Mice carrying a GluN2B protein-truncating variant have altered NMDA receptor subunit composition and their behavior recapitulates patient phenotypes

TL;DR

Mice with a GluN2B mutation show changes in brain receptor function and behavior similar to human patients with intellectual disability.

Contribution

A novel mouse model reveals how a GluN2B protein-truncating variant affects NMDA receptor composition and behavior.

Findings

Reduced full-length GluN2B protein and increased GluN2A were observed in the mouse model.

Behavioral changes included hypoactivity, increased anxiety in males, and impaired sensorimotor gating.

Altered NMDA receptor function was consistent with a shift toward GluN2A-containing receptors.

Abstract

Pathogenic variants in GRIN2B, encoding the NMDA receptor (NMDAR) GluN2B subunit, are linked to intellectual disability (ID) and related neurodevelopmental disorders. While most disease-associated variants are missense, protein-truncating variants (PTVs) may cause haploinsufficiency with less severe phenotypes. Here, we characterize a knock-in mouse model carrying the GluN2B-L825Ffs*15 PTV (Grin2b+/Δ). Proteomic analysis revealed markedly reduced full-length GluN2B protein and no detectable truncated GluN2B, accompanied by a small compensatory increase in GluN2A. Electrophysiology in hippocampal neurons demonstrated reduced NMDA-induced currents, diminished ifenprodil sensitivity, and accelerated NMDAR-mediated EPSC deactivation, consistent with a shift toward GluN2A-containing receptors. AMPAR-mEPSC amplitudes were increased, indicating altered excitatory synaptic function. Behaviorally, Grin2b+/Δ mice exhibited hypoactivity, increased anxiety in males, and impaired sensorimotor gating in both sexes, while learning, memory, and social behaviors remained largely intact. These results demonstrate that a monoallelic GluN2B PTV alters NMDAR subunit composition and function, producing moderate behavioral effects, and provide insight into mechanisms underlying GRIN2B-associated ID.

The online version contains supplementary material available at 10.1007/s00018-025-06057-1.

Figures

Full text

Introduction

Glutamate is the primary excitatory neurotransmitter in the central nervous system, acting on ionotropic NMDA, AMPA, and kainate receptors. NMDA receptors (NMDARs) are characterized by slow gating kinetics and high Ca^2+^ permeability, and they perform key synaptic computations underlying cognitive function and guiding nervous system development [1]. NMDARs are heterotetrameric cation channels comprising two obligatory GluN1 and two GluN2 and/or GluN3 subunits [1]. There are four types of GluN2 subunits (GluN2A-D), with distinct spatial and temporal expression patterns [2–4] and with influence on receptor biophysical and pharmacological properties [5–7]. The GluN2B subunit, whose expression peaks during embryonic and early postnatal development, is critical for synapse development, and various forms of synaptic plasticity [1, 8].

A significant portion of severe neurodevelopmental disorders likely have de novo monogenic causes, with GRIN genes, encoding NMDAR subunits, among the implicated genes [9]. GRIN gene variants have been linked to developmental delay (DD), intellectual disability (ID), epilepsy, autism spectrum disorder (ASD) and schizophrenia [10–15]. Over 700 disease-associated GRIN gene variants have been identified, predominantly in genes encoding GluN2A (43%), GluN2B (35%), and GluN1 (13%) subunits [1, 16, 17]. Most frequent are missense variants, accounting for 65% of the identified disease-associated GRIN gene variants, while 23% are protein-truncating variants (PTVs) [18]. Compared to missense variants, PTVs tend to be associated with a less severe DD/ID presentation in carriers [18–20].

Disease mechanisms underlying symptoms in patients with missense variants vs. PTVs are not known. At the molecular level, missense variants lead to a single amino acid substitution in the subunit protein that, when incorporated into NMDAR complexes, may alter various aspects of receptor function, including channel gating, allosteric modulation, post-translational modifications, protein-protein interactions, or trafficking [16, 17, 21]. In contrast, it is thought that the products of most PTVs (with the exception of PTVs affecting the subunit C-terminal domain (CTD) [22] are degraded at the mRNA or the protein stage, or fail to assemble into stable NMDAR complexes, resulting in haploinsufficiency of the affected subunit [18].

To study the circuit-level functional impact of different disease-causing GRIN gene variants, animal models carrying clinically relevant variants are needed. In recent years, several mouse strains carrying individual patient missense variants have been created and characterized, including two with patient variants in Grin2b, encoding GluN2B-C456Y [23] or GluN2B-L825V [24]. To assess haploinsufficiency, heterozygous individuals of conventional Grin2b knock-out strains have been examined [23], but such data are very limited so far.

To investigate the functional consequences and possible compensatory mechanisms linked to disease-causing GluN2B PTVs, we generated a knock-in mouse model carrying GluN2B-L825Ffs*15 (GluN2B^Δ^) frameshift variant in the Grin2b gene, creating a premature termination codon in the transmembrane domain M4 helix. We observed severely impaired cell surface expression of the GluN2B^Δ^ subunit in heterologous expression systems. In the hippocampal tissue of heterozygous (Grin2b^+/Δ^) mice, we found decreased mRNA levels of Grin1 and Grin2b, accompanied by reduced GluN2B and slightly increased GluN2A protein levels. Cultured hippocampal neurons from Grin2b^+/Δ^ mice exhibited smaller NMDA-induced current amplitudes and accelerated NMDAR-mediated evoked excitatory postsynaptic current (eEPSC) deactivation kinetics. Together, our data suggest that in Grin2b^+/Δ^ mice, the truncated GluN2B^Δ^ protein is present only at negligible levels, and—given the abundance of GluN2A subunits—the relative contribution of GluN2A-containing NMDARs is consequently increased. Behaviorally, Grin2b^+/Δ^ mice showed hypoactivity, increased anxiety, altered sensorimotor gating, but no apparent cognitive impairment. This study provides insights into the biological consequences of Grin2b PTVs and may inform the development of personalized therapies for patients with different variants in GRIN2B.

Methods

Ethics statement

Animals were bred and maintained according to the guidelines of the European Union Directive 2010/63/EU and in full compliance with the Czech Act No. 246/1992 Coll. and Decree No. 419/2012 Coll., regarding the protection of animals used for scientific purposes. All experiments were approved by the Animal Care and Use Committee of the Institute of Molecular Genetics, Czech Academy of Sciences, and by the Resort Professional Commission of the Czech Academy of Sciences for Approval of Projects of Experiments on Animals.

Animals

Heterozygous mice (C57BL/6N background) carrying the GluN2B-L825Ffs15 mutation in the Grin2b gene (Grin2b^+/Δ^) were obtained using CRISPR/Cas9 genome-editing system [24]. Cas9 protein, gRNA, and ssDNA were transfected into the zygote by electroporation. In the process of generating the mouse strain carrying the GluN2B-L825V variant [24] we also obtained mice with the GluN2B-L825Ffs15 variant. Results of genome editing were confirmed by PCR amplification in the founder mice and the presence of the frameshift in individual mice was also confirmed by direct Sanger sequencing of PCR amplicons. Heterozygous mice were obtained from breeding Grin2b^+/Δ^ with wild-type mice. Newborn Grin2b^+/Δ^ pups and their wild-type littermates (Grin2b^+/+^) were used for the preparation of primary neuronal cultures. For behavioral experiments, mice were housed under standard 12:12 h light/dark cycle and received food and water ad libitum. Animals underwent behavioral testing at 9–15 weeks of age.

RNA isolation and RT-qPCR

Total RNA was extracted from the individual hippocampi of 4-month-old male and female mice. Brain tissue was frozen on dry ice immediately after dissection and stored at −80 °C until RNA isolation. The tissue samples were homogenized in TRI Reagent (Merck), and total RNA was extracted according to the manufacturer’s protocol. The RNA pellet was resuspended in 20 µl of RNase-free water and the concentration of RNA was measured using a Nano Drop instrument (Thermo Scientific). We used 1 µg of total RNA for cDNA synthesis. We combined 250 ng of random primers (Promega) with total RNA, the mixture was incubated at 70 °C for 5 min and then cooled on ice for 2 min. Subsequently, cDNA was synthetized (at 37 °C, for 90 min) in a 25 µl reaction containing 200 U of M-MLV Reverse Transcriptase (Promega). The obtained cDNA was additionally diluted by 25 µl of PCR grade water and stored at 4 °C until RT-qPCR analysis.

TaqMan qPCR reactions were performed in 20 µl qPCR reaction mixture containing 1.5 µl cDNA samples, 10 µl 2x TaqMan Gene Expression Master Mix, and 2 µl TaqMan Gene Expression Assay (Thermo Fisher Scientific Inc.) containing the probe and the appropriate forward and reverse primers (Table S1). All qPCR reactions were performed in triplicate on a LightCycler 480 instrument (Roche Diagnostics). Relative gene expression levels (RQ values) of target genes were analyzed using the ΔΔCt method [25, 26], where Ct refers to the threshold cycle value. The mean Ct values obtained from qPCR were normalized to the levels of housekeeping genes glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and beta-2-microglobulin (β2M) (Table S1).

Mass spectrometry analysis

Mass spectrometry analysis was performed as previously described [24]. Hippocampal samples from 4-month-old male mice (n = 5 for Grin2b^+/+^ and n = 5 for Grin2b^+/Δ^) were lysed by multiple rounds of sonication in 2% sodium deoxycholate detergent solution in 100 mM triethylammonium bicarbonate buffer. Cysteines were reduced with Tris(2-carboxyethyl) phosphine and alkylated with chloroacetamide. Proteins were digested with trypsin, labeled with 16-plex tandem mass tags (TMTpro), and fractionated by reverse-phase chromatography at pH 10. The resulting eight fractions were analyzed using nanoLC coupled with a Thermo Orbitrap Lumos mass spectrometer, with TMT quantification performed in triple-stage mass spectrometry (MS3). Data processing was conducted in ProteomeDiscoverer 2.5 (Thermo Scientific, USA), using TMT reporter ion ratios to estimate relative protein abundance. The database search was performed against the Mus musculus UniProt reference database and a common contaminant database. Fixed modifications included TMTpro labeling at the peptide N-terminus and lysine (Unimod #2016) and cysteine carbamidomethylation (Unimod #39), while variable modifications included methionine oxidation (Unimod #1384) and protein N-terminal acetylation (Unimod #1). Proteins and peptides were filtered to a 1% false discovery rate (FDR).

Data were normalized to total peptide amount, and statistical analysis was performed in Perseus software [27]. Contaminants were removed, intensity values were log₂-transformed, and proteins with insufficient quantification (fewer than three valid values in at least one experimental group) were excluded. A Student’s t-test with permutation-based FDR correction was applied to identify significantly altered proteins at a 5% FDR threshold. ASD-related genes from the SFARI database (https://gene.sfari.org) were annotated.

To estimate the protein copy number (CN) per cell, the “proteomic ruler” method [28] was used. This estimation utilized the mass spectrometry signal of histones, which correlates with DNA content and thus the number of cells in a sample. CN calculations were performed using the Proteomics Ruler Perseus plugin, with the scaling mode set to total protein amount.

Calculation of proportions of diheteromeric and triheteromeric NMDA receptors

Mass spectrometry provided the CN for the GluN1, GluN2A and GluN2B subunits of NMDA receptors (CN1, CN2A, CN2B). To estimate the relative proportions of NMDA receptor subtypes based on subunit composition, we assumed that each receptor tetramer comprises two obligatory GluN1 subunits and two additional subunits randomly occupied by either GluN2A or GluN2B. The contribution of other GluN2 or GluN3 subunits was considered negligible for the purpose of this analysis. The total number of assembled receptors depends on whether there are fewer GluN1 or GluN2 subunits. It is equal to the smaller number of CN1/2 or (CN2A + CN2B)/2. Out of all GluN2 subunits (CN2A + CN2B), the proportions of GluN2A and GluN2B are equal to:

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{P}_{2\mathrm{A}}={CN}_{2A}/({CN}_{2\mathrm{A}}+\:{CN}_{2\mathrm{B}})$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{P}_{2B}={CN}_{2B}/({CN}_{2\mathrm{A}}+\:{CN}_{2\mathrm{B}})$$\end{document}

The probability that a receptor with a 1-1-2A-2A subunit composition assembles and therefore the proportion of receptors with a 1-1-2A-2A subunit composition (P2 A − 2A) is equal to P2A^2^:

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{P}_{2\mathrm{A}-2\mathrm{A}}={{P}_{2A}^{2}=\left({CN}_{2\mathrm{A}}\right)}^{2}/{({CN}_{2\mathrm{A}}+\:{CN}_{2\mathrm{B}})}^{2}$$\end{document}

Similarly, out of all assembled receptors, the proportions of 1-1-2B-2B and 1-1-2A-2B receptors, respectively, are given below. The last expression contains the factor of 2 since combinations of 2A-2B and 2B-2A provide receptors with the same subunit composition.

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{P}_{2B-2B}={{P}_{2B}^{2}=\left({CN}_{2B}\right)}^{2}/{({CN}_{2\mathrm{A}}+\:{CN}_{2\mathrm{B}})}^{2}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{P}_{2\mathrm{A}-2B}={P}_{2A}{P}_{2B}={2\:CN}_{2A}\:{CN}_{2B}/{({CN}_{2\mathrm{A}}+\:{CN}_{2\mathrm{B}})}^{2}$$\end{document}

Homology modeling

The initial models of the murine GluN1/GluN2B complex were constructed based on a previously developed rat NMDAR homology model, which was derived from the available crystal structures 4pe5, 4tll, and 4tlm [29, 30]. The template receptor geometry corresponded to the ligand-activated state (RAA) as obtained from previous opening/closing simulations [31]. Three models of the GluN1/GluN2B complex were generated using the automodel function in MODELLER 9.23 [32, 33]. Each model was assembled with two GluN1 subunits and all possible combinations of two GluN2B subunits, either incorporating or omitting the truncation mutation. The GluN1 subunit models encompassed residues 23 to 847, in accordance with the UniProt record P35438 [34], excluding the last 91 residues of the CTD, which are disordered. The GluN2B subunits containing the truncation mutation in the transmembrane helix M4 included residues 30 to 824, followed by the sequence FFGGSHGSQPHHLHL. In contrast, the wild-type GluN2B models incorporated residues 30 to 852. The largely disordered CTD, comprising residues 853 to 1482, was excluded from the modeling process.

Molecular dynamics simulations

Molecular dynamics (MD) simulations of the murine GluN1/GluN2B receptor model were prepared using CHARMM-GUI [35, 36]. The receptor was embedded in a model membrane composed of DOPC and cholesterol (3:1 ratio). The simulation box (160 Å × 160 Å × 220 Å) was solvated with TIP3P water, and Na⁺ and Cl⁻ ions were added to neutralize the system at a final salt concentration of 150 mM. Following the CHARMM-GUI equilibration protocol, 500 ns production MD simulations were conducted using GPU-accelerated GROMACS 2023.2 [37] with the CHARMM36 force field [38]. A 2 fs time step was used, with short-range Coulomb and van der Waals cutoffs set to 1.2 nm, while long-range electrostatics was handled using Particle Mesh Ewald (PME) [39]. Atomic coordinates were recorded every 10 ps. The NPT ensemble simulation was controlled using the Nosé-Hoover thermostat at 303.15 K, with a coupling parameter of 1.0 ps, applying separate thermal baths for the solute, membrane, and solvent. The Parrinello-Rahman barostat maintained the system at 1.0 atm, with a coupling time constant of 5.0 ps and a compressibility factor of 4.5 × 10⁻⁵. MD geometry snapshots were superposed using the GROMACS trjconv -fit rot + trans command, with the first GluN1 chain as the reference for least-squares fitting. Visualization, graphical representation, and geometric analysis of selected regions were performed using PyMOL version 2.6.

Analysis of subunit cell surface expression

DNA Constructs. The following cDNAs encoding human NMDA receptor subunits were used: GluN1-1a (NM_007327), GluN2A (NM_001134407), and GluN2B (NM_000834), all cloned into the pCI-neo expression vector (a generous gift from prof. Stephen Traynelis, Emory University). To analyze receptor surface expression, a green fluorescent protein (GFP) tag was introduced into the extracellular portion of GluN2B directly after the signal peptide using the In Vivo Assembly (IVA) cloning method [40]. The resulting construct, GluN2B-GFP, was subsequently transferred via restriction enzyme cloning into the pLEX-MCS expression vector (OpenBiosystems, Thermo Scientific) for lentiviral particle production and subsequent GluN2B-GFP expression in neurons [22, 24, 41]. To replicate the genetic modification present in our mouse model, insertion of four thymine bases at the position L825 was achieved using IVA cloning, creating a frameshift mutation identical to the murine model. To ensure the same premature termination codon after 15 amino acids, an additional synonymous mutation at C838 was introduced by IVA mutagenesis, changing the triplet TGC (wild-type sequence) to TGT, thereby generating the mutant construct GluN2B-L825Ffs*15-GFP (GluN2B^Δ^-GFP) in both pCI-neo and pLEX-MCS vectors. Primer design was performed manually, and oligonucleotides were purchased from Eurofins Genomics (Germany). Following PCR amplification using Q5 DNA Polymerase (NEB), DpnI-treated products were transformed into ultracompetent XL10-Gold E. coli cells (Agilent, Santa Clara, USA). Clones were grown overnight, plasmid DNA was extracted, and sequencing (Eurofins Genomics) was performed to confirm the presence of the intended variant. Amino acid numbering follows the full-length protein sequence, including the signal peptide, with the initiating methionine designated as position 1.

Cell culture and transfection. Human embryonic kidney 293T (HEK293T) (American Type Culture Collection, ATCC No. CRL-1573, Rockville, MD, USA and monkey COS-7 cells were cultured as described earlier [22]. Surface expression experiments were performed 24 h (microscopy) or 48 h (western blot) after the transfection of HEK293T cells and 48 h after the transfection of COS-7 cells with cDNA encoding GluN1 and GluN2A-GFP, GluN2B-GFP or GluN2B^Δ^-GFP subunits. Transfection was carried out using the GenJet reagent (SignaGen) following the manufacturer’s protocol. Following transfection, cells were maintained in culture medium supplemented with NMDAR antagonists D-(-)-2-amino-5-phosphonopentanoic acid (AP-5, 50 µM), Mg^2+^ (20 mM), and ketamine (1 µg/ml). Electrophysiology experiments were conducted 24–48 h after the transfection of HEK293T cells with cDNA encoding human GluN1 and GluN2B subunits, along with GFP (pQBI 25, Takara, Tokyo, Japan) as a marker of successful transfection, using Matra-A transfection reagent (IBA, Göttingen, Germany) as previously described [42]. For the analysis of recombinant subunit cell surface expression, primary neuronal cultures were prepared from mouse hippocampi on postnatal day 1 (P1) and cultured under standard conditions using Neurobasal A medium supplemented with B27 (Gibco/Invitrogen). Neurons were transduced with lentiviruses containing genes for GluN2B-GFP or GluN2B^Δ^-GFP expression, following previously established protocols [22, 41].

Analysis of surface expression by SDS-PAGE and Western blot in HEK293T cells

Experiments were performed 48 h after the co-transfection of HEK293T cells with GluN1 and either GluN2B-GFP or GluN2B^Δ^-GFP subunits. NMDAR antagonists (AP-5, 50 µM; Mg^2+^, 20 mM; ketamine, 1 µg/ml) were added to the culture medium to prevent receptor activation. Cells were detached using 12.5 mM ethylenediaminetetraacetic acid (EDTA) at 37 °C for 30 min [43], collected by centrifugation, resuspended in phosphate-buffered saline (PBS), and fractionated. One quarter of the cell suspension was used as the total cell lysate (T) and incubated in 0.5% NP-40/Tris buffer with Halt Protease Inhibitor Cocktail (Thermo Scientific). The remaining cells were fractionated using the NEXT protocol [44]. Cells were first treated with EB1 buffer (0.015% digitonin) to isolate the cytosolic fraction (S1), followed by EB2 buffer (0.5% Triton X-100, Benzonase Nuclease) to obtain the cytoplasmic membrane (S2) and nuclear/organelle (P) fractions. Samples were mixed with 4× Laemmli buffer, denatured at 95 °C, separated on 4–20% Tris-glycine SDS-PAGE gels, and transferred to nitrocellulose membranes. Membranes were blocked in 5% milk/PBS-Tween and incubated with primary antibodies: anti-GFP (1:2000; 11–476; EXBIO) for GFP-tagged GluN2B, anti-GAPDH (1:2000; 11–942; EXBIO) as a cytosolic marker, and anti-histone H3 (1:3000; ab1791; Abcam) as a nuclear marker. After washing, a horseradish peroxidase (HRP)-conjugated secondary antibody (1:6000, Thermo Scientific) was applied, signals were developed using SuperSignal West Pico substrate (Thermo Scientific) and acquired with a CCD camera system (Vilber Loumart). Notably strong signals of GAPDH verified cytosolic enrichment of the S1 fraction and histone H3 confirmed nuclear enrichment of the P fraction.

Analysis of surface expression by a colorimetric assay in COS-7 cells

The colorimetric surface expression assay was performed as previously described [22] using COS-7 cells co-transfected with GluN1 and either GluN2B-GFP or GluN2B^Δ^-GFP subunits and maintained in culture medium supplemented with NMDAR antagonists (AP-5, 50 µM; Mg^2+^, 20 mM; ketamine, 1 µg/ml). To assess surface and total receptor expression, two 24-well plates were used for each experiment, with each condition run in quadruplicate. Cells transfected with GluN1 alone served as background controls. Cells were washed and fixed in 4% paraformaldehyde (PFA) before treatment with either the surface antigen blocking solution (PBS containing 50 mM NH_4_Cl, 10% goat serum, and 2% fish gelatin) or the permeabilizing blocking solution (PBS with 50 mM NH₄Cl, 10% goat serum, 2% fish gelatin, 0.1% Triton X-100, and 0.1% Tween-20). After blocking, cells were incubated with a rabbit anti-GFP primary antibody (1:500; 11–476; EXBIO) followed by a goat anti-rabbit HRP-conjugated secondary antibody (1:500; AQ132P; Sigma-Aldrich). Detection was performed using o-phenylenediamine dihydrochloride (SIGMAFAST OPD tablet set; P9187; Sigma-Aldrich), and absorbance was measured at 492 nm using an Infinite PRO 200 multifunction modular reader (Tecan). Surface and total receptor expression values were obtained after background subtraction. The average value of surface labelling was subtracted from the average value of total labelling thus obtaining the value of intracellular labelling. Surface expression was determined as the ratio of surface to intracellular values.

Analysis of surface expression by fluorescence microscopy in HEK293T cells

Fluorescence microscopy-based surface expression analysis was performed as previously described [22]. HEK293T cells were co-transfected with GluN1 along with GluN2A and/or GluN2B-GFP or GluN2B^Δ^-GFP subunits and maintained in culture medium supplemented with NMDAR antagonists (AP-5, 50 µM; Mg^2+^, 20 mM; ketamine, 1 µg/ml). Experiments were performed 24 h post-transfection. Cells were washed with PBS, fixed in 4% PFA, and blocked using the surface antigen blocking solution (PBS containing 50 mM NH₄Cl, 10% goat serum, and 2% fish gelatin). Surface expression was assessed by incubating cells with a rabbit anti-GFP primary antibody (1:1000; AB3080P; Merck) followed by a goat anti-rabbit Alexa Fluor 647 secondary antibody (1:1000; A-21244; Invitrogen). Intracellular receptor expression was evaluated following blocking and permeabilization (PBS containing 50 mM NH₄Cl, 10% goat serum, 2% fish gelatin, 0.1% Triton X-100, and 0.1% Tween-20) using the same primary antibody and a goat anti-rabbit Alexa Fluor 555 secondary antibody (1:1000; A-21428; Invitrogen). 4′,6-diamidino-2-phenylindole (DAPI) was used to visualize cell nuclei. Images were acquired using a Leica DMi8 S microscope equipped with an sCMOS camera (Leica DFC9000 GTC), captured as z-stack images (z-step size: 0.18 μm) using a PlanAPO 63 × (1.47 NA) oil-immersion objective, and analyzed with ImageJ software. Surface expression was quantified as the ratio of surface-to-intracellular fluorescence signals [41].

Analysis of surface expression by fluorescence microscopy in neurons

The analysis of surface expression in neurons was performed as previously described [22, 24, 41]. At 4 days in vitro (DIV), lentiviral particles encoding GluN2B-GFP or GluN2B^Δ^-GFP subunits were introduced into primary neuronal cultures. The surface and intracellular expression of GluN2B-GFP subunits was assessed at 14 DIV, following a protocol similar to that used for HEK293T cells, with minor modifications. Antibodies were diluted to 1:500, and intracellular GluN2B staining was performed using a goat anti-rabbit Alexa Fluor 488 secondary antibody (1:500; AA-11034; Invitrogen). Images were acquired and analyzed as described for the HEK293T cell surface expression analysis, with an additional deconvolution step performed using Huygens software with the Classic Maximum Likelihood Estimation algorithm.

Primary hippocampal microisland cultures

For electrophysiological experiments primary hippocampal microisland cultures were used and prepared as previously described [45]. Glass coverslips (24 mm; Glaswarenfabrik Karl Hecht) were coated with 0.15% agarose (Serva) and allowed to dry. Microdots (~ 200 μm in diameter, spaced ~ 200 μm apart) were stamped using a growth-permissive substrate containing 0.1 mg/ml poly-D-lysine (Sigma) and 0.2 mg/ml collagen (Serva). Following decapitation, cortices and hippocampi were dissected from postnatal day 0–1 (P0–P1) mouse pups in Hank’s Balanced Salt Solution (HBSS; Gibco/Invitrogen). Primary cortical astrocytes were prepared from wild-type mice, dissociated using trypsin (Gibco/Invitrogen) and mechanical trituration, and cultured for two weeks in Dulbecco’s Modified Eagle Medium (DMEM) with GlutaMAX (Gibco/Invitrogen), supplemented with 10% fetal bovine serum (FBS; Gibco/Invitrogen) and 1% penicillin-streptomycin (Sigma), at 37 °C in a humidified atmosphere containing 5% CO₂. Two days before hippocampal dissection, astrocytes were trypsinized and replated onto the microdot-stamped coverslips at a density of ~ 6000 cells/cm^2^. Primary hippocampal neurons were then prepared from pups obtained from Grin2b^+/+^ and Grin2b^+/Δ^ crossings. Hippocampi from each pup (P0–P1; male and female) were processed individually. A sample of brain tissue was collected and stored at −20 °C for subsequent genotyping. Hippocampi were digested using papain (Worthington) at 37 °C for 40 min, followed by mechanical dissociation. Neurons were plated at a density of ~ 300 cells/cm² on top of the astrocyte microislands in Neurobasal A medium supplemented with B27 (Gibco/Invitrogen), GlutaMAX (Gibco/Invitrogen), and penicillin-streptomycin (Sigma). Microisland cultures were maintained at 37 °C in a humidified 5% CO₂ incubator until used for electrophysiological recordings.

Electrophysiology

Whole-cell voltage-clamp recordings were obtained from one-neuron hippocampal microisland cultures (7–21 DIV) at −70 mV using an Axopatch 200B amplifier (Axon Instruments/Molecular Devices). Data were sampled at 10 kHz and filtered at 2 kHz. Series resistance was compensated by 70–80%. Recordings were excluded if series resistance exceeded 20 MΩ or increased by more than 20% during the session. Patch electrodes (4–6 MΩ) pulled from borosilicate glass capillaries were filled with intracellular solution containing (in mM): 125 K^+^ gluconate, 15 KCl, 10 HEPES, 5 EGTA, 0.5 CaCl_2_, 2 ATP Mg salt, 0.3 GTP Na salt, and 10 creatine phosphate, with pH adjusted to 7.2 using KOH. For agonist-evoked whole-cell responses, the extracellular solution contained (in mM): 160 NaCl, 2.5 KCl, 10 HEPES, 10 glucose, 0.2 EDTA, and 0.7 CaCl_2_, with pH adjusted to 7.3 using NaOH. Kainate (100 µM) or NMDA (100 µM) were applied to evoke AMPAR or NMDAR currents, respectively, while tetrodotoxin (TTX, 0.5 µM) was used to block action potentials.

Miniature excitatory postsynaptic currents (mEPSCs) were recorded in extracellular solution containing (in mM): 160 NaCl, 2.5 KCl, 10 glucose, 10 HEPES, 2 CaCl_2_, and 1 MgCl_2_, with pH adjusted to 7.3 using NaOH. Evoked EPSCs (eEPSCs) were elicited by brief depolarization (1 ms at + 10 mV) from the holding potential of −70 mV every 10 s and recorded in Mg²⁺-free extracellular solution supplemented with 10 µM glycine. All extracellular solutions contained bicuculline methochloride (10 µM) to block GABA_A_ receptors. To pharmacologically isolate NMDAR- or AMPAR‑EPSCs, 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide (NBQX, 3 µM) or AP-5 (50 µM) were used, respectively. Ifenprodil (3 µM) was used to inhibit responses mediated by NMDARs containing GluN2B subunits. To assess short-term plasticity, paired-pulse ratio (PPR) was calculated from AMPAR-EPSCs by dividing the peak amplitude of the second eEPSC by that of the first, using an interpulse interval of 50 ms. The amplitudes, the PPR, and the deactivation kinetics of eEPSCs were analyzed from 4 to 5 consecutive individual sweeps and the obtained values were averaged. Data were analyzed with respect to the genotype and the sex of the animals to examine potential sex differences in the studied parameters.

Animal behavioral tests

Behavioral testing of Grin2b^+/Δ^ animals and their Grin2b^+/+^ littermates took place at 9–15 weeks of age, with males and females tested separately on different days. The analyses were conducted in the following sequence: open field test, elevated plus maze, novel object recognition, social preference test, nestlet shredding, prepulse inhibition of the acoustic startle (PPI) test, and contextual and cued fear conditioning. A separate cohort of animals was tested in the IntelliCage system. The details of the open field test, the PPI test, contextual and cued fear conditioning, and the IntelliCage system have been previously described [24].

Open field test

Each mouse was placed in the corner of a cubic (42 × 42 cm) uniformly illuminated (200 lx) open-top box and allowed 10 min of free exploration. Mouse movement was monitored and analyzed using a video tracking system (Viewer software, Biobserven GmbH, Germany).

Prepulse inhibition test

Animals were acclimated to soundproof cabinets (Med Associates Inc., USA) for 10 min before the start of the test. Four prepulse acoustic stimuli (70, 77, 82, 85 dB; 20 ms duration) were used alone and in combination with a startle stimulus (110 dB, 50 ms duration). White noise was used as the stimulus, and all delivery and recordings were made against a constant background noise of 65 dB. Ten trials of startle, prepulse, and combined prepulse with startle were performed in a randomized order, with intertrial intervals randomly varied between 10 and 20 s. All animal responses to the stimuli were detected, quantified, graphically represented, and automatically stored using a sensor platform, amplifier, and computer software.

Cued and contextual fear conditioning test

Mice were placed in a conditioning cage inside a soundproof cabinet (Ugo Basile, Gemonio, Italy) with a stainless-steel rod floor for shock delivery. The test starts with an adaptation period followed by two pairings of a conditioned stimulus (20 s of 4 kHz pure tone at 77 dB) and the unconditioned stimulus (1 s of 0.5 mA constant current to the cage floor). The unconditioned stimulus coincided with the cessation of the conditioned stimulus, with a 2-min interval between pairings. Contextual memory was assessed 24 h after learning in the same environment, with freezing behavior automatically recorded during a 6-min trial using ANY-maze software (Ugo Basile, Gemonio, Italy). For the delayed cued memory test, mice were introduced to a novel context 3 h after the contextual trial. The freezing response to the conditioned stimulus was monitored for 2 min.

IntelliCage tests

IntelliCage system is a fully automated device that monitors the cognitive abilities and emotional states of group-housed laboratory mice. One week before transferring to the IntelliCage, all animals received subcutaneous transponders (ISO-compliant). The IntelliCages were equipped with four operant corners, each containing two door-guided openings for drinking bottles. Each corner was equipped with a radiofrequency identification antenna to detect implanted transponders, allowing identification of animals upon entry. Access to water was contingent upon correct nose pokes at the door of the selected corner. Animals were tested in several phases, with water availability in specific corners regulated to assess spatial learning, spatial reversal learning, learning extinction, and working memory [24].

Elevated plus maze test

The elevated plus maze apparatus consisted of two closed arms and two open arms, with a light intensity of 60 lx at the center of the maze. Animals were placed in the center and allowed to explore the maze for 5 min. The total time spent in the open arms, closed arms, and the center was automatically recorded and analyzed using the Viewer software (Biobserve GmbH, Germany).

Novel object recognition test

The novel object recognition test was employed to evaluate exploratory behavior towards a novel object as a measure of recognition memory. This fully automated test utilized a video tracking system (Viewer, Biobserve GmbH, Bonn, Germany). During a two-day habituation phase the animals were allowed to explore the testing arena for 10 min each day. On the third day, the animals were exposed to two identical objects during a familiarization phase. Following a 3-h retention interval, a probe trial was conducted in which one of the familiar objects was replaced with a novel object. All trials were performed under the illumination of 70 lx at the center of the maze. Objects were placed 7.5 cm from the walls, with their positions alternating across trials to control for side preference. Test subjects were consistently introduced from the wall opposite the object locations to maintain positional balance. The system automatically recorded object visits, exploration time for each object, and the corresponding percentages.

Social preference test

The social preference test was performed using two cylindrical wire cages placed in the open-field arena. Each cage (10.5 cm in diameter, 15 cm in height) was positioned equidistantly from two diagonally opposite corners of the arena, 10 cm from the walls. All procedures were conducted under low-light conditions (70 lx at the center of the open field). The experimental and juvenile target animals were sex-matched. Target mice were pre-habituated to the empty wire cages in the open field for 10 min on the day before testing. On the test day, the target animal was placed in one of the wire cages for 10 min without the experimental mouse present; a Duplo block was placed in the other as a non-social control. The experimental mouse was then introduced into the open field and allowed to explore both cages freely for 10 min. Latency to approach, number of visits, and interaction time with the cage containing the target animal were automatically recorded and analyzed using the Viewer software (Biobserve GmbH, Germany).

Nestlet shredding

Mice were housed individually for one week for acclimation. Following this period, 12 g of nesting material (Animalab) was provided in each cage for five consecutive days. To evaluate repetitive nestlet shredding behavior, the unused nesting material was collected and weighed at the same time each day to determine the amount utilized by the mice.

Statistical analysis

Data are presented in bar graphs showing mean ± standard error of the mean (SEM), or in scatter plots showing the distribution of individual cells overlaid on bar graphs showing mean ± SEM. Unless specified otherwise, the sample size n refers to the number of cells or animals (independent values) per group. Statistical analysis was performed with STATGRAPHICS Centurion 18 (Statgraphics Technologies, Inc., USA). If the original data did not have a symmetric distribution and constant variance, they were power transformed to attain symmetric distribution and homoscedasticity (constant variance). We used absolute values of studentized residuals greater than 3 to identify outliers. Data were further analyzed by Student’s t-test or the Analysis of variance (ANOVA) to compare more than two groups, using the General Linear Model function in STATGRAPHICS. When statistically significant differences were found, post-hoc analyses were conducted using the Dunnett method (for comparisons to the control), the Duncan method (for all pairwise comparisons), or the Least Significant Difference (LSD) method (for selected pairwise comparisons). A p-value ≤ 0.05 was considered statistically significant throughout the study. For experiments where data were collected with respect to cell age in vitro, we divided the DIV into four age groups with approximately the same number of data points in each group, to be able to see the trend with respect to age. Afterwards, we analyzed these groups using the ANOVA (using the General Linear Model function in STATGRAPHICS).

Results

Generation of mice with a protein-truncating variant in the Grin2b gene

Understanding the interplay between genetic variation and possible compensatory mechanisms is critical for developing targeted interventions for GRIN-related disorders [46]. Recently, we have generated a knock-in mouse strain carrying a patient missense variant in the Grin2b gene (Grin2b^+/L825V^) to explore the consequences of this variant for NMDAR function, synaptic transmission, and behavior [24]. A male patient heterozygous for the GluN2B-L825V variant presented with severe ID and ASD [12, 47]. We have also generated a mouse line with a frameshift variant introduced at the same location in Grin2b, leading to a premature termination codon (Grin2b^+/Δ^) (Fig. 1A). In the process of CRISPR-Cas9 genome editing, four thymidines were inserted in the Grin2b gene at the triplet encoding L825. This leads to a leucine to phenylalanine substitution and a frameshift, altering the sequence of the subsequent 15 amino acids and introducing a stop codon (GluN2B-L825Ffs15). The mutation is predicted to produce a truncated GluN2B subunit lacking part of the M4 membrane domain and the entire CTD (GluN2B^Δ^; Fig. 1A-B; see Fig. 1C for Sanger validation of the GluN2B-L825Ffs15 mutation in Grin2b⁺^/Δ^ and Grin2b^Δ/Δ^ mice and confirmation of its absence in Grin2b^+/+^ mice). A female patient heterozygous for a PTV GluN2B-E839* with a premature termination codon at the same position as in the GluN2B-L825Ffs15 variant exhibits mild/moderate ID, fine motor skill deficits, abnormal EEG, sleep problems, and early puberty (Dr. Xavier Altafaj, Fac. de Medicina i Ciències de la Salut, Barcelona, personal communication) [18]. To elucidate possible consequences of a PTV at this location we decided to examine the cellular and neurobehavioral effects of the GluN2B-L825Ffs15 PTV in our mouse model and assess potential compensatory changes at the mRNA and protein levels.Fig. 1Generation of Grin2b**⁺^/Δ**^ mice using CRISPR-Cas9. (A) Schematic representation of the GluN2B subunit structure, showing the extracellular amino-terminal domain (ATD); agonist-binding domain (ABD; comprising S1 and S2 lobes), the transmembrane domain (TMD; consisting of M1, M3, M4, and the re-entrant M2 loop), and the intracellular CTD. The M4 segment contains the target site used to generate both the GluN2B^Δ^ and the GluN2B-L825V variants. (B) Top: Nucleotide sequences of the wild-type and the two Grin2b variants. Note the insertion of four thymidines (blue) resulting in a frameshift and a premature termination codon (TGA; red), corresponding to the GluN2B-L825Ffs15 variant. Bottom: Amino acid sequences of the wild-type and the two GluN2B subunit variants. (C) Chromatograms represent the Sanger sequencing validation of the GluN2B-L825Ffs15 variant in heterozygous (Grin2b⁺^/Δ^) and homozygous (Grin2b^Δ/Δ^) mice, with confirmation of its absence in wild-type (Grin2b^+/+^) controls

