Glutamic acid decarboxylase 65 (GAD65) is the principal autoantigen of type 1 diabetes (T1D): autoantibodies to GAD65 (GADA) are among the earliest and most predictive markers of disease, and GAD65 remains a leading antigen-specific immunotherapy candidate. GAD65’s immunogenicity is inseparable from its biochemistry — it is an unusually unstable enzyme that readily loses its pyridoxal-5′-phosphate (PLP) cofactor to form a conformationally mobile apo species, and the flexible catalytic-loop region implicated in this apo↔holo cycle overlaps known conformational GADA epitopes. Yet the electronic events that drive this chemistry have never been modelled from first principles at the active site: conventional quantum chemistry is too costly for reactive scans at this size, and force fields cannot represent bond making/breaking or cofactor electronics.
We propose to apply RealQM, a parameter-free real-space electronic-structure method (no basis sets, no fitted force field, GPU-accelerated) that resolves reactive chemistry cheaply enough to scan an enzyme active site. In preliminary work RealQM, given only the active-site nuclei, spontaneously reproduced the textbook PLP electron-sink mechanism of GAD65 decarboxylation. We will (Aim 1) put this mechanism on a quantitative footing and extend it to the holo-vs-apo active site; (Aim 2) couple active-site electronics to catalytic-loop conformation and epitope exposure; and (Aim 3) build atomistic conformational-epitope models for GADA recognition — the first first-principles electronic picture linking GAD65 catalysis, cofactor loss, and autoantigenicity.
Convert the qualitative preliminary result into calibrated, reproducible profiles of the decarboxylation coordinate. Compare PLP-bound (holo) vs PLP-free (apo) active sites to test whether loss of the electron sink not only abolishes catalysis but reshapes the local electronic landscape the catalytic loop responds to. Deliverable: validated holo/apo profiles benchmarked against high-level reference calculations.
Map how the holo↔apo electronic difference propagates to the mobility of the catalytic-loop region overlapping conformational GADA epitopes; test whether the apo electronic state correlates with increased exposure of autoantibody-recognized residues. Deliverable: a residue-resolved link between cofactor state and epitope accessibility.
Replace the placeholder linear-peptide dock with real anti-GAD65 CDR sequences and the genuine conformational GAD65 epitope surface; characterize paratope–epitope recognition. Deliverable: atomistic recognition models for the major conformational GADA epitopes, ranked by interface energetics.
Relation to prior computational work. PLP decarboxylation has been characterized by DFT QM-cluster and QM/MM studies (e.g. histidine decarboxylase; the group-II decarboxylase family, anchored by crystallographic Dunathan/quinonoid snapshots; LigW and OMP-decarboxylase benchmarks), which yield calibrated kcal/mol barriers and transition states for the generic mechanism. We do not aim to re-derive that step: we use it as a calibration anchor (Aim 1 benchmarks RealQM against high-level DFT/QM-MM on the catalytic coordinate) and then extend — with a parameter-free method carrying no basis set and no exchange-correlation functional (DFT’s principal systematic-error source) and cheap enough to scan — into the GAD65-specific apo/holo and conformational-epitope regime the DFT/QM-MM literature has left largely unaddressed.
A reduced gas-phase GAD65/PLP active-site model (pyridinium electron-sink mimic + 3-OH + 4′-imine + Dunathan-aligned glutamate external aldimine; 24 atoms, 160³ grid) was scanned along the breaking Cα–CO2 coordinate for two environments — PLP-conjugated vs. an identical free amino acid (no sink). RealQM relaxed the electron density at each clamped geometry.
| R (Å) | ΔE free | ΔE PLP | sink = free − PLP |
|---|---|---|---|
| 1.55 | 0.000 | 0.000 | 0.00 |
| 2.00 | −0.012 | −1.916 | +1.90 |
| 3.30 | −2.560 | −5.426 | +2.87 |
| 4.20 | −3.101 | −6.605 | +3.50 |
| 6.00 | −3.454 | −7.620 | +4.17 |
(model energy units; sign/trend interpreted, magnitudes not yet calibrated — see Rigor)
The free curve saturates (≈ −3.5 beyond ~4 Å) while the PLP curve keeps descending (to −7.6 at 6 Å); the sink stabilization grows monotonically to ~+4. Given only coordinates and no built-in reaction coordinate, the method reproduces PLP’s electron-sink role, and the identical-substrate control attributes the effect specifically to the conjugated ring. This establishes feasibility for Aim 1 and motivates the holo/apo contrast.