Abstract
N-methyl-D-aspartate receptors (NMDARs) are crucial for neuronal development and synaptic plasticity. Dysfunction of NMDARs is associated with multiple neurodevelopmental disorders, including epilepsy, autism spectrum disorder, and intellectual disability. Understanding the impact of genetic variants of NMDAR subunits can shed light on the mechanisms of disease. Here, we characterized the functional implications of a de novo mutation of the GluN2A subunit (P1199Rfs*32) resulting in the truncation of the C-terminal domain. The variant was identified in a male patient with epileptic encephalopathy, multiple seizure types, severe aphasia, and neurobehavioral changes. Given the known role of the CTD in NMDAR trafficking, we examined changes in receptor localization and abundance at the postsynaptic membrane using a combination of molecular assays in heterologous cells and rat primary neuronal cultures. We observed that the GluN2A P1199Rfs*32-containing receptors traffic efficiently to the postsynaptic membrane but have increased extra-synaptic expression relative to WT GluN2A-containing NMDARs. Using in silico predictions, we hypothesized that the mutant would lose all PDZ interactions, except for the recycling protein Scribble1. Indeed, we observed impaired binding to the scaffolding protein postsynaptic protein-95 (PSD-95); however, we found the mutant interacts with Scribble1, which facilitates the recycling of both the mutant and the WT GluN2A. Finally, we found that neurons expressing GluN2A P1199Rfs*32 have fewer synapses and decreased spine density, indicating compromised synaptic transmission in these neurons. Overall, our data show that GluN2A P1199Rfs*32 is a loss-of-function variant with altered membrane localization in neurons and provide mechanistic insight into disease etiology.
Significance Statement
Dysfunction of NMDARs contributes to a variety of diseases, including neurodevelopmental disorders. Trafficking, expression, and localization of NMDARs at the postsynaptic density are tightly regulated mechanisms that profoundly impact synaptic transmission and plasticity. The CTD of NMDAR subunits is important for the fine modulation of these mechanisms. Here, we describe the altered trafficking of NMDARs resulting from a frameshift variant of the GluN2A subunit that truncates the CTD. This loss-of-function mutation, identified in a patient with epilepsy and developmental delay, leads to altered receptor localization and decreased synaptic transmission in neurons.
Introduction
Over the past decade, there has been an explosion of human genomic studies revealing insights into disease. These data have accelerated the identification of genes associated with neurodevelopmental disorders (NDDs), including intellectual disability (ID), autism spectrum disorder (ASD), and epilepsy (Need and Goldstein, 2010; Willsey et al., 2022). NDDs have widely variable presentations and a high degree of comorbidity. So far, more than 100 genes have been identified as causative of NDDs, some of which can be clustered in shared pathways (Krumm et al., 2014; Deciphering Developmental Disorders Study, 2017; Satterstrom et al., 2020). One of the most prominent clusters is the synaptic function group of genes, which encodes many proteins involved in synaptic structure, communication, and plasticity, namely, proteins related to glutamatergic neurotransmission.
N-methyl-D-aspartate receptors (NMDARs) are a subtype of ionotropic glutamate receptor, crucial for normal brain and neuronal function, with important roles in synaptic transmission, plasticity, and learning and memory (Hansen et al., 2021). As such, dysfunction of these receptors has also been found to be associated with acute and chronic disease etiology (Vieira et al., 2020; Haddow et al., 2022). NMDARs are heterotetramers containing two GluN1 subunits, which are obligatory, and different combinations of GluN2 (A–D) or GluN3 (A–B) subunits (Hansen et al., 2021). GluN2A and GluN2B subunits are the most prominent GluN2 subunits in the forebrain, with well-defined developmental expression profiles (Paoletti et al., 2013). NMDAR subunits have a similar structure to other ionotropic glutamate receptors [α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and kainate receptors], with extracellular N-terminal and ligand-binding domains, three transmembrane domains and a reentrant loop (M1–M4), and an intracellular C-terminal domain (CTD) (Hansen et al., 2021).
Rare variants in NMDAR subunit genes, most prominently in GRIN2A and GRIN2B, are highly associated with NDDs with some variants being pathogenic (Endele et al., 2010; Sanders et al., 2015; Benke et al., 2021; Hansen et al., 2021). In fact, both genes are highly intolerant to genetic variation, particularly in the transmembrane and linker regions (Swanger et al., 2016; Ogden et al., 2017; Traynelis et al., 2017; Vieira et al., 2021). Variants in the CTD have been less studied and are generally considered to be less damaging to protein function. The CTD of NMDARs is important for functions such as trafficking, localization, stabilization at the postsynaptic density (PSD), and establishment of protein–protein interactions. These CTD functions impact receptor modulation and contribute to the coupling of NMDAR activity to downstream signaling pathways, a topic still poorly understood (Vieira et al., 2016, 2020; Park et al., 2022). Studies addressing the functional impact of CTD variants of NMDAR subunits have found that some of those mutations lead to alterations in receptor trafficking, the composition of the receptor-associated molecular complex, and ultimately synapse number and neuronal activity (Liu et al., 2017; Mota Vieira et al., 2020; Yong et al., 2021; Li et al., 2022).
In the current study, we report a rare variant of GRIN2A identified in an individual with epileptic encephalopathy with continuous spike-and-wave during slow-wave sleep (CSWS), multiple seizure types, development of severe aphasia, and significant neurobehavioral changes. This presentation is consistent with the more severe phenotypes reported in patients with GRIN2A-related mutations (von Stulpnagel et al., 2017; Strehlow et al., 2019). To better understand the underlying synaptic dysfunction, we examined the impact of a mutation resulting in a frameshift in the GRIN2A gene leading to the truncation of the CTD of GluN2A (Fig. 1). We found that this variant did not have significant effects on the biophysical or pharmacological properties of the receptor. However, the variant resulted in impaired binding to NMDAR interacting proteins, altered synaptic localization, and a reduction in spine density and neuronal activity, suggesting this is a loss-of-function variant.
The frameshift variant of GluN2A (P1199Rfs*32) causes the truncation of approximately half of the subunit's CTD. A, Graphical representation of GluN2A (WT, left; P1199Rfs*32, right). B, Protein sequence of the CTD of GluN2A WT (top) and P1199Rfs*32 (bottom). The frameshift causes an alteration in the amino acid sequence (shaded in purple), starting at position 1199. Created in BioRender.com.
Methods
Case study
Informed consent was obtained for participation in an NIH Institutional Review Board-approved research protocol (18-N-0066). This allowed for remote review of the patient's medical records and for study investigators to conduct a telephone interview to further characterize the patient's developmental and medical history for research purposes.
DNA constructs and antibodies
The PSD-95 intrabody, pCAG_PSD95.FingR-eGFP-CCR5TC, was a gift from Don Arnold (Addgene plasmid #46295; RRID: Addgene_46295). The GFP-PSD-95 (N-terminal tagged) was a gift from Bernardo Sabatini at Harvard University. pKvenus-Scribble was a gift from Ian Macara (Addgene plasmid #58738; RRID: Addgene_58738).
The following antibodies were used: anti-Myc (Cell Signaling Technology, mouse, 9B11), anti-GFP (Invitrogen, rabbit, #A11122—ICC and co-IP; NeuroMab, mouse, N86/8—Western blot); anti-GluN2A N-terminal (Alomone Labs, rabbit; AGC-002), anti-β-tubulin III (Sigma-Aldrich, rabbit, T2200), and Alexa Fluor-conjugated antibodies (Invitrogen).
Molecular biology
We performed site-directed mutagenesis to introduce the frameshift mutation, c.3596delC, p.Pro1199Argfs*32, (hereafter as P1199Rfs*32) on the pCl-Neo-hGluN2A plasmid, cloned the subunit sequence (WT and mutant) onto the pcDNA3.1 plasmid (gift from Marcel Bruchez; Addgene plasmid #145766; RRID Addgene_145766), and introduced the Myc tag after the signal peptide. After introduction of the mutation, the entire open reading frame of the GluN2A cDNA was verified using dideoxy DNA sequencing to verify the presence of the intended mutation and confirm no secondary site mutations were present (Eurofins Genomics). The human WT GluN1-1a (hereafter GluN1; NM_007327; NP_015566) and human GluN2A (NM_000833; NP_000824) cDNA reference sequences were used as WT GluN1-1a and WT GluN2A. The cDNA for WT and variant NMDAR subunits was linearized using FastDigest (Thermo Fisher) restriction digestion at 37°C for 1 h, and the corresponding complementary RNA (cRNA) was synthesized in vitro using the mMessage mMachine T7 kit according to manufacturer's instructions (Invitrogen).
Unfertilized Xenopus laevis oocytes (Stage V–VI) were prepared essentially as previously described (Han et al., 2022) from commercially available ovaries (Xenopus One). Briefly, the ovary was digested with gentle shaking at 23°C for 2 h in Ca2+-free Barth's solution with Collagenase Type 4 (Worthington Biochemical), which contained (in mM) 88 NaCl, 2.4 NaHCO3, 1 KCl, 0.82 MgSO4, and 10 HEPES (pH 7.4 with NaOH), supplemented with 1 U/ml penicillin-streptomycin (Invitrogen #15140122), then rinsed 10 times (5 min each) with fresh Ca2+-free Barth's solution, and then rinsed 4 times (5 min each) with normal Barth's solution that contained (in mM) 88 NaCl, 2.4 NaHCO3, 1 KCl, 0.33 Ca(NO3)2, 0.41 CaCl2, 0.82 MgSO4, and 10 HEPES, adjusted to pH 7.4 with NaOH, and supplemented with 100 mg/ml gentamycin and 1 U/ml penicillin-streptomycin. Xenopus laevis oocytes were injected with cRNA encoding either WT or variant NMDAR subunits (GluN1/GluN2A WT or GluN1/GluN2A P1199Rfs*32 with a ratio 1:1 for 0.30 ng total weight in 50 nl of RNAase-free water per oocyte) and then maintained in normal Barth's solution at 16°C.
