Main

In women, breast cancer is the most common malignant disease in industrialised countries. About 5–10% are the so-called familial cases, which can be mainly attributed to deleterious mutations in BRCA1 and BRCA2 (Dunning et al, 2001; Nathanson and Weber, 2001), and also for the remaining majority of spontaneous breast cancer cases a strong genetic component has been postulated (Lichtenstein et al, 2000). Although finally having a strong impact, the responsible genetic markers may be common, low-penetrance genetic variants that modify susceptibility to breast cancer.

Potential candidates for these markers are single-nucleotide polymorphism (SNPs) that alter the sequence or the expression of a gene product (Lander and Schork, 1994; Cargill et al, 1999; Gray et al, 2000; Risch, 2000; Sunyaev et al, 2001). A large variety of SNPs have already been investigated for their association with breast cancer (Dunning et al, 2001; Nathanson and Weber, 2001). These were SNPs in DNA repair genes, steroid hormone metabolism genes and carcinogen metabolism genes (see also, Goode et al, 2002). We have chosen eight different SNPs from six genes. These were Q356R and P871L in BRCA1, N372 H in BRCA2, R72P inTP53, C112R (E4) and R158C (E2) in ApoE, P187S in NQO1 and C825 T in GNB3.

BRCA1 and BRCA2 are the well-established susceptibility genes for familial breast cancer (Dunning et al, 1997, 2001; Healey et al, 2000; Nathanson and Weber, 2001; Goode et al, 2002), TP53 is a gene involved in apoptosis (Dumont et al, 2003), ApoE influences lipid metabolism and cardiovascular disease (Menzel et al, 1983) and may be involved in tumour proliferation (Moysich et al, 2000; Zunarelli et al, 2000), NQO1 is engaged in carcinogen metabolism (Nebert et al, 2002) and GNB3 is part of a signal transduction pathway (Siffert et al, 1998).

Searching the literature, no association with breast cancer has been found for Q356R and P871L (BRCA1), although the authors claimed that being homozygous for 356R might protect against breast cancer (Dunning et al, 1997). In case of the 372H allele (BRCA2), an increased risk for developing breast cancer has been observed (Healey et al, 2000) together with an association with foetal survival. The 72P allele of the polymorphism in TP53 was only weakly associated with an increased risk for breast cancer (Nathanson and Weber, 2001). No association of breast cancer was observed with the apolipoprotein E polymorphism (Moysich et al, 2000; Zunarelli et al, 2000). For the P187S polymorphism in NQO1, conflicting results were published (Siegelmann-Danieli and Buetow, 2002; Hamajima et al, 2002). The C825 T polymorphism in GNB3 has not been investigated so far.

A central consideration with case–control studies are spurious results due to a large variety of reasons (Lander and Schork, 1994). It is therefore mandatory to repeat published studies in different populations, and also null results should be published to avoid bias (Hemminki and Shields, 2002). One of the reasons for spurious results is a general statistical problem due to multiple testing. This can either be accounted for by applying statistical correction methods (e.g. Bonferoni) or by investigating at least two different populations.

Here we present the results of repetitive SNP association studies in our case–control study and of a new one that was performed in two independent populations, one from Tyrol, Austria and the other from Prague, Czech Republic. In addition, we have analysed the concomitant effect of two polymorphisms in two different genes in order to mimic the situation in vivo where the different genes/gene products do not act as single entities but as members of an ‘orchestra’, as suggested by Risch (2000).

Materials and methods

Control and patient populations

Controls from Tyrol

The controls (400 women) were randomly drawn from a group of 13 000 apparently healthy blood donors from Tyrol. All came from the same geographical area as the patient group. The mean age of the control group was 39±12 years. The control persons were all anonymous, and only their age and gender were known. For the investigation of a possible association of the N372H polymorphism with gender, we randomly chose additional 600 women and 1600 men.

Controls from Prague

The control group enclosed 231 women from Prague with a mean age of 60±23 years. Controls were recruited from the staff of the National Institute of Public Health, nurses and patients of collaborating hospitals in Prague and inhabitants of houses for elderly citizens living in the same urban area as the patients. Controls were interviewed and only those having no personal history neither of breast cancer nor other malignancies were included into the study. The composition of the control group was comparable to cases in terms of age. Controls were asked to read and sign an Informed Consent protocol.

Patient groups

The patient group from Tyrol, 220 women, had a mean age of 56±13 years and the patient group from Prague consisted of 190 women with an average age of 58±13 years. All patients gave Informed Consent. In all cases, the diagnosis of breast cancer was confirmed histological. The cases from Prague were all incident cases, whereas the cases from Tyrol were a mixture of incident and prevalent cases with a median of one survival year (mean 2.5±3.7).

