Gene-level germline contributions to clinical risk of recurrence scores in Black and White breast cancer patients

Abstract

Continuous risk of recurrence scores (CRS) based on tumor gene expression are vital prognostic tools for breast cancer (BC). Studies have shown that Black women (BW) have higher CRS than White women (WW). Although systemic injustices contribute substantially to BC disparities, evidence for biological and germline contributions is emerging. We investigated germline genetic associations with CRS and CRS disparity using approaches modeled after transcriptome-wide association studies (TWAS). In the Carolina Breast Cancer Study, using race-specific predictive models of tumor expression from germline genetics, we performed race-stratified (N=1,043 WW, 1083 BW) linear regressions of three CRS (ROR-S: PAM50 subtype score; Proliferation Score; ROR-P: ROR-S plus Proliferation Score) on imputed Genetically-Regulated tumor eXpression (GReX). Using Bayesian multivariate regression and adaptive shrinkage, we tested GReX-prioritized genes for associations with PAM50 tumor expression and subtype to elucidate patterns of germline regulation underlying GReX-CRS associations. At FDR-adjusted P < 0.10, we detected 7 and 1 GReX-prioritized genes among WW and BW. Among WW, CRS were positively associated with MCM10, FAM64A, CCNB2, and MMP1 GReX and negatively associated with VAV3, PCSK6, and GNG11 GReX. Among BW, higher MMP1 GReX predicted lower Proliferation score and ROR-P. GReX-prioritized gene and PAM50 tumor expression associations highlighted potential mechanisms for GReX-prioritized gene to CRS associations. Among BC patients, we find differential germline associations with CRS by race, underscoring the need for larger, diverse datasets in molecular studies of BC. Our findings also suggest possible germline trans-regulation of PAM50 tumor expression, with potential implications for CRS interpretation in clinical settings. SIGNIFICANCEWe find race-specific genetic associations with breast cancer risk-of-recurrence scores (CRS). Follow-up analyses suggest mediation of these associations by PAM50 molecular subtype and gene expression, with implications for clinical interpretation of CRS.