Affinities of human NMDA receptor autoantibodies: implications for disease mechanisms and clinical diagnostics

Abstract

Anti-N-methyl-D-aspartate receptor (NMDAR) encephalitis is a common autoimmune encephalitis presenting with psychosis, dyskinesias, autonomic dysfunction and seizures. The underlying autoantibodies against the NR1 subunit are directly pathogenic by disrupting synaptic NMDAR currents. However, antibody titers correlate only partially with the clinical outcome, suggesting the relevance of other factors such as antibody affinity. We thus determined the binding curves of human monoclonal autoantibodies and patients' cerebrospinal fluid (CSF) against NR1-expressing HEK293 cells using flow cytometry. Antibody affinity was highly variable with binding constants (half-maximal concentration, c₅₀) ranging from 1 to 74 µg/ml for monoclonal antibodies. Comparing values of individual monoclonal antibodies with human CSF samples suggested that the CSF signal is predominantly represented by higher-affinity antibodies, potentially in a concentration range of NR1 antibodies between 0.1 and 5 µg/ml, roughly reflecting 1-10% of total CSF IgG in NMDAR encephalitis. Binding curves further depended on the CSF composition which must be considered when interpreting established clinical routine assays. Normalization of measurements using reference samples allowed high reproducibility. Accurate and reproducible measurement of NR1 antibody binding suggested that biophysical properties of the antibody might contribute to disease severity. Normalization of the data can be an elegant way to allow comparable inter-laboratory quantification of CSF NR1 antibody titers in autoimmune encephalitis patients, a prerequisite for use as surrogate markers in clinical trials. Based on our calculations, low-affinity antibodies can easily remain undetected in routine cell-based assays, indicating that their relation to clinical symptoms should be analyzed in future studies.

Keywords: Antibody affinity; Cerebrospinal fluid; Flow cytometry; Human monoclonal antibody; NMDA receptor encephalitis.

Fig. 1

Validation of binding assays using monoclonal human anti-NR1 autoantibodies. NR1–EYFP-transfected HEK293 cells (a) were gated for the population with the maximum 20% NR1–EYFP protein expression (dark green rectangle), determined by the fluorescence of EYFP (b). EYFP fluorescence correlated well with NR1 expression as determined by staining with the NR1-reactive antibody #003-102 [c, lowest/median/highest 20% EYFP-fluorescent populations (gated in b) shown in light/medium/dark green, respectively]. In contrast, NR1-reactive antibodies (exemplarily shown for #003-102 at 100 µg/ml) did not bind to control HEK cells transfected with EYFP only (d). The NR1-reactive antibody #003-102 showed a concentration-dependent right shift of the fluorescence curves (blue; 0.1, 1 and 100 µg/ml) compared to the background fluorescence (black; 100 µg/ml) of a human control monoclonal antibody #mGo53 [e, median fluorescence intensity (MFI) is shown by vertical dotted lines]

Fig. 2

Binding curves of monoclonal human anti-NR1 autoantibodies. MFI of all measurements were normalized and plotted (MFI ± SEM) against the concentration of the monoclonal antibodies, and sigmoid functions with the best fit were generated demonstrating large differences in NMDAR binding (a). The MFI_max was not correlated with the number of mutations in the antigen-binding site of the NR1 autoantibodies at the DNA (b; SHM somatic hypermutations) or protein level (c; AA amino acids)

Fig. 3

Binding curves of CSF samples from patients with NMDAR encephalitis. Normalized MFI signals (± SEM) show concentration-dependent binding of human CSF samples to NR1 protein (a). None of the binding curves reached their MFI_max plateau, indicating that CSF NR1 antibody concentrations were clearly below the saturation of the NR1 epitopes and that the binding constant c₅₀ cannot be calculated in these samples. The MFI did not correlate in this small patient cohort with patient age (b), modified Rankin scale at the time of CSF analysis (c) and the duration of the hospital stay (d)

Fig. 4

Binding curves of human monoclonal NR1 autoantibodies diluted in CSF. The binding constant c₅₀ was markedly reduced (blue arrow) or increased (green arrow) depending on which monoclonal NR1 antibody was diluted in the identical control CSF