DNA samples, each sample was run in duplicate. The methylation level of each DNA sample was determined visually by comparing it against the standard curve.

Digital MS-HRM was done by limiting dilution (21, 22). Selected amplified products were bisulfite sequenced using Big Dye Terminator Chemistry v3.1 (Applied Biosystems). Amplified MS-HRM products were diluted 1 in 20 times with sterile water and subjected to standard PCR sequencing protocols followed by ethanol precipitation. Clean sequencing products were analyzed on the 3730 Genetic Analyzer (Applied Biosystems).

Primers were designed to flank 4 BRCA1 promoter region single nucleotide polymorphisms (SNP; rs8176072, rs8176073, rs11655505, and rs799906). Each reaction mix consisted of 1X GeneAmp PCR Buffer II, 3.0 mmol/L MgCl₂, 100 μmol/L dNTPs, 200 nmol/L each of forward and reverse primers, 450 nmol/L Syto9 green-fluorescent nucleic acid stain, 0.375 U AmpliTaq Gold DNA polymerase, and approximately 50 ng DNA extracted from blood spots on Guthrie cards. Cycling conditions on the RotorGene 6000 were as follows: initial polymerase activation at 95°C for 15 minutes followed by 50 cycles of denaturation at 95°C for 15 seconds and annealing at 60°C for 60 seconds. Fluorescent data were acquired to the green channel at the end of the annealing step and melting was analyzed after rapid cooling from 95°C to 70°C.

The cancer family history, pathology, and molecular features listed in Table 3 were assessed for association with BRCA1 promoter methylation in the tumor (some methylation versus none) using simple and multiple unconditional logistic regression. The best-fitting multivariate model was defined to be the one with the lowest Bayesian information criterion (BIC; ref. 23). Unconditional logistic regression was also used to estimate the association of methylation with disease and to assess evidence for differences in prevalences across case groups. All P values were based on the likelihood ratio test and all calculations were made using R version 2.7.2 (24).

We used 2 independent assays to determine the proportion of methylated BRCA1 alleles: MethyLight and MS-HRM. The regions covered by each assay were overlapping but not identical (21) due to the different requirements of primer design for these 2 assays (25). MS-HRM, which provides a semiquantitative measurement of methylation, was used to confirm the levels of methylation as estimated by MethyLight. MS-HRM gave well-defined peaks (Fig. 1) indicating that BRCA1 promoter region methylation was homogeneous within the region of the 9 CpGs analyzed by the 124 bp MS-HRM assay.

Of the 52 peripheral blood DNA samples from group 1 women (whose breast tumors had 5 or more BRCA1 mutation-associated morphologic features), 16 (30.8%) were found to be methylated in the BRCA1 promoter using MethyLight (Table 1). Calculation of relative methylation levels using relative Ct values showed that 7 had low-level methylation (≤1% of DNA methylated), 6 had intermediate-level methylation (more than 1% but less than 5% of DNA methylated), and 3 had moderate-level promoter region methylation (ranging from more than 5% to 17%; Table 1; Supplementary Table S1).

Four of the women in the intermediate group (group 2) had BRCA1 promoter methylation in their peripheral blood (4 of 39; 10.3%; Table 1). Two women had moderate levels of methylation whereas 1 had intermediate and 1 had low levels of methylation, (Table 1; Supplementary Table S2).

The frequency and level of measurable methylation in peripheral blood DNA samples for group 3 women and for unaffected controls were strikingly similar to each other. Of the 164 affected women in group 3, only 8 (4.9%) had detectable BRCA1 promoter methylation in their peripheral blood DNA (Table 1; Supplementary Table S3). Four had low-level methylation and 4 had intermediate-level methylation (Table 1). Interestingly, 2 of those peripheral blood DNAs scored as intermediate level by MethyLight were negative and weakly positive by MS-HRM indicating possible heterogeneous methylation. BRCA1 promoter methylation was detected in only 6 (3.6%) of the 169 peripheral blood DNAs from controls. Three had low-level methylation and 3 had intermediate-level methylation (Table 1; Supplementary Table S4).

