Background
The genetic determinants of breast cancer are under intensive study. Some women with a strong family history of breast cancer inherit
BRCA1 or
BRCA2 mutations, which have a variable penetrance for breast cancer, between 40 to 66% [
1], suggesting that additional factors contribute to cancer risk among
BRCA1 and
BRCA2 carriers. For sporadic cancers, however, many low-penetrant single-nucleotide polymorphisms (SNPs) have been investigated in pathways ranging from growth factor signaling to DNA repair. Yet, it has been difficult to find consistency across study results [
2‐
4], due to differences in study populations, sample sizes and study designs [
5]. However, studies of high risk populations generally help uncover the molecular mechanisms of a disease and provide guidance and direction for studies of sporadic disease. While
BRCA1 and
BRCA2 mutations are highly penetrant [
1], resulting in higher risk for breast cancer, both of these genes are also highly polymorphic. Moreover, several of their variants result in amino acid changes which could ultimately change the structure and function of the genes. It is therefore plausible that the combination of genetic changes in these genes or in genes in their pathway may at least contribute to the disease or the mechanisms associated with the disease in the general population. Another approach to enhancing the chances of identifying true positive associations is to conduct studies based on
a priori hypotheses, for example by studying SNPs known to affect protein functions or levels. Thus, genotype-phenotype associations are needed to reduce the chances of false positive associations in breast cancer risk studies.
The mutagen sensitivity assay (MSA) provides a phenotypic marker of DNA repair capacity and genomic stress response, which has been reported as a heritable trait that affects both familial and sporadic breast cancer risk [
6‐
9]. This assay measures the number of chromosomal breaks in cultured lymphocytes following exposure to DNA damaging agents. Different mutagens have been used, but gamma radiation has been the most widely utilized in breast cancer studies, because it is a direct DNA damaging agent whose effects are not dependent on cell penetration, metabolism, or clearance [
10]. For example, using this assay, DNA from women in high-risk families and sporadic breast cancer cases exhibits about a 2-fold increase in the mean number of breaks per cell compared to DNA from women without cancer from low-risk families [
6,
7,
11]. Thus, mutagen sensitivity may be a biomarker for DNA repair capacity and may specifically reflect differences in an individual's ability to repair DNA through the pathway of interest, homologous repair.
BRCA1, a nuclear protein that contains 24 exons (NC_000017.10), has a role in sensing DNA damage and cell cycle checkpoint control. It has been shown that
BRCA1-deficient cells have DNA repair defects partially rescued by introducing exogenous wild-type
BRCA1 [
12], and that
BRCA1 is a trigger of homology-directed DNA repair [
12‐
14]. Many
BRCA1 polymorphisms with allele frequencies >5% in Caucasians have been identified; however, only six of these (Q356R, D693N, P871L, E1038G, K1183R, and S1613G) result in amino acid changes (BIC-Breast Cancer Information Core;
http://research.nhgri.nih.gov/bic/). These polymorphisms, with the exception of Q356R and D693N, are in significant linkage disequilibrium and are inherited as part of a shared haplotype [
15]. Some studies have associated these SNPs with both familial and sporadic breast cancer risk, although there is a lack of consistency [
16‐
24].
BRCA1 haplotypes have received some attention and haplotype-risk analysis has been done in several small and large case-control studies, but no associations between risk and haplotypes have been found [
21,
23,
24]. Herein, we have chosen to study the
BRCA1 polymorphisms Q356R, D693N, and E1038G because these SNPs could be functional variants associated with risk [
25] and are in linkage disequilibrium with other SNPs of interest.
