Skip to main content
Journal of Neuroinflammation logoLink to Journal of Neuroinflammation
. 2026 Mar 28;23:152. doi: 10.1186/s12974-026-03789-0

D-serine ameliorates cognitive deficits by preserving neuronal and synaptic function in experimental anti-NMDAR encephalitis

Yu Yao 1, Er-Deng E 2, Jian-Hong Yang 1, Chang Liu 2, Tian-Yang Xu 2, Ting Zhang 2, Yi Huang 3, Chao Wei 1, Xiang Li 1, Xiao-Qin Zhang 1,2,✉
PMCID: PMC13151373  PMID: 41896867

Abstract

Anti-N-methyl-D-aspartate receptor (NMDAR) encephalitis is an antibody-mediated autoimmune disorder characterized by neuroinflammation, synaptic dysfunction, and prominent cognitive impairment. While immunotherapies reduce antibody burden, cognitive recovery is often incomplete, suggesting that persistent neuronal dysfunction contributes to disease outcome. Here, we investigated the effects of D-serine, an endogenous NMDAR co-agonist, in a mouse model of anti-NMDAR encephalitis induced by active immunization with the GluN1₃₅₉–₃₇₈ peptide. Immunized mice exhibited sustained learning and memory deficits accompanied by hippocampal neuronal integrity, dendritic disruption, and decreased expression of synaptic proteins. D-serine administration during the established disease phase significantly improved cognitive performance, partially restored NMDAR1 expression, and ameliorated neuronal and synaptic alterations in the hippocampus. Proteomic analysis further revealed that D-serine was associated with the partial normalization of proteins involved in axonal development, synaptic transmission, and receptor-related signaling pathways. In parallel, GluN1₃₅₉–₃₇₈ immunization induced increased microglial density and activation-associated changes, while D-serine had limited effects on microglial abundance and morphology but attenuated TSPO expression within IBA1⁺ microglia. Together, these findings suggest that neuronal and synaptic alterations are key contributors to cognitive deficits in this model, and that D-serine may promote functional recovery primarily through modulation of neuronal and synaptic processes rather than through broad suppression of neuroinflammatory responses. Targeting neuronal resilience may therefore represent a complementary therapeutic strategy alongside immune modulation in anti-NMDAR encephalitis.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-026-03789-0.

Keywords: D-serine, Anti-NMDAR encephalitis, Cognitive deficits, Neuronal dysfunction, Hippocampus

Introduction

Autoimmune encephalitis (AE) is a group of severe inflammatory brain disorders caused by the immune system mistakenly attacking its own central nervous system (CNS) tissues. These responses often involve autoantibodies targeting neuronal or glial antigens [1], leading to a wide spectrum of neuropsychiatric symptoms, including memory loss, behavioral abnormalities, altered mental status, and seizures [2–4]. AE accounts for approximately 20% of all encephalitis cases, with high mortality and disability rates, reflecting its complex pathophysiology and the devastating effects on the nervous system [5]. Among the subtypes of AE, anti-N-methyl-D-aspartate receptor (NMDAR) encephalitis is the most common form, mediated by IgG autoantibodies that specifically target NMDARs [6]. These antibodies induce receptor capping, cross-linking, and internalization, resulting in a significant loss of surface NMDARs [7, 8]. Functionally, this receptor loss impairs synaptic signaling and excitatory neurotransmission, which underlies the cognitive deficits, behavioral disturbances, and neuropsychiatric symptoms characteristic of the disease [9–13].

Immunotherapies remain the standard treatment, including high-dose intravenous corticosteroids, plasma exchange, and intravenous immunoglobulin (IVIG), with second-line use of rituximab or cyclophosphamide [14, 15]. However, many patients show incomplete responses or experience adverse effects. Corticosteroids, for instance, may exacerbate psychiatric symptoms and cause complications such as gastrointestinal bleeding and osteoporosis [16]. Similarly, IVIG carries risks of thromboembolism and hyponatremia, which may worsen cerebral edema and mental status [17]. Moreover, prolonged recovery of neuropsychiatric symptoms, particularly cognitive deficits, remains a major clinical challenge [18, 19].

Emerging evidence indicates that NMDAR dysfunction extends beyond autoimmune encephalitis to other CNS disorders, including schizophrenia, where downregulation of NMDAR activity is a hallmark [20, 21]. Enhancing NMDAR function has therefore been proposed as a promising therapeutic strategy [22]. Notably, anti-NMDAR encephalitis shares significant clinical and mechanistic features with schizophrenia, including catatonia and cognitive impairments [21, 23, 24], suggesting overlapping neurobiological mechanisms. This overlap highlights the therapeutic potential of D-serine in anti-NMDAR encephalitis. D-serine, a selective full agonist of the co-agonist binding site on NMDARs, is synthesized, stored, and released by astrocytes [25]. It plays a pivotal role in modulating NMDAR activity and supporting synaptic plasticity [26, 27]. Preclinical studies have shown that D-serine administration increases cerebral blood flow in schizophrenia mouse models [28] and alleviates electrophysiological, neurochemical, and cognitive deficits in other models of NMDAR hypofunction [22, 29, 30]. However, whether D-serine can effectively counteract the synaptic and behavioral deficits induced by NMDAR autoantibodies in anti-NMDAR encephalitis remains unclear.

In this study, we investigated the therapeutic potential of D-serine in a mouse model of anti-NMDAR encephalitis induced by active immunization with the GluN1₃₅₉–₃₇₈ peptide. Given the central role of NMDAR hypofunction in this disorder, we hypothesized that D-serine may improve cognitive deficits by supporting neuronal and synaptic function. To test this, we evaluated behavioral performance, hippocampal neuronal integrity, and synaptic protein expression following D-serine treatment during the established phase of disease. In addition, we performed proteomic analysis to explore molecular pathways associated with its effects. Considering the contribution of neuroinflammation to disease pathology, we further assessed microglial responses, including cell density, morphology, and activation-associated TSPO expression. By integrating behavioral, histological, and proteomic approaches, this study aimed to clarify whether the potential benefits of D-serine are primarily linked to neuronal and synaptic mechanisms, or whether modulation of microglial responses also contributes to its effects in this model.

Materials and methods

Animals

Eight-week-old male and female C57BL/6J mice were purchased from Shanghai Sino-British SIPPR/BK Lab and housed under standard laboratory conditions (24 ± 1 °C; 50–70% humidity) on a 12-h reversed light/dark cycle, with food and water available ad libitum. All behavioral experiments were conducted during the dark phase. Mice were acclimated to the experimental room for at least 30 min before testing. Body weight was recorded daily throughout the study. Both sexes were included in all experimental groups, and sex distribution was balanced across treatment conditions whenever possible. Sex was recorded for each animal, and individual data points in graphical presentations are annotated by sex. Experimenters were blinded to treatment allocation during data collection and analysis to minimize bias. All experiments were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and these were approved by the Animal Care and Use Committees of Ningbo University (NBU20230347, June 2023), China.

Active immunization

The immunization protocol was adapted from the active immunization model described by Wagnon et al. [31]. In our study, autoimmune responses were induced using the GluN1₃₅₆–₃₈₅ peptide, which spans a slightly extended region of the extracellular domain compared with the GluN1₃₅₉–₃₇₈ peptide used in the original report. The remaining immunization procedures, including adjuvant formulation and pertussis toxin administration, followed the established protocol [32]. Briefly, an autoimmune response was induced in mice by subcutaneous injection of 200 µg GluN1₃₅₆–₃₈₅ peptide (Nanjing Peptide Biotech Co., Ltd.) emulsified in a 1:1 mixture of complete Freund’s adjuvant (Sigma-Aldrich) and 600 µg heat-killed Mycobacterium tuberculosis H37Ra (Becton, Dickinson and Company). The total emulsion volume (200 µL) was administered as four 50 µL injections into the limbs. In addition, mice in the model group received intraperitoneal injections of 200 ng Bordetella pertussis toxin (Sigma-Aldrich) in 200 µL saline at the time of immunization and again 48 h later. Pertussis toxin was used to facilitate immune cell trafficking and transiently increase blood–brain barrier permeability, consistent with established active immunization protocols for autoimmune encephalitis models [33, 34]. For the sham group, both the peptide emulsion and pertussis toxin were replaced with an equal volume of saline. Pertussis toxin was not administered to sham animals in order to avoid nonspecific systemic immune activation unrelated to antigen-specific immunization. Pertussis toxin alone does not induce anti-NMDAR–like pathology in the absence of GluN1-directed immune responses [31, 32].

Detection of anti-NMDAR antibodies

Anti-NMDAR antibody levels in serum were measured at the end of the first or second week post-immunization using an enzyme-linked immunosorbent assay (ELISA) kit (FineTest, Wuhan Fine Biotech Co., Ltd., China, EM2087). All procedures, including incubation and washing, were carried out following the manufacturer’s instructions. Optical density (OD) values were measured at 450 nm using a microplate reader. The antibody positivity threshold was defined as the negative control OD value plus 0.25. Samples with OD values above this threshold were considered positive, while those below were considered negative.

D-serine administration and experimental design

D-Serine (Sigma Aldrich) was dissolved in 0.9% saline and administered intraperitoneally at a dose of 600 mg/kg in a volume of 10 mL/kg body weight per mouse. From days 11 to 26, D-serine or saline was injected once daily, 30 min prior to behavioral testing sessions. The dose and frequency of D-serine administration were based on previously reported studies demonstrating effective modulation of NMDAR hypofunction in mice [8, 22, 35, 36]. The treatment initiation time was selected based on prior experience with this model [31]. Although no reports specify the optimal endpoint for treatment, day 26 was chosen as the cutoff for exploratory treatment based on related studies of other NMDAR modulators [37]. To assess the effects of D-serine on anti-NMDAR antibody, serum antibodies were measured again on day 18 (midpoint of treatment) using the same ELISA kits. Brain tissue analyses were performed on subsets of mice euthanized at days 10 and 18, while the rest mice were euthanized on day 26 (Fig. 1).

Fig. 1.

Fig. 1

Experimental timeline and treatment for the anti-NMDAR encephalitis mouse model. Mice were divided into four groups: Sham + vehicle (blue), Sham + D-serine (green), GluN1₃₅₆–₃₇₈ + vehicle (yellow), and GluN1₃₅₆–₃₇₈ + D-serine (red). Behavioral tests and sampling occurred on days 10, 13, 18, and 26. D-serine or vehicle treatments were administered daily from day 11

Behavioral tests

Behavioral tests were performed on days 10, 13, 18, and 26 in the following order: open field tests (OFTs), Y-maze tests, and novel object recognition (NOR) tests. These tests were designed to evaluate locomotor activity and anxiety-like behavior (OFTs), spatial working memory (Y-maze), and recognition memory (NOR).

