Amyloid beta-protein (Abeta) has been implicated as an early and essential factor in the pathogenesis of Alzheimer's disease. Although its cellular production has been studied extensively, little is known about Abeta clearance. Recently, insulin-degrading enzyme (IDE), a 110-kDa metalloendopeptidase, was found to degrade both endogenously secreted and synthetic Abeta peptides. Surprisingly, IDE-mediated proteolysis of [(125)I]Abeta(1-40) in microglial cell-culture media was accompanied by the formation of (125)I-labelled peptides with higher apparent molecular masses, raising the possibility that the degradation products act as 'seeds' for Abeta oligomerization. To directly address the role of IDE in Abeta degradation and oligomerization, we investigated the action of purified recombinant wild-type and catalytically inactive IDEs. Our data demonstrate that (i) IDE alone is sufficient to cleave purified Abeta that is either unlabelled, iodinated or (35)S-labelled; (ii) the initial cleavage sites are His(14)-Gln(15), Phe(19)-Phe(20) and Phe(20)-Ala(21); and (iii) incubation of IDE with [(125)I]Abeta, but not with [(35)S]-Abeta, leads to the formation of slower migrating species on gels. Since iodination labels N-terminal fragments of Abeta, and (35)S labels C-terminal products, we analysed unlabelled synthetic fragments of Abeta and determined that only the N-terminal fragments migrate with anomalously high molecular mass. These results indicate that IDE alone is sufficient to degrade Abeta at specific sites, and that its degradation products do not promote oligomerization of the intact Abeta peptide.