The genetics of breast cancer may be further elucidated by comprehensive pathway-based candidate gene studies. We have recently completed a large-scale genetic screen within the Mayo Clinic Breast Cancer Study (3,072 SNPs) including assessment of the nuclear factor-kappaB (NF-kappaB) pathway. The NF-kappaB family of transcription factors regulates the expression of several hundred cancer-related genes including those involved in immune response, inflammation, and apoptosis. We sought to assess whether variation in genes encoding NF-kappaB subunits (REL, RELA, RELB, NFKB1, and NFKB2) and their primary inhibitors (NFKBIA and NFKBIB) was associated with breast cancer risk. Our ongoing hospital-based study solicits questionnaire data and blood samples from invasive breast cancer cases and age- and residence-matched controls at the Mayo Clinic in Rochester, MN. European-American participants recruited between 2001 and mid-2005 with residence in the surrounding area were included in the current analysis (798 cases, 843 controls). Gene-specific tagSNP selection used data from the SeattleSNPs PGA (NFKBIA) and the HapMap Consortium's release 21 (the remaining six NF-kappaB genes). 48 NF-kappaB tagSNPs were identified using ldSelect (r2 >= 0.80 and MAF >= 0.05) and genotyped along with three 3' UTR SNPs at the Illumina Corporation using genomic DNA and the Illumina Golden Gate AssayTM. Genotyping of 3,072 SNPs on 1,748 samples (including duplicates) yielded data on all but 165 SNPs and two samples; the remaining genotype data showed an average per-SNP call rate of 99.9%, average per-sample call rate of 99.9%, and duplicate concordance rate of > 99.9%. Associations were assessed by multi-SNP logistic regression (using log-additive main effects) and haplotype score testing within genes, as well as single-SNP logistic regression. Overall, global testing with multiple SNPs or haplotypes did not reveal associations with any NF-kappaB gene (all p > 0.10). However, two RELA tagSNPs yielded suggestive results (multivariable adjusted per allele risk: OR 1.2, 95% CI 1.0 - 1.3, p = 0.04 and OR 1.2, 95% CI 1.0 - 1.4, p = 0.02). Although in independent LD bins, as expected based on HapMap data, LD was present (r2 = 0.69, D' > 0.99), suggesting a common signal. RELA is a strong candidate gene encoding p65 which localizes into the nucleus as a heterodimer with p50 (NFKB1); however the observed association may be due to another regional variant. For example, a tagSNP for MAP3K11, a gene 25 kb 5' of RELA on chromosome 11 and a potential NF-kappaB activator, was in LD with the RELA tagSNPs (mean r2 = 0.73, D' = 0.95) and showed similar results (OR 1.2, 95% CI 0.9 - 1.7, p = 0.2). We now aim to clarify this signal using higher LD thresholds and marker density, CART analysis across multiple NF-kappaB genes, and fine-mapping within the RELA region. This study demonstrates the utility of coordinated analysis of multiple candidate genes in the context of a high-throughput SNP screen.

98th AACR Annual Meeting-- Apr 14-18, 2007; Los Angeles, CA