OFTs

Mice were placed in the center of a white Plexiglas box (40 cm L × 40 cm W × 40 cm H) with a 20 cm × 20 cm central zone. Each mouse was allowed to explore freely for 10 min, and movements were recorded using an infrared camera. The total distance traveled and the distance traveled in the central area were analyzed using ANY-maze software (Stoelting, United Kingdom).

Y-maze

The maze consisted of three arms (40 cm L × 10 cm W × 25 cm H) arranged at 120° angles. Mice were placed at the intersection of the three arms and allowed to explore freely for 8 min. The total number of arm entries and the sequence of entries were recorded. The percentage of spontaneous alternations, an indicator of spatial working memory, was calculated using the following formula: % Alternations = (Number of correct alternations) / (Total arm entries − 2) × 100.

NOR

NOR tests were conducted to assess recognition memory, with minor modifications to previously described protocols [38]. The test consisted of three phases: acclimation, training, and testing. Acclimation Phase: Mice were acclimated to an open field box (25 cm L × 25 cm W × 25 cm H, with one wall specially marked) by allowing free exploration for 10 min per day over two consecutive days. Training Phase: On the third day, two identical objects were placed in two corners of the box on the same side. Mice were allowed to explore the objects for 10 min. The investigation time for each object was recorded, defined as the time a mouse’s nose touched or was oriented toward the object within a 2 cm distance. Testing Phase: 24 h after the training phase, one of the objects was replaced with a new object of a different shape but the same material. Mice were allowed to explore for 5 min, and investigation time for the novel and familiar objects was recorded. The NOR discrimination index, reflecting recognition memory performance, was calculated using the formula: Discrimination Index = (Novel object investigation time - Familiar object investigation time) / (Novel object investigation time + Familiar object investigation time). In addition, exploration time for the novel and familiar objects and the total distance traveled during the test session were analyzed to evaluate object exploration behavior and potential locomotor influences.

Western blot

Hippocampal tissues were homogenized in ice-cold RIPA buffer (Beyotime, Shanghai, China) containing a protease inhibitor cocktail (Beyotime, Shanghai, China) using a glass Teflon homogenizer, followed by incubation on ice for 10 min. The homogenates were centrifuged at 12,000 × g for 12 min at 4 °C, and the supernatants were collected. Protein concentrations were measured, and 40 µg of protein per sample was loaded onto each lane of the gel. After electrophoresis and transfer, membranes were blocked in 5% skim milk prepared in PBST (0.05% Tween-20 in PBS) at room temperature (RT) and incubated overnight at 4 °C with primary antibodies against NMDAR1 (rabbit, 1:1000, AF6406, Affinity Biosciences, RRID: AB_2835236), PSD95 (rabbit, 1:10000, 20665-1-AP, Proteintech, RRID: AB_2687961), SNAP25 (rabbit, 1:1500, A2234, ABclonal, RRID: AB_2753186) and synaptophysin (rabbit, 1:20000, 17785-1-AP, Proteintech, RRID: AB_2271365). The membranes were then incubated with a secondary antibody (HRP-conjugated goat anti-rabbit IgG, 1:5000, S0001, Affinity Biosciences, RRID: AB_2839429) for 2 h at RT. Protein bands were detected using a fluorescence scanner (Odyssey Infrared Imaging System, LI-COR Biotechnology, Lincoln, NE, USA) with automatic exposure. The gray value of the immunoreactive protein bands was analyzed using ImageJ software (version 1.52a, National Institutes of Health). GAPDH (1:5000, AF7021, Affinity Biosciences, RRID: AB_2839421) was used as an internal control.

Immunostaining and quantitative analysis

After removal, brains were post-fixed in 4% paraformaldehyde (PFA) at 4 °C for 2–3 days, cryoprotected in 30% sucrose, and cut into 30 μm coronal sections using a sliding microtome (Leica CM1950, Leica Microsystems, Nussloch, Germany). Sections were stored at − 20 °C in cryoprotectant solution (30% glycerol and 30% ethylene glycol in PBS) until further processing. For immunohistochemistry, free-floating sections were rinsed in PBS and permeabilized with PBS containing 0.5% Triton X-100 (PBS-Tx). Endogenous peroxidase activity was quenched by incubation in 0.3% hydrogen peroxide with 10% methanol for 15 min. Sections were blocked in PBS-Tx containing 10% normal goat serum, 1% milk, and 0.2% gelatin, and then incubated overnight at 4 °C with the following primary antibodies: rabbit anti-NeuN (1:1000, Proteintech, 26975-1-AP, RRID: AB_2880708), rabbit anti-NMDAR1 (1:50, Affinity, AF6406, RRID: AB_2835236), and rabbit anti-IBA1 (1:2000, Abcam, ab178846, RRID: AB_2636859). After washing, sections were incubated for 2 h at room temperature with peroxidase-conjugated goat anti-rabbit IgG (1:200, Proteintech, SA00001-2, RRID: AB_2722564). Immunoreactivity was visualized using the avidin–biotin complex method (ABC Vector Elite, Vector Laboratories) according to the manufacturer’s instructions. For immunofluorescence staining, sections were permeabilized with 0.3% Triton X-100 for 15 min and blocked with 10% normal donkey serum for 1 h at room temperature. Sections were then incubated overnight at 4 °C with the following primary antibodies: goat anti-Iba1 (1:500, Invitrogen, PA5-18039, RRID: AB_10982846), rabbit anti-TSPO (1:500, Abcam, ab109497, RRID: AB_10860844), and mouse anti-MAP2 (1:500, Proteintech, 67015-1, RRID: AB_2882331). After washing, sections were incubated for 2 h with species-appropriate secondary antibodies including donkey anti-goat IgG Alexa Fluor™ Plus 488 and donkey anti-rabbit or goat anti-mouse IgG Alexa Fluor™ 594 (all 1:500, Thermo Fisher Scientific). Sections were mounted using antifade medium containing DAPI. Neurodegeneration was assessed using Fluoro-Jade C staining according to the manufacturer’s protocol (TR-160-FJC, Biosensis, Australia). Given the extremely low number of FJC-positive cells observed across all groups, quantitative statistical analysis was not performed, as such sparse data would not provide reliable statistical inference.

Brightfield and fluorescence images were acquired using a Leica DM3000 upright fluorescence microscope equipped with a digital camera. Images were captured using Leica Application Suite X (LAS X) software under identical acquisition settings within each experiment to ensure comparability across groups. For brightfield and low-magnification fluorescence imaging, images were acquired using a 20× objective. The image resolution was 2736 × 1824 pixels. Spatial calibration (µm/pixel) was determined using the microscope calibration settings in ImageJ based on the objective magnification. Illumination intensity, exposure time, and gain were kept constant across all groups within each experiment. As images were acquired as single optical planes under widefield conditions, z-stack acquisition was not performed. For NeuN, MAP2 and NMDAR1 staining, at least three coronal sections per mouse were analyzed, and optical density within the CA1, CA3, and DG regions of the hippocampus was quantified as mean gray value using ImageJ (NIH, USA). IBA1-positive microglia were manually counted using the ImageJ Cell Counter plugin, and cell density was calculated as the number of cells per unit area (cells/mm²). Fluorescence intensity of TSPO was quantified within predefined regions of interest (ROIs) and expressed as mean gray value. Microglial morphology was assessed using high-magnification IBA1-stained images acquired with a 100× oil-immersion objective. Morphological complexity was analyzed using Sholl-based methods implemented in ImageJ, by quantifying the number of process intersections at increasing radial distances from the soma. At least three microglial cells per hippocampal region per animal were analyzed in a blinded manner. At least three microglial cells per hippocampal region per animal were analyzed. All image analyses were performed by investigators blinded to group allocation. For each animal, measurements obtained from individual sections were averaged to generate a single value per animal for statistical analysis. To facilitate comparisons across groups, all values were normalized to the control group within each experiment.

Proteomics

Protein extraction and digestion

Hippocampal tissue samples were retrieved from a -80 °C freezer and ground into powder in liquid nitrogen using a pre-cooled mortar. The powder was transferred to a new centrifuge tube, and lysis buffer (8 M urea with 1% protease inhibitor) was slowly added. The samples were sonicated using a sonicator and centrifuged at 15,000 × g for 10 min at 4 °C. The supernatants were collected, and protein concentrations were measured using the BCA assay. Equal amounts of protein from each sample were used, and the volume was adjusted with lysis buffer. For digestion, samples were incubated with dithiothreitol (DTT, final concentration 5 mM) at 56 °C for 30 min. Alkylation was performed with iodoacetamide (IAA, final concentration 15 mM) at room temperature in the dark for 15 min. Samples were transferred to ultrafiltration tubes (Millipore) and centrifuged at 12,000 × g for 20 min at room temperature. The buffer was replaced three times with 8 M urea and then three times with 25 mM HEPES. Proteins were digested with trypsin at a 1:50 enzyme-to-protein ratio (w/w) at 37 °C overnight. A second digestion was performed by adding trypsin at a 1:100 ratio (w/w) at 37 °C for 4 h. After digestion, the samples were centrifuged at 12,000 × g for 20 min at room temperature, and the ultrafiltration membrane was rinsed with purified water. The filtrates from both steps were combined. The samples were acidified with 10% TFA to adjust the pH to 2–3, desalted with Strata X columns (Phenomenex), and vacuum-dried. Peptides were reconstituted in water and quantified using the Pierce™ Quantitative Peptide Assay kit (Thermo Scientific).

Liquid Chromatography-Mass Spectrometry (LC-MS/MS) analysis

Desalted peptides were analyzed using an Orbitrap Astral mass spectrometer (Thermo Scientific) with data-independent acquisition (DIA) mode. Chromatographic separation was performed using the Vanquish Neo system (Thermo Scientific) at a nanoliter flow rate. For MS analysis, the following parameters were used: Ionization mode: Positive, Precursor ion scan range: 380–980 m/z, Resolution (MS1): 240,000 at 200 m/z, Normalized AGC target: 500%, Maximum injection time (IT): 5 ms. For MS2 analysis in DIA mode: Number of DIA windows: 299, Isolation window: 2 m/z, Collision energy (HCD): 25 eV, Normalized AGC target: 500%, Maximum IT: 3 ms.

LC-MS/MS data analysis

All DIA raw data were analyzed using DIA-NN software. A sample-specific protein database was constructed based on the origin of the samples. The parameters for DIA-NN were set as follows: Enzyme: Trypsin, Max missed cleavages: 1, Fixed modification: Carbamidomethylation (C), Variable modifications: Oxidation (M) and Acetylation (Protein N-term), False discovery rate (FDR): <1%. Peptide and protein-level quality control (QC) was performed to ensure data reliability, including analysis of peptide mass error distribution, peptide length distribution, amino acid frequency, and signal intensity distribution. Proteins were quantified, and statistical analyses included quantitative distribution and reproducibility evaluations.

