Multi-ancestry methylation-wide association analyses identifies putative risk methylation markers for breast cancer

Introduction: Aberrant DNA methylation is a hallmark of cancer. Identifying methylation sites (CpGs) associated with breast cancer risk may help to elucidate disease mechanisms and inform precision prevention. We conducted a multi-ancestry breast tissue-based methylome-wide association study (MeWAS) to discover breast cancer susceptibility CpGs. Methods: We profiled DNA methylation in normal breast tissue from 152 African ancestry and 267 European ancestry women using the Illumina MethylationEPIC array, with matched genotype data. We trained ancestry-specific genetic prediction models for each methylation marker, and then tested associations with breast cancer risk by applying S-PrediXcan to ancestry-matched breast cancer GWAS summary statistics (European:133,384 cases and 113,789 controls; African :18,044 cases and 22,187 controls)). We also performed stratified analyses by estrogen receptor status (ER+, ER−) and triple-negative breast cancer (TNBC). Then fixed effect inverse variance weighted meta-analysis was conducted using METAL. Methylation set enrichment analysis was evaluated using R missMethyl. Results: We successfully built 160,204 African ancestry and 166,069 European ancestry methylation imputation models (R > 0.1 and P < 0.05). Meta-analysis across ancestries identified 625 CpGs whose genetically predicted methylation levels were significantly associated with overall breast cancer risk (Bonferroni-adjusted P < 0.05), 458 (73.3%) of which have not been reported in prior breast tissue MeWAS. Notably, the significant association between genetically proxied methylation levels of cg23766285 and overall breast cancer risk were only observed among African Ancestry. Among 308 CpGs with valid prediction models in both ancestries, 265 (86.0%) showed concordant directions of association. We additionally identified 32, 11, and 11 CpGs that were exclusively associated with risk of ER+ and ER- breast cancer and TNBC at Bonferroni corrected P <0.05, respectively. CpGs associated with overall risk (odds ratio [OR], 1.19; 95% confidence interval [CI], 1.02-1.39), ER- breast cancer (OR, 1.97; 95% CI, 1.29-3.05), and TNBC (OR, 2.12; 95% CI, 1.27-3.61) were enriched in promoter regions, whereas ER+-associated CpGs were not. The identified CpGs are involved in various biological processes, including hormone, p53, and KRAS signaling pathways. Discussion: This is the first multi-ancestry, large-scale, tissue-based MeWAS in breast cancer. Our findings offer novel insight into breast carcinogenesis and breast cancer disparities.