A lthough glutamic acid decarboxylase (GAD) has been implicated in IDDM, there is no direct evidence showing GAD-reactive T cells are diabetogenic in vivo. To address this issue, 3-wk-old NOD mice received two injections of purified rat brain GAD; one mouse rapidly developed diabetes 3 wk later. Splenocytes from this mouse showed a proliferative response to purified GAD, and were used to generate a CD4+ T cell line, designated 5A, that expresses TCRs encoding Vbeta2 and Vbeta12. 5A T cells exhibit a MHC restricted proliferative response to purified GAD, as well as GAD65 peptide 524-543. After antigen-specific stimulation, 5A T cells secrete IFNgamma and TNFalpha/beta, but not IL-4. They are also cytotoxic against NOD-derived hybridoma cells (expressing I-Ag7) that were transfected with rat GAD65, but not nontransfected hybridoma cells. Adoptive transfer of 5A cells into NOD/SCID mice produced insulitis in all mice. Diabetes occurred in 83% of the mice. We conclude that GAD injection in young NOD mice may, in some cases, provoke diabetes due to the activation of diabetogenic T cells reactive to GAD65 peptides. Our data provide direct evidence that GAD65 autoimmunity may be a critical event in the pathogenesis of IDDM.