Inter-Individual Heterogeneity, Not Mean Abundance, Distinguishes Crinophagy Machinery in Type 2 Diabetic Human Islets

Type 2 diabetes (T2D) is associated with transcriptional and proteomic heterogeneity in human pancreatic islets, yet inter-individual variance is rarely treated as a phenotype in its own right. We tested whether variance captures disease-associated information missed by mean-based analysis, focusing on crinophagy, the direct lysosomal degradation of insulin secretory granules (iSGs). A 31-member crinophagy set and a 14-member macroautophagy comparator were pre-registered before analysis and evaluated in a discovery proteomics cohort of 129 organ donors and an independent RNA-seq cohort of 54 donors (18 non-diabetic, 36 T2D), with adjustment for batch, immune infiltration, and b-cell content. Global proteome dispersion did not differ between groups (centroid-distance ratio 1.05, permutation P = 0.71), and the mean crinophagy-set abundance was unchanged by glycemic status (P = 0.88). In contrast, crinophagy-set variance was greater than expected from size-matched random gene sets (median SD ratio 1.178 vs 1.078, permutation P = 0.036), with no corresponding shift in the comparator set (P = 0.71). UNC13D and STX7 ranked at the 99.38th and 99.09th percentiles of 20,255 expressed genes. UNC13D showed closely concordant SD ratios in discovery proteomics and replication RNA-seq (2.065 and 2.001), whereas RAB3A did not replicate (1.725 and 1.101). Neither UNC13D nor STX7 showed a corresponding mean-expression difference in independent single-cell b-cell data. Inter-individual variance therefore identifies molecular differences in T2D islets not captured by mean-based analysis, with UNC13D providing a reproducible variance phenotype across independent cohorts and molecular platforms.