Homozygous Grin2b^Δ/Δ^ mice, similar to mice with loss-of-function missense mutations Grin2b^C456Y/C456Y^ or Grin2b^L825V/L825V^ [23, 24] die shortly after birth. Our heterozygous mutant mice, obtained by breeding Grin2b^+/Δ^ mice with wild-type animals, were born at approximately the expected Mendelian ratios (25%: 27%: 20%: 28% for male Grin2b^+/+^, female Grin2b^+/+^, male Grin2b^+/Δ^, and female Grin2b^+/Δ^, respectively) and showed normal survival and growth. No differences in weight were observed when adult (9–15 weeks of age) Grin2b^+/Δ^ and Grin2b^+/+^ mice were compared (Grin2b^+/Δ^ vs. Grin2b^+/+^; male: 24.4 ± 0.59 g, n = 11 vs. 25.4 ± 0.97 g, n = 10 (p = 0.48; Student’s t-test); female: 20.1 ± 0.30 g, n = 10 vs. 20.4 ± 0.39 g, n = 12 (p = 0.64; Student’s t-test).

Expression of Grin mRNA in the hippocampus of Grin2b+/Δ mice

Organisms have developed genetic robustness, defined as the ability to minimize the effects of pathogenic nonsense variants at the phenotypic level [48]. First we proceeded to investigate whether the monoallelic PTV in Grin2b^+/Δ^ mice could introduce compensatory mechanisms resulting in transcriptional downregulation of mutant mRNA and the upregulation of homologous genes [49]. The mRNA expression levels of NMDAR subunits were determined in the hippocampus of 4-month-old male and female Grin2b^+/+^ and Grin2b^+/Δ^ mice by RT-qPCR. In male Grin2b^+/Δ^ mice, Grin1 and Grin2b mRNA levels were significantly lower by 20.1 ± 2.5% (p = 0.017; Student’s t-test) and 5.8 ± 2.3% (p = 0.042; Student’s t-test), respectively, compared to samples from male Grin2b^+/+^ controls, while Grin2a expression remained unchanged (p = 0.964; Student’s t-test) (Fig. 2A). In female Grin2b^+/Δ^ mice, Grin1 mRNA levels tended to be lower by 17.8 ± 6.4% (p = 0.079; Student’s t-test) and Grin2b mRNA levels were significantly reduced by 19.3 ± 4.8% (p = 0.004; Student’s t-test), compared to samples from female Grin2b^+/+^ controls, while Grin2a mRNA expression was unchanged (p = 0.552; Student’s t-test), (Fig. 2B).Fig. 2The effect of the Grin2b^Δ^ **mutation on ** Grin1, Grin2a, and Grin2b mRNA expression. Relative mRNA expression levels of Grin1, Grin2a, and Grin2b in the hippocampus of 4-month-old male (A) and female (B) Grin2b^+/+^ and Grin2b⁺^/Δ^ mice, as determined by RT-qPCR. Data were normalized to the corresponding mean Grin2b^+/+^ value and are presented as mean ± SEM and analyzed using Student’s t-test to compare the expression in Grin2b⁺^/Δ^ versus Grin2b^+/+^ mice; asterisks (*) indicate statistically significant differences. (C) Representative chromatograms of Sanger sequencing of PCR amplicons spanning the region of the frameshift mutation (gray box) in male and female Grin2b⁺^/Δ^ mice and the plasmid DNA mixture of Grin2b⁺:Grin2b^Δ^ at a 1:1 ratio; yellow arrowheads indicate the first two of the five analyzed nucleotides, selected to avoid surrounding identical bases that could bias peak intensity measurements. (D) The relative expression levels of Grin2b mRNA show comparable expression of wild-type and mutant alleles. Data are presented as mean ± SEM and were analyzed using one-way ANOVA (no significant differences were found)

It is important to note that Grin2b mRNA levels in Grin2b^+/Δ^ mice reflect transcripts from both the wild-type and the mutant allele. To determine their relative expression, we performed classical end-point PCR using primers spanning multiple exon–exon boundaries to exclude genomic DNA (agarose gel of RT-PCR 1200 nt amplicons from hippocampal RNA using primers spanning exon 7 to 13—see Table S1 for primer sequences—is shown in Fig. S1), followed by direct sequencing of the resulting amplicons. Chromatogram analysis revealed that the peak amplitude ratio of five different nucleotides downstream of the GluN2B-L825Ffs*15 mutation (selected because they are not surrounded by identical nucleotides) was 40.8 ± 0.8% in males (n = 7) and 44.3 ± 1.0% in females (n = 6), indicating near-equal presence of the mutant and wild-type alleles (Fig. 2C, D). These ratios were not significantly different from control amplicons generated from a 1:1 plasmid DNA mixture of Grin2b^+^ and Grin2b^Δ^ alleles (42.0 ± 0.6%, n = 4; male p = 0.328; female p = 0.121). This ratio between wild type and variant mRNA indicates that the variant mRNA does not undergo extensive degradation. This suggests that nonsense-mediated decay (NMD)—a mechanism that typically eliminates faulty mRNA—is not fully engaged in this case [49].

Protein expression in the hippocampus of Grin2b+/Δ mice

To assess the impact of the Grin2b^Δ^ mutation on the brain proteome, we performed mass spectrometry-based proteomic analysis on hippocampi from 4-month-old male Grin2b^+/+^ (n = 5) and Grin2b⁺^/Δ^ (n = 5) mice (see Methods). After applying filtering criteria, we identified 5,454 proteins for further analysis. The presence of the Grin2b^Δ^ allele in Grin2b⁺^/Δ^ mice significantly altered the expression of 23 proteins (4 down-regulated and 19 up-regulated), each showing more than a twofold change compared to Grin2b^+/+^ controls (Fig. 3A). Among these, three up-regulated proteins were neuron specific: ACTL6B (implicated in neurodevelopmental deficits), BACE1 (involved in amyloid precursor protein processing), and BAX (essential for developmental neuronal apoptosis). The remaining proteins lacked a clear link to brain function, including Sect. 62 (endoplasmic reticulum (ER)-associated protein), VMP1 (involved in zymogen granule degradation in pancreatitis), ARMH3 (Golgi organization regulator), H2BU2 (nucleosome component), MT1 (metal homeostasis and detoxification), and GAS8 (structural component of ciliary movement) (Fig. 3A; Table S2 lists proteins with altered expression levels, and their biological functions). Notably, several mildly dysregulated proteins were associated with tumorigenesis. We also performed protein-set enrichment analyses, but these did not identify any significant Gene Ontology term enrichment across molecular function, biological process, or cellular component categories (Tables S3, S4). In addition, we compared the proteomic changes observed in Grin2b⁺^/Δ^ mice with our previously published dataset from Grin2b⁺^/L825V^ mice [24], focusing on all proteins that showed significant differences (p < 0.05) in both genotypes. Although the datasets were positively correlated (Pearson r = 0.939, p < 0.001; n = 461), none of the proteins exhibited a two-fold increase or decrease in both Grin2b⁺^/Δ^ and Grin2b⁺^/L825V^ mice (see Fig. S2).Fig. 3Protein expression analysis in the hippocampus of Grin2b^+/+^ and Grin2b**⁺^/Δ**^ mice. (A) Volcano plot of proteomics analysis comparing protein levels in hippocampal tissue from 4-month-old male Grin2b^+/+^ and Grin2b⁺^/Δ^ mice. The x-axis shows log₂-fold change (negative: downregulated; positive: upregulated), and the y-axis shows –log₁₀ p-values. The horizontal change of color marks the p = 0.05 significance threshold, and vertical dashed lines indicate a two-fold change. Proteins significantly downregulated in Grin2b⁺^/Δ^ mice are shown in blue; significantly upregulated proteins are shown in yellow; non-significant changes are in gray. Statistical significance was determined using a two-sample test (see Methods). (B) The heatmap displays log₂-fold change of selected synaptic protein expression in hippocampal tissue between Grin2b^⁺/Δ^ and Grin2b^+/+^ mice. (C) Protein abundance levels of GluN1, GluN2A, GluN2B-no-CTD, and GluN2B-CTD in Grin2b ^+/+^ (2b^+/+^), Grin2b^⁺/Δ^ (2b^⁺/Δ^), and Grin2b^⁺/L825V^ (2b^⁺/LV^) mice, normalized to the wild-type Grin2b^+/+^ reference. Data are presented as mean ± SEM; asterisks (*) indicate statistically significant differences (one-way ANOVA followed by the Dunnett method) Since disease-associated GluN2B variants have been shown to impact embryonic and early postnatal neurodevelopment, including neuron differentiation, dendritogenesis, synaptogenesis, and synaptic plasticity [50] we sought to determine whether the presence of the Grin2b^Δ^ allele in Grin2b⁺^/Δ^ mice may affect the expression of specific synaptic proteins in the hippocampal tissue. The results of the analysis showed that the expression levels of proteins involved in vesicle exocytosis and synaptic function were not significantly altered in the hippocampus of Grin2b⁺^/Δ^ mice compared to their Grin2b^+/+^ siblings (p = 0.767; Student’s t-test). The average ratio for the analyzed synaptic proteins in Grin2b⁺^/Δ^ mice relative to Grin2b^+/+^ controls was 1.08 ± 0.016 (see Table S5 for protein names of the selected synaptic proteins, along with their ratios in Grin2b^+/Δ^ vs. Grin2b^+/+^ hippocampal tissue), suggesting that the development of synaptic contacts in the hippocampus remains largely unchanged in Grin2b^+/Δ^ mice. A heatmap of the corresponding log₂-fold change values provides a visual overview of these abundance differences across the selected synaptic proteins (Fig. 3B). To assess potential changes in the relative expression of NMDAR subunits, we quantified the protein abundance levels of GluN subunits in Grin2b^+/Δ^ mice, normalizing them to the Grin2b^+/+^ reference level (Fig. 3C). Protein abundance levels of GluN1 and GluN2A were quantified using total peptide signal intensities spanning the entire protein and did not differ significantly between Grin2b^+/Δ^ and Grin2b^+/+^ mice. For GluN2B, two separate protein abundance measurements were obtained: (i.) GluN2B-no-CTD, determined by the total intensity of unique GluN2B peptides from the N-terminal domain to Asp768, representing the combined abundance of both wild-type and GluN2B^Δ^ subunits and (ii.) GluN2B-CTD, determined by the total intensity of unique CTD peptides, reflecting only the wild-type GluN2B subunit. The protein abundance level of GluN2B-no-CTD was significantly reduced in Grin2b^+/Δ^ mice (0.64 ± 0.04-fold; n = 5) compared to Grin2b^+/+^ (Fig. 3C). Similarly, the GluN2B-CTD abundance was significantly reduced in Grin2b^+/Δ^ mice (0.63 ± 0.03-fold; n = 5) compared to Grin2b^+/+^. The small difference (1.0%) between GluN2B-CTD and GluN2B-no-CTD abundance suggests that the truncated GluN2B^Δ^ subunit protein is present at negligible levels in the proteome (Fig. 3C). To determine whether the observed changes in the abundance of the GluN2B subunit were specific to the frameshift variant, we performed proteomic analysis on hippocampi from heterozygous mice carrying the missense GluN2B-L825V mutation at a similar site (Grin2b^+/L825V^) (see [24], for electrophysiological, molecular, and behavioral characterization). The protein abundance levels of GluN1, GluN2A, GluN2B-no-CTD, and GluN2B-CTD in Grin2b^+/L825V^ mice did not significantly differ from the Grin2b^+/+^ reference levels (Fig. 3C), indicating that, unlike the frameshift Grin2b^Δ^ allele, the missense Grin2b^L825V^ allele likely produces significant quantities of the protein product.

Possible consequences of the Grin2bΔallele on the NMDAR subunit composition

Next, we used the “proteomic ruler” method [28], which relies on the abundance levels of histone proteins (see Methods) to estimate the CN per cell for each GluN subunit and then we normalized the obtained values to the CN of GluN1. In Grin2b^+/+^ mice, the mean relative CN of GluN2A and GluN2B subunits was 0.79 and 1.02, respectively, of the CN of the GluN1 subunit (Fig. 4A). Given the tetrameric structure of NMDARs—containing two GluN1 and two GluN2 subunits—it may be the abundance of GluN1, rather than the abundance of GluN2A/B subunits, that primarily determines the total number of NMDARs in the 4-month-old mouse hippocampus. The relative CN of NMDAR subunits allows for an estimation of the distribution of diheteromeric and triheteromeric NMDARs, given several assumptions: First, NMDARs consist of four subunits: two GluN1 and either two GluN2A, two GluN2B (diheteromeric receptors), or one of each (triheteromeric receptors) [51–54]. Second, GluN1 subunits randomly associate with GluN2 subunits with equal preference. Third, the majority of the available subunits assemble into functional NMDARs [52, 55]. These assumptions, together with the relative CN per cell of GluN1, GluN2A, and GluN2B, were used to calculate the expected proportions of receptors composed of different subunit combinations (see Methods). In Grin2b^+/+^ mice, assuming all GluN1 subunits are incorporated into NMDARs, together 44.8% of GluN2A and GluN2B subunits remain unassembled. The estimated relative distribution of GluN1/GluN2A, GluN1/GluN2B, and GluN1/GluN2A/GluN2B receptors in Grin2b^+/+^ mice is 19.2%, 31.5%, and 49.2%, respectively (Fig. 4B).Fig. 4Relative GluN subunit copy number per cell and estimated NMDAR subunit composition in Grin2b mutant mice. (A) Relative CN calculated as the CN of GluN1, GluN2A, and GluN2B subunits in Grin2b^+/+^, Grin2b⁺^/Δ^, and Grin2b⁺^/L825V^ mice, normalized to the CN of GluN1 subunit in Grin2b^+/+^ mice. Arrows indicate the reference GluN subunit levels in Grin2b^+/+^ mice. Data are presented as mean ± SEM; asterisks (*) indicate significant differences in CN between genotypes (one-way ANOVA followed by the Dunnett method). (B) Estimated relative abundance of GluN1/GluN2B diheteromers (1-1-2B-2B), GluN1/GluN2A/GluN2B triheteromers (1-1-2A-2B), and GluN1/GluN2A diheteromers (1-1-2A-2A) in Grin2b^+/+^, Grin2b⁺^/Δ^, and Grin2b⁺^/L825V^ mice, based on the CN data in (A) and calculation methods described the Methods. Dotted lines indicate total NMDAR content

In the hippocampus of Grin2b^+/Δ^ mice, the mean relative CN of GluN1, GluN2A, and GluN2B subunits was 0.91, 0.95, and 0.63, respectively, of the relative CN of the GluN1 subunit in Grin2b^+/+^ mice (Fig. 4A). Based on protein abundance estimates, the GluN2B signal predominantly represents the wild-type GluN2B subunit (~ 99%). In Grin2b^+/Δ^ mice, again assuming all the available GluN1 subunits are incorporated into NMDARs, 42.2% of GluN2A and GluN2B subunits remain unassembled. Calculations indicate an 8.6% reduction in the total number of NMDARs in Grin2b^+/Δ^ mice compared to Grin2b^+/+^ controls (Fig. 4B, dotted lines). Notably, this reduction is likely driven by a small decline in GluN1 levels rather than a decrease in GluN2B. On the other hand, the subunit composition of NMDARs was affected by the relative increase in GluN2A and particularly the decrease in GluN2B levels in Grin2b^+/Δ^ mice compared to Grin2b^+/+^ controls. In Grin2b^+/Δ^ mice, the estimated relative distribution of diheteromeric GluN1/GluN2A, GluN1/GluN2B, and triheteromeric GluN1/GluN2A/GluN2B receptors was 32.9%, 14.6%, and 43.9%, respectively (Fig. 4B).

Next, we analyzed the CN of NMDAR subunits in Grin2b^+/L825V^ mice [24]. In these mice, the relative CN of GluN1, GluN2A, and GluN2B were 1.02, 0.85, and 0.99, respectively, of the CN of the GluN1 subunit in Grin2b^+/+^ mice (Fig. 4A). Since proteomic analysis cannot distinguish between wild-type GluN2B and the GluN2B-L825V variant, the value reflects the combined abundance of both wild-type and variant subunits. Minor CN variations in Grin2b^+/L825V^ resulted in modest changes in the total number of NMDARs and the relative distribution diheteromeric of GluN1/GluN2A, GluN1/GluN2B, and triheteromeric GluN1/GluN2A/GluN2B receptors compared to Grin2b^+/+^ mice (Fig. 4B). Notably, while all NMDARs in Grin2b^+/Δ^ mice should exhibit normal functional properties, a subset of NMDARs in Grin2b^+/L825V^ mice is expected to have reduced function due to a reduced probability of opening of receptors containing the GluN2B-L825V subunit [24].

Together, these findings highlight the possible role of the GluN1 subunit availability in determining the total number of NMDARs in hippocampal neurons. Our results suggest that the reduced availability of the GluN2B subunit in Grin2b^+/Δ^ mice may result in an altered distribution of GluN2A-containing vs. GluN2B-containing NMDARs while not substantially affecting the overall NMDAR content.

Surface expression of the GluN2BΔ subunit

While our RT-qPCR results show no evidence of Grin2b^Δ^ mRNA degradation (Fig. 2D), our proteomic analysis strongly suggests that the truncated GluN2B^Δ^ subunit is virtually absent in the hippocampi of adult Grin2b⁺^/Δ^ mice (Fig. 3C). To better understand possible cellular fates of the GluN2B^Δ^ protein, we sought to determine whether NMDARs containing the GluN2B^Δ^ subunit would be able to assemble, reach the cell surface, and retain any functional activity. In addition, we aimed to elucidate whether GluN2B^Δ^ subunit could affect surface expression of wild type GluN2B subunits. NMDARs containing the GluN2B subunit truncated before or within the M4 domain, or with mutations in the four amino acids following the M4 segment (HLFY(840–843)), exhibit defects in surface expression [22, 56, 57]. We performed patch-clamp recordings from GFP-labeled HEK293T cells transfected with GluN1 and GluN2B^Δ^ and found no detectable responses to 1 mM glutamate (0/10; response > 10 pA was considered the minimum threshold), whereas all GFP-labeled HEK293T cells transfected with GluN1 and GluN2B exhibited responses to 1 mM glutamate (10/10; mean peak response − 596 ± 134 pA). There are several possible explanations for the absence of responses in GluN2B^Δ^-transfected cells: GluN2B^Δ^ subunits may not be expressed, or GluN2B^Δ^ subunits may fail to assemble with GluN1, or NMDARs containing GluN2B^Δ^ subunits may assemble but are not trafficked to the cell surface, of the GluN2B^Δ^ subunit-containing NMDA receptors are non-functional.

To determine whether NMDARs containing the GluN2B^Δ^ subunit are expressed and trafficked to the cell surface, we performed SDS-PAGE followed by western blot analysis on HEK293T cells co-transfected with GluN1 and either wild-type GluN2B-GFP or GluN2B^Δ^-GFP. Figure 5 shows a western blot probed with anti-GFP to detect GluN2B subunits, and anti-GAPDH or anti-histone H3 antibodies to confirm the appropriate enrichment of cytosolic (S1), cytoplasmic membrane (S2) and nuclear/organelle (P) crude fractions (see Methods). In the cytoplasmic membrane fraction (S2), a band at ~ 194 kDa corresponding to wild-type GluN2B fused to GFP (167 + 27 kDa) was observed in HEK293T cells expressing GluN2B-GFP, although this band was also prominent in the nuclear/organelle (P) fraction. Similarly, a band at ~ 122 kDa corresponding to GluN2B^Δ^-GFP (95 + 27 kDa) was detected in HEK293T cells expressing the truncated subunit, suggesting that in the heterologous expression system significant amounts of the truncated subunit protein are produced. Since NMDARs are trafficked to the membrane only as assembled complexes [58], the presence of GluN2B^Δ^-GFP in the cytoplasmic membrane (S2) fraction supports the existence of NMDARs containing the truncated subunit and their trafficking to the cell surface in HEK293T cells.Fig. 5Expression of GluN2B (WT) and GluN2B^Δ^ in HEK293T cells and the presence of GluN2B subunits in the cytoplasmic membrane. HEK293T cell cultures were co-transfected with GluN1 and wild-type or mutant GluN2B subunits tagged with GFP and harvested 48 h after transfection. From each culture, a sample of the total cell lysate was taken (T). Remaining cells were fractionated to obtain enriched cytosolic (S1), cytoplasmic membrane (S2) and nuclear/organelle (P) fraction. Samples of all fractions from both cultures were analyzed by western blot. An anti-GFP antibody was used to detect GFP-tagged GluN2B subunits, and anti-GAPDH and anti-histone H3 antibodies were used to verify the enriched cytosolic and nuclear/organelle cell fractions, respectively. The image shows the presence of both GluN2B (WT) and GluN2B^Δ^ in the S2 cytoplasmic membrane fraction suggesting the presence of GluN1/GluN2B tetramer in transfected cells. Expected molecular weights are presented on the right side of blot; M denotes the lane containing the protein molecular weight marker (see also Fig. S3)

To evaluate the impact of the GluN2B^Δ^ variant on NMDAR surface expression more quantitatively, we performed immunofluorescence microscopy in HEK293T cells transfected with equal amounts of expression vectors encoding GluN1 and either wild-type or mutant GFP-tagged GluN2B subunits. Compared to NMDARs containing GluN2B, receptors incorporating the GluN2B^Δ^ variant exhibited a 78% reduction in surface expression (p < 0.001, Student’s t-test) (Fig. 6A). This reduction was further confirmed by a colorimetric assay in COS-7 cells, which revealed a 90% decrease in surface expression (p < 0.001, Student’s t-test) (Fig. S4).Fig. 6The GluN2B^Δ^ variant disrupts receptor cell surface expression. (A) Left: Representative immunofluorescence images showing surface and intracellular expression of GluN2B-GFP subunits in HEK293T cells co-transfected with GluN1 and either wild-type (2B-GFP) or mutant (2B^Δ^-GFP) constructs. DAPI staining was used as a cell viability marker. Right: Quantification of the normalized surface/intracellular fluorescence ratio in HEK293T cells (n ≥ 100 cells per group, four independent experiments). (B) Left: Representative images of surface and intracellular expression of GluN2B-GFP or GluN2B^Δ^-GFP in HEK293T cells also co-transfected with GluN1 and GluN2A (2A). Right: Quantification of the normalized surface/intracellular fluorescence ratio (n ≥ 180 cells per group, four independent experiments). (C) Left: Representative fluorescence images showing surface and intracellular localization of GluN2A-GFP (2A-GFP) in HEK293T cells also co-transfected with GluN1 and either GluN2B (2B) or GluN2B^Δ^ (2B^Δ^). Right: Quantification of the normalized surface/intracellular fluorescence ratio for GluN2A-GFP (n ≥ 100 cells per group, four independent experiments). (D) Left: Representative images of surface and intracellular immunostaining of GluN2B-GFP and GluN2B^Δ^-GFP in hippocampal neurons from Grin2b^+/+^ mice. Yellow outlines indicate neuronal somata used for quantification. Right: Normalized surface/intracellular fluorescence ratios for GluN2B-GFP and GluN2B^Δ^-GFP in the somata of hippocampal neurons (n ≥ 20 cells per group, four independent experiments). Data are shown as mean ± SEM. Asterisks (*) indicate statistically significant differences (p ≤ 0.05, Student’s t-test)

Given that NMDARs can form triheteromeric complexes [51–54], we next assessed whether co-expression of GluN2A could enhance the surface localization of GluN2B^Δ^. Co-transfection of HEK293T cells with GluN1, GluN2B^Δ^-GFP, and GluN2A resulted in a twofold increase in GluN2B^Δ^-GFP surface expression compared to cells expressing GluN1 and GluN2B^Δ^-GFP alone (p < 0.001, Student’s t-test) (Fig. 6A, B). However, co-expression of GluN2B^Δ^ negatively affected the surface expression of GluN2A-GFP (Fig. 6C), suggesting impaired assembly and trafficking of NMDARs containing the GluN2B^Δ^ compared to the wild-type GluN2B subunit.

