Transcriptomic profiling of dysfunctional adipose tissue defines biomarker signatures that improve diabetes risk assessment

Background Adipose tissue dysfunction has diverse cardiometabolic sequelae, yet in clinical practice is inferred using flawed metrics, such as body mass index. We reasoned that analysis of adipose tissue from donors with diabetes could define circulating biomarkers of high-risk adiposity. Methods Using bulk RNA sequencing and histology from the GTEx project (168 DM vs 478 non-DM SAT; 129 DM vs 407 non-DM VAT), we defined depot-specific transcriptomic and structural alterations associated with DM. BMI-stratified analyses were performed in normal weight, overweight, and obese groups. Differentially expressed genes (DEGs) were functionally profiled. Concordant plasma protein signatures were validated in 52,743 UK Biobank participants, and an adipose-derived protein risk score was developed to predict incident DM. Results DM was associated with 1140 DEGs in subcutaneous adipose tissue (SAT) and 314 in visceral adipose tissue (VAT), with limited overlap (n=69). Enriched pathways included extracellular matrix remodelling, inflammation, metabolism, axon guidance, and endoplasmic reticulum stress. Histology demonstrated larger SAT adipocytes in DM, restricted to the overweight group (Fig 1A-B, p=0.014), while VAT adipocyte size differed only in normal-weight individuals. BMI-stratified transcriptomics revealed minimal DEG overlap across BMI strata (Fig 1C), with the strongest SAT signal in overweight donors (425 DEGs). Of these, 38 plasma proteins showed directionally concordant differential abundance in UK Biobank (Fig 1D). LASSO regression identified a 12-protein panel (ACP5, CDCP1, CHI3L1, EPHA1, ESM1, FBP1, GLA, IFI30, IL1RN, KYNU, MMP7, PCSK9) independently predictive of incident DM (Fig 1E) which improved 10-year DM prediction beyond the Leicester Diabetes Risk Score (Fig 1F, AUC 0.736 vs 0.805, p<0.001) and were associated with incident myocardial infarction, heart failure, atrial fibrillation, and hepatic fat accumulation. Conclusions DM is associated with depot-specific and BMI-dependent adipose dysfunction. Overweight individuals exhibit a distinctive diabetic SAT phenotype characterised by adipocyte hypertrophy and inflammatory signalling. A transcriptomics-informed circulating protein panel improves prediction of incident DM and may enable targeted metabolic intervention.