Several risk factors for the development of breast cancer, such as family history, reproductive factors and exposure to radiation, have been identified. Out of all breast cancers, 5 to 10% can be attributed to germline mutations in familial high-risk genes such as BRCA1 or BRCA2 that result in a lifetime breast cancer risk of about 45 to 65% 1. The penetrance varies between families, depending on 'risk modifiers' such as hormonal factors, mutation type and also exposure to radiation 234. Mutations in low-penetrance genes may account for a larger proportion (10 to 30%) of all breast cancers 5. The contribution of these genes might be explained by their role in the DNA-damage control pathway. For instance, ATM heterozygous carriers have a relative risk of breast cancer of 2.2 compared with the general population and a relative risk of 4.9 for those younger than 50 years of age 67. It has been shown that particular alterations in the ATM gene are associated with increased in vitro chromosomal sensitivity to radiation 8910. One particular sequence variant in CHEK2 (CHEK2*1100delC) has been implicated in a twofold increased risk of breast cancer, functioning as a low-penetrance breast cancer susceptibility allele 1112131415. Mutation frequencies reported in the literature indicate that about 10% of all women with breast cancer diagnosed before the age of 50 years have a germline mutation in BRCA1, BRCA2, ATM or CHEK2 1615.

Exposure to ionising radiation is a strong risk factor for breast cancer. A pooled analysis of eight cohorts by Preston and colleagues showed that the increased risk is directly proportional to the radiation dose received and inversely related to age at irradiation 16. For women exposed at the age of 25 years the excess relative risk per Gy was estimated to be 1.8. Radiation-induced DNA damage initiates a complex series of overlapping responses responsible for the maintenance of genome integrity. The increased incidence of breast cancer after exposure to ionising radiation might be restricted to a genetically defined radiosensitive subpopulation 17. Candidate genes (for example BRCA1, BRCA2, CHEK2, ATM, MDM2 and TP53) are implicated in the maintenance of genome integrity; their involvement in breast cancer susceptibility and their role in DNA-damage repair signalling make them excellent candidates as genes with a role in radiation-induced breast cancer 18.

The use of diagnostic X-rays has not been associated with increased risk of breast cancer in the general population, with the exception of frequent chest fluoroscopies in tuberculosis 1920. However, low-dose ionising radiation has recently been shown to increase the risk of breast cancer significantly among BRCA1 and BRCA2 mutation carriers 4. In a retrospective cohort study of 1,601 female BRCA1 and BRCA2 mutation carriers we found an association with reported chest X-ray exposure and significantly increased risk of breast cancer (hazard ratio 1.54). Furthermore, a recent publication also showed a strong association (odds ratio (OR) 3.21) between CHEK2*1100delC carrier status, breast cancer risk and a history of chest X-rays 21.

Women with breast cancer in general have a threefold to fourfold increased risk of developing a new primary cancer in the opposite breast 22. The increased risk may be explained by the same genetic and hormonal factors that caused the first breast cancer. In addition, radiotherapy for primary breast cancer may also contribute to the development of cancer in the contralateral breast. The contralateral breast can receive a dose of several Gy of leakage and scattered radiation during radiation treatment (RT) 2324. Several studies have shown that exposure of the contralateral breast to RT increases the risk of developing second primary breast cancer among young women 2325.

To evaluate the association between germline mutations in DNA-damage repair pathway (DDRP) genes (BRCA1, BRCA2, CHEK2 and ATM) and radiation-induced contralateral breast cancer (CBC), we conducted a case-only study. We assessed the germline mutation frequency in women who developed CBC, according to their history of RT for the first breast cancer, and assessed whether DDRP mutation carriers have an increased risk of radiation-associated breast cancer compared with that for non-carriers.

This study is, to the best of our knowledge, the first to suggest that women with a germline mutation in a gene involved in DDRP have an increased risk of developing radiation-associated CBC, compared with women who are not carrying such a mutation. We examined the mutation frequency of BRCA1, BRCA2, CHEK2 and ATM in women who had developed a second primary breast tumour at least 1 year after a first breast cancer diagnosed before the age of 50 years, with or without RT for their first breast cancer. Such a case-only design is particularly well suited to the assessment of gene-environment interactions, although the independent effects of the genes involved or of radiation cannot be determined from these data 2627. ORs from such a design can be interpreted as the risk of developing radiation-associated disease for carriers compared with non-carriers. We present evidence for an interaction between mutations in the DDRP genes and radiation for the first breast cancer in the development of CBC.