Therefore, the prevalence of BRCA1 promoter region methylation in DNA from peripheral blood differed between cases and controls (P = 0.004) as calculated by logistic regression. Within cases, it differed by group (P = 2 × 10⁻⁶) consistent with increasing prevalence with greater number of BRCA1 mutation-associated morphology features.

Tumor tissue was available for 20 women in group 1 and 32 women in group 3. Nine (45%) of the tumor-enriched DNA specimens from the 20 available group 1 tumors had methylation in the BRCA1 promoter region. All were highly methylated (50%–100% methylated alleles). With the exception of 2 samples, all tumor-enriched DNAs that displayed BRCA1 promoter region methylation were from women who also displayed BRCA1 promoter methylation in their corresponding peripheral blood DNA (Table 2). Significantly, there were no instances in group 1 where peripheral blood with detectable methylation was not associated with a tumor-enriched DNA sample with markedly increased methylation (Supplementary Table S5).

Three of the available 32 tumor-enriched DNA samples from group 3 women had detectable methylation at the BRCA1 promoter region. For 1 of the samples, there was moderate methylation (more than 1% and less than 5% methylated alleles) in the corresponding peripheral blood DNA. In contrast to group 1, this case showed no increase of methylation in its tumor-enriched DNA (Table 2), indicating that the cancer did not arise from a BRCA1 methylated clone and that the level of methylation seen represented a background level. The remaining 2 group 3 tumor-enriched DNA specimens were strongly methylated but there was no detectable methylation in the corresponding peripheral blood DNA (Supplementary Table S6).

Simple logistic regression (Table 3) showed that the following pathology features were individually predictive of BRCA1 promoter methylation: syncytial, trabecular, and circumscribed growth patterns; high mitotic index; and necrosis. ER and PR status were predictive of BRCA1 promoter methylation but family history variables were not (see Table 3). Multivariate logistic regressions identified high mitotic index and circumscribed growth pattern as the best predictors of methylation, with estimated odds ratios (OR) of 5.8 (95% CI, 2.3–15.0; high mitotic index) and 3.4 (95% CI, 1.3–8.7; circumscribed growth pattern). Once these 2 features were taken into account, however, no other features added predictive power to the regression model (i.e., this 2-feature model had the lowest BIC of all possible models; in addition, adding any other feature to the model gave P > 0.1 for the third feature). In particular, although ER and PR status were individually predictive of BRCA1 promoter methylation (P = 0.006 and 0.03, respectively), neither was predictive once adjustment had been made for high mitotic index and circumscribed growth patterns (both P > 0.05). This is consistent with the finding that all 4 features were strongly associated with each other (all P < 0.01). In a separate analysis in which methylation was regressed against the total number of morphology features present, the odds of methylation were estimated to increase by a factor of 1.63 (95% CI 1.33 − 2.03) for each additional feature present (p-trend = 0.0001). This model does not differentiate between morphology features so the odds ratio per feature is analogous to an average over all nine morphology features of the odds ratios given above for the two most predictive features. Last, in an analysis based on control and case groups, BRCA1 promoter region methylation was estimated to be associated with a 3.5-fold (95% CI, 1.4–10.5) increased risk of breast cancer before age 40 years (P = 0.004).

We screened peripheral blood DNA from the family members of 9 women (8 from group 1 and 1 from group 3) who had moderate to high levels of methylation (1% or greater) at the BRCA1 promoter region in their peripheral blood DNA (Table 4, Relatives of probands with detectable peripheral blood methylation). Neither MethyLight nor MS-HRM analysis detected methylation at any level in any of the family members. We also identified methylation-negative individuals with family members who had been diagnosed with either breast and/or ovarian cancer (Table 4, Other relatives with breast or ovarian cancer). Neither MS-HRM nor MethyLight detected methylation at any level in the peripheral blood DNA of any of these affected family members.

All group 1 women were genotyped for 4 SNPs (rs8176072: A>T, rs8176073: A>G, rs11655505: A>G, and rs799906: C>T) in the BRCA1 promoter region to assess if there was an association with detectable BRCA1 promoter methylation in the peripheral blood (Table 5). There was a trend (significant at the 5% level in the case of rs799906 and rs11655505) for detectable methylation to be associated with the presence of the minor allele. All 3 peripheral blood DNAs displaying moderate levels of methylation (at least 5% methylated alleles) were heterozygous for both rs799906 and rs11655505.

It is reasonable to consider that constitutional epimutations, when present, can act as the first step in carcinogenesis in a manner analogous to germline mutations. It is also reasonable to consider that cancers with a driver lesion in a given gene should have a similar pathology, regardless of whether the driver lesion is genetic or epigenetic.

This study sought to determine whether detectable constitutional epimutations in the BRCA1 gene were frequent in the peripheral blood using a population-based study of women with early-onset breast cancer who did not have BRCA1 germline mutations, but differed according to the extent to which their tumors exhibited BRCA1 mutation-associated pathology features. This would not only have implications for our understanding of breast cancer pathogenesis but also would raise the possibility of using BRCA1 directed prevention measures for unaffected women with constitutional epimutations.

The BRCA1 exon 1a promoter comprises 11 CpG sites between BRCA1 exon 1a and the first exon of NBR2 (26). Dobrovic and Simpfendorfer (27) first showed that the BRCA1 promoter region was methylated in approximately 20% of breast tumors. This finding was confirmed and BRCA1 promoter region methylation was also shown for cases of ovarian cancer (26–33). The identification of BRCA1 methylation in 2 cancer types that are driven by germline BRCA1 mutations, and the lack of BRCA1 methylation in other cancers such as colorectal cancer and leukemia that were not associated with germline BRCA1 mutations, argued that BRCA1 methylation could play a pathogenic role in breast and ovarian cancers (29–30). Thus, constitutional methylation of BRCA1, in which the promoter is methylated throughout the somatic tissues, could be a predisposing factor if present.

Peripheral blood is a convenient tissue to assay for constitutional methylation. We previously reported a pilot study of BRCA1 methylation in the peripheral blood of a small group of BRCA1 mutation-negative women from a familial cancer registry (12). Of the 7 women whose tumors showed BRCA1 mutation-associated pathology, 3 had detectable methylation of the BRCA1 exon 1a promoter (at levels between 1% and 12%) in their peripheral blood. All had an early age of onset of breast cancer. The corresponding tumors from these women were heavily methylated at the BRCA1 promoter, suggesting that BRCA1 epimutations could act as an alternative mechanism leading to breast cancer predisposition and that they might be particularly important in early-onset disease (21). We therefore set to test these initial observations by assessing BRCA1 promoter methylation in a much larger set of 255 women with early-onset breast cancer from a population-based study (13, 15).

The prevalence of BRCA1 promoter region methylation in peripheral blood DNA from group 1 individuals with 5 or more BRCA1 mutation-associated morphology features was 31%, considerably higher than the 10% seen in the intermediate group with 4 features (P = 0.02), the 5% seen in the group 3 individuals with 3 or less features (P = 2 × 10⁻⁶), and the 4% seen in the controls (P = 2 × 10⁻⁷). The prevalence of BRCA1 promoter methylation did not significantly differ between the group 3 women and the unaffected controls (P = 0.5).

The amount of BRCA1 methylation in the peripheral blood DNA (when present) was not at the expected 50% level expected of a germline epimutation indicating that mosaicism was present. It varied markedly between the samples examined (0.1%–17% of BRCA1 alleles). There was good concordance between the MethyLight and the MS-HRM results (Supplementary Tables S1–S4). The 2 methodologies take very different approaches to the quantification of methylation, giving confidence in the accuracy of the results.

It could be argued that our findings were due to circulating breast cancer cells that were methylated for BRCA1. This is unlikely because circulating tumor cells are present at very low levels and can only be isolated from the blood after enrichment, as studies from our laboratory and other laboratories have shown (34). The sensitivity of our methylation assays is 0.1%. It is unlikely that cancer cells are present in the peripheral circulation at levels equal to or higher than this in any breast cancer patient let alone any normal control woman.

