While both isoforms of glutamic acid decarboxylase (GAD) function as important autoantigens in autoimmune diabetes mellitus-GAD65 in humans and GAD67 in the NOD mouse-GAD67 is not synthesized in human pancreatic islets and is thought not to be an autoantigen in human diabetes. We have recently shown, however, that human islets contain a GAD67 splice variant: GAD25. Given the evidence that GAD67 could be a key diabetogenic autoantigen in the NOD mouse and the high prevalence of GAD65 autoantibodies in human type 1 diabetes, it became important to ask whether there is also immune reactivity to GAD25 in type 1 diabetes-possibly implicating it in the pathogenesis of the disease-and whether GAD25 reactivity could, like GAD65 reactivity, function as a clinically useful marker for the disease. We also hypothesized that the presence of autoantibodies to the smaller splice variant could be a cause of the up to 30% prevalence of GAD67 autoreactivity associated with type 1 diabetes. We therefore analyzed GAD25 reactivity in 105 newly-diagnosed children with type 1 diabetes and 74 control subjects. While 14 (13%) of the diabetic subjects were positive for GAD67 autoantibodies, only 3 (3%) were positive for GAD25 reactivity, none of which were GAD67 antibody-positive. Analysis of reactivity to a GAD67 chimera was consistent with GAD67 binding activity being due to cross-reactive GAD65 antibodies. Immunostaining confirmed the presence of GAD25 in human islets, revealing GAD25-positive cells to be sparse. Our results indicate that autoreactivity to GAD25 is rare in newly diagnosed type 1 diabetes and does not underlie GAD67 reactivity.