Although apoptosis of podocytes has been widely reported in in vitro studies, it is less frequently and less definitively documented in in vivo situations. To investigate this discrepancy, we analyzed the dying process of podocytes in vitro and in vivo using LMB2, a human (h) CD25-directed immunotoxin. LMB2 induced cell death within 2 days in 56.8 ± 13.6% of the cultured podocytes expressing hCD25 in a caspase-3, Bak1 and Bax-dependent manner. LMB2 induced typical apoptotic features, including TdT-mediated deoxyuridine triphosphate nick-end labeling (TUNEL) staining and fragmented nuclei without LDH leakage. In vivo, LMB2 effectively eliminated hCD25-expressing podocytes in NEP25 mice. Podocytes injured by LMB2 were occasionally stained for cleaved caspase-3 and cleaved lamin A but never for TUNEL. Urinary sediment contained TUNEL-positive podocytes. To examine the effect of glomerular filtration, we performed unilateral ureteral obstruction in NEP25 mice treated with LMB2 one day before sacrifice. In the obstructed kidney, glomeruli contained significantly more cleaved lamin A-positive podocytes than those in the contralateral kidney (50.1 ± 5.4% vs. 29.3 ± 4.1%, P value < 0.001). To further examine the dying process without glomerular filtration, we treated kidney organoids generated from nephron progenitor cells of NEP25 mice with LMB2. Podocytes showed TUNEL staining and nuclear fragmentation. These results indicate that upon activation of apoptotic caspases, podocytes are detached and lost in the urine before nuclear fragmentation and that the physical force of glomerular filtration facilitates detachment. This phenomenon may be the reason why definitive apoptosis is not observed in podocytes in vivo.