Introduction and Objective: Pancreatic islet function is closely related to cell-cell communication and paracrine signaling, suggesting that the spatial arrangement of endocrine cells could play a critical role in islet physiology. However, quantitative analyses linking islet spatial organization to functional outcomes are limited. This study systematically characterized the distribution and spatial arrangement of endocrine cells and their relationship to islet function. Methods: We analyzed the phenotyping data of isolated human islets from 578 organ donors without diabetes obtained through Integrated Islet Distribution Program (IIDP) and Human Pancreas Analysis Program (HPAP), representing different age, sex, BMI, and race/ethnicity. Immunofluorescence images of islet histological sections were processed with machine learning-based pipelines to annotate cell types and extract spatial organization features from 54,974 islets (diameter ≥ 50 μm). Statistical modeling and graph-based deep learning identified associations between cell organization and secretory function. Results: Endocrine cell organization was non-random, with significant enrichment of homotypic (i.e., α-α, β-β, and δ-δ cell) contacts (p10-100). The enrichment level varied substantially between donors (p10-100). Deep learning models demonstrated that the spatial arrangement of α, β, and δ cells provided predictive power for hormone secretion beyond endocrine cell composition alone. Despite being the smallest of the three endocrine cell types, δ cells showed the strongest associations with stimulated insulin secretion. Both higher δ cell number (p0.001) and larger per-δ-cell cytoplasmic area (p0.01) correlated with reduced insulin secretory responses. Conclusion: The spatial organization of endocrine cells in human islets exhibits substantial inter-donor heterogeneity yet follows non-random patterns. Neighborhood relationships among endocrine cells, particularly δ cells, are key determinants of human islet function in vitro. Disclosure F. Feng: None. T. Bate: None. A.L. Hopkirk: None. D.C. Saunders: None. S. Mei: None. A. Coldren: None. J.R. Taylor: None. H. Durai: None. C. Davis: None. C. Reihsmann: None. A. Eskaros: None. R. Jenkins: None. A.C. Powers: None. J.C. Niland: None. C. Evans-Molina: Advisory Panel; Ended; Sanofi. Speaker's Bureau; Ended; Sanofi. Other - Shared grant and project; Current; Neurodon. Advisory Panel; Current; Diogenyx. M. Brissova: None.
FENG et al. (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: