Ketosis-prone diabetes is heterogeneous. Its causes could include novel beta-cell functional defects. To characterize such defects, 103 patients with diabetic ketoacidosis were evaluated for beta-cell autoimmunity and human leukocyte antigen (HLA) class II alleles, with longitudinal measurements of beta-cell function and biochemical and clinical parameters. They were classified into four A beta groups, based on the presence of glutamic acid decarboxylase (GAD)65, GAD67, or IA-2 autoantibodies (A+ or A-) and beta-cell functional reserve (beta+ or beta-). The group distribution was: 18 A+beta-, 23 A-beta-, 11 A+beta+, and 51 A-beta+. Collectively, the two beta- groups differed from the two beta+ groups in earlier onset and longer duration of diabetes, lower body mass index, less glycemic improvement, and persistent insulin requirement. HLA class II genotyping showed that the A-beta- group differed from the A+beta- group in having lower frequencies of two alleles strongly associated with autoimmune type 1 diabetes susceptibility: DQA*03 and DQB1*02. Similarly, the A-beta+ group differed from the A+beta+ group in having a lower frequency of DQB1*02. Ketosis-prone diabetes comprises at least four etiologically distinct syndromes separable by autoantibody status, HLA genotype, and beta-cell functional reserve. Novel, nonautoimmune causes of beta-cell dysfunction are likely to underlie the A-beta+ and A-beta- syndromes.