Roeles, J.; Giesler, A.; Schmitz, J.; Koenig, J.; Friedersdorff, F.; Busch, J.; Greite, R.; Krappitz, M.; Layne, I.; Koettgen, M.; Kuehn, W. E.; Schmid, J.; Schmidt, K.; Hoeft, K.; Kramann, R.; Eckardt, K.-U.; Kocks, C.; Braesen, J. H.; Halbritter, J.; Schmitt, R.; Haerteis, S.; Hopp, K.; Schmidt-Ott, K. M.; Hinze, C.

Autosomal dominant polycystic kidney disease (ADPKD) exhibits substantial interpatient variability in disease course and therapeutic response, but the cellular basis for this variability remains poorly understood. Here, we combine single-nucleus RNA sequencing of human cyst epithelia with machine learning-based histological analysis of >1,800 cysts to resolve three epithelial cyst types-proximal tubule-like, collecting duct-like, and mixed. These cyst types display distinct injury states, metabolic programs, and stromal microenvironments, including a mixed-cyst niche enriched for CCL2-associated inflammatory signaling. Expression of key therapeutic targets was highly cell-type specific with CFTR enriched in proximal-like epithelia, whereas AVPR2 expression was confined to AQP2-positive collecting duct-like cells. Cyst-type composition varied widely across patients and in an orthologous mouse model (Pkd1RC/RC) in which the burden of AQP2-positive cysts correlated with responsiveness to tolvaptan. These findings identify cyst-type heterogeneity as a major determinant of molecular pathway activation and predictability of therapeutic response in ADPKD.