Since protein overexpression in HEK293T cells can engage trafficking and quality-control pathways that differ from those operating in neurons, we also examined cell surface expression of the GluN2B^Δ^-GFP subunit introduced by lentiviral vector transduction in wild-type primary hippocampal neurons. In this system the GluN2B^Δ^-GFP subunit also showed a markedly lower surface expression on the soma compared to the wild-type GluN2B-GFP (86% reduction, p < 0.001, Student’s t-test) (Fig. 6D). Although our analyses of receptor cell surface expression in HEK293T cells and wild-type neurons were performed under nonphysiological overexpression conditions driven by the strong CMV promoter and using constructs lacking introns that cannot be processed by NMD, and despite possible tissue- or cell-type specificity of protein degradation mechanisms, these assays consistently demonstrated a pronounced trafficking defect of the GluN2B^Δ^ subunit. This impairment is fully compatible with our in vivo finding that the truncated GluN2B^Δ^ protein is not detectable under endogenous expression conditions in Grin2b⁺^/Δ^ mice.

MD analysis of GluN1/GluN2B receptors with truncated GluN2B subunits

Despite the truncation of the M4 helix and the introduction of polar residues within the membrane bilayer, MD simulations of GluN1/GluN2B complexes carrying one or two truncated GluN2B subunits suggest that the receptors remain stable and embedded in the model membrane. During the 500 ns simulation—a timescale typically sufficient to capture TMD rearrangements—no significant structural changes associated with receptor closure were observed. Most interactions between the remaining transmembrane helices and the intact portion of M4 are preserved. A notable exception is the loss of van der Waals contact between the GluN2B M824 residue and the GluN1 M3 helix, which is accompanied by a rearrangement of the truncated GluN2B M4 helix toward the GluN1 M1 helix. This shift further induces rearrangements in GluN2B pre-M1 interactions with the upper part of the truncated M4 helix; notably, this region retains its original amino acid sequence. The weakened M824-GluN1 M3 interaction observed in our truncated constructs may disrupt coordination between the M1, pre-M1, and M3 helices. This could impair mechanical coupling between ligand binding and channel gating, ultimately reducing channel open probability.

In addition to changes in the TMD, the ATD and ABD exhibited substantial separation and loss of inter-domain contacts, potentially resembling desensitized states observed in related ionotropic glutamate receptors [59]. It should be noted, however, that the complete functional cycle of receptor opening and closing involves numerous residues forming transient, state-dependent contacts that are further mediated or modulated by endogenous factors; thus, the effects of M4 truncation on receptor structure and function may be more complex than captured here. Figure 7 presents structural snapshots at the beginning and at the end of the 500 ns simulation for the GluN1/GluN2B receptor with both GluN2B subunits truncated; further details are provided in the supplementary PyMOL session available on Zenodo under the 10.5281/zenodo.17911364 DOI.Fig. 7Structural model of the GluN1/GluN2B receptor with both GluN2B subunits truncated. The left side shows the initial geometry and the right side displays a snapshot after 500 ns of MD simulation; grey, GluN1; orange, GluN2B. (A) The ATD and ABD regions exhibited substantial separation and loss of inter-domain contacts, resembling desensitized-like states. (B) Enlarged TMD snapshots show the rearrangement of the truncated GluN2B M4 helix toward the GluN1 M1 helix. The GluN2B M824 residue (blue sticks) loses van der Waals contact with the GluN1 M3 helix over the course of the simulation

Diminished NMDA-induced whole-cell responses in Grin2b+/Δ hippocampal neurons

To evaluate the functional impact of the GluN2B^Δ^ subunit in its native context, primary hippocampal microisland cultures were prepared from newborn Grin2b^+/+^ and Grin2b^+/Δ^ mice and maintained for 7–21 DIV. It is widely accepted that cultured neurons undergo developmental processes similar to those observed in vivo, encompassing both morphological and functional aspects [60]. Developmental increase in dendritic arborization is evident in hippocampal neurons immunostained for β3-tubulin at 7 and 21 DIV (Fig. 8A). Using whole-cell voltage-clamp recordings, we examined the amplitude of responses induced by 100 µM kainic acid or 100 µM NMDA (with 10 µM glycine) at a holding potential of −70 mV. Recordings were restricted to excitatory neurons, identified by the occurrence of autaptically evoked AMPAR-eEPSCs. To account for variations in neuronal size, agonist-induced peak currents—reflecting synaptic and extrasynaptic receptor activation—were normalized to cell capacitance. The AMPA/kainate receptor current density was not consistently affected by genotype; in different DIV age groups, mean current density in Grin2b^+/Δ^ neurons ranged from − 44% to + 12% compared to Grin2b^+/+^ neurons (ANOVA: p = 0.011 for genotype, p = 0.015 for age, p = 0.02 for genotype × age interaction; n = 57 and n = 40 for Grin2b^+/Δ^ and Grin2b^+/+^, respectively) (Fig. 8B). In contrast, the analysis of NMDAR current densities showed a consistent significant reduction in Grin2b^+/Δ^ neurons, ranging from − 31% to −59% compared to Grin2b^+/+^ neurons in individual age groups. The ANOVA revealed significant effects of genotype (p < 0.001) and age (p = 0.002), while the genotype × age interaction was not significant (p = 0.253); n = 59 for Grin2b^+/Δ^ and n = 54 for Grin2b^+/+^ neurons (Fig. 8C).Fig. 8NMDAR whole-cell currents are reduced in Grin2b^+/Δ^ neurons. (A) Primary hippocampal microisland cultures stained for β3-tubulin. Representative images show neuronal morphology at 7 DIV (left) and 21 DIV (right). Insets display the same microislands at a lower magnification, with neurons labeled by β3-tubulin (red), astrocytes by glial fibrillary acidic protein (GFAP) (green), and nuclei by DAPI (blue). Scatter plots show the current density distribution in individual hippocampal neurons by DIV. Currents were induced by 100 µM kainate (B) or 100 µM NMDA in the presence of 10 µM glycine (C). Insets show representative whole-cell responses evoked by kainate or NMDA in neurons prepared from Grin2b^+/+^ (black) or Grin2b^+/Δ^ (red) animals and cultured for 21 DIV. (D) The scatter plot shows the distribution of ifenprodil (3 µM) inhibition of NMDA-evoked currents recorded in individual neurons by DIV. Inset shows the effect of 3 µM ifenprodil on responses to 100 µM NMDA in neurons prepared from Grin2b^+/+^ (black) or Grin2b^+/Δ^ (red) animals cultured for 21 DIV. Blue lines indicate the expected sum of voltage-dependent and voltage-independent inhibition of diheteromeric GluN1/GluN2B (1-1-2B-2B; 87.2%), and GluN1/GluN2A (1-1-2A-2A; 14.8%), and triheteromeric GluN1/GluN2A/GluN2B (1-1-2A-2B; 41.4%) receptors by 3 µM ifenprodil at −80 mV (with IC_50_ values and Hill slopes from [7]). Data were power transformed and tested using ANOVA, followed by pairwise comparisons (LSD method); asterisks (*) indicate significant differences with respect to: Gen, genotype; Age, age (DIV); Int, genotype × age interaction

Subsequently, we used ifenprodil, a potent inhibitor of NMDARs containing the GluN2B subunit [61], to investigate the source of the reduced NMDAR current densities observed in Grin2b⁺^/Δ^ neurons (Fig. 8D). In Grin2b⁺^/Δ^ neurons, mean NMDAR current responses were inhibited by ifenprodil (3 µM) by an average of −31% to −47% compared to Grin2b^+/+^ neurons across different age groups (ANOVA: p < 0.001 for genotype, p < 0.001 for age, p = 0.792 for genotype × age interaction; n = 58 and n = 50 cells for Grin2b⁺^/Δ^ and Grin2b^+/+^, respectively) (Fig. 8D). These findings indicate that Grin2b⁺^/Δ^ hippocampal neurons exhibit smaller NMDA-induced current densities and a reduced contribution of GluN2B-containing NMDARs. Additionally, the degree of ifenprodil block of NMDA-induced currents declines with age in both genotypes, consistent with the developmental switch from GluN2B to GluN2A subunits [62]. We also analyzed sex differences in AMPA/kainate and NMDAR current densities, as well as the degree of NMDAR current inhibition by ifenprodil. No significant differences were observed between female and male neurons (data are summarized in Table S6).

Accelerated deactivation of NMDAR eEPSC in Grin2b+/Δ mice

To evaluate the impact of the Grin2b^Δ^ variant on excitatory synaptic transmission, we recorded AMPAR- and NMDAR-mediated components of eEPSCs, pharmacologically isolated using AP-5 or NBQX (see Methods), in hippocampal microisland cultures. We first reliably detected eEPSCs at 6–7 DIV. AMPAR-eEPSC peak current densities increased with age in both Grin2b⁺^/Δ^ and Grin2b^+/+^ neurons, with no significant genotype effect (ANOVA: genotype p = 0.774, age p < 0.001, genotype × age interaction p = 0.589; n = 54 and n = 58 for Grin2b⁺^/Δ^ and Grin2b^+/+^, respectively; Fig. 9A). NMDAR-eEPSC peak amplitudes also increased with age in both genotypes, with no statistically significant overall difference between genotypes (ANOVA: genotype p = 0.051, age p < 0.001, genotype × age interaction p = 0.062; n = 50 and n = 55 for Grin2b⁺^/Δ^ and Grin2b^+/+^, respectively; Fig. 9B). Notably, the mean deactivation time course of NMDAR-eEPSCs was significantly faster in Grin2b⁺^/Δ^ neurons—by 22% to 41%—compared to Grin2b^+/+^ neurons across age groups, consistent with a reduced contribution of GluN2B-containing NMDARs. Deactivation kinetics accelerated with age in both genotypes, as expected during synaptic maturation (ANOVA: genotype p < 0.001, age p = 0.015, genotype × age interaction p = 0.438; n = 50 and n = 57 for Grin2b⁺^/Δ^ and Grin2b^+/+^, respectively; Fig. 9C). Results summarized in Table S6 revealed no significant differences in eEPSC parameters between sexes.Fig. 9NMDAR-eEPSCs in Grin2b^+/Δ^ neurons have faster deactivation. (A–B) Scatter plots show the distribution of peak current densities for AMPAR-eEPSCs (A) and NMDAR-eEPSCs (B) recorded from individual Grin2b^+/+^ (gray) and Grin2b⁺^/Δ^ (red) neurons, plotted as a function of DIV. Insets: representative traces from Grin2b^+/+^ (black) and Grin2b⁺^/Δ^ (red) neurons at 8 DIV. (C) Weighted deactivation time constants of NMDAR-eEPSCs plotted by DIV. Inset: scaled NMDAR-eEPSCs from Grin2b^+/+^ (black) and Grin2b⁺^/Δ^ (red) neurons at 8 DIV. (D) Ifenprodil (3 µM) inhibition of NMDAR-eEPSC peak amplitudes plotted by DIV. (E) Weighted deactivation time constants of NMDAR-eEPSCs recorded in the presence of ifenprodil (3 µM) and plotted by DIV. Inset: example traces from Grin2b^+/+^ (black) and Grin2b⁺^/Δ^ (red) neurons recorded at 15 DIV, showing the effect of ifenprodil (gray). The green lines in C, D, and E indicate mean weighted deactivation time constants of NMDAR-eEPSCs recorded from mouse hippocampal pyramidal neurons at −60 mV (C), mean ifenprodil (3 µM) inhibition of NMDAR-eEPSCs at + 40 mV (D), and mean weighted deactivation time constants of NMDAR-eEPSCs recorded in the presence of ifenprodil (3 µM) after conditional knockout of GluN2A (Δ2A) or GluN2B (Δ2B) subunits [65]. Data were power transformed and analyzed by ANOVA followed by pairwise comparisons (LSD method); asterisks (*) indicate significant differences for: Gen, genotype; Age, age (DIV); Int, genotype × age interaction

To further probe the contribution of GluN2B-containing receptors, we applied ifenprodil (3 µM), a selective GluN2B antagonist. Despite the observed genotype difference in NMDAR-eEPSC deactivation kinetics (Fig. 9C), ifenprodil inhibited NMDAR-eEPSC peak amplitudes to a similar degree in both genotypes, with inhibition decreasing with age (ANOVA: genotype p = 0.755, age p = 0.001, genotype × age interaction p = 0.843; n = 40 and n = 44 for Grin2b⁺^/Δ^ and Grin2b^+/+^, respectively; Fig. 9D). The age-dependent effect is consistent with previous reports showing developmental decrease in ifenprodil inhibition of EPSCs likely reflecting increasing synaptic incorporation of GluN2A subunits and a corresponding reduction in GluN2B subunits [63–65]. The decay kinetics of NMDAR-eEPSC recorded in the presence of ifenprodil were on average accelerated by 30% in Grin2b⁺^/Δ^ neurons and by 33% in Grin2b^+/+^ neurons when compared to corresponding controls recorded without ifenprodil. The mean deactivation time course of NMDAR-eEPSCs recorded in the presence of ifenprodil was significantly faster in Grin2b⁺^/Δ^ neurons than in Grin2b^+/+^ neurons across all age groups (ANOVA: genotype p < 0.001, age p = 0.459, genotype × age interaction p = 0.965; n = 33 and n = 46 for Grin2b⁺^/Δ^ and Grin2b^+/+^, respectively; Fig. 9E). Our eEPSC data suggest that the overall synaptic NMDAR content may be largely preserved in Grin2b⁺^/Δ^ neurons (Fig. 9B), despite the observed reduction in whole-cell NMDAR current density (Fig. 8C). However, NMDAR subunit composition is likely altered in both the synaptic and the extrasynaptic compartment, with a decrease in GluN2B contribution to NMDAR signaling in Grin2b⁺^/Δ^ compared to Grin2b⁺^/^⁺ neurons.

