Introduction

Germline mutations in the two major breast cancer susceptibility genes, BRCA1 and BRCA2 (MIM113705 and 600185), are frequently found in families containing multiple individuals affected by breast and ovarian cancer [1]. However, BRCA1 and BRCA2 mutations are only identified in about 15–20% of multiple-case families affected by breast cancer alone [1, 2]. Other breast cancer-predisposing genes might account for a proportion of the remaining cases.

The ATM (ataxia–telangiectasia mutated) gene (MIM 208900, 607585), whose product has a central role in DNA repair after damage induced by ionising radiation [3, 4] and phosphorylates BRCA1 [5, 6], has been proposed as one such candidate breast cancer-predisposing gene. Individuals harbouring homozygous (or compound heterozygous) deleterious mutations in the ATM gene develop ataxia–telangiectasia (A–T), which is characterised by progressive cerebellar degeneration, telangiectasia, immunodeficiency, extreme sensitivity to ionising radiation and a predisposition to lymphoid malignancies [7]. Most pathogenic ATM mutations described so far (now in excess of 300) result in truncating mutations that produce little or no detectable ATM protein, but pathogenic missense mutations have also been described. About 1% of the population are asymptomatic heterozygous carriers. However, because the ATM protein forms high-molecular-mass complexes, it is conceivable that heterozygous missense mutations (or certain truncating mutations) could have a dominant-negative effect, interfering with the function of the normal allele and thereby resulting in an increased predisposition to cancer [8].

Clinical observation and epidemiological studies since the 1970s have shown that blood relatives of A–T patients seem to be at increased risk for breast cancer, suggesting that heterozygosity for mutations in the ATM gene might predispose to breast cancer [9, 10]. After the identification of the gene in 1995 [7], several studies demonstrated an association between ATM heterozygosity and increased risk for breast cancer [11–15]. However, other studies have failed to demonstrate an association [16–18], concluding that the contribution of ATM heterozygosity to familial breast cancer is minimal. It is difficult to reconcile the different outcomes of these studies because they differed in sample size, in subject selection criteria, and in the mutation detection methods used.

A recent study of multiple-case breast cancer families in Australia suggested that two ATM variants, ATM 7271T→G and IVS10-6T→G, confer a substantial risk for breast cancer in families with multiple cases [19]. The ATM 7271T→G missense variant was subsequently not detected in two large independent cohorts of breast cancer families [20, 21], indicating that the variant is of limited clinical significance. The ATM IVS10-6T→G variant leads to incorrect splicing at exon 11 and results in a truncated ATM product [22]. In the Australian study, this variant was detected in 2 of 76 (2.6%) breast cancer patients with a strong family history of breast cancer, 0 of 268 breast cancer patients without a family history of the disease, and 0 of 68 women with no personal or family history of breast cancer [19]. In the two families with this variant, the penetrance (in terms of breast cancer risk) was estimated to be 78% (95% confidence interval 36–99%) by 70 years of age. An earlier study had also demonstrated an association between this variant and breast cancer risk [22]. However, three subsequent studies in North America and Europe failed to identify a significant role for this variant in hereditary or sporadic breast cancer [20, 21, 23]. Nevertheless, it has been argued that the ATM IVS10-6T→G allele represents a biologically and clinically significant variant because ATM kinase activity is reduced to 25–40% in heterozygous cells [19].

The possibility that ATM IVS10-6T→G could be a high-penetrance breast cancer susceptibility allele, or even a low-penetrance risk-modifying allele, has major ramifications for genetic testing and clinical management of individuals with a hereditary predisposition to breast and ovarian cancer. Because the association had been made in a cohort of (ethnically diverse) multiple-case Australian families, we determined the frequency and relevance of this variant in an independent series of families from Australian familial cancer clinics. We compared the frequency of the ATM IVS10-6T→G variant in affected individuals undergoing genetic testing for BRCA1 and BRCA2 at four major Australian familial cancer clinics with the carrier rate in Australian women without a family history of breast cancer.

Materials and methods

Samples

Patients who had been affected by either breast and/or ovarian cancer and were index cases within a family that had initiated BRCA1 and BRCA2 mutation detection were ascertained from four Australian familial cancer clinics: Genetic Health Services Victoria (GHSV), Melbourne, the Royal Melbourne Hospital (RMH), Melbourne, the South Australian Familial Cancer Service (SAFCS), Adelaide, and Westmead Hospital (WH), Sydney. For GHSV and RMH, the series included peripheral blood leucocyte genomic DNA samples collected between July 1999 and June 2003 (inclusive), for which retrospective patient (or next of kin) consent was obtained [48 of 59 (81%) and 85 of 97 (88%), respectively]. For SAFCS and WH, a consecutive series from probands, for whom testing at that stage had not detected a BRCA1 or BRCA2 mutation, was screened for the ATM IVS10-6T→G variant (201 and 130 samples, respectively). Testing was subsequently extended at WH to include BRCA1 affected mutation carriers (31 samples). Approval to conduct ATM testing was obtained from each relevant institutional ethics committee, unless the standard clinical genetic testing consent form covered such testing.

