Insulin degrading enzyme (IDE), a zinc metalloprotease, can specifically recognize and degrade insulin, as well as several amyloidogenic peptides such as amyloid beta (Abeta) and amylin. The disruption of IDE function in rodents leads to glucose intolerance and cerebral Abeta accumulation, hallmarks of type 2 diabetes and Alzheimer's disease, respectively. Using limited proteolysis, we found that human IDE (113kDa) can be subdivided into two roughly equal sized domains, IDE-N and IDE-C. Oligomerization plays a key role in the activity of IDE. Size-exclusion chromatography and sedimentation velocity experiments indicate that IDE-N is a monomer and IDE-C serves to oligomerize IDE-N. IDE-C alone does not have catalytic activity. It is IDE-N that contains the crucial catalytic residues, however IDE-N alone has only 2% of the catalytic activity of wild type IDE. By complexing IDE-C with IDE-N, the activity of IDE-N can be restored to approximately 30% that of wild type IDE. Fluorescence polarization assays using labeled insulin reveal that IDE-N has reduced affinity to insulin relative to wild type IDE. Together, our data reveal the modular nature of IDE. IDE-N is the catalytic domain and IDE-C facilitates substrate recognition as well as plays a key role in the oligomerization of IDE.