Increased amplitude of AMPAR-mEPSC in Grin2b+/Δ mice

NMDARs regulate glutamatergic transmission by limiting AMPAR incorporation at developing synapses [66]. Grin2b deletion increases synaptic AMPAR localization and AMPAR-mEPSC amplitudes [67, 68] or increases the number of functional synapses as suggested by increased AMPAR-mEPSC frequency [65].

To investigate possible alterations of AMPAR-mediated synaptic transmission in a disease-relevant model, we recorded action potential-independent AMPAR-mEPSCs from Grin2b⁺^/Δ^ and Grin2b^+/+^ single-neuron hippocampal microisland cultures at different DIV [24]. The amplitude of AMPAR-mEPSCs in Grin2b⁺^/Δ^ neurons was 32% higher than in age-matched Grin2b^+/+^ neurons from littermates (ANOVA: p < 0.001 for genotype, p = 0.841 for age, p = 0.243 for genotype × age interaction; n = 52 and n = 43 for Grin2b⁺^/Δ^ and Grin2b^+/+^, respectively) (Fig. 10A, B). The frequency of AMPAR-mEPSCs in Grin2b⁺^/Δ^ neurons was not consistently different from Grin2b^+/+^ neurons (ANOVA: p = 0.161 for genotype, p < 0.001 for age, p = 0.024 for genotype × age interaction; n = 52 and n = 43 for Grin2b⁺^/Δ^ and Grin2b^+/+^, respectively; see Fig. 10C).Fig. 10Properties of AMPAR-mEPSCs and glutamate release in Grin2b^+/Δ^ and Grin2b^+/+^ neurons. (A) Representative recordings of AMPAR-mEPSCs in neurons from Grin2b^+/+^ and Grin2b⁺^/Δ^ animals (DIV = 8). (B) Scatter plot showing the distribution of mean AMPAR-mEPSC amplitudes recorded from individual Grin2b^+/+^ and Grin2b⁺^/Δ^ neurons (each point represents the mean AMPAR-mEPSC amplitude recorded over a 2-min interval). (C) Scatter plot showing the distribution of AMPAR-mEPSC frequencies recorded from individual Grin2b^+/+^ and Grin2b⁺^/Δ^ neurons. (D) Scatter plot showing the distribution of the AMPAR PPR in individual Grin2b^+/+^ and Grin2b⁺^/Δ^ neurons. Data were power transformed and analyzed using ANOVA followed by pairwise comparisons (LSD method); asterisks (*) indicate significant differences for: Gen, genotype; Age, age (DIV); Int, genotype × age interaction

Furthermore, the PPR of AMPAR-eEPSCs was similar between Grin2b⁺^/Δ^ and Grin2b^+/+^ neurons (ANOVA: p = 0.679 for genotype, p = 0.016 for age, p = 0.560 for genotype × age interaction) (Fig. 10D) [69]. Together these findings indicate that that a monoallelic PTV in Grin2b, and the consequent reduction in GluN2B-mediated NMDAR signaling, promotes an increase in the density or function of synaptic AMPARs without affecting the number of functional AMPAR-containing glutamatergic synapses in excitatory hippocampal neurons or the properties of presynaptic glutamate release.

Behavioral impact of GluN2BΔ in Grin2b+/Δ mice

Reduced NMDAR function in mice is associated with a range of behavioral abnormalities, including motor symptoms, altered anxiety levels, impaired sensorimotor gating, changes in social behavior, and cognitive deficits [70]. Mice heterozygous for the conventional Grin2b deletion [8] show a relatively mild behavioral phenotype characterized by impaired prepulse inhibition (PPI) of the acoustic startle [71], hypoactivity in the open field, and reduced anxiety [23]. To assess the behavioral effects of a monoallelic Grin2b PTV, we conducted a series of behavioral tests comparing Grin2b⁺^/Δ^ mice with their Grin2b^+/+^ littermates at 9-15 weeks of age.

Open field and PPI tests in Grin2b⁺/Δ mice

In the open field test, which evaluates general activity and anxiety in a novel environment, adult male Grin2b⁺^/Δ^ mice exhibited reduced activity, covering 22% less total distance and spending significantly more time resting compared to Grin2b^+/+^ males. In contrast, no activity differences were observed between female Grin2b⁺^/Δ^ and Grin2b^+/+^ mice, whose activity levels were similar to Grin2b^+/+^ males (Fig. 11A, Fig. S5, Table S7). The analysis of thigmotaxis also revealed a sex-dependent behavioral phenotype. Male Grin2b⁺^/Δ^ mice exhibited 27% fewer entries into the center zone and covered 36% less distance in the center zone compared to Grin2b^+/+^ males, indicating increased anxiety. No significant differences were observed between female Grin2b⁺^/Δ^ and Grin2b^+/+^ mice (Fig. 11A, Fig. S5, Table S7). In contrast, no genotype-dependent differences in anxiety were observed in the elevated plus maze test, as both male and female Grin2b⁺^/Δ^ mice spent similar time in the closed and open arms as their Grin2b^+/+^ littermates (Fig. S6, Table S7).Fig. 11Male Grin2b**⁺^/Δ**^ mice exhibit reduced activity and increased anxiety,** while both sexes show impaired sensorimotor gating.** (A) Quantification of total distance traveled, whole-arena resting time, number of center entries, and center distance traveled in Grin2b⁺^/Δ^ and Grin2b^+/+^ mice (9–15 weeks old) assessed in the open field test. Data are presented as mean ± SEM. Sample sizes (Grin2b^+/+^/Grin2b⁺^/Δ^): open field test, n = 10/11 males, n = 12/10 females. Asterisks () indicate significant genotype differences, as analyzed by Student’s t-test. (B) Quantification of the percentage of the inhibition of the startle response as a function of prepulse intensity in the PPI test. Sample sizes (Grin2b^+/+^/Grin2b⁺^/Δ^) PPI, n = 11/11 males, n = 12/11 females. Asterisks () indicate significant genotype differences, as analyzed by ANOVA followed by post-hoc tests using the LSD method. Behavioral assays were performed using age-matched animals. The numerical data underlying this figure can be found in Table S7

The PPI test was used to assess sensorimotor gating, a process often impaired in connection with NMDA hypofunction in animal models of schizophrenia and ASD [72], obsessive-compulsive disorder [73] and other neuropsychiatric conditions [70]. PPI measures the suppression of the startle reflex when a weak, non-startling prepulse precedes a strong startle stimulus. The PPI ratio was lower in Grin2b⁺^/Δ^ males compared to Grin2b^+/+^ males across all prepulse intensities. However, significant reductions were observed only at 70 and 82 dB. Similarly, Grin2b⁺^/Δ^ females showed a significantly lower PPI ratio at 70 and 78 dB but an increase at 85 dB (Fig. 11B, Table S7).

Cognitive performance in Grin2b⁺/Δ mice

Homozygous mice with a conditional postnatal deletion of Grin2b in the forebrain exhibit deficits in learning and memory [74]. To determine whether heterozygous Grin2b⁺^/Δ^ mice show an impairment of cognitive function, we evaluated their performance in associative memory and recognition tasks. In the contextual and cued fear conditioning test, Grin2b⁺^/Δ^ mice exhibited normal activity levels during the baseline period and showed equivalent foot shock reactivity compared to Grin2b^+/+^ controls. During the post-training memory test, Grin2b⁺^/Δ^ mice displayed context freezing, with no significant difference from Grin2b^+/+^ mice (Fig. 12A, Table S7). Cognitive performance was further assessed using the novel object recognition test. The preference for the novel object did not differ significantly between Grin2b⁺^/Δ^ males and Grin2b^+/+^ controls; however, it was reduced in Grin2b⁺^/Δ^ compared to Grin2b^+/+^ females (Fig. 12B, Table S7).Fig. 12Normal learning and memory in Grin2b^+/Δ^ mice. (A) Fear conditioning test showing normal baseline, context, and cue freezing responses in male and female Grin2b^+/Δ^ mice. Group sizes Grin2b^+/+^/Grin2b^+/Δ^: n = 10/11 males; n = 12/10 females. (B) Novel object recognition performance was reduced in Grin2b^+/Δ^ females. Group sizes Grin2b^+/+^/Grin2b^+/Δ^: n = 15/8 males; n = 16/12 females. (C–G) IntelliCage-based behavioral testing of female Grin2b^+/Δ^ and Grin2b^+/+^ mice. (C) Number of visits during the 4-day free adaptation phase (all corners open, unrestricted water access). (D–G) Session scores (% visits to the correct corner) for place preference (D), place preference reversal (E), place preference extinction (F), and patrolling behavior (G). Trial design: C = unconditionally closed; W = water available at designated corners during drinking sessions (9–11 PM and 3–5 AM). Data are shown as mean ± SEM and were analyzed by Repeated Measures ANOVA. Group sizes Grin2b^+/+^/Grin2b^+/Δ^: n = 9/8 females. No significant genotype differences were observed in any test. See Table S7 for numerical data

To further assess cognitive performance, we tested Grin2b⁺^/Δ^ mice in IntelliCages, an automated system that evaluates behavior in group-housed animals without human intervention [24]. The IntelliCages use door-guarded water reservoirs located in the four corners of the cage, with individual drinking responses tracked via subcutaneous transponders. With the start of restricted access to water, male Grin2b⁺^/Δ^ mice developed aggressive behavior, necessitating the termination of their participation for animal welfare reasons. Therefore, only females were tested in IntelliCages. Circadian rhythmicity assessed during the free adaptation phase remained intact in Grin2b⁺^/Δ^ females, as their visits to the drinking corners followed the normal day/night cycle, with no genotype differences (p = 0.927, Repeated Measures ANOVA) (Fig. 12C, Table S7). During place preference learning phase, when water availability is restricted to one specific corner, Grin2b⁺^/Δ^ females performed similarly to Grin2b^+/+^ females (p = 0.443, Repeated Measures ANOVA) (Fig. 12D). Likewise, during place preference reversal phase, when water availability is changed to a corner diagonally opposite to the one learned previously, Grin2b⁺^/Δ^ females showed comparable place preference reversal learning to Grin2b^+/+^ females, demonstrating similar cognitive flexibility (p = 0.257, Repeated Measures ANOVA) (Fig. 12E). There was also no difference between Grin2b⁺^/Δ^ and Grin2b^+/+^ females in place preference extinction trials with water availability again unrestricted (p = 0.113, Repeated Measures ANOVA), or in patrolling trials testing working memory, where the rewarded corner changes in the clockwise direction following each correct drinking response (p = 0.314, Repeated Measures ANOVA) (Fig. 12F, G, Table S7). Together these results indicate that cognitive function in Grin2b⁺^/Δ^ mice is largely intact.

Social behavior in Grin2b⁺/Δ mice

To assess social and repetitive behavior in Grin2b⁺^/Δ^ mice, we conducted the social interaction test and the nestlet shredding test [75, 76]. Grin2b⁺^/Δ^ males exhibited normal social behavior, with no significant differences in time spent sniffing the juvenile target compared to Grin2b^+/+^ controls. In contrast, Grin2b⁺^/Δ^ females spent relatively more time exploring the juvenile compared to the object-containing chamber (Fig. S7, Table S7). Nestlet shredding performance did not differ significantly between Grin2b⁺^/Δ^ and Grin2b^+/+^ mice (Fig. S8, Table S7).

While homozygous Grin2b deletion is lethal [8], heterozygous conventional Grin2b deletion is associated with relatively moderate behavioral phenotype characterized by hypoactivity and impaired sensorimotor gating [23, 71]. Similarly, Grin2b⁺^/Δ^ mice exhibit only moderate behavioral changes. These include hypoactivity and increased anxiety in males, and impaired sensorimotor gating in both sexes. Learning and memory, and social behavior remain largely unaffected.