Affected women's families were categorised as being at potentially high, moderate or average risk for developing breast cancer on the basis of family history, according to the National Health and Medical Research Council (Australia) guidelines [24], as follows.

Potentially high risk was defined as the presence of at least two first-degree or second-degree relatives on one side of a family diagnosed with breast or ovarian cancer (including the patient) plus at least one of the following high-risk features: additional relative(s) with breast or ovarian cancer, breast cancer diagnosed before the age of 40 years, ovarian cancer diagnosed before the age of 50 years, bilateral breast cancer, breast and ovarian cancer in the same woman, Jewish ancestry, or breast cancer in a male relative.

Moderate risk was defined as one or two first-degree relatives diagnosed with breast cancer (including the patient) before the age of 50 years, two first-degree or second-degree relatives on the same side of the family with breast or ovarian cancer, or one first-degree relative with ovarian cancer before the age of 50 years (in all cases without additional high-risk features). Standard risk was defined as the affected woman falling outside either of the other two risk categories.

Mutation screening

Samples were screened for the ATM IVS10-6T→G variant essentially as described previously [19]. Briefly, a 193-base-pair fragment spanning the IVS10-6 region was amplified from genomic DNA by polymerase chain reaction with the primers 5'-ACAGCGAAACTCTGGCTCAAA-3' and 5'-TGATCTTTTATTACTTCCCAGCCTAGT-3'. Because the T→G variant creates a novel RsaI restriction site, heterozygous carriers were detected by RsaI restriction digest, which produces 58-base-pair and 135-base-pair variant-specific fragments in addition to the wild-type 193-base-pair fragment. Fragments were separated and detected on either a 12% SDS-polyacrylamide gel or a 1.5% agarose gel. Alternatively, the restriction fragments were separated by denaturing high-performance liquid chromatography (details available from the authors on request). A positive control genomic sample was kindly provided by Dr G Chenevix-Trench (Queensland Institute of Medical Research, Brisbane, Australia). For sequencing, independently amplified polymerase chain reaction products were generated and confirmed by automated sequencing with the ABI Prism BigDye Terminator Cycle Sequencing Kit (Perkin Elmer).

Statistical analysis

Two-tailed P values were generated with Fisher's Exact Test. In accordance with convention, P < 0.05 was considered significant.

Results and discussion

To investigate the frequency of the ATM IVS10-6T→G variant in patients at high probability of carrying a gene predisposing them to breast cancer, we ascertained 495 patients who had attended one of four familial cancer clinics located in three Australian States. These patients had been affected by either breast or ovarian cancer and had undergone genetic testing for BRCA1 and BRCA2 mutations. In most cases (90.7%), the family was classified as 'potentially high risk' for breast and/or ovarian cancer predisposition [24]. A small proportion (7.5%) were categorised as being in the moderate risk group, although these generally exhibited unusual features (such as breast and ovarian cancer in a single individual in the absence of a family history), and insufficient information on risk was available for the remaining 1.8%. The proportion of families in the different risk categories was similar at the four centres (Table 1; data not shown). For patients from GHSV, RMH and WH, the average age of breast cancer diagnosis of the proband was 45 years (range 22–78) and an average of 3.4 breast cancers and 0.42 ovarian cancers were documented per family. Of the potentially high-risk kindreds, the majority (71%) of families contained breast-only cancer cases, whereas breast and ovarian cancers were noted in the remaining families. For SAFCS families, the average age of breast cancer diagnosis for the proband was 48 years, and an average of 3.2 breast cancers and 0.25 ovarian cancers were documented per family. The mean predictions for GHSV and RMH patients carrying a BRCA1 or BRCA2 mutation were 0.22 and 0.19, respectively, as determined by either the BRCAPRO or the MYRIAD II statistical model [25].

Table 1 Characteristics of index cases and families
Full size table

We screened probands from non-BRCA1/BRCA2 families in whom no mutation had been identified at the time (n = 464) and probands with known BRCA1 mutations (WH cohort, n = 31). Of the 495 probands screened, 7 (1.4%) were found to be heterozygous for the ATM IVS10-6T→G variant. Four of these were ascertained through SAFCS and one from each of the other centres (Table 2). None of these ATM heterozygotes had been reported in the study by Chenevix-Trench and colleagues [19]. Their age at diagnosis, family histories and family risk status were quite diverse (Table 3). Their mean age at first cancer diagnosis was 53 years (range 35–62). None had a history of prior radiation exposure. Three of the seven individuals had a family history of breast/ovarian cancer, one a family history of breast/thyroid cancer, and the remaining three families had a history of breast cancer alone. Three heterozygous individuals were from moderate-risk families. No heterozygous carriers had clinical features attributable to A–T such as ataxia, telangiectasia or immunodeficiency.