Nine of the 20 tumors examined from women in group 1 had high-level (defined as more than 50%) BRCA1 methylation whereas none of the 32 tumors examined from women in group 3 had high-level or even moderate-level methylation. Importantly, the tumors were strongly methylated for each of the 7 group 1 women with detectable peripheral blood methylation for whom a tumor could be obtained. This contrasted dramatically with the 1 woman from group 3 with detectable peripheral blood methylation for whom we were also able to test a tumor. Her tumor showed no increase of methylation relative to that seen in peripheral blood. These results are consistent with BRCA1 methylation driving the development of tumors with BRCA1-associated pathology.

Reports of the relationship between BRCA1 promoter region methylation and clinicopathologic features have been conflicting. However, most studies show a clear tendency for BRCA1 methylated tumors to occur in young women with high-grade, ER-negative, PR-negative tumors (33, 35). Several studies pointed out the resemblance between BRCA1 methylated and BRCA1 mutated tumors. Esteller et al. (30) reported that BRCA1 methylation was associated with the less common medullary (67% methylated) and mucinous (55% methylated) subtypes, which are overrepresented in families carrying BRCA1 mutations. Turner et al. (36) reported that BRCA1 methylation occurred in 63% of metaplastic breast cancers.

We hypothesize that there are likely to be 2 groups of BRCA1 promoter region methylated tumors. In “first hit” tumors, BRCA1 methylation predisposes to and initiates tumorigenesis and these tumors accordingly resemble BRCA1 mutation-positive tumors morphologically. In tumors in which BRCA1 methylation occurs later during tumor progression, the loss of BRCA1 would not be expected play a major role in determining the tumor's morphology. This may be more common in postmenopausal tumors. Interestingly, population-based studies of women with BRCA1 mutations indicate that they are primarily at risk of early-onset breast cancer, with risk falling as the women age (13).

Two tumors from women in group 3 showed high-level BRCA1 methylation without corresponding peripheral blood methylation. It is likely that these are examples of BRCA1 methylation that occurred during a later stage of tumor evolution and thus did not drive the breast cancer.

The origin of constitutional methylation of the BRCA1 promoter remains obscure. There is very little evidence of transmission of an epimutation in human heredity. In most cases, what is transmitted is genetic variation, both in cis and in trans sequences, which alter the propensity of a locus to methylate (8). Cis-acting sequences that affect methylation in normal tissues have been identified for the RIL and MGMT loci among others (37). For the MGMT gene, we have previously shown that a promoter SNP plays a major role in determining the probability of the methylation of the MGMT promoter CpG island in normal tissues (38). The same SNP affects the probability of MGMT methylation in colorectal cancer and adjacent colonic tissues (39).

We thus assessed whether genetic variants in the promoter region of the BRCA1 gene bore any relation to the probability of BRCA1 promoter region methylation in the women with tumors showing BRCA1 mutation-associated features (Table 5). Moderate associations were observed for the rs11655505: A>G and rs799906: C>T SNPs. Women with the minor alleles (T allele of rs11655505/G allele of rs799906) were more likely to be methylated. The association of methylation with both SNPs was significant at the 5% level.

These 2 genetic variants have been reported to alter promoter activity (40). Surprisingly, carriers of the T allele of rs11655505 were reported as having a reduced risk of breast cancer in a Chinese population, although in that study the protective effect was most strongly seen for women aged more than 45 years (40). However, a recent genotyping study of Caucasian populations did not find any association between rs11655505 and breast cancer risk (41). The study included the ABCFS data set from which our cases derived, and although there was a lack of significant association, there was a weak trend for the T allele to have an increased risk of breast cancer, concordant with our results.

