Allogeneic islet transplantation in Type I diabetic patients is considerably hampered by the variable outcome of islet isolation and purification. After collagenase digestion of the pancreas, islet isolation is traditionally performed under hypothermic conditions in physiological solutions such as Hanks and RPMI. The University of Wisconsin solution (UWS) has been shown superior for hypothermic preservation of the pancreas. We, therefore, compared the UWS and RPMI for canine islet isolation and subsequent purification in either a conventional hyperosmotic density gradient of dextran in Hanks, or a novel normosmotic density gradient of Percoll in UWS. The isolation solution did not affect islet yield before purification (51% of the native islet mass). Loss of amylase (30%) and swelling of the acinar cells were observed in RPMI. In contrast, no loss of amylase and slight shrinkage of the acinar cells were observed in the UWS. Cell swelling affected the density separation and viability of the cells. Dextran density separation resulted in a 15% purity and 41% recovery of the islets isolated in RPMI, as compared to a 93% purity and 52% recovery of islets isolated in UWS. Percoll density separation improved the purity (99%) and recovery (74%) of islets isolated in UWS. Islets isolated in UWS demonstrated a superior basal and glucose stimulated insulin release during perifusion. Electron microscopy demonstrated a well-preserved islet ultrastructure after isolation in both solutions--except for slightly swollen mitochondria after isolation in RPMI. Autotransplantation of islets in pancreatectomised dogs was successful both after isolation in UWS and RPMI. We conclude that prevention of cell swelling during isolation and purification in the UWS resulted in an improved yield of viable and consistent virtually pure islets. Prevention of cell swelling during islet isolation should facilitate the analysis and control of other factors affecting outcome in man.