Data visualization

Differential Protein Analysis: Differentially expressed proteins were identified using statistical tests, with results visualized as volcano plots and heatmaps. Functional Classification: Differential proteins were analyzed for functional classification, including GO secondary categories, KEGG pathway classification, subcellular localization, and COG/KOG classification. Enrichment Analysis: Fisher’s exact test was used to identify significantly enriched biological processes or pathways among the differential proteins.

Statistical analysis

For all experiments, statistical analyses were performed using GraphPad Prism version 7 (GraphPad Software, San Diego, CA, United States). Prior to statistical testing, data distribution and variance characteristics were examined within GraphPad Prism to guide the selection of appropriate statistical methods. When data approximated a normal distribution and exhibited no marked variance heterogeneity, parametric tests were applied. Comparisons between two groups were analyzed using unpaired two-tailed Student’s t-tests. For experiments involving two independent variables (immunization: Sham vs. GluN1₃₅₉–₃₇₈; treatment: vehicle vs. D-serine), two-way ANOVA followed by Tukey’s post hoc test was used. When data deviated from normality or showed unequal variance, nonparametric tests were employed, including the Mann–Whitney U test for two-group comparisons. Sex was initially included as a biological variable in the analysis and no obvious sex-dependent differences or sex × treatment interactions were observed; therefore, data from male and female mice were pooled for subsequent analyses. All data are presented as mean ± SEM. Statistical significance was set at p < 0.05. Degrees of freedom, test statistics, and exact p-values are reported in the Results section and figure legends.

Results

Cognitive deficits were found in the anti-NMDAR encephalitis mouse model induced by active immunization with the peptide GluN1359–378

Active immunization with GluN1 peptides has been previously suggested as a method to establish an anti-NMDAR encephalitis mouse model [31, 32, 39]. To further explore the potential pathogenesis and therapeutic strategies for anti-NMDAR encephalitis, we developed a mouse model using the GluN1₃₅₆–₃₈₅ peptide. After immunization, serum IgG antibody levels and behavioral assessments were conducted to confirm the successful construction of the model. The presence of serum IgG antibodies was evaluated at the end of the first week post-immunization using a commercial ELISA kit. Serum autoantibodies were identified in peptide-immunized mice but were absent in the sham group (Fig. 2A; unpaired two-tailed t-test, t (12) = 11.71, p < 0.0001). To investigate the impact on hippocampal NMDAR1 expression, we performed western blot analysis. A significant reduction in total hippocampal NMDAR1 protein levels was observed in peptide-immunized mice compared to sham controls (Fig. 2B-C; unpaired two-tailed t-test, t (8) = 4.923, p = 0.0012), consistent with previous findings [32]. Immunohistochemical staining further confirmed these findings, revealing a significant decrease in the optical density of NMDAR1 in the hippocampal regions, including CA1, CA3, and DG, in anti-NMDAR encephalitis mice compared to the sham group (Fig. 2D-G; unpaired two-tailed t-test, CA1: t (8) = 7.246, p < 0.0001; CA3: t (8) = 5.367, p = 0.0007; DG: t (8) = 7.084, p = 0.0001).

Fig. 2.

Fig. 2

NMDAR1 expression was reduced in the hippocampus of GluN1₃₅₆–₃₇₈-immunized mice. A ELISA results showing significantly increased serum anti-NMDAR1 IgG levels in GluN1₃₅₆–₃₇₈-immunized mice compared to sham controls (unpaired two-tailed t-test, t (12) = 11.71, p < 0.0001). B Western blot analysis of NMDAR1 and GAPDH levels in hippocampal tissues from sham and GluN1₃₅₆–₃₇₈-immunized mice. C Quantification of NMDAR1/GAPDH ratios demonstrating a significant reduction in NMDAR1 expression in GluN1₃₅₆–₃₇₈-immunized mice (unpaired two-tailed t-test, t (8) = 4.923, p = 0.0012). D Representative immunohistochemistry images of NMDAR1 in the CA1, CA3, and DG regions of the hippocampus. Scale bar = 100 µm. E–G Quantification of NMDAR1 area optical density (AOD), normalized to sham controls. NMDAR1 expression was significantly reduced in the CA1 (unpaired two-tailed t-test, t (8) = 7.246, p < 0.0001), CA3 (t (8) = 5.367, p = 0.0007), and DG (t (8) = 7.084, p = 0.0001) regions of GluN1₃₅₆–₃₇₈-immunized mice compared to sham mice. Each data point represents an individual mouse. Male mice are indicated by filled circles, and female mice by open circles

Behavioral assessments revealed cognitive dysfunction in the anti-NMDAR encephalitis model. In the NOR test, peptide-immunized mice exhibited a significantly lower discrimination index, characterized by reduced exploration of novel objects compared with sham mice (Fig. 3A; unpaired two-tailed t-test, t (36) = 5.903, p < 0.0001). Similarly, in the Y-maze test, peptide-immunized mice showed a reduced alternation percentage compared with sham controls (Fig. 3B; unpaired two-tailed t-test, t (37) = 3.103, p = 0.0037), indicating impaired working memory. These findings further support the memory deficits in the model, consistent with the cognitive impairments reported in patients with anti-NMDAR encephalitis [40, 41]. Additionally, the total number of arm entries was significantly lower in the model group compared to the sham group (Fig. 3C; unpaired two-tailed t-test, t (37) = 3.747, p = 0.0006). Total distances traveled (Fig. 3D; unpaired two-tailed t-test, t (37) = 5.054, p < 0.0001) and distances traveled in the central area (Fig. 3E; unpaired two-tailed t-test, t (37) = 3.925, p = 0.0004) in the OFTs were significantly reduced, indicating decreased locomotor activity. Furthermore, reduced time spent in the central area (Fig. 3F; unpaired two-tailed t-test, t (37) = 3.003, p = 0.0048) and shorter distances traveled in the OFTs suggest the presence of anxiety-like behaviors, which align with core psychiatric symptoms of anti-NMDAR encephalitis, such as hallucinations, mania, and schizoaffective episodes [41–43]. Importantly, supplementary analyses of the NOR task confirmed comparable exploration of the two identical objects during the training phase and showed reduced exploration of the novel object in peptide-immunized mice during the test phase, along with decreased locomotor distance during the test session (Supplementary Fig. 1).

Fig. 3.

Fig. 3

Behavioral impairments in GluN1₃₅₆–₃₇₈-immunized mice compared to sham controls. A Novel object recognition (NOR) test showing a significantly reduced discrimination index in GluN1₃₅₆–₃₇₈-immunized mice compared to sham controls (unpaired two-tailed t-test, t (36) = 5.903, p < 0.0001). B Y-maze test showing a significant reduction in spontaneous alternation percentage in GluN1₃₅₆–₃₇₈-immunized mice (unpaired two-tailed t-test, t (37) = 3.103, p = 0.0037). C Total number of arm entries in the Y-maze was significantly lower in GluN1₃₅₆–₃₇₈-immunized mice (unpaired two-tailed t-test, t (37) = 3.747, p = 0.0006). D Open field test showing a significant reduction in total distance traveled by GluN1₃₅₆–₃₇₈-immunized mice (unpaired two-tailed t-test, t (37) = 5.054, p < 0.0001). E Distance traveled in the center of the open field arena was significantly lower in GluN1₃₅₆–₃₇₈-immunized mice (unpaired two-tailed t-test, t (37) = 3.925, p = 0.0004). F Time spent in the center of the open field arena was significantly reduced in GluN1₃₅₆–₃₇₈-immunized mice (unpaired two-tailed t-test, t (37) = 3.003, p = 0.0048). Each data point represents an individual mouse. Male mice are indicated by filled circles, and female mice by open circles

In summary, we established an active immunization-based mouse model of anti-NMDAR encephalitis using the GluN1₃₅₆–₃₈₅ peptide. This model demonstrated impaired cognition, diminished locomotor activity, and anxiety-like behaviors, providing a robust basis for investigating the underlying mechanisms and potential therapeutic approaches for anti-NMDAR encephalitis.

Dysfunctions in both neurons and microglial cells in the hippocampus of the anti-NMDAR encephalitis mouse model

To assess neuronal damage in the anti-NMDAR encephalitis mouse model, NeuN levels were evaluated by Immunohistochemical staining in the hippocampus. A significant reduction in the optical density of NeuN was observed in peptide-immunized mice compared to sham controls (Fig. 4A-D; unpaired two-tailed t-test, CA1: t (8) = 22.75, p < 0.0001; CA3: t (8) = 7.776, p < 0.0001; DG: t (8) = 7.320, p < 0.0001), indicating neuronal dysfunction or phenotypic alterations in the hippocampal regions. This suggests that active immunization with the GluN1₃₅₆–₃₈₅ peptide leads to structural or functional impairments in neurons, consistent with the cognitive deficits observed in this model.

Fig. 4.

Fig. 4

NeuN and IBA1 staining revealed neuronal damage and microglial activation in GluN1₃₅₆–₃₇₈-immunized mice. A Representative images of NeuN immunostaining in the CA1, CA3, and DG regions of the hippocampus. Scale bar = 100 µm. B–D Quantification of NeuN area optical density (AOD) normalized to sham controls. NeuN expression was significantly reduced in the CA1 (unpaired two-tailed t-test, t (8) = 22.75, p < 0.0001), CA3 (t (8) = 7.776, p < 0.0001), and DG (t (8) = 7.320, p < 0.0001) regions of GluN1₃₅₆–₃₇₈-immunized mice compared to sham controls. E Representative images of IBA1 immunostaining in the CA1, CA3, and DG regions of the hippocampus. Scale bar = 100 µm. F–H Quantification of IBA1-positive cell density normalized to sham controls. Microglial activation was significantly increased in the CA1 (unpaired two-tailed t-test, t (8) = 4.225, p = 0.0029), CA3 (t (8) = 2.773, p = 0.0242), and DG (t (8) = 3.021, p = 0.0165) regions of GluN1₃₅₆–₃₇₈-immunized mice compared to sham controls. Each data point represents an individual mouse. Male mice are indicated by filled circles, and female mice by open circles

To further investigate the inflammatory response, microglial activation was assessed by immunostaining for IBA1, a marker of microglial cells. In the hippocampus of peptide-immunized mice, a marked increase in IBA1+ cell density was detected compared to sham controls, indicating significant activation of microglial cells (Fig. 4E-H; unpaired two-tailed t-test, CA1: t (8) = 4.225, p = 0.0029; CA3: t (8) = 2.773, p = 0.0242; DG: t (8) = 3.021, p = 0.0165). This excessive activation of microglial cells is consistent with neuroinflammatory processes reported in previous studies of anti-NMDAR encephalitis [44].

Together, these findings demonstrate that active immunization with the GluN1₃₅₆–₃₈₅ peptide leads to neuronal damage and microglial activation in the hippocampus, highlighting the link between neuronal dysfunction and neuroinflammation in the pathogenesis of anti-NMDAR encephalitis.

D-serine improved cognitive function in anti-NMDAR encephalitis mice

Four groups of mice were actively immunized with the GluN1₃₅₆–₃₈₅ peptide or control antigen from days 1–10, followed by D-serine or vehicle treatment from days 11–26. Behavioral assessments were performed on days 13, 18, and 26 (Fig. 1). A total of 38 mice (8–11 per group) were included.

Compared to controls, peptide-immunized mice exhibited significant recognition memory deficits as early as day 10, which persisted until day 18, as shown in the NOR test (Fig. 5D). However, peptide-immunized mice treated with D-serine from day 11 showed rapid recovery of recognition memory, which became comparable to that of controls by days 13 (Fig. 5A; two-way ANOVA, GluN1×D-serine F (1,34) = 10.06, p = 0.0032, GluN1 F (1,34) = 12.41, p = 0.0012, D-serine F (1,34) = 5.021, p = 0.0317) and 18 (Fig. 5D; two-way ANOVA, GluN1×D-serine F (1,34) = 3.625, p = 0.0654, GluN1 F (1,34) = 7.875, p = 0.0082, D-serine F (1,34) = 5.243, p = 0.0284). By day 26, two-way ANOVA revealed a significant main effect of GluN1₃₅₆–₃₈₅ immunization (Fig. 5G; F (1,34) = 7.853, p = 0.0083) and a significant main effect of D-serine treatment (Fig. 5G; F (1,34) = 8.260, p = 0.0069), while the GluN1 × D-serine interaction was not significant (Fig. 5G; F (1,34) = 0.7666, p = 0.3874). These findings indicate that immunization continued to impair recognition memory and that D-serine exerted an overall beneficial effect; however, the magnitude of improvement did not differ significantly between groups at this time point.

Fig. 5.

Fig. 5

D-serine improved cognitive deficits in GluN1₃₅₆–₃₇₈-immunized mice. Day 13: A NOR discrimination index. Two-way ANOVA revealed a significant GluN1 × D-serine interaction (F (1,34) = 10.06, p = 0.0032), as well as significant main effects of GluN1 immunization (F (1,34) = 12.41, p = 0.0012) and D-serine treatment (F (1,34) = 5.021, p = 0.0317). B Y-maze spontaneous alternation percentage. A significant GluN1 × D-serine interaction was detected (F (1,34) = 6.170, p = 0.0181), whereas the main effects of GluN1 (F (1,34) = 3.597, p = 0.0664) and D-serine (F (1,34) = 2.208, p = 0.1465) were not significant. C Total arm entries. A significant main effect of GluN1 immunization was observed (F (1,34) = 22.64, p < 0.0001), whereas neither the main effect of D-serine nor the interaction reached significance (both p > 0.16). Day 18: D NOR discrimination index. Two-way ANOVA revealed a significant main effect of GluN1 immunization (F (1,34) = 7.875, p = 0.0082) and D-serine treatment (F (1,34) = 5.243, p = 0.0284), while the interaction was not significant (F (1,34) = 3.625, p = 0.0654). E Y-maze spontaneous alternation percentage. Significant main effects of GluN1 immunization (F (1,34) = 5.734, p = 0.0223) and D-serine treatment (F (1,34) = 7.026, p = 0.0121) were observed, whereas the interaction was not significant (F (1,34) = 3.144, p = 0.0851). F Total arm entries. A significant main effect of GluN1 immunization was observed (F (1,34) = 9.838, p = 0.0035), with no significant effect of D-serine or interaction (both p > 0.05). Day 26: G NOR discrimination index. Significant main effects of GluN1 immunization (F (1,34) = 7.853, p = 0.0083) and D-serine treatment (F (1,34) = 8.260, p = 0.0069) were observed, while the interaction was not significant (F (1,34) = 0.7666, p = 0.3874). H Y-maze spontaneous alternation percentage. A significant main effect of D-serine treatment was detected (F (1,34) = 6.292, p = 0.0171), whereas neither the main effect of GluN1 (F (1,34) = 0.0626, p = 0.8039) nor the interaction (F (1,34) = 2.991, p = 0.0928) was significant. I Total arm entries. Neither GluN1 immunization (F (1,34) = 2.114, p = 0.1552) nor D-serine treatment (F (1,34) = 0.3249, p = 0.5724) showed significant main effects, and no significant interaction was observed. Each data point represents an individual mouse. Male mice are indicated by filled circles, and female mice by open circles

We also analyzed spontaneous alternation in the Y-maze to assess working memory. On day 13, two-way ANOVA revealed a significant GluN1 × D-serine interaction (Fig. 5B; F (1,34) = 6.170, p = 0.0181), whereas neither the main effect of GluN1 (Fig. 5B; F (1,34) = 3.597, p = 0.0664) nor D-serine (Fig. 5B; F (1,34) = 2.208, p = 0.1465) reached significance. By day 18, significant main effects of GluN1 immunization (Fig. 5E; F (1,34) = 5.734, p = 0.0223) and D-serine treatment (Fig. 5E; F (1,34) = 7.026, p = 0.0121) were observed, while the interaction was not significant (Fig. 5E; F (1,34) = 3.144, p = 0.0851) On day 26, the main effect of GluN1 was no longer significant (Fig. 5H; F (1,34) = 0.0626, p = 0.8039), whereas D-serine treatment showed a significant main effect (Fig. 5H; F (1,34) = 6.292, p = 0.0171), with no significant interaction (Fig. 5H; F (1,34) = 2.991, p = 0.0928). Together, these findings indicate that working memory deficits were most pronounced at earlier time points and that D-serine exerted beneficial effects over time, although the magnitude of treatment effects varied across stages of disease progression.

To determine whether these cognitive effects were confounded by alterations in locomotor activity, open field test parameters were analyzed. Two-way ANOVA revealed no significant GluN1 × D-serine interactions at any time point (all p > 0.05). On days 13 and 18, significant effects of GluN1 immunization were observed (Fig. 5C, F; F (1,34) = 22.64, p < 0.0001; F (1,34) = 9.838, p = 0.0035), whereas no significant D-serine effects were detected (both p > 0.16). By day 26, neither GluN1 nor D-serine showed significant main effects (Fig. 5I; GluN1 F (1,34) = 2.114, p = 0.1552; D-serine F (1,34) = 0.3249, p = 0.5724). Similar patterns were observed in total distance traveled in the OFT (Supplementary Fig. 1A, D, G), with significant main effects of GluN1 at earlier time points (p ≤ 0.0021) but no consistent D-serine effects. These findings indicate that D-serine did not significantly alter locomotor activity. Anxiety-like behavior, assessed by time and distance traveled in the center of the OFT, showed significant main effects of GluN1 immunization on day 13 (Supplementary Fig. 2B, C; F (1,34) = 20.47, p < 0.0001; F (1,34) = 22.36, p < 0.0001), whereas no significant D-serine effects or interactions were detected (all p > 0.59). On days 18 and 26, no significant interactions were observed (all p > 0.35). Although GluN1 main effects remained significant in some measures (Supplementary Fig. 2E, H; p < 0.05), D-serine treatment did not significantly modify anxiety-related parameters at any time point.

Detailed analyses of exploration behavior during the NOR task are provided in the Supplementary Materials. Peptide-immunized mice exhibited reduced exploration of the novel object compared with sham controls, indicating impaired recognition memory. Although a modest reduction in total distance traveled was observed in the model group, exploration of the familiar object remained largely preserved, suggesting that the decreased novel object exploration was not solely attributable to reduced locomotor activity but rather reflected impaired object recognition memory. Notably, D-serine treatment partially restored exploration of the novel object and modestly increased total distance traveled at later time points (Supplementary Fig. 3).

Reduction of NMDAR1 in the hippocampus of anti-NMDAR encephalitis mice was partially reversible with D-serine treatment

To investigate whether the D-serine affects NMDAR1 levels in anti-NMDAR mice, we examined serum IgG antibodies and total hippocampal NMDAR1 protein levels on days 18 and 26. Serum IgG antibody levels were significantly elevated in peptide-immunized mice compared to controls on day 18 (Fig. 6A; F (1,16) = 119.4, p < 0.0001). However, D-serine treatment did not significantly reduce serum IgG levels at either time point (Fig. 6A; F (1,16) = 0.07946, p = 0.7816). Western blot analysis of total hippocampal protein lysates revealed a marked reduction in NMDAR1 protein levels in peptide-immunized mice on days 18 and 26 (Fig. 6B-C; F (1,16) = 25.56, p = 0.0001 and Supplementary Fig. 4A-B; F (1,16) = 24.90, p = 0.0001). Immunohistochemical staining confirmed these findings, showing a significant decrease in the optical density of NMDAR1 in hippocampal regions, including CA1, CA3, and the DG, in anti-NMDAR encephalitis mice compared to sham controls on day 18 (Fig. 6D-G; CA1: F (1,16) = 17.29, p = 0.0007; CA3: F (1,16) = 56.56, p < 0.0001; DG: F (1,16) = 37.63, p < 0.0001). By contrast, D-serine treatment significantly increased the optical density of NMDAR1 in all hippocampal regions at day 18, indicating partial restoration of NMDAR1 levels (Fig. 6D-G; CA1: F (1,16) = 5.720, p = 0.0294; CA3: F (1,16) = 6.968, p = 0.0178; DG: F (1,16) = 5.723, p = 0.0294). However, by day 26, while a reduction in NMDAR1 optical density persisted in the model group, D-serine treatment did not result in significant improvements in NMDAR1 levels compared to untreated model mice (Supplementary Fig. 4C-F). These findings suggest that D-serine can partially restore hippocampal NMDAR1 expression during the early stages of anti-NMDAR encephalitis, but its efficacy diminishes over time, possibly due to sustained immune-mediated damage or other compensatory mechanisms.

Fig. 6.

Fig. 6

Effects of D-serine on NMDAR1 expression in the hippocampus of GluN1₃₅₉–₃₇₈-immunized mice on Day 18. A Serum anti-NMDAR1 IgG levels. Two-way ANOVA revealed a significant main effect of GluN1 immunization (F (1,16) = 119.4, p < 0.0001), whereas D-serine treatment had no significant effect (F (1,16) = 0.07946, p = 0.7816). B Representative Western blot images of NMDAR1 and GAPDH in total hippocampal protein lysates. C Quantification of NMDAR1/GAPDH ratios. A significant main effect of GluN1 immunization was observed (F (1,16) = 25.56, p = 0.0001), indicating reduced NMDAR1 expression in peptide-immunized mice. D Representative immunohistochemistry images of NMDAR1 in the CA1, CA3, and DG regions of the hippocampus. Scale bar = 100 µm. E–G Quantification of NMDAR1 area optical density (AOD) normalized to sham + vehicle mice. On day 18, two-way ANOVA revealed significant main effects of GluN1 immunization in CA1 (F (1,16) = 17.29, p = 0.0007), CA3 (F (1,16) = 56.56, p < 0.0001), and DG (F (1,16) = 37.63, p < 0.0001). A significant main effect of D-serine treatment was observed in CA1 (F (1,16) = 5.720, p = 0.0294), CA3 (F (1,16) = 6.968, p = 0.0178) and DG (F (1,16) = 5.723, p = 0.0294). Each data point represents an individual mouse. Male mice are indicated by filled circles, and female mice by open circles

D-serine preserved neuronal integrity and rescued synaptic protein expression in the hippocampus

To evaluate whether D-serine mitigates neuronal impairment in GluN1₃₅₉–₃₇₈-immunized mice, neuronal integrity was assessed by NeuN and MAP2 immunostaining in the hippocampus at day 18. NeuN staining revealed a marked reduction in neuronal signal intensity across CA1, CA3, and DG in immunized mice (Fig. 7A–D). Two-way ANOVA demonstrated a significant main effect of GluN1 immunization (CA1: F (1,16) = 76.40, p < 0.0001; CA3: F (1,16) = 27.45, p < 0.0001; DG: F (1,16) = 48.74, p < 0.0001), along with significant effects of D-serine treatment and GluN1 × D-serine interactions across all regions (all p < 0.05). Post hoc analysis confirmed a robust reduction in NeuN expression in GluN1₃₅₉–₃₇₈ + vehicle mice (***p < 0.001; ****p < 0.0001), which was robustly restored by D-serine (**p < 0.01; ***p < 0.001). At day 26, NeuN reduction persisted in CA3 and DG (Supplementary Fig. 5A–D; CA3: F (1,8) = 13.19, p = 0.0067; DG: F (1,8) = 36.55, p = 0.0003), while D-serine showed a significant effect in CA1 (F (1,8) = 9.784, p = 0.0141), indicating partial recovery.

Fig. 7.

Fig. 7

Effects of D-serine on neuronal integrity and synaptic protein expression in the hippocampus of GluN1₃₅₉–₃₇₈-immunized mice at day 18. A Representative images of NeuN immunostaining in the CA1, CA3, and DG regions of the hippocampus. Scale bar = 100 μm. B–D Quantification of NeuN area optical density (AOD), normalized to sham + vehicle mice. Two-way ANOVA revealed a significant main effect of GluN1 immunization (CA1, F (1,16) = 76.40, p < 0.0001; CA3, F (1,16) = 27.45, p < 0.0001; DG, F (1,16) = 48.74, p < 0.0001), D-serine treatment (CA1, F (1,16) = 36.44, p < 0.0001; CA3, F (1,16) = 8.906, p = 0.0088; DG, F (1,16) = 5.943, p = 0.0268), and their interactions (CA1, F (1,16) = 41.50, p < 0.0001; CA3, F (1,16) = 23.01, p = 0.0002; DG, F (1,16) = 13.08, p = 0.0023). NeuN levels were reduced in GluN1₃₅₉–₃₇₈ + vehicle mice (***p<0.001; ****p<0.0001) and restored by D-serine (**p<0.01; ***p<0.001). E Representative images of MAP2 immunostaining. Scale bar = 100 μm. F–H Quantification of MAP2 expression. Two-way ANOVA showed significant effects of GluN1 immunization (CA1, F (1,12) = 23.94, p = 0.0004; CA3, F (1,12) = 20.34, p = 0.0007; DG, F (1,12) = 23.78, p = 0.0004). D-serine treatment (CA1, F (1,12) = 3.930, p = 0.0708; CA3, F (1,12) = 6.117, p = 0.0293; DG, F (1,12) = 9.059, p = 0.0109), and interaction effects (CA1, F (1,12) = 8.282, p = 0.0139; CA3, F (1,12) = 5.303, p = 0.0400; DG, F (1,12) = 3.187, p = 0.0744). MAP2 levels were decreased in GluN1₃₅₉–₃₇₈ + vehicle mice (**p<0.01; ***p<0.001) and partially restored by D-serine (*p<0.05). I Representative Western blots of PSD95, SNAP25, synaptophysin and GAPDH. J-L Quantification of protein levels. Two-way ANOVA revealed significant effects of GluN1 immunization (PSD95: F (1,20) = 24.89, p < 0.0001; SNAP25: F (1,20) = 9.575, p = 0.0057; synaptophysin: F (1,20) = 21.10, p = 0.0002) and significant interactions (PSD95: F (1,20) = 5.587, p = 0.0283; SNAP25: F (1,20) = 10.11, p = 0.0047; synaptophysin: F (1,20) = 9.776, p = 0.0053). D-serine treatment partially restored synaptic protein levels (*p<0.05; **p<0.01). Each data point represents an individual mouse. Male mice are indicated by filled circles, and female mice by open circles

To determine whether the reduced NeuN signal reflected neuronal degeneration or alterations in neuronal integrity, Fluoro-Jade C staining was performed. Only sparse FJC-positive cells were detected across all groups, with no apparent group differences (Supplementary Fig. 6), indicating the absence of overt neurodegeneration. These findings suggest that the decrease in NeuN immunoreactivity reflects altered neuronal integrity or phenotype rather than irreversible cell loss at this stage. Consistent with neuronal structural impairment, MAP2 immunostaining showed reduced dendritic integrity in all hippocampal subregions (Fig. 7E–H). Two-way ANOVA revealed a significant main effect of GluN1 immunization (CA1: F (1,12) = 23.94, p = 0.0004; CA3: F (1,12) = 20.34, p = 0.0007; DG: F (1,12) = 23.78, p = 0.0004), with significant effects of D-serine treatment in CA3 and DG and significant interaction effects in CA1 and CA3. Post hoc analysis confirmed that MAP2 expression was significantly reduced in GluN1₃₅₉–₃₇₈ + vehicle mice (**p < 0.01; ***p < 0.001) and partially restored by D-serine (*p < 0.05).

To further assess synaptic alterations, synaptic protein levels were examined by Western blot (Fig. 7I–L). GluN1₃₅₉–₃₇₈ immunization significantly decreased the expression of PSD95 (F (1,20) = 24.89, p < 0.0001), SNAP-25 (F (1,20) = 9.575, p = 0.0057), and synaptophysin (F (1,20) = 21.10, p = 0.0002), with significant interaction effects observed for all proteins (p < 0.05). D-serine treatment significantly restored their expression levels (*p < 0.05; **p < 0.01), indicating improved synaptic integrity.

Collectively, these findings demonstrate that D-serine preserves neuronal integrity and partially rescues dendritic and synaptic deficits in the hippocampus, supporting neuronal functional recovery as a key mechanism underlying its therapeutic effects in this model.

D-serine attenuated microglial activation–associated changes without altering microglial density or morphology

Given that D-serine restored neuronal integrity and synaptic protein expression, we next assessed microglial responses to evaluate whether they were associated with these changes. We first assessed microglial density in the hippocampus following GluN1₃₅₉–₃₇₈ immunization. At day 18, IBA1 immunostaining revealed a significant increase in microglial density across CA1, CA3, and DG (Supplementary figure S7A–D). Two-way ANOVA showed a significant main effect of GluN1 immunization (CA1: F (1,16) = 57.77, p < 0.0001; CA3: F (1,16) = 49.37, p < 0.0001; DG: F (1,16) = 29.37, p < 0.0001), with no significant effect of D-serine treatment or interaction (all p > 0.05). This increase persisted at day 26, with elevated microglial density in CA1 and CA3 (CA1: F (1,8) = 44.35, p = 0.0002; CA3: F (1,8) = 20.99, p = 0.0018), while no significant difference was observed in DG (Supplementary figure S7E–H). Although significant effects of D-serine and interaction terms were detected in CA1 and CA3 (both p < 0.01), post hoc analyses indicated only partial attenuation. Microglial morphology was further assessed using Sholl analysis at day 18 (Supplementary figure S8). A significant effect of distance was observed (p < 0.0001), consistent with normal branching patterns, whereas no significant effects of GluN1 immunization or D-serine treatment were detected, indicating that overall morphological complexity remains largely unchanged.

In contrast, microglial activation–associated changes were evident. Double immunofluorescence analysis revealed increased TSPO expression within IBA1⁺ microglia in GluN1₃₅₉–₃₇₈-immunized mice (Fig. 8A–F). Two-way ANOVA demonstrated significant effects of GluN1 immunization (CA1: F (1,16) = 39.75, p < 0.001; CA3: F(1,16) = 85.68, p < 0.0001; DG: F(1,16) = 33.17, p < 0.001) and D-serine treatment (CA1: F(1,16) = 5.327, p = 0.0347; CA3: F(1,16) = 12.90, p = 0.0024; DG: F(1,16) = 5.665, p = 0.0301), with significant or near-significant interaction effects. Post hoc analysis confirmed that TSPO intensity was markedly elevated in GluN1₃₅₉–₃₇₈ + vehicle mice (CA1: ***p < 0.001; CA3: ****p < 0.0001; DG: ***p < 0.001) and significantly attenuated by D-serine (CA1: *p < 0.05; CA3: **p < 0.01; DG: *p < 0.05), although levels remained higher than in sham controls.

Fig. 8.

Fig. 8

D-serine reduces microglial activation–associated TSPO expression in the hippocampus. A, C, E Representative images of IBA1 (green) and TSPO (red) double immunofluorescence in CA1 (A), CA3 (C), and DG (E) at day 18. Merged images are shown below. Scale bar = 100 μm. B, D, F Quantification of TSPO fluorescence intensity (AOD), normalized to sham + vehicle. Two-way ANOVA revealed significant effects of GluN1 immunization (CA1: F (1,16) = 39.75, p < 0.001; CA3: F (1,16) = 85.68, p < 0.0001; DG: F (1,16) = 33.17, p < 0.001), D-serine treatment (CA1: F (1,16) = 5.327, p = 0.0347; CA3: F (1,16) = 12.90, p = 0.0024; DG: F (1,16) = 5.665, p = 0.0301), and their interactions (CA1: F (1,16) = 4.638, p = 0.0469; CA3: F (1,16) = 6.681, p = 0.0200; DG: F (1,16) = 3.832, p = 0.0680). Post hoc analysis showed that TSPO intensity was significantly increased in GluN1₃₅₉–₃₇₈ + vehicle mice (CA1: ***p<0.001; CA3: ****p<0.0001; DG: ***p<0.001) and attenuated by D-serine treatment (CA1: *p<0.05; CA3: **p<0.01; DG: *p<0.05). Each data point represents an individual mouse. Male mice are indicated by filled circles, and female mice by open circles

Taken together, these results indicate that GluN1₃₅₉–₃₇₈ immunization induces robust microglial recruitment without overt structural remodeling, while D-serine selectively attenuates activation-associated changes in microglia rather than altering their abundance. These observations raise the possibility that modulation of microglial functional states may contribute to the synaptic and neuronal protection conferred by D-serine. To further explore the molecular pathways potentially involved, we next performed proteomic analysis of hippocampal tissue.

D-serine partially normalized synaptic- and axon-related proteomic alterations in anti-NMDAR encephalitis mice

Proteomic analysis of hippocampal tissues revealed significant molecular alterations in GluN1₃₅₉–₃₇₈-immunized mice. A total of 155 differentially expressed proteins (DEPs) were identified between the GluN1₃₅₉–₃₇₈ + vehicle and sham + vehicle groups, including 83 upregulated and 72 downregulated proteins, indicating widespread proteomic disruption (Fig. 9A). Compared with immunized mice, D-serine treatment resulted in 111 DEPs, suggesting partial normalization of immunization-induced protein alterations (Fig. 9B). Venn diagram analysis revealed both shared and unique DEPs among experimental groups (Fig. 9C). Heatmap analysis of the top 21 DEPs identified several proteins, including Prr12, Colgalt1, Tmem200a, Shpk, and Klhl2, that were markedly dysregulated in GluN1₃₅₉–₃₇₈ + vehicle mice but exhibited expression levels shifted toward sham values following D-serine treatment (Fig. 9D). Functional enrichment analysis demonstrated that DEPs in immunized mice were significantly enriched in pathways associated with synaptic organization, synaptic plasticity, axonal development, and immune-related processes (Fig. 9E). In contrast, D-serine-treated mice exhibited reduced enrichment of these dysregulated pathways compared with vehicle-treated immunized mice, indicating partial normalization of synaptic- and immune-related proteomic signatures. Taken together, these findings indicate that GluN1₃₅₉–₃₇₈ immunization induces widespread synaptic- and immune-related proteomic alterations in the hippocampus. D-serine treatment was associated with partial normalization of these molecular changes.

Fig. 9.

Fig. 9

The protective role of D-serine in axonal growth, synaptic transmission, and receptor function. A Volcano plot showing differentially expressed proteins (DEPs) between the GluN1₃₅₉–₃₇₈ + vehicle and sham + vehicle groups. Red indicates high expression of DEPs, while blue denotes low expression. A total of 155 DEPs were identified, including 83 upregulated and 72 downregulated proteins. B Volcano plot showing DEPs between the GluN1₃₅₉–₃₇₈ + D-serine and GluN1₃₅₉–₃₇₈ + vehicle groups. A total of 111 DEPs were identified, including 60 upregulated and 51 downregulated proteins. C Venn diagram of DEPs across four comparison groups. Overlapping proteins between groups highlight shared and unique targets affected by D-serine treatment. D Heatmap of the top 21 DEPs ranked by fold change. The GluN1₃₅₉–₃₇₈ + vehicle group displayed significant alterations in proteins such as Prr12, Colgalt1, Tmem200a, Shpk, and Klhl2, which were attenuated with D-serine treatment. E Gene Ontology (GO) enrichment analysis showing significantly altered pathways between the GluN1₃₅₉–₃₇₈ + vehicle and sham + vehicle groups (left) and the GluN1₃₅₉–₃₇₈ + D-serine and GluN1₃₅₉–₃₇₈ + vehicle groups (right). Synaptic organization, plasticity, and immune response pathways were prominently affected. Data with statistical significance determined using appropriate bioinformatic analyses

Discussion

Cognitive dysfunction in anti-NMDAR encephalitis is a major concern, with substantial evidence linking NMDAR dysfunction to these impairments [18]. Previous studies have highlighted the therapeutic potential of NMDAR modulators, such as SGE-301 and pregnenolone sulfate, in reversing cognitive deficits by restoring synaptic plasticity [45, 46]. Consistent with these reports, our study shows that D-serine significantly improved both recognition and working memory in a mouse model of anti-NMDAR encephalitis. Recognition memory deficits, assessed using the novel object recognition (NOR) test, were evident in peptide-immunized mice by day 10 and persisted until day 18. Treatment with D-serine starting on day 11 led to a rapid recovery of recognition memory by days 13 and 18, reaching levels comparable to controls. By day 26, however, spontaneous recovery in the model group reduced the observable treatment effect, and no significant differences between groups were detected. Supplementary analyses further indicated that exploration of the familiar object remained largely preserved, while total distance traveled was only modestly reduced, suggesting that the decreased exploration of the novel object primarily reflected impaired recognition memory rather than reduced locomotor activity. Working memory, evaluated by spontaneous alternation in the Y-maze, was significantly impaired in immunized mice at earlier time points and progressively improved following D-serine treatment, with significant effects observed on days 18 and 26. In contrast, locomotor activity was largely unaffected by D-serine treatment, consistent with previous studies showing that anti-NMDAR encephalitis predominantly disrupts cognitive rather than motor functions [46, 47]. Anxiety-like behavior, assessed using the open field test, was increased in immunized mice at day 13 but showed partial spontaneous recovery at later time points. Notably, D-serine did not significantly alter anxiety-related measures, further supporting a selective role for D-serine in improving cognitive deficits rather than affective symptoms.

Previous study found a single intracerebroventricular injection of anti-NMDAR antibodies followed by three consecutive days of intraperitoneal D-serine administration successfully replicated LTP impairments but did not produce significant cognitive deficits in tasks like the Morris water maze or novel object recognition [29]. This suggests that the duration and concentration of antibody exposure are crucial factors in the development of cognitive symptoms. In contrast, our model, which maintains high antibody levels in the CNS for 10 days, consistently demonstrated substantial cognitive impairments [46]. These results emphasize the need to optimize antibody delivery protocols to more accurately reflect the chronic nature of anti-NMDAR encephalitis observed in humans. Our study also further highlights the central role of NMDAR hypofunction in the cognitive deficits associated with this condition. The rapid recovery of recognition memory following D-serine treatment illustrates its ability to enhance NMDAR activity and synaptic function, consistent with prior findings in schizophrenia and dementia models [48–50]. However, the delayed improvement in working memory suggests that different neural circuits or mechanisms may be involved in various cognitive deficits. This warrants further exploration of the region-specific and time-dependent effects of D-serine on synaptic plasticity to better understand the complexities of anti-NMDAR encephalitis.

Our results showed that D-serine treatment did not significantly reduce serum IgG levels, indicating that its therapeutic effects are not driven by immune modulation. Instead, D-serine effectively alleviated cognitive dysfunction, as evidenced by increased NMDAR1 expression. However, its effects on NMDAR1 expression were time-dependent, diminishing with prolonged treatment. On day 18, D-serine was associated with increased NMDAR1 levels in total hippocampal protein lysates and partial restoration of receptor expression in key hippocampal regions, including CA1, CA3, and DG. By day 26, these improvements were no longer statistically significant, suggesting that the effect of D-serine on total hippocampal NMDAR1 levels was not sustained at the later stage of the disease. Previous studies have highlighted NMDAR internalization as a key pathological mechanism in anti-NMDAR encephalitis, driven by the crosslinking effect of autoantibodies [2]. While D-serine enhances synaptic function by optimizing the activity of remaining NMDARs, it has limited capacity to reverse antibody-induced receptor internalization and degradation. This emphasizes the need for adjunct therapies to stabilize NMDAR expression or promote receptor recycling, complementing D-serine’s functional benefits.

Our findings also suggest that the effects of D-serine in this model are primarily associated with the preservation of neuronal and synaptic integrity. D-serine restored NeuN expression and partially improved dendritic structure, as reflected by MAP2 immunostaining, together with increased levels of synaptic proteins including PSD95, synaptophysin, and SNAP25. These changes are consistent with stabilization of neuronal and synaptic components rather than direct regulation of synapse number, given that MAP2 reflects dendritic architecture rather than spine density. This interpretation aligns with current understanding that anti-NMDAR encephalitis is largely driven by antibody-mediated NMDA receptor internalization, leading to synaptic dysfunction without extensive neuronal loss [2, 51]. In parallel, GluN1₃₅₉–₃₇₈ immunization induced a marked increase in microglial density without detectable changes in morphological complexity, suggesting recruitment or proliferation rather than overt structural remodeling. Although D-serine did not significantly affect microglial abundance or morphology, it attenuated TSPO expression within IBA1⁺ microglia, indicating a selective influence on activation-associated cellular states. It has been reported that TSPO expression in microglia can serve as an indirect marker of activation status and may be associated with changes in microglia-mediated synaptic engulfment activity [52]. These findings suggest that microglial responses are present but may not represent the primary target of D-serine, and are more likely to reflect adaptive or secondary responses to neuronal dysfunction, consistent with the recognized role of microglia synaptic remodeling in maintaining synaptic homeostasis [53, 54]. Future studies should explore the potential benefits of combining D-serine with anti-inflammatory agents or glial modulators. Additionally, the selective improvement in cognitive function without significant effects on anxiety-like behaviors or locomotor activity suggests that D-serine primarily targets neuronal and synaptic pathways. Since glial dysfunction is implicated in anxiety and mood disorders [55], combining D-serine with therapies that modulate glial activity may offer a broader therapeutic strategy for anti-NMDAR encephalitis.

Proteomic analysis further supports the above interpretations by providing convergent molecular evidence. GluN1₃₅₉–₃₇₈ immunization induced broad alterations in proteins associated with synaptic organization, axonal development, and immune-related processes, whereas D-serine treatment was associated with partial normalization of synaptic- and axon-related protein networks. Several synaptic-associated proteins, including Prr12, Colgalt1, and Tmem200a, were dysregulated following immunization and showed partial restoration after D-serine administration [26, 27, 30]. Functional enrichment analysis indicated that pathways related to synaptic organization and plasticity were prominently affected, whereas immune-related pathways exhibited comparatively modest changes, consistent with the limited effects observed on microglial measures. Notably, our proteomic analysis revealed a reduction in complement-related proteins, including C1q, following D-serine treatment. Given the established role of the complement cascade in tagging synapses for microglia-mediated elimination, this observation raises the possibility that complement-dependent synaptic remodeling may be altered in this model [56]. C1q has also been implicated in activity-dependent synaptic pruning and pathological synapse loss under neuroinflammatory conditions [57]. However, as the present study did not directly assess microglia–synapse interactions or phagocytic activity, these findings should be interpreted with caution and considered as indirect, associative evidence rather than definitive proof of altered synaptic engulfment. Future studies incorporating high-resolution imaging or functional assays of microglial phagocytosis will be necessary to determine whether D-serine influences complement-mediated synaptic pruning in this model.

It is important to note that, these findings are consistent with clinical observations that cognitive deficits often persist in patients with anti-NMDAR encephalitis despite effective immunotherapy [11, 58]. In the present study, D-serine was administered without prior immunotherapeutic intervention to isolate its direct effects on neuronal and synaptic function under sustained antibody exposure. While this design does not fully replicate the clinical setting, it demonstrates that enhancing neuronal resilience may confer functional benefits independently of antibody reduction. These results support the concept that neuronal-targeted interventions could complement immune-based therapies, although the potential additive or synergistic effects of combining D-serine with immunotherapy remain to be determined. Future investigations incorporating immune modulation followed by D-serine treatment will be necessary to determine whether additive or synergistic cognitive recovery can be achieved, thereby more closely reflecting clinical practice.

Several additional limitations should also be acknowledged. First, although the active immunization mouse model of anti-NMDAR encephalitis is widely used, it does not fully capture the complexity of the human disease, particularly regarding immune heterogeneity and cognitive manifestations. Moreover, pertussis toxin was administered only to immunized mice to facilitate autoimmune induction; inclusion of a PTX-only control group in future studies would further strengthen the experimental design and help exclude potential nonspecific effects. Second, the duration and dosage of D-serine treatment may have influenced its efficacy. Its limited effects on microglial activation suggest that alternative dosing strategies or longer treatment periods might yield different outcomes. We also did not evaluate long-term effects or potential interactions with standard immunotherapies. Third, although proteomic and protein expression analyses suggest partial normalization of synaptic- and axon-related pathways, direct functional assessments (e.g., electrophysiology or structural analyses) were not performed, and dendritic spine density was not directly quantified. Finally, microglia–synapse interactions were not directly assessed. While TSPO expression within IBA1⁺ microglia was used as an indicator of activation-associated changes, it does not provide direct evidence of synaptic engulfment or phagocytic activity. More specific approaches, such as synaptosome phagocytosis assays or high-resolution imaging of microglia–synapse colocalization, were not performed due to technical constraints. Therefore, whether D-serine modulates microglia-mediated synaptic remodeling remains to be determined.

Conclusions

In summary, this study demonstrates that D-serine treatment was associated with improved cognitive performance and partial restoration of neuronal and synaptic integrity in a mouse model of anti-NMDAR encephalitis. These findings suggest that neuronal and synaptic alterations are key contributors to cognitive deficits in this model and highlight the potential of targeting NMDAR-related signaling to promote neuronal resilience. Importantly, our results support the concept that interventions aimed at restoring neuronal and synaptic function may complement immune-based therapies in anti-NMDAR encephalitis.

Supplementary Information

Supplementary Material 1. (11.4MB, docx)
Supplementary Material 2. (10.1MB, docx)

Authors’ contributions

Y.Y.: Investigation, Formal analysis, Data curation, Methodology, Writing–original draft. E.D.: Investigation, Formal analysis, Validation, Writing–original draft. J.H.: Investigation, Formal analysis, Writing–original draft. C.L.: Formal analysis, Data curation. T.Y.: Formal analysis, Data curation. T.Z.: Formal analysis, Data curation. Y.H.: Formal analysis. C.W.: Formal analysis. X.L.: Writing – review & editing. X.Q.: Conceptualization, Funding acquisition, Supervision, Writing – original draft, Writing – review & editing.

Funding

This work was supported by grants from the Natural Science Foundation of Zhejiang Province (LY24H090001), the Municipal Key R&D Program of Ningbo (2023Z175), the Zhejiang Province Applied Basic Research Program (2026C02A1256(SD2)) and the Ningbo Top Medical and Health Research Program (2022020304, 2024010317).

Data availability

The data that support the findings of this study are available from the corresponding author with reasonable request.

Declarations

Ethics approval and consent to participate

The animal study protocol was approved by the Committee of Ningbo University on the Ethics of Animal Experiments (NBU20230347, June 2023).

