Integrative network analysis reveals different pathophysiological mechanisms of insulin resistance among Caucasians and African Americans

Background: African Americans (AA) have more pronounced insulin resistance and higher insulin secretion than European Americans (Caucasians or CA) when matched for age, gender, and body mass index (BMI). We hypothesize that physiological differences (including insulin sensitivity [SI]) between CAs and AAs can be explained by co-regulated gene networks in tissues involved in glucose homeostasis. Methods: We performed integrative gene network analyses of transcriptomic data in subcutaneous adipose tissue of 99 CA and 37 AA subjects metabolically characterized as non-diabetic, with a range of SI and BMI values. Results: Transcripts negatively correlated with SI in only the CA or AA subjects were enriched for inflammatory response genes and integrin-signaling genes, respectively. A sub-network (module) with TYROBP as a hub enriched for genes involved in inflammatory response (corrected p= 1.7E-26) was negatively correlated with SI (r= -0.426, p= 4.95E-04) in CA subjects. SI was positively correlated with transcript modules enriched for mitochondrial metabolism in both groups. Several SI-associated co-expressed modules were enriched for genes differentially expressed between groups. Two modules involved in immune response to viral infections and function of adherens junction, are significantly correlated with SI only in CAs. Five modules involved in drug/intracellular transport and oxidoreductase activity, among other activities, are correlated with SI only in AAs. Furthermore, we identified driver genes of these race-specific SI-associated modules. Conclusions: SI-associated transcriptional networks that were deranged predominantly in one ethnic group may explain the distinctive physiological features of glucose homeostasis among AA subjects. We performed integrative gene network analyses of transcriptomic data in subcutaneous adipose tissue of 99 CA and 37 AA subjects metabolically characterized as non-diabetic, with a range of SI and BMI values.