Autoantibodies to the islet-cell 65-kDa variant of glutamate decarboxylase (GAD65) are found in most insulin-dependent diabetes mellitus (IDDM) patients many years before the appearance of clinical symptoms of the disease. As IDDM-preventive therapies may be available in the future, an international effort is taking place to develop widely applicable anti-GAD immunochemical tests. These tests would help to detect individuals at risk before the full installation of the disease and to enroll them in prevention programs. Autoantibodies to GAD65 are mostly directed to conformational epitopes, and the enzyme is a complex molecule with a prosthetic group and 15 cysteine residues. Thus, the conformational integrity of GAD65 is essential for an appropriate anti-GAD assay. Isolation of large amounts of GAD65 from pancreas or other tissues is impractical, and no successful production of properly folded GAD65 has been reported in bacteria. Native recombinant GAD65 for immunochemical tests is usually obtained from eukaryotic expression systems. Since the large-scale production of a recombinant protein in an eukaryotic system is expensive and technically difficult, we investigated the expression of GAD65 in Escherichia coli as an alternative. A number of DNA constructs intended to export the enzyme to the periplasmic space or to improve its cytoplasmic solubility were designed and tested. Our results provide a solution to the two main problems associated with the expression of GAD65 in E. coli: misfolding, leading to the formation of inclusion bodies; and the presence of alternative initiation sites for translation that causes the preferential production of truncated variants of GAD65. We describe here the production of properly folded, fully active, and immunochemically competent GAD65 as an N-terminal fusion protein with thioredoxin. An account of the reactivity of the produced protein with sera of six IDDM patients is also presented.