Autoimmune reactions, in which the body's white blood cells harm its own tissues, cause many diseases including diabetes, multiple sclerosis and arthritis. It is not known why immune cells target certain organs, and in particular for childhood diabetes, why only insulin-producing cells are killed. Nakayama et al. now report that this may be because insulin itself is a primary autoantigen for autoimmune diabetes. In NOD mice, the standard animal model for diabetes, when the part of the insulin molecule that gives rise to autoantibodies is altered, autoimmune diabetes disappears. This also suggests that deletional immune therapy could be a practical proposition. The possible clinical relevance of this work is confirmed by a separate study by Kent et al. of human patients with type 1 diabetes. T lymphocytes found in the draining lymph nodes around the pancreas specifically recognize part of the insulin protein. This has implications for antigen specific therapies and islet-cell transplantation in diabetes. A fundamental question about the pathogenesis of spontaneous autoimmune diabetes is whether there are primary autoantigens. For type 1 diabetes it is clear that multiple islet molecules are the target of autoimmunity in man and animal models1,2. It is not clear whether any of the target molecules are essential for the destruction of islet beta cells. Here we show that the proinsulin/insulin molecules have a sequence that is a primary target of the autoimmunity that causes diabetes of the non-obese diabetic (NOD) mouse. We created insulin 1 and insulin 2 gene knockouts combined with a mutated proinsulin transgene (in which residue 16 on the B chain was changed to alanine) in NOD mice. This mutation abrogated the T-cell stimulation of a series of the major insulin autoreactive NOD T-cell clones3. Female mice with only the altered insulin did not develop insulin autoantibodies, insulitis or autoimmune diabetes, in contrast with mice containing at least one copy of the native insulin gene. We suggest that proinsulin is a primary autoantigen of the NOD mouse, and speculate that organ-restricted autoimmune disorders with marked major histocompatibility complex (MHC) restriction of disease are likely to have specific primary autoantigens.