Women with breast cancer have in general a threefold to fourfold increased risk of developing a new primary cancer in the contralateral breast 22. The 15-year cumulative risk of developing CBC amounts to 10 to 13% 343536. BRCA1 and BRCA2 mutation carriers have an estimated CBC cumulative risk of 50 to 60% at the age of 70 years. The highest incidence is seen in the first 5 years after diagnosis of the primary breast cancer (12 to 33% per year), in contrast with 0.4 to 1% per year for breast cancer patients in general 3738. In a recent study, based on a consecutive series of breast cancer patients under 50 years of age, unselected for family history, we showed that CHEK2*1100delC carriers have a twofold increased risk of second breast cancer 39.

In the present study we found that mutation carriers have an additional excess risk, compared with non-carriers, of developing CBC during follow-up if they received RT for their first breast tumour. The increased risk due to RT exposure in carriers was mainly seen from 5 years after RT (Table 5). This is in accordance with the analysis of Boice and colleagues, who suggested previously that cancers resulting from an exposure to radiation would develop within predictable time windows; a latency period of at least 5 to 10 years was suggested for second primaries 32. Metcalf and colleagues have shown that RT actually protects against local recurrences and ipsilateral breast cancer among BRCA1 and BRCA2 carriers 36.

Age at time of exposure is known to be a major determinant of the risk of breast cancer 40. Our data also suggest that carriers in the younger age group at the time of exposure (less than 40 years old) showed the highest risk of developing radiation-associated CBC, although the difference from those above 40 years of age at RT was not significant (Table 4). Importantly, the mean and median ages at diagnosis of the first primary breast cancer were the same in the CBC patients with and without prior RT and in CBC patients with and without a germline mutation, resulting in the same median age at time of treatment in both groups (Table 1). Indeed, adjustment for age at first diagnosis did not affect our risk estimates.

The association between breast cancer risk and low-dose radiation has been a subject of debate. Recently we showed that diagnostic ionising radiation exposure from chest X-rays may be associated with a significantly increased breast cancer risk among women who carry BRCA1 or BRCA2 pathogenic germline mutations 4. Comparable findings were reported for CHEK2*1100delC carriers by Bernstein and colleagues 21. In the present study we included women exposed to higher doses from therapeutic RT, in whom the contralateral breast is exposed to about 1 to 10% of the total dose (namely 1 to 6 Gy), increasing the likelihood of detecting a gene-radiation interaction. The average (mean and median) maximum radiation dose (about 1.5 Gy) and the dose at the site of the tumour (about 1.8 Gy) measured in this study were found to be comparable between carriers and non-carriers. According to Preston and colleagues, the excess relative risk at 1 Gy is 2, implying that mutation carriers may have an excess relative risk of about 2.5 times this risk 16.

We increased the power of our analysis by investigating the effect of four breast cancer susceptibility genes together, considering them as a DNA repair gene group. These genes are all involved in DNA damage response triggered by damage induced by ionising radiation. An efficient response to DNA damage is essential for cellular life. The most detrimental form of damage is DNA double-strand breaks, which can be induced by ionising radiation and are lethal to the cell if not repaired. If repaired incorrectly, as a result of improper functioning of the repair machinery caused by mutations in the genes implicated, double-strand breaks can lead to carcinogenesis through translocation, inversions or deletions of genetic material. Although the relative impact of germline mutations in the separate genes studied in response to RT is unknown, all of these genes have been identified as crucial to the repair of RT-induced double-strand breaks. We therefore considered it justified to regard these genes together as a DNA repair gene group in our analyses. Although we screened for only a subset of all mutations in the four genes, there is no biological evidence that undetected mutations would have occurred more often in the non-RT group.

