γ-Radiation-Induced Chromosomal Mutagen Sensitivity Is Associated with Breast Cancer Risk in African-American Women: Caffeine Modulates the Outcome of Mutagen Sensitivity Assay

Breast cancer is a disease of mixed etiology. The best documented and most common risk factors are related to endogenous and exogenous hormonal exposure. Although a positive family history is the strongest risk factor, familial cases caused by highly penetrant genes (BRCA1 and BRCA2) account for only ∼2% of all cases (1). Therefore, for most women, low penetrant susceptibility genes elevate their risk for breast cancer, including those for DNA repair (2, 3). There are both genotypic and phenotypic assays available to assess genetic susceptibility from low penetrant genes, and genotypic methods are more commonly used (3, 4). However, there remains a role for phenotypic assays because these methods represent complex genetic traits that can account for the net result of several genetic pathways and the cumulative effects of low-risk genetic variants whose effects may be difficult to see in single candidate gene studies (4).

The mutagen sensitivity assay (MSA) represents a complex genetic trait that measures chromosomal aberrations (i.e., chromatic breaks), and this assay is considered to be a marker of unrepaired DNA damage. Virtually every study of the MSA that has been published reports statistically significant risk estimates for increased sensitivity (5-22). Methods vary across laboratories depending on the mutagen used, e.g., γ-radiation (15), radiomimetic chemotherapeutic agent bleomycin (6), or benzo(a)pyrene-diol-epoxide (23, 24), and others. The results for different types of mutagens do not yield similar results in the same individuals, indicating that they are measuring different DNA repair capacities (24). For breast cancer risk, the MSA, or similar assays, have been applied only in small studies, but elevated chromosomal radiosensitivity has been reported (16-22, 25). Helzlsouer et al. studied 17 high-risk cancer-free women who had close relatives with breast cancer, 4 breast cancer cases, and 19 controls (26). They showed that the women at high-risk and the cases with breast cancer were five times likely to have suboptimal DNA repair compared with controls. Only 32% of the controls were sensitive, compared with 71% of high-risk women. In another study, 14 breast cancer cases, 19 first-degree family members, and 17 controls were tested, with identical results (19). They found that 35% of controls were sensitive versus 79% of family relatives. A third study examined 50 breast cancer cases and compared them to 111 controls (27). Mean breaks per cell were 109 (SD, 26.8) and 94 (SD, 13.6), respectively (P = 0.001). These results, however, still need further replication. Also, none has specifically addressed the relationship between MSA and breast cancer risk in African-American women. Although African-American women have a lower risk of breast cancer overall, they have a higher risk of presenting with higher stage disease, or tumors that are more aggressive by stage of disease (28-30). Although reasons for this include issues around access to health care. In fact, there is only one study that we could find that assessed African-Americans for MSA and cancer risk of any type, which was for lung cancer risk, and sensitivity was associated with a 3-fold risk (Caucasians were not compared with African-Americans (6)). In this study, we have focused exclusively on risk in African-American women.

A variation of the γ-radiation-induced MSA involves the introduction of radiosensitizers at a critical time point postirradiation. We conducted the MSA in the presence of caffeine because the effect(s) of caffeine on mutagen sensitivity is considered an important model for assessing mutagenesis (31). This modification allows for the separate consideration for the contribution of DNA repair and cell cycle control in response to mutagen exposure, where it is not known which contribution affects the number of chromosomal breaks (32). Caffeine decreases a naturally occurring shoulder on the survival curve for cultured lymphocytes following radiation exposure (radiation induces G₂ arrest) by decreasing time for G₂ and phosphorylation of G₂-M checkpoint proteins (33, 34). It is a known radiation sensitizer (35). Alterations in cell cycle checkpoints, which allow time for DNA repair, could also affect mutation rates (36). Thus, the use of caffeine may enhance the predictivity for DNA repair capacity in the MSA because it inhibits DNA repair time by eliminating G₂ arrest, and hence variability in the population would be due to differences in repair efficiency of the radiation damage without regard to cell cycle variables. We hypothesized that the difference in the number of breaks between radiation and radiation plus caffeine reflects the maximum dose which cannot be repaired without G₂ arrest. It should not be inferred or expressly concluded that this is a study of caffeine and cancer risk because the in vitro treatment is not analogous to human exposure; caffeine is not considered to be a mutagen or human carcinogen (37-39).

In this study, using a case-control study design, we have assessed if MSA is predictive of breast cancer risk in African-American women. Although there has been limited application of the MSA for breast cancer in general, there has been no prior study in African-American women. We also have investigated the outcomes of caffeine-modulated mutagen sensitivity to assess if reducing the time for DNA repair could result in a more predictive MSA, through elimination of “repair-time” variability, and thus directly assessing DNA repair capacity. To our knowledge, this is the first report on the evaluation of caffeine-modified mutagen sensitivity in women with breast cancer.

