Insights into autoimmunity: structural basis of autoantigenicity of GAD

Glutamic acid decarboxylase (GAD), the pyridoxal 5’-phosphate (PLP) dependent enzyme that catalyses the synthesis of the inhibitory neurotransmitter !-aminobutyric acid (GABA) occurs as two isoforms, GAD65 and GAD67. GAD is found in the brain and substantial amounts of GAD65 and GAD67 are also found in the !-islet cells of the pancreas, epithelial cells of oviduct and spermatocytes in the testes. The 67 kDa isoform, GAD67, is constitutively active and produces >90% of GABA in the central nervous system, whereas GAD65 is transiently activated to produce GABA in response to a sudden increase in demand. According to their linear sequence, the first 100 amino acids in the two isoforms differ substantially, but the isoforms show 74% sequence identity within the middle and the Cterminal domain. The association and dissociation of PLP plays an important role in the regulatory function of the GAD enzymes and GAD65 exists predominantly as an apoenzyme (no PLP attached in the enzyme catalytic site), but GAD67 exists predominantly in its active holoform. There are structural differences between the apo- and holo-forms of the enzyme. Antibodies to GAD65 (anti-GAD65) occur frequently in type 1 diabetes (T1D) but GAD67 is rarely autoantigenic in T1D and these antibodies (anti-GAD67) may represent a cross reactive population of anti-GAD65. The recently elucidated crystal structures of GAD65 and GAD67, N-terminally truncated by 87 (GAD65) or 92 (GAD67) amino acids revealed flexible disordered regions at the C termini on GAD65 but not in GAD67, and B-cell epitopes for various human monoclonal antibodies (hmAb) are located on the flexible regions of GAD65. The hypothesis for this work is that differences in the two isoforms in structural flexibility, or structural changes resulting from interaction with the PLP co-factor provide the basis for the difference in autoantigenicity of the GAD isoforms. Based on the hypothesis the project is designed to have three different aims; 1. To confirm that GAD65 is the major autoantigen in T1D, and that anti-GAD67 represents a cross-reactive subset of anti-GAD65. 2. To design mutants of GAD65 to stabilize the C-terminal domain of GAD65, and to compare their reactivity with T1D sera with that of wild type GAD65. 3. To use computational modelling and computational docking to compare the reactivities of the hMabs b96.11 and b78 with apo- and holo-GAD. The levels, frequency, affinity and nature of cross reactivity of anti-GAD67 in sera from patients with T1D or LADA, that contained anti-GAD65 were measured by radioimmunoprecipitation (RIP) using 125I-labeled recombinant GAD65 and GAD67. For each antibody population, the specificity of the binding was measured by incubation with 100-fold excess of unlabeled GAD in homologous and heterologous inhibition assays, and the affinity of binding with GAD65 and GAD67 was measured in selected sera. Sera were also tested for reactivity to GAD65 and GAD67 by immunoblotting. Of the 85 sera that contained anti-GAD65, 28 contained anti–GAD67 measured by RIP. Inhibition with unlabeled GAD65 substantially or completely reduced antibody reactivity with both 125I GAD65 and with 125I GAD67. In contrast, unlabeled GAD67 reduced autoantibody reactivity with 125I GAD67 but not with 125I GAD65. Both populations of antibodies were of high affinity. Immunoblotting experiment indicated that the epitopes for anti-GAD67 are conformational, like that of anti- GAD65. Competitive inhibition assay results suggest that for most sera anti-GAD67 represents a minor population of anti-GAD65 reactive with an epitope shared between GAD65 and GAD67. So our findings lead to the conclusion that the anti-GAD67 represents a cross reactive population of anti-GAD65. The effect of flexibility on the autoantigenicity of holo-GAD65 was tested using a combination of molecular dynamic simulation and mutagenesis techniques. Mutants were designed based on molecular dynamic simulations performed on the wild type GAD isoforms GAD65 and GAD67. These simulations predicted a flexible α–helix 14 on GAD65 in the Cterminal domain. Potential amino acid residues in and around the α–helix 14 which could be the main cause for the flexibility were identified. Two single mutants (GAD65H432A and GAD65P531A) and two double mutants in which the C-terminus was stabilized by disulphide bonds (GAD65H432C_P531C and GAD65I509C_V529C), were tested for structural and functional stability by CD spectroscopy