Are ATM Mutations 7271T→G and IVS10-6T→G Really High-Risk Breast Cancer-Susceptibility Alleles?

Mutations of the high-risk breast cancer-susceptibility genes BRCA1 and BRCA2 (MIM 113705 and 600185) account for less than half of the families with breast cancer that include many cases of early onset breast cancer and only about one-quarter of clinically ascertained families (1, 2). It is thus likely that other breast cancer-susceptibility genes exist. Biallelic mutations of the ATM gene cause the neurological disorder ataxia-telangiectasia (A-T; MIM 208900). The ATM gene has been considered a candidate breast cancer-susceptibility gene because the observation of an increased breast cancer incidence in otherwise healthy female relatives of patients with A-T (3). Notwithstanding a decade of intensive research, controversy still exists about ATM-related breast cancer risks, with estimates ranging from no increase to a 13-fold increase in risk (4, 5, 6). A recent study on Australian families with breast cancer has renewed this debate by suggesting that two A-T-related mutations of the ATM gene (5, 6, 7) confer a high breast cancer risk (8). According to this study, the mutations ATM 7271T→G (also known as T7271G) and ATM IVS10–6T→G confer cumulative breast cancer risks of 55 and 78% by 70 years of age (8) and would thus compare with mutations of the high-risk breast cancer genes BRCA1 and BRCA2. We and others have shown that high-risk women often opt for genetic testing for BRCA1 and BRCA2 and, if a mutation is identified, proceed with risk-reducing interventions, including prophylactic bilateral mastectomy (9, 10). The clinical impact of the reported findings on the two ATM mutations (8) could thus be considerable. We therefore sought to replicate these findings.

We ascertained 961 families with breast cancer that did not carry a pathogenic mutation of the BRCA1 or BRCA2 genes through an international collaborative effort by five centers (Table 1). All families were originally referred to clinics for medical care and were later selected for research purposes for including at least two cases of invasive breast cancer in first- or second-degree relatives, with at least one of them diagnosed before age 60 years. The families were classified according to the following clinical criteria: (a) three or more cases of breast cancer diagnosed before age 60 years but no ovarian cancer or male breast cancer (high-risk families, representing 44% of all families); (b) one or two cases of breast cancer diagnosed before age 60 years but no ovarian cancer or male breast cancer (moderate-risk families, representing 45%); and (c) also including cases of ovarian cancer and/or male breast cancer (high-risk families; representing 11%). On the basis of the family history, the odds of identifying a BRCA1 or BRCA2 mutation are substantial for the high-risk families (chances of 25–80%) and lower for the moderate-risk families (chances of <25%; Ref. 11).⁹ We also ascertained a series of 211 families in which a pathogenic mutation of the BRCA1 or BRCA2 gene had been identified. Informed consents to screen for breast cancer-susceptibility genes were obtained from all individuals that participated in this study. Population controls were ascertained in the Netherlands (12, 13) and Austria (Table 1).

The ATM 7271T→G mutation was detected by a PCR-based allele-specific oligonucleotide hybridization assay (Ref. 12; Rotterdam, the Netherlands), denaturing high-performance liquid chromatography (Lyon, France, and Vienna, Austria), denaturing gradient gel electrophoresis (Amsterdam, the Netherlands), or fluorescence-based direct sequencing (Sainte-Foy, Quebec, Canada). The ATM IVS10–6T→G mutation was detected by a mutation-specific RsaI restriction endonuclease assay (Ref. 13; Rotterdam, Lyon, and Amsterdam), denaturing high-performance liquid chromatography (Vienna) or fluorescence-based direct sequencing (Sainte-Foy). All mutant samples were confirmed by direct sequencing of an independently amplified template. For each family, at least the index case was screened for the ATM mutations, defined as the youngest case with invasive breast cancer in the family from whom DNA was available. All index cases had been screened for mutations of the BRCA1 and BRCA2 genes by extensive analysis of the complete coding sequence and splice junctions of both genes, using a variety of techniques (12, 14, 15).

Descriptive statistics were used to determine the frequencies of index cases and control individuals that carried either ATM mutation. Population specific and pooled odds ratios and 95% confidence intervals were calculated, using the Mantel-Haenszel estimator to allow for differences in population frequencies. For the ATM mutation-positive families with non-BRCA1/BRCA2 breast cancer from which multiple individuals were tested, we calculated the probability of causality using the Bayesian method of Petersen et al. (16), assuming a prior probability of causality of 0.5 to be comparable with the analysis of Chenevix-Trench et al. (8), and extended the method to incorporate general models of genotype, phenotype (including unaffected individuals), and pedigree structure using a modified version of the program LINKAGE (17). All statistical tests were two sided.

The ATM 7271T→G mutation was not detected in any of 1504 tested individuals from 840 families with non-BRCA1/BRCA2 breast cancer (1025 cases tested), 204 BRCA1/BRCA2-positive families, and 275 control individuals, thus precluding any assessment of the breast cancer risk conferred by this mutation in our series (Table 1). Thus far, however, this mutation has only been detected in a single Australian family with breast cancer (8), suggesting that the mutation bears no clinical significance.

