Cardiovascular disease progression is characterised by the dysregulation of lipid metabolism and pro-atherogenic effects of adipose tissue signalling. Recent findings from the analysis of transcriptomic data in bulk tissue has enabled these insights and revealed important changes in gene expression. However, few studies have explored these molecular mechanisms before the onset of cardiovascular disease. We explore associations between future lipid-regulating drug use and cardiometabolic traits (n = 103), including DXA scans of body composition at baseline and follow-up 5–10 years later, in a cohort of British twins (n up to 6963). Utilising transcriptomic profiles from a subset of twins (n = 766), we explore the associations between baseline adipose tissue gene expression, clinical traits, and future lipid-regulating drug usage. We then test the joint predictive capacity of clinical traits plus gene expression compared to traditional risk scores using an automated machine learning approach. We find 44 traits are associated with lipid-regulating drug usage including measurements of abdominal fat tissue, cardiovascular health, and lipid metabolism (FDR 5%). Then, we present that adipose tissue gene expression levels at baseline are associated cross-sectionally with 19 of these 44 traits (FDR 5%). By comparing adipose gene expression levels between individuals prescribed lipid-regulating drugs in the future and controls, we discover that genes associated with 16 of these 19 traits produced greater log2-fold changes, suggesting shared mechanisms. We reveal 15 differentially expressed genes comparing future lipid-regulating drug users and controls at baseline (FDR 10%), including some implicated in angiogenesis: ESM1, RCAN2, and SOCS3. Functional enrichment with 1212 significantly differentially expressed genes (p < 0.05) included molecular mechanisms related to abnormal cardiovascular system electrophysiology (p = 1.89 × 10−3), arrhythmia (p = 4.02 × 10−3), and mitochondrial pathways (p = 1.12 × 10−3). Finally, we confirm inclusion of gene expression levels as features in machine learning models achieves a better AUC (0.919) compared to traditional risk predictors. These findings highlight the potential of bulk transcriptomic data to improve risk stratification for lipid-regulating drug use, offering new insights into the RNA biology of adipose tissue and advancing approaches for cardiovascular disease prevention.
原文链接
分享
引用