Approval for this study was obtained from the Institutional Review Boards of the NIH, the Georgetown University Medical Center, and Howard University. All cases and controls included in the study were African-American women.

All cases recruited in this study were African-American women born in the U.S., residing in the Washington, DC metro area and diagnosed with breast cancer at the Howard University Hospital (Washington, DC). After the initial identification of the cases from the biopsy report and the confirmation of diagnosis via pathology reports, consent was obtained from the surgeon. This was followed by contacting the patient by a formal letter and a follow-up telephone call to discuss the willingness to participate in the study and schedule an appointment. The participation rate was 70%. On the day of interview, a formal consent form was signed, anthropometric measurements were taken, phlebotomy was conducted, and a survey questionnaire was completed.

All women had a working telephone at home and were able to communicate in English. Severely ill or institutionalized women were excluded, as were those diagnosed with HIV or chronic hepatitis. Also excluded were women suffering from drug abuse or unable to give informed consent. Our study included a total of 61 women diagnosed with breast cancer (unilateral or bilateral). Women who had received chemotherapy or radiation treatment were not included in this study, as these treatments might affect mutagen sensitivity.

The controls for our study were randomly selected from the Washington, DC Voters Registration List obtained from the Board of Election. The prospective controls were contacted by a letter followed by telephonic call to discuss the study and willingness to participate. On the day of interview, the participants were requested to sign the consent form and complete a set of questionnaires, after which blood sample was taken. Eligibility criteria were the same as those for cases, except that the controls must have had no personal history of breast cancer.

Other exclusion criteria (for both cases and controls) were previous diagnosis of other cancer(s), radiation treatment for other illnesses, surgery, or intake of specific medications (antibiotics, steroids, and immunosuppressants) at least 2 months prior to drawing blood for MSA. Blood samples drawn from cases and controls were handled similarly and the laboratory personnel were blinded to the case's status.

Whole blood cultures were established from each subject within 24 hours of phlebotomy as described elsewhere (40). Briefly, 1 mL of fresh venous blood was inoculated into a T-25 flask with 9 mL of RPMI 1640, supplemented with 20% fetal bovine serum and phytohemagglutinin (112.5 μg/mL). These cultures were then incubated at 37°C for 67 hours, at which time they were exposed to 1 Gy of γ-radiation from a ¹³⁷Cs irradiator (JL Shepherd Mark I). The cultures were incubated for an additional 4 hours to allow time for DNA repair. Subsequently, colcemid (0.04 μg/mL) was added for 1 hour to arrest the cells in metaphase. The cells were then spun down and treated with a hypotonic solution of 0.06 mol/L potassium chloride for 25 minutes, spun down again, and the supernatant was discarded. The cells were fixed with methanol and glacial acetic acid in a 3:1 ratio. Slides were prepared by adding two to three drops of the fixed cells on clean cold slides to obtain metaphase spreads. Slides were air-dried for 24 to 48 hours and then stained with 4% Giemsa (Sigma Aldrich Corp., St. Louis, MO). For each subject, 50 well-spread metaphases were screened and the mean chromatid break per cell (MBPC) was assessed per individual. Breaks were defined as lesions in which the distance between the “broken” pieces was larger than the width of the chromatid; otherwise, they were considered as gaps. Gaps were not included in the present study. If both chromatids had a break, it was designated as a chromosomal break. Each chromatid break was counted as one break, whereas each chromosomal break was counted as two breaks. Cells with more than 12 chromatid breaks were excluded from this study.

Concurrently, fresh phytohemagglutinin-stimulated lymphocyte cultures were established as above, except that after 67 hours of culture and treatment with γ-radiation, caffeine (Sigma Chemicals) was added to a final concentration of 1 mg/mL and incubated for an additional 4 hours. The subsequent steps of colcemid arrest, harvesting, fixing, slide preparation, staining, and chromatid break assessment methods followed were the same as described above for MSA.

All analyses were conducted with SigmaStat version 3.0.1 (SPSS, Inc.). The mean and SDs of MBPC was determined for cases and control groups under separate categories, i.e., 1 Gy and 1 Gy + C. Nonparametric t test, i.e., Mann-Whitney test was used to compare cases and controls. A significance level of α = 0.05 was used throughout. The cutoff MBPC value for defining radiation sensitivity in cases was set at 75th percentile value of the MBPC for control group. To analyze the correlation between radiation-treated (1 Gy) and radiation plus caffeine-treated (1 Gy + C) scores for each group, the Spearman's correlation coefficient was used. The relationships between variables were analyzed using Pearson product moment correlation. For estimation of cancer risk adjusting for influence of other variables, unconditional logistic regression was done. Models were adjusted for age, menopausal status, age at menarche, family history of breast cancer, and alcohol use. All P values were based on two-sided tests.

