Decreased parietal epithelial cell density is linked to podocyte depletion and predictors of kidney disease progression in human kidneys.

Parietal epithelial cells (PECs) have been implicated in the pathogenesis of glomerulosclerosis in rodent models, and novel technologies are beginning to unravel their contributions to human glomerular disease. Here, we report the development, validation, and application of a deep learning approach to analyze the PEC population in over 14,000 glomeruli from nephrectomy samples from patients with minimal overt chronic kidney disease (CKD). This analysis revealed a striking correlation between PEC density and podocyte density. Reduced PEC density was also associated with aging and the presence of diabetes. Furthermore, the PEC density in normal-appearing glomeruli was associated with the frequency of glomerular pathology, including global and segmental glomerulosclerosis, in the same patient sample. Patients with low PEC density had gene expression changes consistent with cellular stress in PECs. These observations support a link between PEC population and the progression of CKD.NEW & NOTEWORTHY Little is known about the contribution of PECs to human chronic glomerular disease. We developed, validated, and deployed deep learning image analysis tools to analyze the relationship between PECs and clinical and histopathologic risks of glomerular disease progression in human nephrectomy samples. These analyses revealed a novel link between PEC depletion and early evidence of chronic glomerular disease. The tools developed can be applied to many applications in the study of human kidney disease.