Dynamic disorganization of synaptic NMDA receptors triggered by autoantibodies from psychotic patients

The identification of circulating autoantibodies against neuronal receptors in neuropsychiatric disorders has fostered new conceptual and clinical frameworks. However, detection reliability, putative presence in different diseases and in health have raised questions about potential pathogenic mechanism mediated by autoantibodies. Using a combination of single molecule-based imaging approaches, we here ascertain the presence of circulating autoantibodies against glutamate NMDA receptor (NMDAR-Ab) in about 20% of psychotic patients diagnosed with schizophrenia and very few healthy subjects. NMDAR-Ab from patients and healthy subjects do not compete for binding on native receptor. Strikingly, NMDAR-Ab from patients, but not from healthy subjects, specifically alter the surface dynamics and nanoscale organization of synaptic NMDAR and its anchoring partner the EphrinB2 receptor in heterologous cells, cultured neurons and in mouse brain. Functionally, only patients’ NMDAR-Ab prevent long-term potentiation at glutamatergic synapses, while leaving NMDAR-mediated calcium influx intact. We unveil that NMDAR-Ab from psychotic patients alter NMDAR synaptic transmission, supporting a pathogenically relevant role.

(a) Immunostaining of HEK293 cells expressing GluN1-GFP and GluN2B subunits with the sera of healthy subjects or schizophrenic patients (1/10, 3h incubation). Note the overlap between serum reactivity (red) and GFP-positive HEK cells (green) for seropositive samples. Scale bar, 10 µm. Lower panel: Immunostaining of IgLON5-GFP+ HEK293 cells with the sera of Healthy-, PSY+ or IgLON5+ subjects. Note the absence of staining in presence of serum from Healthy-and PSY+ samples. Scale bar, 10 µm. (b) Surface co-immunostaining for GluA1-SEP containing AMPAR (red) and human IgG's target (5µg/ml, green) in live hippocampal neurons (11 div). Scale bar, 2µm. Right panel: corresponding linescans plotting GluA1-SEP and human IgG fluorescence intensities over distance (pixel). No overlap in any of the conditions was observed. (c) Representative dendritic areas of cultured hippocampal neurons labelled with purified IgG (5 µg/ml, green) from Healthy+ or PSY+ samples and Homer-1c (red) which localizes glutamate postsynaptic densities. Both staining reveal a clustertype distribution with good co-localization with the synaptic area. Scale bar, 20 µm. (d) Immunostaining of mice hippocampal slices (P22) incubated with a commercial αGluN1 N-term antibody (20 µg/ml, 3h) or purified IgG from Healthy+ (20 µg/ml), PSY+ (20 µg/ml) or an anti-human Alexa 488 without any primary antibody (no primary Ab). Note the similar staining pattern between all conditions in presence of primary antibodies. Scale bar, 50 µm. Representative fluorescence staining after incubation with a PSY+ NMDAR-Ab (IgG1, green) at different concentrations (0.5, 5, or 50 µg/ml, overnight, 4°C) followed by incubation with the same IgG (IgG2 red, 5 µg/ml, overnight, 4°C). Scale bar, 400 nm. Note that IgG1 were first coupled with a secondary anti-human Alexa 488 and then remaining antigen binding sites were blocked using anti-human Fab fragments to reduce unspecific labelling. The decrease in IgG2 staining indicates the lack of binding sites for IgG2. (c) Quantification of mean fluorescence intensity of IgG2 within IgG1 cluster areas and comparison using different concentrations of IgG1: 0.5 µg/ml (n = 25 cluster fields, N = 14 neurons), 5 µg/ml (n = 30, N = 16) and 50 µg/ml (n = 33, N = 14).
After 4h incubation with the DNA mix, medium was replaced by fresh, equilibrated and heated supplemented DMEM medium. The following day, either transfected or non-transfected HEK cells were incubated for 1h with

Primary cell culture and protein expression. Cultures of hippocampal neurons were prepared from E18
Sprague-Dawley rats. Cells were plated at a density of 50 x 10 3 cells per ml on poly-lysine pre-coated coverslips.
Coverslips were maintained in a 3% horse serum containing Neurobasal medium (Invitrogen). After a few days in vitro (div), the original plating medium was replaced by a serum-free medium. Cultures were maintained at 37°C in 5% CO2 for 15 div at maximum. For exogenous protein expression, 7-10 div hippocampal cultured neurons were transfected at least 48h before each experiment using either the Effectene (Qiagen) or phosphate calcium transfection 3 .

Immunochemistry
Immunohistochemistry. The pattern staining obtained with human IgG was assessed using immunohistochemistry on hippocampal slices. For this, 22 postnatal day-old mice were perfused with 4% PFA.
Coronal sections of 50 μm were obtained on a vibratome (Leica) and incubated overnight at 4°C with either a polyclonal antibody against the N-terminal part of the GluN1 subunit (αGluN1N-term Alomone Labs, 20 μg/ml) or purified IgG from healthy (Healthy+, 20 μg/ml) and psychotic (PSY+, 20 μg/ml) subjects. Fluorescent revelation was carried out with secondary anti-rabbit or anti-human Alexa 488 antibodies (Life Technologies, 1/1000) for 2h at room temperature. Images were obtained using a Nanozoomer and a confocal microscope (SP8, Leica).
In another series, brains (provided by A. Ramsey) from wild-type and GluN1-KD animals at 14 weeks were perfused and stored at -20°C. Coronal tissue sections of hippocampal areas (20 µm thick) were cut on a microtome-cryostat, thaw-mounted onto Thermo Scientific, SuperFrost Ultra Plus adhesion slides, and stored at -20°C until further processing. Sections were fixed at 4°C in 4% paraformaldehyde. Blocking was carried out in a 1xTBS solution in 0.3M glycine containing 10% normal goat serum (Sigma) and incubation with PSY+ purified IgG (5 μg/ml, pooled from 2 different patients) was done in a 1xTBS solution containing 10% normal goat serum overnight at 4°C. Staining with secondary antibody anti-human Alexa 568 (Invitrogen, 1/500) was performed for 1h and slides were mounted with Vectashield + Dapi (Vector Laboratories). Image acquisition was done on a video spinning-disk system (Leica DMI6000B, 40X).
Immunocytochemistry. The pattern staining obtained with human IgG was also assessed in dissociated cells.

Immunocompetition
Brain tissue. Snap frozen, non-perfused 7 µm thick sagittal sections of Sprague-Dawley rat brains were used for this experiment. Sections of tissue were fixed (4% PFA, 15min), washed in PBS, and incubated with undiluted serum from patients with schizophrenia, anti-NMDAR encephalitis, or blood donors overnight at 4°C.
Sections were then extensively washed with cold PBS and incubated for 1h at 4°C with biotinylated IgG from a representative patient with anti-NMDAR antibodies. After washing, the binding of biotinylated IgG was demonstrated with a standard avidin-biotin-peroxidase method (Vectastain ABC kit Elite, PK-6100, Vector).
Slides were then mildly counterstained with hematoxylin, mounted, and results photographed with a digital camera (AxioCam MRc) adapted to a confocal microscope (Zeiss LSM710).
Samples were illuminated in TIRF mode and images were obtained with an exposure time of 10.85 ms with up to 100,000 consecutive frames. Imaging was carried out at room temperature in a closed Ludin chamber (Life Imaging Services) using pH-adjusted extracellular solution containing oxygen scavengers and reducing agents.
Image acquisition was controlled by the Leica LAS software. An initial high power 642 nm laser was used to convert the fluorescence into the dark state in order to reach a desired density for single molecule detection. Multicolor fluorescent microbeads (Tetraspeck, Invitrogen) were used as fiduciary markers to correct for lateral drifts. GluN2A subunit or PSD-95 protein clusters were identified on the respective epifluorescence images.
Structures with a higher intensity than a respective value were identified as clusters. GluN2A subunit nanoobjects area and shape were quantified after segmentation of GluN2A subunit dSTORM images (MetaMorph software, Molecular Devices). Morphological features, such as surface area, length and shape of each segmented structure, were exported to calculate their respective distributions. The dimensions were computed by 2D anisotropic Gaussian fitting, from which the principal and the auxiliary axes were extracted as 2.3σ long and 2.3σ short, respectively. The shape factor was calculated as a ratio between the auxiliary and the principal axes. The epifluorescence image of PSD-95 was superimposed on the GluN2A subunit dSTORM image to identify the synaptic nano-objects. movies (3000 frames) were successively recorded: 1) "Pre" (baseline period), 2) "Post" (5min after bath application of purified IgG, 5 µg/ml) and 3) "D-AP5" (5min after bath application, 50 µM). Time-lapse movies were concatenated and realigned in ImageJ (PoorMan3DReg plugin, Michael Liebling). Fluorescence from calcium transients vs. time was measured within individual ROIs manually defined by the experimenter (ImageJ). All pixels within each ROI were averaged to give a single value time course associated to the ROI.