We detected the ATM IVS10–6T→G mutation in 8 of 961 tested families with non-BRCA1/BRCA2 breast cancer (1287 cases tested) and in 1 of the 211 BRCA1/BRCA2-positive families (Table 1). Two of these 9 families were ascertained through the University of Vienna and the remaining through Dutch centers. The BRCA1-positive family included 4 cases with ovarian cancer but no cases with breast cancer. The index case, diagnosed with ovarian cancer (age 62 years), also carried the BRCA1 2138delA mutation. The ATM IVS10–6T→G mutation frequency among Dutch and Austrian families with non-BRCA1/BRCA2 breast cancer was similar to its frequency among population-matched control series (Dutch series: 1.0 versus 0.7%; odds ratio = 1.46; 95% confidence interval = 0.36–5.87; Austrian series: 2.3 versus 1.1%; odds ratio = 2.12; 95% confidence interval = 0.19–23.78; Table 1). There was no significant frequency variation between the Dutch and Austrian series of families with non-BRCA1/BRCA2 breast cancer, nor between the respective control series. The pooled Mantel-Haenszel odds ratio for the Dutch and Austrian series was 1.60 (95% confidence interval = 0.48–5.35; P = 0.44), thus providing no evidence for an increased breast cancer risk conferred by ATM IVS10–6T→G. The ATM IVS10–6T→G mutation frequency was even somewhat lower among the high-risk families than the moderate-risk families, additionally negating that this mutation is associated with a high breast cancer risk (0.4 versus 1.4% but neither significantly different from the controls, Table 1). Our findings are consistent with a German study (7), where similar carrier frequencies of ATM IVS10–6T→G were observed among unselected cases with breast cancer (3 of 500, 0.7%) and control individuals (7 of 1000, 0.7%).

To further investigate the causality associated with the ATM IVS10–6T→G mutation, we examined statistically its pattern of cosegregation in the 5 families with non-BRCA1/BRCA2 breast cancer for which multiple individuals were tested (Fig. 1). Three cases with invasive breast cancer, a single case with lobular carcinoma in situ (LCIS), 5 unaffected women and 4 men were tested in addition to the index cases. Of the 3 affected women, 1 carried the ATM IVS10–6T→G mutation (family NKI-F423), whereas the other 2 did not (families UV-M27 and EMC-10098, respectively). The index case from family NKI-F117 was diagnosed at age 52 years with invasive lobular carcinoma with a LCIS component and carried the ATM IVS10–6T→G mutation, whereas her sister was diagnosed with LCIS at age 47 years and did not carry the mutation. Given the increased familial risk reported for LCIS, this observation also argues against the ATM IVS10–6T→G mutation being causal in this family (18). Of the 5 additionally typed unaffected women, 2 were shown to be carriers at ages 73 and 59 years, whereas the others were noncarriers at ages 62, 57, and 25 years. Using the hazard ratio of 26 estimated for ATM IVS10–6T→G (8) and age-specific incidence rates in the Netherlands and Austria, the overall evidence for or against this level of risk in these 5 families was assessed, compared with the hypothesis that this mutation is not associated with breast cancer. On the basis of this analysis, the overall posterior probability of causality for these 5 families is 0.008 if the case of LCIS from family NKI-F117 is classified as unaffected and 0.0004 if she is classified as affected with breast cancer. Thus, these 5 families are 125-2500 times more likely under the hypothesis of noncausality for this mutation, again suggesting that the ATM IVS10–6T→G mutation does not confer a significant breast cancer risk.

Our data thus refute those of Chenevix-Trench et al. (8) who proposed that the ATM IVS10–6T→G and 7271T→G mutations are high-risk breast cancer-susceptibility alleles. They based their estimates of the breast cancer risks conferred by these two mutations on only 2 and 1 single family, respectively, together including 14 cases with breast cancer. The total likelihood of disequilibrium score for linkage of breast cancer to the ATM locus from these three families was 1.18 (odds of 15:1 in favor of linkage), which does not meet conventional criteria for significant linkage. Given the relatively little linkage information/family (likelihood of disequilibrium scores of 0.14, 0.64, and 0.40), precise estimates of the breast cancer risks conferred by the two mutations could not be derived from their dataset, and hence, their Bayes factors should be viewed with caution. Combining the Bayes factors reported for the two Australian ATM IVS10–6T→G-positive families (8) with those of the 5 families in this study gives total Bayes factors of 0.04 (LCIS case classified as unaffected) and 0.0025 (LCIS case classified as affected). These results imply overall odds of 25:1 and 400:1 against causality, respectively. On the basis of the published frequency data among cases with breast cancer and control individuals, as well as our data reported here, a much lower prior probability of causality seems justified, which results in even lower posterior probabilities of causality. The expectation that many of the breast cancer-susceptibility alleles yet to be identified will confer low breast cancer risks (2) underlines the need for stringent thresholds of statistical significance, large sample sizes, and independent replication before results should be considered convincing (19, 20).

In summary, our results do not support an increased breast cancer risk for the ATM IVS10–6T→G mutation, although a slightly increased risk cannot be formally excluded. Neither the ATM IVS10–6T→G mutation nor the ATM 7271T→G mutation is likely to have a substantial contribution to familial breast cancer. No evidence currently exists that any mutation of the ATM gene confers a high risk of breast cancer (3, 4, 5, 6, 7, 8, 13, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36). In contrast to others (37, 38), we believe that carrier screening in clinical settings for the purpose of breast cancer risk assessment is as yet not indicated for any ATM allele.

Fergus Couch, Mayo Clinic, Rochester, MN; Dominique Stoppa-Lyonnet, Institute Curie, Paris, France; Maria de los Angeles Rios, National Institute of Oncology and Radiobiology, Havana, Cuba; Ana Osario and Javier Benitez, Centro Nacional de Investigaciones Oncologicas, Madrid, Spain; Trinidad Caldes, San Carlos Hospital, Madrid, Spain; Olga Sinilnikova, International Agency for Research on Cancer, Lyon, France; Henry Lynch, Creighton University, Omaha, NE; Gilbert Lenoir, Institute Goustave Roussy, Villejuif, France; Steven Narod and Cathy Phelan, Toronto Women’s Hospital, Toronto, Ontario, Canada; Elaine Ostrander, Fred Hutchinson Cancer Research Center, Seattle, WA; Hoda Anton-Culver, University of California, Irvine, CA; Jan Lubinski, Hereditary Cancer Centre, Szczecin, Poland; and Lenka Foretova, Masaryk Memorial Cancer Institute, Brno, Czech Republic.

Daniela Seminara, National Cancer Institute, Bethesda, MD; Saundra Buys, Huntsman Cancer Institute, Salt Lake City, UT; Irene Andrulis, Mount Sinai Hospital/Cancer Care Ontario, Toronto, Ontario, Canada; Dee West, Northern California Cancer Registry, Culver City, CA; John Hopper, University of Melbourne, Melbourne, Australia; Mary Daly, Fox Chase Cancer Institute, Philadelphia, PA; and Ruby Seine, Columbia University, New York, NY.

Carina Bartels, Renate van den Bos, Ellen Crepin, Bert van Geel, Dicky Halley, Conny van der Meer, Marian Menke-Pluymers, Caroline Seynaeve, Anja de Snoo, Madeleine Tilanus-Linthorst, Margreethe van Vliet, and Anja Wagner, The Rotterdam Family Cancer Clinic, Erasmus MC, Rotterdam, the Netherlands.

Christi J. van Asperen, Leiden University Medical Center, Leiden, the Netherlands; Frans B. Hogervorst, The Netherlands Cancer Institute, Amsterdam, the Netherlands; Nicoline Hoogerbrugge and Marjolein J. Ligtenberg, University Hospital, Nijmegen, the Netherlands; and Fred H. Menko, Free University of Amsterdam, and The Netherlands Cancer Institute, Amsterdam, the Netherlands.

Peter J. Oefner, Adriane Roxas, Tierney L. Wayne, and Kristine M. Yu, Stanford Genome Technology Center, Palo Alto, CA; and Thomas Bachrich, Daniela Mühr, and Regina Kroiβ, University of Vienna, Vienna, Austria.

Paul Bessette, Centre Hospitalier de l’Université Sherbrooke, Hôpital Fleurimont, Sherbrooke, Quebec, Canada; Peter John Bridge, Alberta Children’s Hospital, Calgary, Alberta, Canada; Jocelyne Chiquette-Gagnon, Centre Hospitalier Affilé de l’Université de Quebec - Hôpital du St-Sacrement, Québec, Laval, Canada; Rachel Laframboise, Centre Hospitalier de l’Université Quebec - Centre Hospitalier de l’Université de Laval, Ste-Foy, Québec, Canada; Jean Lépine, Centre hospitalier régional de Rimouski, Rimouski, Quebec, Canada; Bernard Lespérance, Hôpital du Sacré-Coeur de Montréal, Montréal, Quebec, Canada; Marie Plante, Hôtel-Dieu de Québec, Laval, Québec, Canada; Louise Provencher, Centre Hospitalier Affilé de l’Université de Quebec - Hôpital du St-Sacrement, Laval, Québec, Canada; Roxane Pichette, Hôpital du Sacré-Coeur de Montréal, Montréal, Quebec, Canada; and Patricia Voyer, Carrefour de Santé de Jonquière, Québec, Canada.

We thank all individuals and families who participated in this study. We also thank Dr. Georgia Chenevix-Trench for positive control samples of the ATM variants.