Structural determinants of GAD antigenicity.

Our aim was to ascertain structural determinants of autoantigenicity based on the model of the diabetes autoantigen glutamic acid decarboxylase 65 kDa isoform (GAD65) in comparison with that of the non-autoantigenic isoform GAD67. This difference exists despite the two isoforms having the same fold and high sequence identity. Autoantibodies to GAD65 precede the development of type 1 diabetes and are clinical markers of this and certain neural autoimmune diseases. To date, epitope mapping has been based on particular amino acid differences between the two isoforms, and there is no explanation as to why autoantibodies that react with GAD65 only infrequently cross-react with GAD67. To characterize each isoform of the enzyme and gain insights into their contrasting autoantigenic properties, we have used the recently determined crystal structures of GAD65 and GAD67 to compare their structure, hydrophobicity, electrostatics, flexibility and physiochemical properties. The results revealed striking differences which appear almost exclusively at the C-terminal domain of the isoforms. Whereas GAD65 displayed a highly charged and flexible C-terminal domain containing numerous patches of high electrostatic and solvation energies, these characteristics were absent in the GAD67 molecule. Additionally, analysis indicated potential N-terminal and PLP domain binding sites surrounding the C-terminal domain of GAD65, a major region of autoantigenic activity, but not of GAD67. These features agree with good accuracy with published epitope-mapping data. Our analysis suggests that the high flexibility and charge of GAD65 in the C-terminal domain is coupled with the mobility of its catalytic loop, a property that is absolutely required for its enzymatic function. Thus, the structural features that distinguish GAD65 from GAD67 as a B cell autoantigen are related to functional requirements for its enzymatic mechanism. This could well apply to the various other enzyme autoantigens and, if so, these features could be used as the basis of future predictive strategies.