The frequency of antibodies to GAD (anti-GAD) in insulin-dependent diabetes mellitus (IDDM) varies greatly according to the type of assay employed. We therefore examined the immunoassay characteristics of diabetic sera using GAD purified from porcine brain and shown to contain both isoforms. Sera from 38 patients with IDDM, including 1 patient with both stiff-man syndrome (SMS) and IDDM, were studied for anti-GAD by radioimmunoprecipitation (RIP), Western blotting, and dot-blotting. The sera were selected according to reactivity in the RIP assay. There was a good correlation between potency of the RIP reaction at the screening dilution of 1:2 and the endpoint dilution in the assay which ranged from 1:2 to 1:30,000 for IDDM sera, and 1:300,000 for the SMS serum. Of the 38 sera positive for anti-GAD by RIP, only 6 had antibodies detectable by Western blotting, and all gave an RIP titer of at least 1:250. The low frequency of antibodies by Western blotting was explicable by denaturation of the antigen. Thus, using a dot-blotting assay in which reactivity to untreated "native" GAD was compared with reactivity to GAD after denaturation by reduction with 2-mercaptoethanol and boiling, 20 of the 38 IDDM sera reacted unequivocally with the native GAD compared with only 2 that reacted with denatured GAD after reduction and boiling. The sera were tested for their capacity to inhibit the catalytic activity of GAD, but only the high-titer serum from the patient with SMS did so. Our study further validates the RIP assay for anti-GAD and establishes that anti-GAD exists in IDDM over a wide range of titers that correlate with other assay characteristics, and also indicates that the conformational autoantibody epitope on GAD is susceptible to alteration under denaturing conditions.