Quantum dot (QD) tracking and surface diffusion calculation.
Mean normalized fluorescence (ΔF/F) was calculated by subtracting each value with the mean of the previous 5s values lower than P50 (µ) and dividing the result by µ. Positive calcium transients were identified following a two-step procedure: initially, ΔF/F traces were smoothen by convoluting the raw signal with a 10s squared kernel. True positives (with minimum 1s between transients) were then defined on an automated basis using custom-written MATLAB routines where the threshold was set at 5*SD of D-AP5 average trace.

Chemically induced potentiation (cLTP).
Live hippocampal neurons transfected with GluA1-SEP were incubated overnight with human IgG (Healthy+ or PSY+, 5 µg/ml, 37°C). After washing thoroughly, chemically induced long-term potentiation (cLTP) was elicited by a bath co-application of glycine (200 µM) and picrotoxin (5 µM) for 4min 4 . cLTP was always applied after a period of baseline acquisition and the medium was carefully replaced by fresh equilibrated and heated medium after induction. GluA1-SEP fluorescence signal was then recorded every 5min during the 30min following the stimulus. Glutamate synapses were defined using the synaptic protein Homer-1c DsRed. Homer-1c clusters were outlined and GluA1-SEP intensity was measured over time within the synaptic areas. Synaptic GluA1-SEP clusters intensity and area values were normalized to the baseline values. All images were collected on a video confocal spinning-disk system (Leica DMI6000B, 63X) and a CoolSNAP HQ2 camera (Photometrics).
In vivo hippocampal injection. Surgical procedures were conducted in accordance with the guidelines of the to allow diffusion and reduce reflux up the needle. The incision was then mechanically sutured and rat pups were allowed to recover on a heating blanket before being placed back with their littermates.
Whole-cell voltage clamp recordings of CA1 pyramidal cells were performed using infrared differential interference contrast microscopy under continuous perfusion of heated ACSF (32°C) saturated with 95% O2 / 5% CO2. Electrodes (4-5 MΩ) were prepared from borosilicate pipettes (GC150T-10, Harvard Apparatus, UK) with a horizontal micropipette puller (P-97, Sutter Instrument, USA) and filled with a solution containing (in mM): 120 cesium methanesulfonate, 4 NaCl, 4 MgCl2, 10 HEPES, 0.2 EGTA, 4 Na2ATP, 0.33 Na3GTP, and 5 phosphocreatine adjusted to pH 7.3 with CsOH. EPSCs were evoked at a rate of 0.05 Hz using an ACSF-filled glass micropipette positioned in the stratum radiatum to stimulate Schaffer collaterals. Currents were recorded using a Multiclamp 700B amplifier and a Digidata 1550B interface controlled by Clampex 10.7 (Molecular Devices) at -60 mV in the presence of the GABAA and GABAB receptor blockers SR95531 (10 µM) and CGP55845 (5 µM), respectively. Under these conditions and after a stable eEPSC recording had been maintained for 10min, tetanic stimulation (four trains of 100 stimuli at 100 Hz, delivered at 20s interval) of Schaffer collaterals was used to induce LTP. The access resistance was monitored throughout the experiment and data were discarded when it changed by > 20%.