The absence of strongly predisposing genetic variants is consistent with the lack of somatic methylation in family members. We analyzed 2 groups of immediate family members of the affected individuals in this study (Table 4). In the first group, we were interested in determining if family members (irrespective of affected status) of cases with moderate to high levels of germline methylation were also methylated in BRCA1. In the second group, we were interested in evaluating BRCA1 promoter methylation levels for cancer affected family members where proband methylation levels were not detected. We were unable to detect any methylation in either group. The results indicate that the BRCA1 promoter methylation detected in the index cases was not inherited but rather arose de novo in the women.

This study has shown that the BRCA1 peripheral blood methylation identifies a small group of women in the population (our best estimate for Australian women is 4%) who have approximately a 3.5-fold increased risk of breast cancer before the age of 40 due to epigenetic loss of BRCA1. This hypothesis is further supported when BRCA1 promoter methylation of both tumor-enriched DNA samples and tumor pathology were considered. If this methylation were causal, then our study predicts that approximately 11% of early-onset breast cancers would be attributable to this phenomenon. Given that less than 10% of early-onset breast cancers arise in women with germline mutations in BRCA1, constitutional epimutations of BRCA1 might explain as much, if not more, of early-onset breast cancer as inherited germline defects.

In conclusion, we have described a group of women with early-onset breast cancer who have detectable methylation of the BRCA1 promoter region in the peripheral blood and have tumors that are strongly methylated at the BRCA1 promoter and possess morphologic features consistent with carrying BRCA1 germline mutations. We postulate that BRCA1 epimutations are also present in the breast tissue of these young women and drive breast carcinogenesis in a manner analogous to BRCA1 germline mutations.

BRCA1 methylation thus may be an important new biomarker for breast cancer predisposition. Our findings, if replicated by further studies, could have major implications for prevention by screening women for epigenetic changes that increase risk of early-onset breast cancer. It also raises the possibility for cancer prevention in that women with epimutations might have their risk of developing breast cancer reduced by appropriate pharmaceutical prevention (42) or even by dietary manipulation.

There are no potential conflicts of interest involved.

We thank the participants in this study. We also thank Elena Takano for assistance in the collection of tumor blocks and slides, Mukta Rayoo and Max Yan for assisting the pathology review of the tumor sections for microdissection, Minh Bui for assistance with the statistical analysis, and Andrea Tesoriero and Letitia Smith for technical input.

This study was supported by grants to A. Dobrovic from the U.S. Department of Defense Breast Cancer Research Program under award number W81XWH-06-1-0670, the Cancer Council of Victoria, Susan Komen for the Cure, and the National Breast Cancer Foundation of Australia, and by a grant to M.C. Southey from the Cancer Council of Victoria. The ABCFR was supported by the National Health and Medical Research Council of Australia (NHMRC) [145604], the United States National Institutes of Health (NIH) [CA102740–01A2], and by the United States National Cancer Institute, National Institutes of Health [CA-06–503] through cooperative agreements with members of the BCFR and principal investigators Cancer Care Ontario [CA69467], Columbia University [CA69398], Fox Chase Cancer Center [CA69631], Huntsman Cancer Institute [CA69446], Northern California Cancer Center [CA69417], University of Melbourne [CA69638]. The ABCFR was initially supported by the NHMRC, the New South Wales Cancer Council and the Victorian Health Promotion Foundation. The ABCFR has also received support from the Victorian Breast Cancer Research Consortium of which M.C. Southey and J.L. Hopper are group leaders. J.L. Hopper is an Australia Fellow of the NHMRC. M.C. Southey is a Senior Research Fellow of the NHMRC. E.M. Wong was supported by the Dora Lush postgraduate biomedical scholarship from the NHMRC.

Views and opinions of, and endorsements by, the authors do not reflect those of the U.S. Army or the U.S. Department of Defense. The content of this article does not necessarily reflect the views or policies of the National Cancer Institute or any of collaborating centers in the BCFR, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government or the BCFR.

The costs of publication of this article were defrayed in part by the payment 656 of page charges. This article must therefore be hereby marked advertisement in 657 accordance with 18 U.S.C. Section 1734 solely to indicate this fact.