Common BRCA1 Variants and Susceptibility to Breast and Ovarian Cancer in the General Population

Alison M. Dunning 2 Mathias Chiano 2 7 Neil R. Smith 2 Joanna Dearden 2 Martin Gore 1 Suzy Oakes 0 Charles Wilson 6 Michael Stratton 5 Julian Peto 4 Doug Easton 3 David Clayton 7 Bruce A. J. Ponder 2 0 Department of Community Medicine 1 Department of Medicine, Royal Marsden Hospital , Fulham Road, London SW3 6JJ, UK 2 CRC Human Cancer Genetics Research Group , Box 238 3 CRC Genetic Epidemiology Unit, Department of Community Medicine 4 CRC Section of Epidemiology, Institute of Cancer Research , Sutton, Surrey SN2 5NG, UK 5 CRC Section of Molecular Carcinogenesis 6 Department of Clinical Oncology, Addenbrooke's Hospital , Hills Road, Cambridge , CB2 2QQ UK 7 MRC Biostatistics Unit, Institute of Public Health, University of Cambridge, University Forvie Site , Robinson Way, Cambridge CB2 2SR, UK Most multiple case families of young onset breast cancer and ovarian cancer are thought to be due to highly penetrant mutations in the predisposing genes BRCA1 and BRCA2. However, these mutations are uncommon in the population and they probably account for only a few percent of all breast cancer incidence. A much larger fraction of breast cancer might, in principle, be due to common variants which confer more modest individual risks. There are several common polymorphisms in the BRCA1 gene which generate amino acid substitutions. We have examined the frequency of four of these polymorphisms: Gln356Arg, Pro871Leu, Glu1038Gly and Ser1613Gly in large series of breast and ovarian cancer cases and matched controls. Due to strong linkage disequilibrium, these four sites generate only three haplotypes with a frequency >1.3%. The two most common haplotypes, defined by the alleles Gln356Pro871Glu1038Ser1613 and Gln356Leu871Gly1038Gly1613, have frequencies of 0.57 and 0.32 respectively, and these frequencies do not differ significantly between patient and control groups. Thus the most common polymorphisms of the BRCA1 gene do not make a significant contribution to breast or ovarian cancer risk. However, our data suggest that the Arg356 allele may have a different genotype distribution in breast cancer patients from that in controls (Arg356 homozygotes are more frequent in the control groups, P = 0.01), indicating that it may be protective against breast cancer. If this finding can be confirmed, it may provide an insight into the - *To whom correspondence should be addressed structural features of the BRCA1 protein that are important for its function. Breast cancer is a common disease with a lifetime risk of 8% in women in the UK, while ovarian cancer is approximately one-tenth as frequent but remains the fifth most common cancer of women. The first human familial breast and ovarian cancer susceptibility gene, BRCA1, on chromosome band 17q21, was cloned in 1994 (1). The normal function of the 1863 residue protein encoded by this gene remains unknown, although two potentially functional motifs have been identified: a ring-finger domain encoded by exons 2, 3 and 5 (1), and a granin consensus sequence (residues 12141223) which indicates that the protein may be secreted (2). More than 100 distinct highly penetrant mutations in BRCA1 have been described (3,4). These mutations confer a 90% risk of breast or ovarian cancer by age 70 years. The majority are predicted to result in a truncated BRCA1 protein. This is consistent with other findings which suggest that BRCA1 acts as a tumour suppressor gene (5,6). Epidemiological data suggest that the frequency of highly penetrant BRCA1 mutations in the population is between 1 in 500 and 1 in 2000 individuals (7), which implies that these mutations account for only ~ 2% of all breast cancer diagnosed before age 70. The proportion of breast cancer due to BRCA2 is likely to be of a similar magnitude (7). Germline p53 mutations account for a much smaller proportion of cases (8,9), whilst mutations in the ATM gene probably account for <5% of all breast cancer (10). In principle, a greater proportion of breast cancer incidence could be accounted for by common but less highly penetrant predisposing genes. Currently the only well-established example is theHRAS1 mini-satellite locus (11). So-called rare alleles at this locus, which have a frequency of ~ 0.06 in the general population, are associated with a 2-fold relative risk of breast and some other cancers; despite this small relative risk, the predisposing alleles are estimated to account for 9% of total breast cancer incidence since they are relatively common. We have used a casecontrol design to assess whether BRCA1 gene variants have any effect on low-penetrance predisposition to breast and ovarian cancer. Individuals in the population series were genotyped for four polymorphisms in BRCA1 which result in amino acid substitutions: Gln356Arg, Pro871Leu, Glu1038Gly and Ser1613Gly. There is strong linkage disequilibrium between the alleles of all these sites (data not shown) and, as a result of this, haplotypes can be deduced with accuracy even in unrelated individuals. There is close to complete allelic association between the alleles at residues 871, 1038 and 1613. Furthermore, 94% of the alleles encoding Arg356 are carried on a single haplotype. Thus, only three of the observed haplotypes have an overall frequency of >1.3%. The estimated haplotype frequencies are given in Table 1. As a consequence of this strong allelic association, the effects of all three common haplotypes can be investigated by considering solely the alleles at the Gln356Arg and Pro871Leu polymorphic sites. The genotype distribution of the Pro871Leu polymorphism does not show any consistent frequency differences between patient and control groups (Table 2). The Leu871 allele is marginally more frequent in the UK patients aged 3645 years than in their matched controls [0.37 (238/642) versus 0.31 (220/700), P = 0.05], but not in any of the other groups [all breast cancer patients versus all controls: 0.34 (548/1602) versus 0.31 (362/1144), P = 0.35]. The estimated relative risk when all the breast cancer series are compared with all the controls is 1.15 [95% confidence interval (CI) 0.921.44] to the heterozygotes and 1.24 (95% CI 0.841.79) to the Leu homozygotes. The Arg356 allele tends to be more frequent among control individuals than among breast cancer cases (all controls versus all breast cancer cases: 0.07 versus 0.05, P=0.06) and, furthermore, Arg homozygotes are only found among the controls (overall difference in genotype distribution P = 0.010, Table 3). This is suggestive that Arg356 may be associated with reduced susceptibility to breast cancer, the relative risk to the heterozygotes being 0.88 (95% CI 0.63, 1.23) and that to the Arg homozygotes being 0 (95% CI 0, 0.56) compared with the Gln homozygotes. Some of the differences in genotype distribution between cases and controls may be due to the unusually large number of Arg356 homozygotes in controls (seven compared with 3.1 which would be expected under HardyWeinberg (...truncated)