Genotyping

All samples were genotyped by the 5′exonuclease assay with fluorescent MGB-probes on an ABI PRISM 7000 Sequence Detection System™ from Applied Biosystems. In addition, some samples were also genotyped by conventional methods (PCR and digestion) (Q356R, patients and controls from Tyrol and P187S, patients and controls from Prague) and by Pyrosequencing™ (P871L, controls from Tyrol). All methods gave identical results and not a single deviation was observed. The sequences of the primers designed for the analysis are given in appendix.

Statistical analysis

The χ2-test was used to compare the distribution of genotypes between cases, controls and expected genotypes assuming a Hardy–Weinberg equilibrium. The risk attributed to individual alleles or genotypes for breast cancer was calculated as odds ratio from 2 × 2 tables. A possible association of genotypes with age and survival was analysed by the Kruskal–Wallis test.

Results

The genotype frequencies of the eight SNPs investigated in the control and case groups from Tyrol and Prague are given in Tables 1 2, respectively. In every group or subgroup, the genotype frequencies were in accordance with the assumption of a Hardy–Weinberg equilibrium. No association between genotype frequencies and age was observed except for the C112R (E4) polymorphism in the controls from Prague. Also, no association between genotypes and survival was discovered in the patient group from Tyrol, which is in agreement with the paper of Goode et al (2002). When allele frequencies were compared between cases and controls, only one significant deviation was observed: the P187S SNP in the NQO1 gene (see Tables 1 and 2). The 187S allele was found significantly more frequently in breast cancer patients than in controls. The same deviation could be observed in both populations from Tyrol and Prague. Regarding all other SNPs no significant differences were observed (Tables 1 and 2).

Table 1 Tyrol
Full size table
Table 2 Prague
Full size table

With respect to this association, we also tested for possible deviations of genotype frequencies of the P187S SNP and compared the homozygote ratio between cases and controls. In the group from Prague, there were in addition to the significant allele frequency difference also highly significant differences of all genotypes (P=0.0025) and of the ratio of the two homozygous genotypes (PP/SS) (odds ratio (OR)=6.9; 95% confidence interval (CI) 1.8–30.6; P=0.0006).

Between the respective groups from Tyrol and Prague, there was no significant difference of the allele frequencies and genotype frequencies of all SNPs under investigation. In order to increase power, we therefore combined the respective cases and controls groups from the two middle European populations. Again, in case of the P187S SNP, there was a significant difference of the allele frequencies (OR=1.46; 95% CI 1.16–1.85; P=0.001), the genotype frequencies (P=0.0003) and the homozygote ratio between patients and controls (OR=3.8; 95% CI 1.73–8.34; P=0.0001). In addition, also the R72P SNP in the TP53 gene showed a borderline significant difference regarding allele frequencies (OR=1.27; 95% CI 1.00–1.61; P=0.044) and the homozygote ratio (OR=1.77; 95% CI 1.00–3.13; P=0.04). No significant difference was observed between cases and controls when we compared frequencies of heterozygotes, both P187S and R72P, and of homozygotes for the common allele, P187 and R72, respectively.

We also analysed the H372 H polymorphism in the BRCA2 gene to test for a previously found association with gender (Healey et al, 2000) in 2442 controls (1530 men and 912 women). No difference of genotype frequencies between men and women (P=0.73) was found and also no deviation from expected frequencies assuming Hardy–Weinberg equilibrium (P=0.99).

Since we had analysed eight different SNPs in six different genes, we also investigated for the presence of linkage disequilibrium and for pair-wise locus association. The two SNPs of BRCA1 and ApoE, respectively, were in linkage disequilibrium as expected.

The calculation of association for pair-wise loci as outlined by the two-locus genetic model of Risch (2000) was performed for the polymorphism at the NQO1 and TP53 gene loci, since these were the only two polymorphic sites that showed a significant association if the two studies were combined. Further, the associations of the two gene loci with breast cancer are the only ones that showed in both studies a trend in the same direction and a gene dosage effect (Hemminki and Shields, 2002) in contrast to the other polymorphic sites. When single-locus heterozygote and common allele homozygote frequencies were compared between patients and controls, there was no significant difference. The double heterozygotes (P187S/R72P) were more frequent in patients (40%) than in controls (28%) compared to the common allele double homozygotes (P187/R72) (67 vs 71%; OR=1.88; 95% CI 1.12–3.15; P=0.011).

Individuals with two or more “deleterious” alleles (Table 3) were more frequent in the patients group than in the control group (OR=1.97; 95% CI 1.31–2.97; P=0.0006). On the other hand, those with only one deleterious allele or those with no deleterious allele at all (Table 3) had comparable ratios in both patients and controls (OR=1.03; 95% CI 0.75–1.53; P=0.7).

Table 3 Genotype combinations at the NQO1 and TP53 loci for the combined populations from Tyrol and Prague
Full size table

Discussion

This association study of breast cancer patients analysing eight different SNPs in two different populations has shown in part accordance with previously published papers, in part divergence with published results and finally new results. The investigated SNPs in BRCA1 and BRCA2 showed no association with breast cancer in the two populations from Austria and the Czech Republic. This is in agreement with previous results for the P871L polymorphism (Dunning et al, 1997), but there are discrepancies for the Q356R polymorphism in BRCA1 (Dunning et al, 1997) and the N372H polymorphism in BRCA2 (Healey et al, 2000). One paper (Dunning et al, 1997) claimed a protective effect for the R356 allele in the homozygous state. In our population from Tyrol, the homozygotes were found in the same ratio in both the control and the patient groups. There were no homozygotes in the population from Prague, but because of the small group size this was still in accordance with Hardy–Weinberg equilibrium. It remains doubtful if the absence of the 356R homozygotes is real or spurious.

The association of the N372H polymorphism with breast cancer observed by Healey et al (2000) could not be confirmed in this study. It is, however, striking that their observation was only found in an English population but not in the German and Finnish population in the later publication. It remains unclear if this is a special condition in the English population. The other published results (Healey et al, 2000) concerning the deviation from Hardy–Weinberg equilibrium and the differences between men and women of the N372H polymorphism could not be confirmed by this study. When we tested for deviation from Hardy–Weinberg by a χ2-test using the published frequencies of the control groups, we obtained no significant deviation. Our own results gave an excellent conformity with Hardy–Weinberg equilibrium.

For the R72P polymorphism in the TP53-gene, only a weak association was observed and only in the combined population. In the literature, there are some publications stating the same result but with smaller group sizes (Sjalander et al, 1996; Wang-Gohrke et al, 1998). Most likely the effect is very small even in the homozygous state of the 72P allele.

The C825T polymorphism in GNB3 apparently does not influence the carcinogenesis of breast cancer, which is also true for the Apo E polymorphism. The latter is in agreement with the literature (Moysich et al, 2000; Zunarelli et al, 2000).

The association of the 187S allele of NQO1 could be observed in both populations from Prague and Tyrol. The effect was more significant in the population from Prague than in the Tyrolean population, which might be explained by the different recruiting of the control groups. The Czech group are age-matched women, partly from a selected environment, with no evidence of breast cancer, whereas the women from Tyrol are blood donors between the age of 18 and 67 years incorporating a substantial number of women who might still get breast cancer in their later years.

Choosing the right control population is a very critical aspect in every case–control study. It has been a longstanding prerequisite that controls should be age matched to patients. Unfortunately, this might lead in the case of late-onset diseases like breast cancer to stratification due to selection bias. At the age in which most of the cases of breast cancer occur, also a lot of circulatory diseases occur that might change profoundly the composition of an age-matched control group. The fact that similar results have been gained, although different recruiting strategies have been used, further confirms that the actual findings of this study are real.

Comparison of our data on the P187S polymorphism in NQO1 with those in the literature shows inconsistency (Hamajima et al, 2002; Siegelmann-Danieli and Buetow, 2002), particularly with the one from Japan where the frequency for the 187S homozygotes was 16.5% in the controls and 14.3% in the breast cancer patients group. In contrast, a group from Philadelphia (Dunning et al, 1997) observed no significant difference, but the frequency of the 187S allele was 19% in the case group and 15% in the control group, which is in the same range as in our groups and shows the same trend as this study. The fact that the American study (Siegelmann-Danieli and Buetow, 2002) showed no significant difference in NQO1 polymorphism may be due to the low numbers of 187S homozygotes in the patient group and high numbers of 187S homozygotes in the control group that might have occurred by chance. Combining the numbers of the studies for this polymorphism from Tyrol, Prague and Philadelphia, very significant differences in allele frequency, genotypes and homozygotes ratio (data not shown) are observed.

NAD(P)H: quinone oxireductase (NQO1) is an enzyme that is involved in metabolising numerous endogenous and environmental quinones. Exchange of the Proline at position 187 by Serine leads to a nonfunctional enzyme (Nebert et al, 2002). Individuals homozygous for the 187S allele have a high risk for aplastic anaemia and leukaemia (Nebert et al, 2002). Reports about the association of the P187S polymorphism with lung cancer are inconsistent (Chen et al, 1999; Lewis et al, 2001). The association of the P187S polymorphism with breast cancer found in this study is the first reported in the literature and should be further investigated.

The increased risk of the double heterozygotes (P187S/R72P) to develop breast cancer is a new finding, which is similar to the observation that double heterozygotes for the Factor V “Leiden” and the Prothrombin mutation G20210>A have a 20-fold risk for developing thrombosis, whereas the risk for single heterozygotes is only five-fold and four-fold, respectively (Emmerich et al, 2001), and agrees with the two-locus genetic model of Risch (2000). Whereas the interactions of the F5 and F2 gene products in thrombosis are well understood, the way of interaction of the NQO1 and TP53 products in breast cancer can only be speculated. Hydroquinone, a substrate for detoxification by the NQO enzyme, can induce apoptosis (Moran et al, 1999), which is less stimulated by the 72P variant of p53 (Dumont et al, 2003). How these pathways are actually interwoven remains open for further investigations.