Demographic characteristics by case-control status are given in Table 1. The mean ages of the cases and controls were 59.4 and 52.7 years, respectively. Family history of breast cancer was present in 38% of the cases compared with 22% in controls; otherwise there was no significant difference.

The MBPC for the patients with breast cancer (0.34 ± 0.15) was significantly greater than that for the control group (0.24 ± 0.12; P < 0.0001; Fig. 1). As evident from the graph, the distribution of cases is skewed towards the high sensitivity end of the graph. To determine the unadjusted odds for breast cancer risk based on mutagen sensitivity, cases and controls were dichotomized at the 75th percentile MBPC of the controls (0.29 chromatid breaks per cell). Analysis based on this value revealed that the sensitive individuals had an overall odds ratio of 4.5 (95% confidence intervals, 2.2-9.1).

In order to confirm whether variables of interest including age, menopausal status, family history of cancer, family history of breast cancer, smoking, and alcohol intake had an effect on the MSA and chromatid breaks per cell, we determined the Pearson product moment correlation. This analysis showed that age was not statistically correlated with chromatid breaks per cell (r = 0.678, P = 0.414) and neither were menopausal status (r = 0.0313, P = 0.707), family history of cancer (r = 0.0408, P = 0.624), alcohol intake (r = 0.0658, P = 0.428), and smoking (r = −0.0646, P = 0.437).

Breast cancer risk was associated with age (r = 0.276, P < 0.001), menopausal status (r = 0.243, P < 0.05), and family history of breast cancer (r = 0.17, P < 0.05). All other variables including family history of other cancer, smoking, and alcohol failed to show any association with cancer status. MSA in unconditional logistic regression, adjusted for age, menopausal status, family history of breast cancer, and alcohol drinking showed a strong association with case-control status (r = 5.799, P < 0.001), in which the odds ratio of chromatid breaks per cell was 3.3 (0.14-74.0) for cases compared with controls (P < 0.001).

Data for the caffeine-modulated assay was available for 59 cases and 80 controls. Unlike the analysis for the MSA, the caffeine-modulated assay did not reveal statistically significant differences for cases (1.6 ± 0.9) versus controls (1.5 ± 0.8; P = 0.9; Fig. 2).

We analyzed if the MSA and caffeine-modulated assay MBPC were correlated in cases and controls; that is, if a woman was sensitive to MSA, would she also be sensitive for caffeine-modulated assay? For this correlation analysis, we included only those cases and controls for which both assays were available. It was found that there was a statistically significant correlation for controls (n = 79; Spearman r = 0.43; P < 0.01; Fig. 3), but not for cases (n = 58; Spearman r = 0.07; P = 0.62; Fig. 4). Table 2 gives a summary of the statistics of mutagen sensitivity with and without caffeine by case-control status. Using the cutoff of the 75th percentile in controls at 0.29 MBPC, 87% of cases were above this cutoff, compared with only 25% of controls. Thus, the 75th percentile is both sensitive and specific.

This study indicates that MSA is an independent risk factor for breast cancer in African-American women. This is consistent with findings for other cancers in Whites and African-Americans, as shown in Table 3. Given that twin studies indicate that the MSA is a heritable trait (41), and numerous studies that show positive associations, the results are not likely confounded by case status, age, or unknown variables. In this study, conducting caffeine-modulated assay did not improve the risk prediction for the MSA, but did indicate that the phenotypic MSA is likely related to both DNA repair and cell cycle G₂-M arrest. Although this epidemiologic study is relatively small, it had sufficient power to show the utility of the MSA for breast cancer risk, as has been done in studies of similar or slightly larger size (22, 25, 42, 43). An odds ratio of 4.5 in our study is higher than that of an earlier report wherein radiation-sensitive individuals had an odds ratio of 2.83 for developing breast cancer (25). However, this study was conducted on a fewer number of cases and controls (n = 44) and the subjects were likely not African-American.

Phenotypes, such as MSA, represent combined genotypes in a pathway, in this case, nonhomologous repair capacity and cell cycle control. The association for phenotypic studies and cancer risk often show main effects in small studies, such as for the MSA shown in Table 3. Single candidate gene studies are challenged because the effect of a single gene variant within a pathway might not be sufficiently penetrant to show the increased risk, even in large studies, especially if the frequency of the variant is small. This would also hold true for haplotypes. However, phenotypes have the capacity to show the result of combined low-penetrant genes or haplotypes of both low and high frequency, and there is some evidence that DNA repair enzyme polymorphisms affect chromosomal aberrations and MSA (44-46). Thus, the penetrance of phenotypes could be higher and observable in small studies. Separately, phenotypic markers may be better cancer risk biomarkers compared with genetic markers because the phenotype would account for differences in genetic diversity among different ethnic and racial groups. Although specific genetic traits might vary widely among different ethnic and racial groups, how these would combine to affect DNA would be assessed jointly in the MSA.

