The subcellular site where insulin is degraded by rat hepatocytes in vivo is controversial. While several potential insulin-degrading enzyme systems, each with its own characteristic cellular location, are known to exist in the liver, questions remain about which of them participates in the degradation of physiologic doses of insulin. These studies examine the proteases that degrade physiologic doses of [125I]-insulin in vivo to determine (1) when and where initial degradation occurs, and (2) which of the potential degradative enzymes is active. Following injection into the mesenteric veins of male rats, intact [125I]-insulin and its labeled degradation products were analysed by reverse-phase high-performance liquid chromatography (RP-HPLC) of biopsy homogenates. [125I]-insulin was rapidly degraded in vivo; the t 1 2 of degradation was approximately 2.7 minutes. To test for extracellular protease activity, an isolated perfused liver system was employed. [125I]-insulin (or [125I]-glucagon) uptake was controlled by changing the temperature of the perfusion medium. Five minutes after [125I]-insulin injection, surface-bound label was recovered in an acidic (pH 3.5) wash. In perfusions at 15°C, both the internalization and degradation of [125I]-insulin were inhibited; 7.2% of unbound hormone was degraded and 5.1% of surface-bound insulin was degraded. Only 11.4% of unbound insulin and 17.4% of surface-bound insulin were degraded at 35°C. In contrast, 95.5% of unbound glucagon and 89.9% of surface-bound glucagon were degraded at 35°C. Thus, although glucagon degradation occurs at the sinusoidal plasmalemma of perfused livers, the same membrane does not mediate the rapid degradation of insulin observed in vivo. Analysis of the RP-HPLC [125I]-insulin elution profiles from liver biopsy homogenates, and comparison of them to profiles produced by purified proteases, suggested that insulin protease is responsible for most hepatic degradation of physiologic doses of insulin. Some cathepsin D-like activity was also observed in vivo, confirming that two pathways exist for insulin metabolism. The time course over which insulin was degraded was more rapid than previous studies in vitro would have predicted. This suggests that most insulin was receptor-bound at the time of its initial degradation, and that the active protease was soluble and was introduced into endocytic peripheral endosomes within seconds after their formation.