Table 2 Frequency of ATM IVS10-6T→G variant carriers
Full size table
Table 3 ATM IVS10-6T→G variant carriers: features of family history and BRCA1 and BRCA2 status
Full size table

BRCA1 and BRCA2 genetic testing results were subsequently available for all seven individuals. Two were found to harbour pathogenic truncating mutations in BRCA1 (both from families with a history of breast/ovarian cancer), and one a BRCA1 variant of uncertain significance (from a moderate-risk family). For both BRCA1 and ATM IVS10-6T→G carriers (probands 1 and 7 in Table 3), the histopathological features of the primary breast and ovarian cancers were consistent with the phenotype associated with BRCA1 tumours [26, 27]. The breast cancer from proband 1 was a grade III oestrogen receptor-negative and progesterone receptor-negative infiltrating ductal carcinoma with lymphocytic infiltrate and pushing margins. The ovarian cancer from proband 7 was a high-grade serous carcinoma with marked nuclear pleomorphism, together with a large cyst containing a serous cystadenoma and smaller cysts containing borderline malignancy. There were no remarkable histopathological features in the tumours from the other five cases, including proband 4 (with the BRCA1 unclassified variant). Thus, the ATM IVS10-6T→G allele was not apparently associated with a distinct tumour phenotype.

The frequency (1.4%) of the ATM IVS10-6T→G variant is somewhat greater than the reported frequency of 6 of 725 (0.8%) for a cohort of Australian women ascertained without a personal or family history of breast cancer [28]; however, this difference is not significantly different (P = 0.4). This frequency is also similar to control cases reported in Dutch, German and Austrian populations [21, 22, 29, 30], and a recent study in the USA [23], although minor differences might be due to population heterogeneity (Table 4). These results are consistent with the observation that the ATM IVS10-6T→G variant is an ancient mutation that might be widely distributed across Europe, the Middle East and western Asia [30]. In the study by Chenevix-Trench and colleagues, analysis was performed on non-BRCA1/BRCA2 cases containing more than three breast cancers per family, and families with male breast cancer(s) were excluded [19]. By restricting the analysis of our samples (GHSV, RMH, WH) to those with more than three breast or ovarian cancers within a family and no cases of male breast cancer, the frequency was 2 of 197 (1.0%), but both cases also carried pathogenic BRCA1 mutations. The basis for the increased frequency reported for kindreds in the Kathleen Cuningham Foundation Consortium for Research into Familial Breast Cancer is unclear, although a subsequent report now suggests a lower frequency at 6 of 385 (1.6%) [28].

Table 4 Studies on the ATM IVS10-6T→G variant in breast cancer
Full size table

It was not feasible to evaluate the co-segregation of the ATM IVS10-6T→G allele with the breast or ovarian cancers in each kindred in our cohort. In the two ATM IVS10-6T→G families reported by Chenevix-Trench and colleagues, carriers had an estimated disease penetrance of 78% (confidence interval 36–99%) to age 70 years, equivalent to a 26-fold increased risk [19]. However, it is noteworthy that the LOD (logarithm of odds) score of 1.18 for linkage with ATM fell well short of standard criteria for significance.

A recent study by Szabo and colleagues evaluated five ATM IVS10-6T→G families with non-BRCA1/BRCA2 breast cancer and concluded that the variant did not confer an increase in risk for breast cancer. That study also did not observe an increase in the frequency of the ATM IVS10-6T→G variant among BRCA1/BRCA2-positive families (0.5%) [21]. In our study, two of the seven ATM IVS10-6T→G heterozygotes harboured pathogenic BRCA1 mutations and a further proband carried a BRCA1 unclassified variant. This possible association is intriguing because ATM is known to phosphorylate BRCA1 [5, 6], and BRCA1 has recently been shown to be required for certain ATM functions, including the phosphorylation of substrates such as p53 and Chk2 that influence cell cycle arrest and apoptosis after DNA damage [31]. Further investigation of this variant as a low-penetrance modifier allele might be warranted.

In contrast, it has been argued that ATM IVS10-6T→G is a high-penetrance allele but our study has failed to support this view. Although this truncating variant has been postulated to act in a dominant-negative manner [19], the identification of a truncated product in heterozygotes has not been reported, suggesting that the product undergoes rapid degradation. It is known that the variant allele produces less than 10% of full-length ATM mRNA [29]; it therefore remains possible that heterozygous carriers simply display reduced ATM levels and that the residual ATM kinase activity is sufficient for maintaining normal cellular responses under most circumstances.

Conclusion

The frequency of the ATM IVS10-6T→G splice-site variant was similar in affected individuals from our clinic-based cases with a strong family history to that reported for individuals with no personal or family history of breast or ovarian cancer. Screening for the IVS10-6T→G variant in a familial cancer clinic setting is therefore unlikely to be of clinical significance. The relatively high prevalence of this variant among unaffected Europeans [30] further undermines its relevance in a clinical context. However, it remains plausible that the ATM IVS10-6T→G variant could function as a low-penetrance breast cancer risk-modifier under certain circumstances.