Rad51 was chosen for investigation because of their interactions with BRCA1 during homologous recombinational (HR) DNA repair [
26]. Rad51 (RecA homolog, E. coli; NC_000015.9) has 10 exons that code for a 339 amino acid protein which forms a helical nucleoprotein filament on DNA [
27]. Several studies of
Rad51 among
BRCA1 mutation carriers have found positive associations with cancer risk [
28‐
32]. Cells deficient in BRCA1 are also defective in Rad51 irradiation-induced foci formation [
26,
33]. Experimental studies show that the loss of Rad51 may drive genetic instability, chromosomal aberrations, and carcinogenesis by facilitating an accumulation of genetic changes [
34‐
36]. Rad51 is over-expressed in a
BRCA1 mutant cell line and rescues cells from apoptosis [
37]. Studies have reported that the
Rad51 5'UTR variant 135C allele (rs1801320) was associated with a decreased risk of breast cancer in
BRCA1 5382insC mutation carriers [
29] and other mutation carriers [
32], while no association was found in a case-control study of sporadic breast cancer [
38,
39]. Antoniou et al. reported that the SNP modified breast cancer risk among
BRCA2 mutation carriers and
BRCA1 loss-of-function mutation carriers [
40]. Although the functional consequences of the 135G>C polymorphism is unknown, it is speculated that because it alters a CpG island in the promoter, it may regulate expression and affect mRNA levels [
40,
41]. Additionally, there is some evidence of an association between this variant and decreased Rad51 protein expression in
BRCA1/2 mutation carriers [
42]. Although, there have been some reports of
Rad51 haplotypes associated risk in high-risk families [
31,
43,
44] and with sporadic breast cancer risk [
31,
43,
44], these haplotypes are composed only of SNPs in the Rad51 putative promoter, introns, or the 3' un-translated region. To date, there are no reported Rad51 haplotypes composed of SNPs in the coding region, indicating the coding region is well conserved [
45].
In order to identify SNPs and haplotypes in BRCA1and Rad51 that might affect familial and sporadic breast cancer risk, we conducted a study of genotype-phenotype relationships. First, we compared and validated the mutagen sensitivity assay in Epstein Barr Virus (EBV)-immortalized lymphocyte cell lines, comparing the lymphoblast results to freshly cultured lymphocytes from whole blood. We, then, used the MSA to study associations between genotypes and haplotypes as they relate to DNA-repair capacity in EBV-immortalized lymphocytes from 138 women with known BRCA1 mutations. We then applied these results to a population-based case-control study of breast cancer.
Discussion
In this study, we found an association between the MSA and breast cancer, and some genotype-phenotype relationships, in subjects from high risk breast cancer families. While there was no overall association for the MSA with BRCA1 Q356R, D693N, and E1038G genotypes in unrelated individuals, associations were found among family members using the polygenic model where the E1038G and Q356R BRCA1 SNPs were significantly associated with MBPC. Furthermore, the rare CTC (356R, 693N, and 1038G) haplotype also was found to be associated with the MSA. As for Rad51, those with the 5'UTR 135C SNP had statistically significantly higher MBPC than those who had the wild-type allele. When stratified by 185delAG or 5382insC mutation carriers, only 185delAG carriers with the Rad51 5'UTR 135C allele were marginally associated with higher MBPC. On the other hand, the Rad51 SNP was not associated with risk in a population-based study of sporadic breast cancer. These data indicate that mutagen sensitivity, and therefore, DNA repair capacity, might be a useful biomarker for determining penetrance among women with BRCA1 mutations, and that the mutagen sensitivity phenotype is partially explained by genetic variants in BRCA1 and Rad51.
The MSA, as a phenotypic assay, has been generally applied to freshly collected peripheral blood lymphocytes. Because we aimed to identify SNPs in genotype-phenotype relationships from EBV-immortalized studies and relate them to breast cancer risk, we first assessed the correlation of MSA results between EBV-immortalized and fresh peripheral blood lymphocytes. Since EBV-immortalized lymphoblasts are primarily derived from B-lymphocytes, while PHA-stimulated whole blood cultures yield primarily T-lymphocytes, it was possible that these lymphocyte subpopulations would yield quantitative differences in MBPC, and that the classification of women by high and low MBPC could be different. In this report, we demonstrate that the results for both assays were statistically related and quantitatively similar.
Mutagen sensitivity is regarded as a heritable trait as reviewed by Wu et al [
63]. Given that our sample included several families with women of differing relationships, the correlation of MBPC between family members was calculated and the sibling (sister:sister) correlation was found to be consistent with previous reports of mutagen sensitivity in dizygotic twins; the correlation coefficient for sisters in this study (r = 0.33), albeit not statistically significant due to a small sample size, was very similar to the study by Wu et al. (r = 0.27) [
64].
