N‐methyl‐D‐aspartate receptor dysfunction by unmutated human antibodies against the NR1 subunit

Recombinant Monoclonal NMDAR Antibodies

Recombinant monoclonal human NR1 IgG autoantibodies were generated as described.7, 8 The study was approved by the Charité University hospital Review Board, and informed consent was obtained from each subject. The control antibody (mGo53) is a nonreactive isotype‐matched human antibody.9 Immunostaining, using primary hippocampal neurons, unfixed mouse brain sections, HEK cell‐expressed NR1 N368Q mutants, and brain sections after intravenous antibody injection, followed our established protocols.7 We generated germline counterpart versions from maturated monoclonal NR1 antibodies with best matching variable V(D)J genes and elimination of somatic hypermutations.10 Relative affinity curves were calculated based on concentration‐dependent antibody binding to hippocampal sections, adapted from previous work using HEK cells transfected with the NR1 subunit.11 For bilateral intracerebroventricular injection, 200 μg of antibodies were infused over 14 days using osmotic minipumps.12

Super‐Resolution Imaging

Direct stochastic reconstruction microscopy (dSTORM) was performed in primary hippocampal neurons (DIV 14) as described.12 Human NR1 autoantibody (clone #003‐109; 4 μg/ml) and polyclonal guinea pig anti‐Homer‐1 (1:300; Synaptic Systems GmbH, Goettingen, Germany) were used as primary antibodies followed by AlexaFluor‐647 goat/anti‐human (1:200; Life Technologies, Carlsbad, CA) and CF‐568 donkey/anti–guinea pig (1:200; Biotium, Fremont, CA) as secondary antibodies.

Electrophysiological Recordings

Autaptic murine hippocampal neurons (DIV 14‐17) were incubated with 1 or 5 μg/ml human NR1 (#003‐109) or control antibody at 37°C for 3 or 24 hours. Data were acquired as described.13 Cells were recorded in standard intra‐ and extracellular solutions, except for chemically induced NMDA responses, measured in extracellular solution containing 0 mM of Mg²⁺, 0.2 mM of CaCl₂, and 10 μM of glycine, and evoked NMDA responses, measured in extracellular solution containing 0 mM of Mg²⁺, 2 mM of CaCl₂, 10 μM of glycine, and 10 μM of NBQX (Tocris Bioscience, Bristol, UK). For kinetics of synaptic NMDA responses, non‐silent traces from each cell were averaged and rise time and decay time constant (τ) measured from 10% to 90% or 90% to 10% of the peak, respectively. Decays were fitted with a double exponential and decay time constants for each of the fits converted to a weighted decay.

Homology Modeling

The homology model of the ATD of the human NMDA receptor was generated using the crystal structure of the rat NMDA receptor subunit, zeta‐1. The homology modeling application of MOE 2014.09 (“Molecular Operating Environment [MOE], 2014.09”, 2015) was used with 10 main chain models, each with one side chain. Samples were built using the amber12 force field.14

Germline NR1 Autoantibodies Target the NMDAR in vitro and in vivo

The CSF autoantibody repertoire in NMDAR encephalitis contains NR1‐binding and non‐NR1‐binding antibodies.7 Across all 8 patients, NR1‐binding antibodies had significantly lower numbers of somatic hypermutations (SHM) in the Ig heavy (5.1 ± 4.0 versus 11.9 ± 8.3) and corresponding Ig light chains (3.9 ± 4.8 versus 7.2 ± 5.4) than non‐NR1‐binding antibodies (total, 9.0 ± 7.9 versus 19.1 ± 12.6 mean ± SD; p = 0.018, unpaired t test). Individual NR1 antibodies were even completely unmutated (#003‐109) or contained only silent SHM (#007‐142, #007‐169).7

The germline antibody, #003‐109, accounted for 1 of 41 (2.4%) of antibody‐secreting cells analyzed of this patient and showed the characteristic NR1 pattern on unfixed mouse brain sections (Fig 1A) and the NMDAR cluster distribution on primary hippocampal mouse neurons previously observed for mutated NR1 antibodies7 (Fig 1B). dSTORM of hippocampal neurons demonstrated NMDAR distribution at synapses, opposed to Homer1‐positive postsynaptic densities (Fig C). Intravenous injection of #003‐109 resulted in binding to cerebellar granule cells in vivo (Fig 1D), which was not detectable with the isotype control (Fig 1E). Intracerebroventricular injection of maturated (Fig 1G) and germline (Fig 1H) NR1 antibodies showed similar neuropil binding in the hippocampus.

Binding of Mutated, Germline, and Reverted Antibodies to NR1

Binding of #003‐109 was prevented by the single‐amino‐acid mutation, N368Q, in the ATD of NR1 (Fig 2A,B).7 To model the possible antibody/ATD interaction in silico, protein‐protein docking was performed against ATD using ClusPro (Fig 2C). ATD residues N368 and G369 are embedded in the protein‐protein interface forming a network of H‐bonds with the antibody residues (Fig 2D). N368 is the only amino acid on the receptor side that stabilizes the binding to both, the antibody heavy and the light chains (Fig 2D). Binding curves of patient‐derived germline NR1 antibodies showed generally lower relative affinity to hippocampal sections than mutated antibodies (Fig 2E–G). However, reverting mutated patient antibodies to germline changed the binding strengths only minimally, suggesting a similar functional role already of the naïve antibodies (Fig 2H).

Unmutated NR1 Antibodies Selectively Reduced Total and Synaptic NMDAR Currents

We expected smaller electrophysiological changes induced by antibody #003‐109 compared to mutated NR1 antibodies, given the lower binding to murine brain (Fig 2E,F). Indeed, incubation of autaptic mouse hippocampal neurons with 5 μg/ml of germline antibody #003‐109 for 24 hours resulted in ~30% reduction of the total NMDA currents compared to the isotype control antibody (Fig 3A). The antibody effect was specific, given that the total gamma‐aminobutyric acid (GABA)‐ and kainate‐mediated currents remained unaffected after application of 5 μM of GABA or 20 μM of kainate, respectively (Fig 3B,C). Measuring synaptic responses in the presence of 10 μM of glycine,10 μM of NBQX, and 0 mM of Mg²⁺, NR1 antibody treatment also reduced synaptic NMDA currents by ~45%, while synaptic alpha‐amino propionic acid (AMPA) currents did not differ from controls (Fig 3D,E). In contrast to higher‐affinity mutated NR1 antibodies,7 the effects were not detectable with lower antibody concentrations (1 μg/ml of #003‐109; Fig 3F–J). In addition, shorter incubation for 3 hours with germline antibody was not sufficient to reduce synaptic or whole‐cell NMDA currents (Fig 3K–O). The kinetics of synaptic NMDA responses were not altered by NR1 antibody treatment (Fig 3P–R).