There are some limitations to this study, including its small size and possible chance effect. Also, we could not stratify where the MSA could have different risks according to menopausal status. Our cases and controls that provided a blood sample were not well matched, but logistic regression with adjustment for age still indicated the increased risk, and age was not correlated with MSA. Another potential confounder might be case status, and it would be more informative to show the risk for the MSA in a cohort study. However, this assay is technically difficult and hence not feasible for cohort studies, but provides evidence for identifying associations with, and application of, other less technically difficult assays (e.g., the COMET assay). One study suggests that the MSA might be confounded by case status because persons with higher stage head and neck cancers were more likely to be sensitive (47), but the reports are conflicting (9). We believe that our results are not confounded by case status because there is no clear a priori reason to believe that the cultured B lymphocytes would behave differently from cases and controls, and that there is consistency across cancer sites. Another argument against case bias is that age, gender, smoking, and micronutrient consumption were not associated with bleomycin or benzo(a)pyrene sensitivity (24, 48, 49). Separately, there is no reason to suspect laboratory variability because the laboratory staff was blinded to case status, and samples were analyzed concurrently. A replicate analysis in our laboratory provided acceptable quality control results (see below),³

The evidence that MSA is a heritable trait comes from studies of high risk breast cancer families (19, 26, 50, 51) and twin studies (41), demonstrating heritability in noncancer subjects. Specifically, Cloos et al., tested bleomycin MSA in 135 healthy volunteers from 53 pedigrees, including monozygotic and dizygotic twins (41). There was no evidence for shared family environments, but 75% of the variance was accounted for by heredity. The correlation coefficient in monozygotic twins was 0.79 (0.65-0.88), whereas for dizygotes was 0.42 (0.00-0.71).

In this study, we chose γ-radiation as the mutagen for the MSA. This was done for several reasons, and because most prior breast cancer studies also used this mutagen (19, 26, 27). Also, γ-radiation is a direct DNA-damaging agent that is not dependent on cell penetration, metabolism, or clearance. In vitro breast cell culture studies of immortalized normal mammary epithelial cells show similar chromatid breaks after G₂ phase irradiation (52). BRCA1 and BRCA2 knockout mice are more sensitive to radiation and have deficient repair for DNA damage (53), and radiation down-regulates BRCA1 (54). With an intact p53, ionizing radiation affects the G₁-S checkpoint and probably the G₂-M checkpoint as well (55). All of these reasons support the hypothesis that the MSA is a good biomarker for decreased DNA repair capacity for different repair pathways, and hence, would also be associated with genetic polymorphisms of different types of DNA repair genes.

Compared with previous reports for γ-radiation-induced MSA (Table 4), the MBPC for MSA from our study was relatively lower. Such differences could arise from a number of causes including differences in study population, radiation dosages, irradiator calibrations, postradiation incubation time, experimental handling, and slide-reading criteria such as inclusion/exclusion of gaps. Significant differences in scoring of breaks and gaps can exist even among experienced cytogeneticists (56). However, provided that the analyst maintains a constant criteria and the study is “blinded,” the results from a given analyst would be valid. In this investigation, cell culture, harvesting, slide preparation, staining, and slide scoring was done by a single analyst (T.G. Natarajan) for the entire period of the study. Prior to performing this study, assay reproducibility was tested in our laboratory with samples from five donors and repeated five times each at intervals of 2 to 4 weeks (Wilks' Lambda P = 0.44).³

The addition of caffeine to the cultures, a priori were believed to have improved the risk estimated, but this did not happen. The differential effect in cases and controls suggest that there may be a difference by case status for G₂M arrest, but that the phenotype did not predict breast cancer risk. There is another evidence that G₂M arrest, specifically mediated by cyclin-dependent kinase 1 / cyclin-B activity, affects chromosomal radiosensitivity (20). One reason for this may be that cells, in the presence of caffeine, were too radiosensitive, and so this would be similar to conducting MSA at much higher radiation doses. This might result in the loss of specificity for the assay.

Our study shows that the MSA may be a clinically useful biomarker of breast cancer risk, pending replication in other studies. It represents a complex phenotype of both DNA repair and cell cycle control. To date, this complex phenotype remains a stronger predictor of sporadic breast cancer risk than any known individual genotypes. The MSA is probably a complex genotype representing both DNA repair and cell cycle control. This study also contributes to other existing data on breast cancer risk in African-American women.

We thank Dr. Theodore Puck for his many insightful discussions.