In this study, MBPC was significantly higher in the group of affected cases compared to women without cancer when the variance-component model developed for family-based association was applied [
56‐
58]. However, when related cases were removed from the analysis, the OR remained elevated but was not statistically significant. It may be that including related cases biased the results because the MSA is a heritable trait (although only with a correlation of 0.33), or that removing the related subjects resulted in lower statistical power due to a smaller sample size. Although the study has a small number of families and the pedigrees are sparse, the variance-component model is considered more informative than the unrelated family member analysis because of its ability to simultaneously estimate residual and multi-factorial (polygenic, familial, marital or sibling) variance components. In addition this approach uses the quantitative trait as is without dichotomizing, thus possibly increasing the power to correctly detect allelic association. Also, the method combines the original association method by George and Elston, with the pedigree TDT-type analysis in such a way as to maximize power [
58,
65]. Although familial components can be incorporated into the equation, the random polygenic effect was the only variance component included when analyzing the two populations. Moreover, when unrelated affecteds were compared to true negatives (unrelated unaffected women without the BRCA1 mutation), the difference was borderline statistically significant (p = 0.07), further supporting a true relationship. Our results are consistent with other studies showing decreased DNA repair capacity in both comparisons of breast cancer cases to controls from population studies and from studies of high risk families [
6‐
9]. However, differences in findings could also reflect how DNA repair capacity was measured.
BRCA1 genotypes as a predictor of mutagen sensitivity have not been previously studied. Although we found no overall association for the
BRCA1 E1038G, D693N, and Q356R genotypes with MBPC, in a polygenic model, the 1038G and 356Q
BRCA1 SNPs predicted higher MBPC. These SNPs, however, have had only limited study for breast cancer risk, and null results were reported for women from high risk families [
16‐
20] and sporadic breast cancer [
21‐
23] possibly because of their low minor allele frequencies in the general population. The Q356R SNP has been studied, but both positive association and null results have been reported in sporadic breast cancer [
21‐
23], and a positive association reported in one study of familial breast cancer [
17], but not in another [
16]. For the Q356R and E1038G SNPs,
in silico analysis indicated that these could have adverse effects due to their location in the
BRCA1 gene [
25]. Recently, the 1038G polymorphism, which was in LD with 1183R, 871L, and 1613G in our study set, as well as in the HapMap CEU data (Utah residents with ancestry from northern and western Europe), was associated with increased BRCA1 protein expression in a small case-control study of breast cancer risk [
66]. However, the same study did not find an association with K1183R, P871L, and S1613S, indicating that population admixture may have contributed to differences in haplotype frequencies.
The
BRCA1 haplotypes examined herein were constructed from the 35 sequenced subjects and HapMap data, using the E1038G SNP, Q356R and D693N SNPs. Given the presumed detrimental effects of these SNPs on
BRCA1 based on the
in silico analysis [
25], it would seem that the 3 SNP haplotype, CTC, would be associated with increased MBPC in our study. To the contrary, we found that the CTC haplotype was significantly associated with decreased MBPC. This association, though, is based on simulated data that only found significant associations in 4 of 10 simulations, using a frequency >3. Given that this haplotype is made up of the minor alleles of our SNPs, it is rare. Other studies evaluating BRCA1 haplotypes have used a combination of the SNPs used in our study to evaluate risk in sporadic breast cancer [
22] and interactions with hormonal therapy [
21], but we are the first to have used the 3-SNP haplotype found to be associated with low MBPC in our population. For example, Freedman and coworkers genotyped 28 BRCA1 SNPs, including the E1038G and Q356R SNPs, and observed 13 common haplotypes, but, none were associated with risk [
22]. The Marie-Genica group studied a haplotype consisting of the Q356R (rs1799950), P871L (rs799917), and K1183R (rs16942) SNPs and found that carriers of one haplotype were at a higher risk of developing breast cancer after estrogen therapy use compared to those with the common haplotype [
21]. Although the results are mixed for these genotypes and haplotypes for breast cancer risk, the genotype-phenotype associations indicate that further study is warranted.
For
Rad51, sequencing was completed for 92 women, namely those with the highest and lowest MBPC. In this study, the
Rad51 5'UTR 135G→C was found to be associated with decreased DNA repair capacity among unrelated subjects, in agreement with other findings [
42,
67]. Several
Rad51 variants were identified and confirmed by reverse sequencing. However, consistent with the NCBI databases (
http://www.ncbi.nlm.nih.gov/sites/entrez), these were found in very low frequency (<1%), with the exception of the
Rad51 5'UTR 135G→C. According to HapMap,
Rad51 has only a single haplotype block and SNP frequencies are low (<10%). Thus, haplotypes for
Rad51 were not studied. For this DNA repair gene, the lack of observed genetic variation in functional components of the gene and homology across species [
45] indicate that genetic variation in this gene might have detrimental effects. The 5'UTR 135G>C SNP, studied herein, can affect mRNA stability and/or translation efficiency, leading to altered product levels [
41,
42]. One study examined the effects of a
Rad51 genotypes in
BRCA1/2 carriers and reported that
Rad51 135G>C genotype association with breast cancer risk was greater in
BRCA1 carriers with truncating mutations (i.e. 185delAG) [
40].