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

  • 1.Guasp M, Dalmau J. Autoimmune encephalitis. Med Clin North Am. 2025;109:443–61. [DOI] [PubMed] [Google Scholar]
  • 2.Varley JA, Strippel C, Handel A, Irani SR. Autoimmune encephalitis: recent clinical and biological advances. J Neurol. 2023;270:4118–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Graus F, Titulaer MJ, Balu R, Benseler S, Bien CG, Cellucci T, Cortese I, Dale RC, Gelfand JM, Geschwind M, et al. A clinical approach to diagnosis of autoimmune encephalitis. Lancet Neurol. 2016;15:391–404. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Hansen N, Lipp M, Vogelgsang J, Vukovich R, Zindler T, Luedecke D, Gingele S, Malchow B, Frieling H, Kühn S, et al. Autoantibody-associated psychiatric symptoms and syndromes in adults: A narrative review and proposed diagnostic approach. Brain Behav Immun Health. 2020;9:100154. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Harutyunyan G, Hauer L, Dünser MW, Karamyan A, Moser T, Pikija S, Leitinger M, Novak HF, Trinka E, Sellner J. Autoimmune encephalitis at the neurological intensive care unit: etiologies, reasons for admission and survival. Neurocrit Care. 2017;27:82–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Zandi MS, Paterson RW, Ellul MA, Jacobson L, Al-Diwani A, Jones JL, Cox AL, Lennox B, Stamelou M, Bhatia KP, et al. Clinical relevance of serum antibodies to extracellular N-methyl-D-aspartate receptor epitopes. J Neurol Neurosurg Psychiatry. 2015;86:708–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Michalski K, Abdulla T, Kleeman S, Schmidl L, Gómez R, Simorowski N, Vallese F, Prüss H, Heckmann M, Geis C, Furukawa H. Structural and functional mechanisms of anti-NMDAR autoimmune encephalitis. Nat Struct Mol Biol. 2024;31:1975–86. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Moscato EH, Peng X, Jain A, Parsons TD, Dalmau J, Balice-Gordon RJ. Acute mechanisms underlying antibody effects in anti-N-methyl-D-aspartate receptor encephalitis. Ann Neurol. 2014;76:108–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Andrzejak E, Rabinovitch E, Kreye J, Prüss H, Rosenmund C, Ziv NE, Garner CC, Ackermann F. Patient-derived anti-NMDAR antibody disinhibits cortical neuronal networks through dysfunction of inhibitory neuron output. J Neurosci. 2022;42:3253–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Gibson LL, McKeever A, Coutinho E, Finke C, Pollak TA. Cognitive impact of neuronal antibodies: encephalitis and beyond. Transl Psychiatry. 2020;10:304. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Brenner J, Ruhe CJ, Kulderij I, Bastiaansen AEM, Crijnen YS, Kret CN, Verkoelen JCP, Tolido AAG, Thomassen B, Kersten LP, et al. Long-term cognitive, functional, and patient-reported outcomes in patients with anti-NMDAR encephalitis. Neurology. 2024;103:e210109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Wright SK, Rosch RE, Wilson MA, Upadhya MA, Dhangar DR, Clarke-Bland C, Wahid TT, Barman S, Goebels N, Kreye J, et al. Multimodal electrophysiological analyses reveal that reduced synaptic excitatory neurotransmission underlies seizures in a model of NMDAR antibody-mediated encephalitis. Commun Biol. 2021;4:1106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Moscato EH, Jain A, Peng X, Hughes EG, Dalmau J, Balice-Gordon RJ. Mechanisms underlying autoimmune synaptic encephalitis leading to disorders of memory, behavior and cognition: insights from molecular, cellular and synaptic studies. Eur J Neurosci. 2010;32:298–309. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Titulaer MJ, McCracken L, Gabilondo I, Armangué T, Glaser C, Iizuka T, Honig LS, Benseler SM, Kawachi I, Martinez-Hernandez E, et al. Treatment and prognostic factors for long-term outcome in patients with anti-NMDA receptor encephalitis: an observational cohort study. Lancet Neurol. 2013;12:157–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Hoshina Y, Smith TL, Mstat AD, Wong KH, Peterson LK, Zekeridou A, Aboseif A, Coffey C, Wright MA, Banwell B, et al. An updated, comprehensive meta-analysis of the treatment of anti-NMDAR encephalitis: Analysis, equipoise, and the urgent need for evidence over anecdote. J Neuroimmunol. 2025;405:578651. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Abboud H, Probasco JC, Irani S, Ances B, Benavides DR, Bradshaw M, Christo PP, Dale RC, Fernandez-Fournier M, Flanagan EP, et al. Autoimmune encephalitis: proposed best practice recommendations for diagnosis and acute management. J Neurol Neurosurg Psychiatry. 2021;92:757–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Nguyen MK, Rastogi A, Kurtz I. True hyponatremia secondary to intravenous immunoglobulin. Clin Exp Nephrol. 2006;10:124–6. [DOI] [PubMed] [Google Scholar]
  • 18.McKeon GL, Robinson GA, Ryan AE, Blum S, Gillis D, Finke C, Scott JG. Cognitive outcomes following anti-N-methyl-D-aspartate receptor encephalitis: A systematic review. J Clin Exp Neuropsychol. 2018;40:234–52. [DOI] [PubMed] [Google Scholar]
  • 19.Liu X, Zhang L, Chen C, Gong X, Lin J, An D, Zhou D, Hong Z. Long-term cognitive and neuropsychiatric outcomes in patients with anti-NMDAR encephalitis. Acta Neurol Scand. 2019;140:414–21. [DOI] [PubMed] [Google Scholar]
  • 20.Saeed S, Wang H, Jia M, Liu TT, Xu L, Zhang X, Hu SH. The spectrum of overlapping anti-NMDAR encephalitis and demyelinating syndromes: a systematic review of presentation, diagnosis, management, and outcomes. Ann Med. 2025;57:2517813. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Stein H, Barbosa J, Rosa-Justicia M, Prades L, Morató A, Galan-Gadea A, Ariño H, Martinez-Hernandez E, Castro-Fornieles J, Dalmau J, Compte A. Reduced serial dependence suggests deficits in synaptic potentiation in anti-NMDAR encephalitis and schizophrenia. Nat Commun. 2020;11:4250. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Tsai GE. Ultimate translation: developing therapeutics targeting on N-methyl-d-aspartate receptor. Adv Pharmacol. 2016;76:257–309. [DOI] [PubMed] [Google Scholar]
  • 23.Kayser MS, Dalmau J. Anti-NMDA receptor encephalitis, autoimmunity, and psychosis. Schizophr Res. 2016;176:36–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Guasp M, Rosa-Justicia M, Muñoz-Lopetegi A, Martínez-Hernández E, Armangué T, Sugranyes G, Stein H, Borràs R, Prades L, Ariño H, et al. Clinical characterisation of patients in the post-acute stage of anti-NMDA receptor encephalitis: a prospective cohort study and comparison with patients with schizophrenia spectrum disorders. Lancet Neurol. 2022;21:899–910. [DOI] [PubMed] [Google Scholar]
  • 25.Coyle JT, Balu D, Wolosker H. D-Serine, the Shape-Shifting NMDA Receptor Co-agonist. Neurochem Res. 2020;45:1344–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Bodner O, Radzishevsky I, Foltyn VN, Touitou A, Valenta AC, Rangel IF, Panizzutti R, Kennedy RT, Billard JM, Wolosker H. D-serine signaling and NMDAR-mediated synaptic plasticity are regulated by system A-type of glutamine/D-serine dual transporters. J Neurosci. 2020;40:6489–502. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Wolosker H. D-serine regulation of NMDA receptor activity. Sci STKE 2006, 2006:pe41. [DOI] [PubMed]
  • 28.Ju J, Liu L, Yang X, Men S, Hou ST. Distinctive effects of NMDA receptor modulators on cerebral microcirculation in a schizophrenia mouse model. Biochem Biophys Res Commun. 2023;653:62–8. [DOI] [PubMed] [Google Scholar]
  • 29.Luo H, Yang X, Yang J, Han Z, Huang D, Gui J, Ding R, Chen H, Cheng L, Ma J, Jiang L. D-serine may ameliorate hippocampal synaptic plasticity impairment induced by patients’ anti-N-methyl-D-aspartate Receptor antibodies in mice. Biomedicines 2024, 12. [DOI] [PMC free article] [PubMed]
  • 30.Zhang XQ, Xu L, Zhu XY, Tang ZH, Dong YB, Yu ZP, Shang Q, Wang ZC, Shen HW. D-serine reconstitutes synaptic and intrinsic inhibitory control of pyramidal neurons in a neurodevelopmental mouse model for schizophrenia. Nat Commun. 2023;14:8255. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Wagnon I, Hélie P, Bardou I, Regnauld C, Lesec L, Leprince J, Naveau M, Delaunay B, Toutirais O, Lemauff B, et al. Autoimmune encephalitis mediated by B-cell response against N-methyl-d-aspartate receptor. Brain. 2020;143:2957–72. [DOI] [PubMed] [Google Scholar]
  • 32.Ding Y, Zhou Z, Chen J, Peng Y, Wang H, Qiu W, Xie W, Zhang J, Wang H. Anti-NMDAR encephalitis induced in mice by active immunization with a peptide from the amino-terminal domain of the GluN1 subunit. J Neuroinflammation. 2021;18:53. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Weber MS, Benkhoucha M, Lehmann-Horn K, Hertzenberg D, Sellner J, Santiago-Raber ML, Chofflon M, Hemmer B, Zamvil SS, Lalive PH. Repetitive pertussis toxin promotes development of regulatory T cells and prevents central nervous system autoimmune disease. PLoS ONE. 2010;5:e16009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Zamvil SS, Steinman L. The T lymphocyte in experimental allergic encephalomyelitis. Annu Rev Immunol. 1990;8:579–621. [DOI] [PubMed] [Google Scholar]
  • 35.Hughes EG, Peng X, Gleichman AJ, Lai M, Zhou L, Tsou R, Parsons TD, Lynch DR, Dalmau J, Balice-Gordon RJ. Cellular and synaptic mechanisms of anti-NMDA receptor encephalitis. J Neurosci. 2010;30:5866–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Kruse JL, Lapid MI, Lennon VA, Klein CJ, Toole OO, Pittock SJ, Strand EA, Frye MA, McKeon A. Psychiatric autoimmunity: N-methyl-D-aspartate receptor IgG and beyond. Psychosomatics. 2015;56:227–41. [DOI] [PubMed] [Google Scholar]
  • 37.Lancaster E, Dalmau J. Neuronal autoantigens–pathogenesis, associated disorders and antibody testing. Nat Rev Neurol. 2012;8:380–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Xu L, Zhou Y, Hu L, Jiang H, Dong Y, Shen H, Lou Z, Yang S, Ji Y, Ruan L, Zhang X. Deficits in N-methyl-D-aspartate receptor function and synaptic plasticity in hippocampal CA1 in APP/PS1 mouse model of Alzheimer’s disease. Front Aging Neurosci. 2021;13:772980. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Yang X, Sun A, Kong L, Yang X, Zhao X, Wang S. Inhibition of NLRP3 inflammasome alleviates cognitive deficits in a mouse model of anti-NMDAR encephalitis induced by active immunization. Int Immunopharmacol. 2024;137:112374. [DOI] [PubMed] [Google Scholar]
  • 40.Kelleher E, Mothersill D, Hargreaves A, Barry H, Smyth S, Chaila E, Boers P, McCabe DJ, Sweeney B, Costello D, et al. Cognitive outcomes and performance of patients diagnosed and treated for N-Methyl-D-Aspartate receptor antibody-mediated (NMDAR) encephalitis compared with patients with schizophrenia and healthy controls. Psychiatry Res Neuroimaging. 2025;349:111983. [DOI] [PubMed] [Google Scholar]
  • 41.Dalmau J, Armangué T, Planagumà J, Radosevic M, Mannara F, Leypoldt F, Geis C, Lancaster E, Titulaer MJ, Rosenfeld MR, Graus F. An update on anti-NMDA receptor encephalitis for neurologists and psychiatrists: mechanisms and models. Lancet Neurol. 2019;18:1045–57. [DOI] [PubMed] [Google Scholar]
  • 42.Dalmau J, Lancaster E, Martinez-Hernandez E, Rosenfeld MR, Balice-Gordon R. Clinical experience and laboratory investigations in patients with anti-NMDAR encephalitis. Lancet Neurol. 2011;10:63–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Forrester A, Latorre S, O’Dea PK, Robinson C, Goldwaser EL, Trenton A, Tobia A, Aziz R, Dhawan S, Brennan A, et al. Anti-NMDAR encephalitis: A multidisciplinary approach to identification of the disorder and management of psychiatric symptoms. Psychosomatics. 2020;61:456–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Maudes E, Planagumà J, Marmolejo L, Radosevic M, Serafim AB, Aguilar E, Sindreu C, Landa J, García-Serra A, Mannara F, et al. Neuro-immunobiology and treatment assessment in a mouse model of anti-NMDAR encephalitis. Brain. 2025;148:2023–37. [DOI] [PubMed] [Google Scholar]
  • 45.Warikoo N, Brunwasser SJ, Benz A, Shu HJ, Paul SM, Lewis M, Doherty J, Quirk M, Piccio L, Zorumski CF, et al. Positive allosteric modulation as a potential therapeutic strategy in anti-NMDA receptor encephalitis. J Neurosci. 2018;38:3218–29. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Mannara F, Radosevic M, Planagumà J, Soto D, Aguilar E, García-Serra A, Maudes E, Pedreño M, Paul S, Doherty J, et al. Allosteric modulation of NMDA receptors prevents the antibody effects of patients with anti-NMDAR encephalitis. Brain. 2020;143:2709–20. [DOI] [PubMed] [Google Scholar]
  • 47.Planagumà J, Leypoldt F, Mannara F, Gutiérrez-Cuesta J, Martín-García E, Aguilar E, Titulaer MJ, Petit-Pedrol M, Jain A, Balice-Gordon R, et al. Human N-methyl D-aspartate receptor antibodies alter memory and behaviour in mice. Brain. 2015;138:94–109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Dong B, Yue Y, Dong H, Wang Y. N-methyl-D-aspartate receptor hypofunction as a potential contributor to the progression and manifestation of many neurological disorders. Front Mol Neurosci. 2023;16:1174738. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Orzylowski M, Fujiwara E, Mousseau DD, Baker GB. An overview of the involvement of D-serine in cognitive impairment in normal aging and dementia. Front Psychiatry. 2021;12:754032. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Zhang T, Liu C, Zhong N, Wang Y, Huang Y, Zhang X. Advances in the treatment of cognitive impairment in schizophrenia: targeting NMDA receptor pathways. Int J Mol Sci. 2024;25:10668. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Huang Q, Xie Y, Hu Z, Tang X. Anti-N-methyl-D-aspartate receptor encephalitis: A review of pathogenic mechanisms, treatment, prognosis. Brain Res. 2020;1727:146549. [DOI] [PubMed] [Google Scholar]
  • 52.Shi Y, Cui M, Ochs K, Brendel M, Strübing FL, Briel N, Eckenweber F, Zou C, Banati RB, Liu GJ, et al. Long-term diazepam treatment enhances microglial spine engulfment and impairs cognitive performance via the mitochondrial 18 kDa translocator protein (TSPO). Nat Neurosci. 2022;25:317–29. [DOI] [PubMed] [Google Scholar]
  • 53.Xie M, Wang T, Feng J, Ma D, Feng L, Hao Y. Roles of microglia in synaptogenesis, synaptic pruning, and synaptic plasticity in physiological conditions and central nervous system disorders. Curr Neuropharmacol. 2026;24:2–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Liu Y, Shen X, Zhang Y, Zheng X, Cepeda C, Wang Y, Duan S, Tong X. Interactions of glial cells with neuronal synapses, from astrocytes to microglia and oligodendrocyte lineage cells. Glia. 2023;71:1383–401. [DOI] [PubMed] [Google Scholar]
  • 55.Yang XY, Wang HQ, Wang ZZ, Chen NH. Linking depression and neuroinflammation: Crosstalk between glial cells. Eur J Pharmacol. 2025;995:177408. [DOI] [PubMed] [Google Scholar]
  • 56.Stevens B, Allen NJ, Vazquez LE, Howell GR, Christopherson KS, Nouri N, Micheva KD, Mehalow AK, Huberman AD, Stafford B, et al. The classical complement cascade mediates CNS synapse elimination. Cell. 2007;131:1164–78. [DOI] [PubMed] [Google Scholar]
  • 57.Wu X, Gao Y, Shi C, Tong J, Ma D, Shen J, Yang J, Ji M. Complement C1q drives microglia-dependent synaptic loss and cognitive impairments in a mouse model of lipopolysaccharide-induced neuroinflammation. Neuropharmacology. 2023;237:109646. [DOI] [PubMed] [Google Scholar]
  • 58.Heine J, Kopp UA, Klag J, Ploner CJ, Prüss H, Finke C. Long-term cognitive outcome in anti-N-methyl-D-aspartate receptor encephalitis. Ann Neurol. 2021;90:949–61. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material 1. (11.4MB, docx)
Supplementary Material 2. (10.1MB, docx)

Data Availability Statement

The data that support the findings of this study are available from the corresponding author with reasonable request.


Articles from Journal of Neuroinflammation are provided here courtesy of BMC

RESOURCES