In a case-only study there must be independence between exposure (RT) and genotype (mutation) 26. By selecting the RT-exposed and non-RT-exposed groups of CBC patients we might have consistently selected for those clinico-pathological factors determining the use of radiotherapy. If mutation carriers were more likely to receive RT for primary breast cancer, for example because of the presence of a family history of the disease, alerting the physician to a potentially more aggressive course of the disease, selection bias might arise in our case-only study. However, this was not the case, mutation carriers did not receive RT more often. Information about a family history of breast cancer was obtained by a mailed questionnaire and from medical records. In our study population 53% of the carriers reported breast cancer in the family (33% with an affected first-degree relative), in contrast with 40% in the non-carriers (29% with an affected first-degree relative). From the data collected from medical records and risk factor questionnaires of all women, we concluded that DDRP mutation carriers had clinico-pathological features (such as size and stage) similar to those of 'sporadic' cases. Adjustment for each of these factors in logistic regression analysis did not materially affect the risk estimates for mutation status. In addition, the percentages of patients receiving chemotherapy were comparable for mutation carriers and non-carriers. Consequently, the genetic factors (the genetic status was not known at time of treatment) under study did not affect treatment choice, justifying our assumption of independence between radiation and a mutation carrier status, which is needed for a case-only study.

In addition, in a related project studying the effect of BRCA1, BRCA2 and CHEK2*1100delC germline mutations on survival and disease outcome in a large cohort of unselected breast cancer patients (diagnosis before 50 years of age), we found that the tumour characteristics of CHEK2*1100delC carriers did not differ significantly from those of non-carriers. Importantly, in that same study we also found that the proportion of breast cancer patients receiving radiotherapy did not significantly differ between the CHEK2*1100delC carriers and non-carriers 39.

Although the effect of pathogenic DDRP gene mutations on breast cancer risk seems evident, the role of most ATM missense mutations remains unclear. Generally, the idea is that there are two groups of ATM heterozygotes in the general population, each with different cancer risks: those with a wild-type and a truncating mutation allele and those with a wild-type and a missense mutation allele. Several epidemiological studies have shown that relatives of patients with ataxia-telangiectasia heterozygotic for an ATM germline mutation have an increased risk of breast cancer 6. Carriers of mutations predicted to encode a full-length ATM protein had cancer risks similar to those of people carrying truncating mutations, predicting that both type of mutation in these families are pathogenic for ataxia-telangiectasia as well as for breast cancer 641. Most studies try to predict the relevance of a particular mutation on the basis of co-segregation in breast cancer families, the location in a functional domain or interference with the splicing machinery 42. Nevertheless, only a few studies have presented the functional analyses necessary to assess the biological impact of unidentified variants found frequently in ATM 43. We found that ATM missense mutations, those detected in our study, do not by themselves contribute to an increased risk of breast cancer after exposure to radiation. Although not significant, there might be an increased risk of developing CBC after RT when patients have both a pathogenic DDRP and an ATM missense mutation (Table 3). This trend suggests that ATM missense mutations might have a risk-modifying effect after radiation exposure.

It has been hypothesised that a multigenic model might explain breast cancer susceptibility in a large part of the population. It has been shown that, although mutations in certain genes had marginal or no associations with risk when studied in isolation, they showed significant association when combined with variant alleles in other genes 44. It will be necessary to confirm the potential pathogenic-missense mutation (gene-gene) interaction findings of this study by a larger-scale study designed specifically to examine the joint effects of radiation exposure and genetic susceptibility of breast cancer risk (for example the Women's Environmental Cancer and radiation epidemiology ('WECARE') study 45).

Our data imply that there is a subgroup in the female population with increased susceptibility to radiation-induced breast cancer. The characterisation and identification of such a radiosensitive subgroup, if confirmed by others, will have implications for both diagnostic testing and the need for tailored treatment strategies. It provides a scientific basis for mutation analysis and subsequently, following the mutation analysis, an intensified follow-up protocol for mutation carriers. Identification of women susceptible to radiation damage will contribute to the risk/benefit assessment of radiation therapy versus alternative therapeutic options (such as breast-conserving surgery plus local radiotherapy or mastectomy).

CBC = contralateral breast cancer; CI = confidence interval; DDRP = DNA-damage repair pathway; OR = odds ratio; RT = radiation treatment.

The authors declare that they have no competing interests.

AB was involved in coordination of the study, performing the experiments and analysing results, and wrote the manuscript. LMB, AH and JU were involved in the mutation analysis experiments. AN was involved in patient selection, medical record and questionnaire abstraction and database management. FH participated in mutation identification management. MKS participated in the SPSS statistical and database analysis. JGMK participated in patient identification and selection. NSR participated in patient selection and provided radiation dosimetry. FEvL participated in the design and epidemiological analysis and LJV participated in the design and coordination of the study. All authors read and approved the final manuscript.