The
Rad51 5'UTR variant C allele was then tested in the WEB case-control study because of the
a priori hypotheses developed from the MSA assays. For the case-control analysis, we used the most parsimonious model because none of the other covariates, such as education, body mass index, age at first birth, age at menarche, age at menopause (for post-menopausal women only), number of births, and previous benign breast disease, changed the impact the genotype has on the odds of disease by greater than 5%. Including non-genetic variables that do not affect the impact of the genotype on odds of disease may actually mute the effect. Our results, however, did not indicate that the SNP was associated with breast cancer risk as a main effect, which is consistent with other studies [
39,
68].
The strength of this study lies using MBPC, a validated intermediate phenotype for breast cancer, as the dependent variable in EBV-immortalized cell lines from a large number of BRCA1 mutation carriers in order to assess genotype-phenotype associations. The advantage of evaluating an intermediate phenotype rather than the actual clinical phenotype, such as disease, is that the number of genetic and environmental factors influencing the intermediate is probably smaller than the number of factors affecting the disease resulting in a better powered study. Because the clinical outcome is taken out of the equation, the risks of spurious associations are limited and in fact, we can better explore the mechanisms of the disease. And, while this study assessed the effect of BRCA1 and Rad51 genetic variation on MBPC and risk of sporadic breast cancer, other genes in the HR pathway, such as BRCA2, PALB2, MERIT40, and others, could potentially be assessed for effects on the breast cancer intermediate phenotype, DNA repair capacity, as well as risk.
The use of a priori hypotheses also helped identify the most plausible SNPs to be examined in our family- and population-based epidemiological studies. Furthermore, the present study used data from a large case-control study of environmental exposures in the etiology of sporadic breast cancer. These types of studies can provide corroborative evidence to epidemiological studies of breast cancer.
This study does have some limiting factors. The FCR study subjects were small in number, limiting statistical power for detecting genotype-phenotype relationships. Additionally, while removing BRCA1/2 mutation carriers that had received prophylactic mastectomy and oophorectomy was reasonable, it is also possible that this may have introduced selection bias and that the reason that some women chose prophylactic surgery may have been because their perceived risk was greater, perhaps due to higher family penetrance. If Rad51 variants were an underlying factor for the increased penetrance and the subjects were excluded from the present study, it is possible that these exclusion criteria would decrease power.
The MBPCs in this study were lower than in other studies [
6‐
8,
69‐
71]. However, this difference may be due to the lower radiation dose and lower post-radiation incubation time used in this study. Although lower, the dose of 1 Gy for γ-irradiation and post-irradiation incubation time (4 hours vs. 0.5-1.5 hours) [
6‐
8,
69‐
71] conditions were chosen herein based on experiments identifying the optimal cell survival at the highest dose for these cell lines; dose-response evaluations showed 100% cell death at 2 Gy (data not shown). Radio-sensitivity due to germ-line
BRCA1 mutations could also result in a lower dose response. Another explanation for lower MBPCs could be that only frank chromatid breaks were counted and all gaps were excluded, whereas other studies counted gaps as well. Ultimately, several associations were found making it 1) important to explore these associations among women without BRCA1 mutations and 2) essential to replicate in our larger case-control study.
Authors' contributions
LJR-S conceived the study, carried out the molecular genetic studies, and drafted the manuscript. LES participated in the statistical design of the family study and carried out the statistical analysis of the family data. YY participated in the statistical design of the study and carried out the statistical analysis of the family data. JLF participated in the design of the study and coordinated the WEB study. CJI participated in the design of the family study, provided subject epidemiologic data, and coordinated the statistical analysis of the family data. MDS participated in the design of the family study, provided subject epidemiologic data, and helped carry out mutation analyses of BRCA1 and BRCA2 in the subjects. RGD carried out the mutagen sensitivity assay. CM carried out the haplotype analysis. JN carried out the final statistical analysis of the WEB study. DV participated in the design of the study and coordinated the WEB study. SBE participated in the design of the study and coordinated the WEB study. PGS conceived the study and participated in the design of the study. All authors read and approved the manuscript.