Discussion

In this study, we used in vitro and in vivo methods to analyze mice heterozygous for a frameshift Grin2b mutation, with the goal to uncover molecular, cellular, and behavioral aspects relevant to human GluN2B PTVs. The GluN2B^Δ^ subunit truncated from the M4 transmembrane helix shows severely impaired surface expression in heterologous expression systems and may be largely degraded in Grin2b⁺^/Δ^ brain tissue. Proteomic analysis revealed reduced GluN2B levels predicting a shift toward an increased proportion of diheteromeric GluN1/GluN2A receptors. Hippocampal neurons from Grin2b⁺^/Δ^ mice showed reduced NMDA-induced currents less sensitive to ifenprodil, faster NMDAR-eEPSC deactivation, and increased AMPAR-mEPSC amplitudes. Behaviorally, Grin2b⁺^/Δ^ males exhibited reduced locomotion in the novel environment and increased anxiety-like behavior, while both sexes showed impaired sensorimotor gating. Learning, memory, and social behavior was not affected. Together our results suggest that a monoallelic Grin2b PTV alters NMDAR subunit composition and leads to relatively moderate behavioral effects.

Aberrant Grin2b mRNA is not selectively downregulated in Grin2b⁺/Δ mice

The RT-qPCR analysis of hippocampal tissue from Grin2b⁺^/Δ^ mice revealed only a modest reduction in Grin2b mRNA levels (Fig. 2 A, B). In many prokaryotic and eukaryotic systems, the presence of a premature termination codon typically engages the NMD pathway to degrade aberrant transcripts and prevent the synthesis of truncated proteins [77, 78]. However, in Grin2b⁺^/Δ^ mice, the relative proportions of wild-type and the premature stop codon-containing Grin2b mRNAs were equal (Fig. 2 C, D), challenging the involvement of canonical NMD mechanisms [49]. Generally, NMD does not target transcripts with a premature stop codon located within the penultimate exons if they occur less than 50–55 nucleotides upstream of the final exon-exon junction, or with a premature stop codon located anywhere within the last exon, unless an unusually long 3’ UTR is present [77, 78]. The stop codon in the Grin2b^Δ^ allele lies approximately 85 nucleotides upstream of the final exon-exon junction, thus just out of range for the transcript to escape NMD. This suggests that the aberrant Grin2b^Δ^ transcripts may instead be spared due to suppressed NMD activity in neurons. Neurons exhibit reduced NMD efficiency compared to other cell types, potentially increasing their susceptibility to truncated protein expression from premature stop codon-bearing transcripts [78], in addition to a broader tolerance for some frameshift variants [79]. The parallel modest downregulation of Grin1 and Grin2b (wild-type and PTV; Fig. 2 C, D) mRNA (Fig. 2 A, B) suggests a shared regulatory mechanism affecting mRNA processing.

Consequences of subunit availability for NMDAR subunit composition in Grin2b⁺/Δ mice

Despite only a modest reduction in Grin2b mRNA levels in hippocampal tissue, proteomic analysis revealed that the wild-type GluN2B subunit levels were substantially reduced and the GluN2B^Δ^ subunit was undetectable by mass spectrometry in the hippocampal proteome of male Grin2b⁺^/Δ^ mice. This suggests that GluN2B^Δ^ is efficiently eliminated by protein quality control mechanisms such as the unfolded protein response, ER-associated degradation, or ER-phagy [80]. Interestingly, GluN2A protein levels were slightly elevated in the Grin2b⁺^/Δ^ hippocampus despite unchanged Grin2a mRNA levels, potentially reflecting reduced internalization or degradation of GluN2A-containing NMDARs.

Changes of relative CN of NMDAR subunits in the presence of disease-associated variants may provide insights into disease pathophysiology. Proteomic analysis of Grin2b^+/+^ hippocampal tissue revealed that the relative CN values of GluN1, GluN2A, and GluN2B subunits are approximately equivalent. Given that NMDARs are tetramers composed of two GluN1 and two GluN2 subunits [81], this suggests that the availability of GluN1 may be the limiting factor in determining total NMDAR numbers in hippocampal neurons. In the Grin2b⁺^/Δ^ hippocampus, the relative CN values of GluN1, GluN2A, and GluN2B were altered by − 9%, + 16%, and − 38%, respectively, compared to Grin2b^+/+^ controls. If GluN1 availability limits the total number of NMDARs, the 9% reduction in GluN1 would lead to a small decrease in overall receptor abundance. However, the total surplus of GluN2 subunits above GluN1 would buffer the loss of GluN2B, maintaining high receptor numbers but shifting receptor subunit composition. Our calculations predict a 2.2-fold reduction in GluN1/GluN2B diheteromers, a 1.7-fold increase in GluN1/GluN2A diheteromers, and a modest increase in triheteromeric GluN1/GluN2A/GluN2B receptors. It is important to note that in the Grin2b^+/Δ^ mice, all the assembled NMDARs will retain their normal function. In contrast, the Grin2b^+/L825V^ mice showed no change in the relative CN for any of the NMDAR subunits. In the case of the missense variant, total receptor numbers and receptor subunit compositions remain unchanged, but receptors containing the GluN2B(L825V) variant are hypofunctional due to a reduced channel open probability [24].

Some evidence argues against GluN1 availability being the limiting factor in NMDAR assembly. Notably, GRIN1 haploinsufficiency in humans is not clearly pathogenic [18], and NMDA-evoked currents in hippocampal and cerebellar neurons from GRIN1^+/‒^ mice are not reduced compared with wild-type [82]. In cerebellar granule neurons, overexpression of GluN2—but not GluN1—increases NMDAR current amplitudes [83]. Conversely, overexpression of GluN2 subunits in cultured hippocampal or cortical neurons does not increase maximal NMDA currents or alter the synaptic GluN2B/GluN2A ratio [84–86], although overexpression of GluN2B in the forebrain enhances NMDAR activation and synaptic potentiation in vivo [87]. Synaptic NMDARs also follow distinct regulatory rules: GluN2B-containing receptors are not recruited in greater numbers by GluN2B overexpression, whereas GluN2A-containing receptors replace GluN2B in an activity-dependent manner, a process accelerated by elevated GluN2A expression [88]. Further studies will be needed to clarify how subunit availability shapes NMDAR assembly under physiological vs. overexpression conditions.

Comparing proteomic and functional profiles of NMDARs in Grin2b⁺/Δ mice

The interpretation of the observed relative subunit abundance is constrained by the nature of mass spectrometry-based proteomics, which measures both intracellular and surface-expressed NMDAR subunits. In contrast, electrophysiological recordings reflect the functional surface (extrasynaptic and/or synaptic) receptor populations. Previous studies have shown that NMDAR subunit composition regulates ER export and synaptic targeting [56, 88, 89]. As receptor trafficking is subunit dependent, the intracellular and surface receptor pools may differ in subunit composition.

Nevertheless, our electrophysiological results generally support the reduced relative GluN2B content suggested by the proteomic analysis. Grin2b^+/Δ^ neurons exhibited significantly decreased whole-cell NMDAR current densities with reduced sensitivity to ifenprodil, a selective GluN2B antagonist [61]. Peak NMDAR-eEPSC amplitudes were preserved but their deactivation kinetics were faster in Grin2b^+/Δ^ than in Grin2b^+/+^ neurons, suggesting a reduced GluN2B subunit contribution, although NMDAR-eEPSC sensitivity to ifenprodil was not reduced. Different impact of the Grin2b^Δ^ variant on whole-cell vs. synaptic responses may reflect distinct subunit compositions at synaptic vs. extrasynaptic sites, consistent with previous studies [89, 90].

For further insight into NMDAR subunit compositions in Grin2b^+/+^ and Grin2b^+/Δ^ neurons we compared our data with published results from studies where GluN2A and GluN2B content was experimentally manipulated. For whole-cell NMDA-induced currents (Fig. 8) we considered the data from [7], where recombinant diheteromeric GluN1/GluN2A or GluN1/GluN2B, or triheteromeric GluN1/GluN2A/GluN2B receptors were studied in a heterologous expression system. The degree of ifenprodil (3 µM) inhibition of NMDA-induced currents in Grin2b^+/Δ^ neurons across all DIV age groups is lower than the expected inhibition of triheteromeric GluN1/GluN2A/GluN2B receptors, consistent with the presence of a high proportion of diheteromeric GluN1/GluN2A receptors. In the older Grin2b^+/Δ^ neurons (18–21 DIV) ifenprodil inhibition reaches the low level expected for a pure population of diheteromeric GluN1/GluN2A receptors.

Comparing NMDAR-eEPSCs in Grin2b^+/+^ and Grin2b^+/Δ^ neurons with NMDAR-eEPSCs in neurons with a conditional deletion of the GluN2A (ΔGluN2A) or the GluN2B (ΔGluN2B) subunit [65] (Fig. 9) reveals that the deactivation time course and ifenprodil inhibition of NMDAR-eEPSCs in Grin2b^+/Δ^ neurons approaches the values observed in neurons completely lacking GluN2B (ΔGluN2B) but does not quite reach them even in the oldest age group (Fig. 9 C, D). This indicates that the GluNB content in synaptic NMDARs is at least partially preserved, perhaps more so than in the extrasynaptic receptor populations (Fig. 8). NMDAR-eEPSC deactivation time course is accelerated by ifenprodil in both genotypes (Fig. 9E) and in Grin2b^+/Δ^ neurons the remaining NMDAR-eEPSCs resemble diheteromeric GluN1/GluN2A eEPSCs recorded in ΔGluN2B neurons [65], consistent with fast-deactivating GluN1/GluN2A-mediated responses [91, 92]. Ifenprodil-treated Grin2b^+/+^ neurons showed slower and more variable NMDAR-eEPSC deactivation kinetics, consistent with some remaining contribution of GluN2B-containing NMDARs to the synaptic response.

Comparison between human GRIN2B variant phenotypes and mouse model outcomes

Both mouse strains, Grin2^+/Δ^ (this study) and Grin2b^+/L825V^ [24], enabled system-level assessment of the effects of a PTV and a missense loss-of-function variant, respectively, each affecting the GluN2B subunit at a similar site. The Grin2b⁺^/Δ^ mice exhibited a milder phenotype: males showed hypoactivity and increased anxiety, and both sexes had impaired sensorimotor gating, while learning, memory, and social behavior were preserved. In contrast, Grin2b^+/L825V^ mice were more severely affected—males displayed hypoactivity, heightened anxiety, and sensorimotor gating deficits, while females exhibited cognitive inflexibility and working memory impairments without anxiety-related changes [24]. These results are consistent with the broader observation that PTVs, which account for approximately 23% of disease-associated GRIN gene variants, are generally linked to less severe DD/ID phenotypes than missense variants [18–20]. The comparison of the two mouse strains also generally aligns with the specific case reports of the severely affected patient carrying the missense GluNB-L825V variant [12, 47] and the relatively more moderate case of the patient carrying the GluN2B-E839* variant [18], which is similar to the PTV studied here (GluN2B-L825Ffs*15).

The behavioral phenotypes observed in Grin2b⁺^/Δ^ and Grin2b^+/L825V^ mice, however, were generally milder than the more substantial clinical symptoms associated with human GRIN2B variants. Several potential explanations can be considered. First, while the mouse models isolate the Grin2b mutations on an unaltered wild-type genetic background, affected patients may carry additional genetic variants that contribute to disease severity [93]. Second, if, in the case of GluN2B PTVs, disease pathology involves the altered distribution of functionally intact GluN2A- and/or GluN2B-containing NMDARs, and since GluN2B expression declines postnatally [94], one might expect symptom severity to lessen with age. Further study of the developmental trajectories of GRIN variant-associated phenotypes is necessary. Finally, it may be challenging to capture the complexity of human behavior in mouse models. Another mouse model of an ASD-associated GRIN2B loss-of-function variant (GluN2B-C456Y [23]) also showed a relatively mild behavioral phenotype including hypoactivity in the open field and age-dependent changes in anxiety-like behaviors, but normal social interactions. Despite some limitations, the existing mouse models of human GRIN2B variant-associated neurodevelopmental disorders are consistent with the general observations from clinical data, give insights into possible disease mechanisms, and provide the platform for the screening of pharmacological or genetic therapies.

In conclusion, the frameshift Grin2b variant studied here reveals a spectrum of molecular and phenotypic alterations, highlighting the complex consequences of Grin2b PTV. Neuron-specific regulatory mechanisms underscore the remarkable plasticity of the NMDAR system and the need to better understand the processes that shape functional outcomes. The comparison of the Grin2b PTV and the previously characterized Grin2b missense variants suggests that Grin2b haploinsufficiency may be better tolerated than the presence of functionally altered GluN2B subunits, with implications for the possibilities of gene therapy.

Supplementary Information

Supplementary Material 1(DOCX 989 KB)