BMI- and subtype-specific molecular clusters reveal distinct obesity mechanisms in breast cancer across Nigerian and U.S. cohorts

Background: Obesity is a major breast cancer risk factor, yet its biological mechanisms remain poorly characterized in African populations. We hypothesized that distinct obesity-associated molecular signatures would differentiate obese from normal-weight Nigerian women and reveal subtype-specific patterns in U.S. women. Methods: Targeted gene expression profiling (NanoString, 785 genes) was performed on tumors from 46 Nigerian women (MEND), and publicly available gene expression data from Women's Circle of Health Study (WCHS; n=367; African and European ancestry) were analyzed. Enrichment scores for obesity-related pathways (glycolysis, inflammation, Extracellular matrix (ECM) remodeling, adipokine signaling, insulin/IGF1 signaling, hypoxia, cholesterol biosynthesis) were computed using Gene Set Variation Analysis in R (v4.5.0) with KEGG and Reactome gene sets. Bipartite network clustering identified co-occurring mechanism clusters. Associations were tested for cluster membership with BMI using Firth logistic regression for MEND, and with subtype and ancestry using standard logistic regression for WCHS, adjusting for covariates (MEND: age, subtype, menopausal status; WCHS: age). Results: MEND participants averaged 48.7 (10.6) years; WCHS averaged 54.07 (11.99) years. In MEND, three clusters emerged (Q=0.14, p<0.05): Cluster 1 (hypoxia, cholesterol biosynthesis), Cluster 2 (ECM remodeling, adipokine, insulin/IGF1 signaling), and Cluster 3 (glycolysis, inflammation). Overweight/obese women had higher odds of Cluster 2 (aOR=5.4, 95% CI: 1.3-31.3), while normal-weight women were enriched for Cluster 1 (0.12, 0.02-0.54). In WCHS, six enrichment scores yielded three clusters (Q=0.10, p<0.05): Cluster 1 (inflammation, hypoxia), Cluster 2 (ECM remodeling, insulin/IGF1 signaling), and Cluster 3 (adipokine signaling, glycolysis). Cluster membership did not differ by ancestry, but subtype associations (Luminal A reference) were striking: Cluster 1 strongly associated with Luminal A (Luminal B: 0.33, 0.18-0.61; HER2: 0.15, 0.06-0.34; Triple Neg: 0.19, 0.11-0.34), Cluster 2 with Luminal B (3.47, 1.8-6.9) and HER2 (4.09, 1.9-8.99), and Cluster 3 with Triple Negative (3.15, 1.78-5.73). Conclusions: Obesity-linked mechanisms differ by BMI and subtype. ECM remodeling and growth factor signaling dominate obesity-related tumors, while hypoxia and lipid metabolism characterize normal-weight tumors. Subtype-specific clustering suggests heterogeneity beyond BMI. Findings highlight opportunities for precision prevention and treatment, including metabolic pathway targeting and potential use of GLP-1 receptor agonists in high-risk subtypes. Disclosure: Generative AI (Microsoft Copilot) was used to assist with language editing; authors reviewed and verified all content.