and GAD enzyme assay, and tested for reactivity with anti- GAD65 in sera by RIP. All the four mutants of GAD65 along with the wild type GAD65 and GAD67 were tested for conformational changes and thermal stability using circular dichroism on a Jasco J-815 spectrometer equipped with a Jasco Petlier type temperature controller (JASCO CDF-426S/15). The mutants GAD65H432A, GAD65H432C_P531C and GAD65I509C_V529C were expressed and purified similarly to WT-GAD65 and showed similar conformations to WT-GAD65. The mutants retained the same fold and secondary structure composition as that for the native WT GAD65 with the expected increased conformational stability, measured as increased thermal melting temperatures (Tm) on CD spectroscopy, and enzyme activity measurements showed that each retained some activity under the reducing conditions of the assay. Nonetheless, anti-GAD65 reactivity by RIP with each of the mutants was substantially or completely lost. The fourth mutant GAD65P531A expressed poorly, lacked enzymatic activity and reactivity by RIP, and was thought to be miss-folded. Taken together these results indicate that C-terminal flexibility is an important component of the antigenicity of GAD65. The flexibility of GAD65 could include changes in structure between holo- and apo- GAD65, and the recent publication of a crystal structure for the apo- form of the related PLPdependent decarboxylase, dopa-decarboxylase, DDC, has provided the opportunity to develop an homology model for apo-GAD65. To determine whether structural changes during binding of PLP are likely to affect antigenicity computational docking was used to compare the reactivity of human monoclonal antibody (hmAb) b78 and b96.11 with holo- and apo- GAD65. Homology models of Fv b78 and Fv b96.11 were docked to structures of holo- GAD65 and modelled apo-GAD65 using the rigid-body docking program ZDOCK, followed by refinement with RDOCK. The 30 most favourable complexes from RDOCK and 10 most favourable from ZDOCK were further analysed based on previously published mutagenesis data and antibody CDR interactions. For comparison, the reactivity of the two mAbs and sera from 14 patients with diabetes were also tested by RIP with GAD65 in the presence of PLP (holo-GAD65), or 5mM glutamate (apo-GAD65). For each mAb, b78 and b96.11, docking predicted epitope locations on both holo-GAD65 and apo-GAD65, but there were differences. For b78, on holo-GAD65 the most favourable epitope region predicted included C-terminal and catalytic loop residues, consistent with its enzyme inhibitory nature. On apo-GAD65, the most favourable complexes involved interactions of three or four acidic C-terminal residues with basic residues in CDR L1 and CDR L2. More extensive interactions were rare, but were similar to those on holo-GAD65, suggesting that two-stage antibody docking could occur with apo-GAD65, initially involving a few strong interactions, followed by more complex interactions occurring with main chain movements and induced fit. For b96.11, interactions were similar with both holo- and apo-GAD65 and involved docking a hydrophobic knob within a pocket on GAD65. RIP results showed that anti-GAD65 reactivity of diabetes sera, or of the mAbs b78 and b96.11, is not affected by the presence or absence of co-factor PLP in the active site of GAD65. These results provide strong evidence that GAD65 is the autoantigenic isoform in T1D, and that anti-GAD67 results from the presence of a minor population of cross-reactive autoantibodies. The structural basis for the antigenicity of GAD65 can be related to flexibility in the C-terminal domain and the adjacent catalytic loop that is associated with the autoinactivation of GAD65 and formation of apo-GAD65 during enzymatic activity. Nonetheless, the formation of apo-GAD65 does not apparently alter the antigenicity of GAD65, and it is likely that the flexibility in the C-terminal domain is retained in the apo- GAD. Flexibility of antigen and/or antibody is frequently required for strong antibody binding, and this study suggests that it plays a major part in the development of an autoimmune response to GAD65 in T1D. It is notable that many autoantigens in autoimmune diseases are enzymes, and that autoantibodies to these enzymes are frequently strongly enzyme inhibitory, and bind within the catalytic domain, a region in an enzyme which is likely to be the most mobile. A requirement for flexibility and mobility for antigenicity could explain the development of such autoimmune responses.