Heterologous cell culture
HEK293T cells were cultured in 6 or 10 cm dishes and transfected 1 d after plating using Lipofectamine 2000 (Life Technologies) as previously described (Won et al., 2016). For co-immunoprecipitation (co-IP) assays, we co-expressed Myc-GluN2A (WT or mutant, as indicated), GluN1, and scaffold proteins (PSD-95—ratio 10:5:1; SAP102 or Scribble1—ratio 2:1:1). Cells were maintained in 20 mM MgCl2 to avoid excitotoxicity. Two days after transfection, we harvested the HEK293T cells in cold PBS, centrifuged at 15,000 × g for 20 min, and resuspended in Tris buffer (pH 8.8) containing 1% sodium deoxycholate (DOC). Next, we sonicated and incubated the lysates for 30 min, at 37°C (Won et al., 2016). The samples were neutralized with pH 7.5 Tris with 1% Triton X-100, at 4°C for another 30 min, and then centrifuged to remove nonsolubilized fractions. Using 1 μg of GFP antibody, we incubated the supernatant overnight at 4°C and then added protein-A beads to the samples. After 4 h incubation at 4°C, beads were washed 3 times with 1% Triton X-100 buffer, and sample buffer was added to the samples. Proteins were denatured at 42°C for 20 min. Samples were resolved in SDS-PAGE and analyzed using immunoblotting with the appropriate antibodies.
Surface biotinylation assay
We performed biotinylation assays as previously described (Nguyen et al., 2020). We co-transfected hGluN2A constructs with hGluN1 and PSD-95 or pKvenus-Scribble1 into HEK293T cells plated on 10 cm dishes, as indicated. Two days post-transfection, cells were washed with cold PBS buffer with 2 mM CaCl2 and 1 mM MgCl2 (PBS++). We then added biotin (1 mg/ml; EZ-Link Sulfo-NHS-LC-Biotin, Thermo Scientific, 21335) in PBS++ to the cells and incubated cells for 30 min on ice. Next, we incubated the cells with 100 mM glycine in PBS++ for 20 min on ice. Three washes with cold PBS buffer were performed between incubations. After the last wash, we collected the cells in cold PBS and centrifuged them at 10,000 × g for 10 min. The pelleted cells were lysed using RIPA buffer with 150 mM NaCl, 50 mM Tris-HCl pH 7.4, 1 mM EDTA, 0.1% SDS, 0.1% DOC, and 1% Triton X-100, protease inhibitors, and phosphatase inhibitor cocktails for 1 h at 4°C. After incubation, we briefly sonicated the lysates and centrifuged them at 20,000 × g for 15 min to remove insoluble debris. We collected a portion of the lysates for the input and incubated the remaining lysates with streptavidin agarose beads overnight at 4°C. Finally, we washed the streptavidin beads three times with cold PBS and eluted the proteins with a 2× SDS loading buffer. To denature the proteins, we incubated the samples at 42°C for 20 min and then analyzed them by immunoblotting with the appropriate antibodies.
We used ImageJ to analyze the immunoblots, by calculating the area under the curve. The integrated intensity was normalized to total GluN2A or tubulin, as indicated in the figure legends; values were then normalized to the control (WT).
Neuronal cultures
The NINDS Animal Care and Use Committee approved all procedures and animal use (protocol #1171). We prepared primary cultures of hippocampal neurons from male and female E18 Sprague–Dawley rats (Envigo), as previously described (Roche and Huganir, 1995), and following the NIH Guide for the Care and Use of Laboratory Animals. In brief, we euthanized pregnant females with CO2 followed by decapitation, and removed the embryos. We collected the brains, dissected the hippocampi, and dissociated the neurons before plating them in poly-D-lysine (Sigma, P7280)-coated coverslips. For imaging experiments, we plated hippocampal neurons on 12-well dishes at 100,000 cells/well; for electrophysiology experiments, hippocampal neurons were plated on 24-well dishes at 200,000 cells/well. We maintained the mixed gender cultures in Neurobasal medium (Life Technologies, Cat#21103-049) supplemented with 2% B27 (Life Technologies, Cat#17504-044) and 2 mM ʟ-glutamine (Sigma-Aldrich, Cat#G-7513) at 37°C and 5% CO2 for 17 d. At DIV 13, we transfected the neurons using Lipofectamine 2000 (Invitrogen, 11668-019) with 2.5 µg Myc-GluN2A constructs, plus 0.75 µg PSD-95.FingR-GFP (for puncta and colocalization analysis), 0.5 µg pCAG-GFP (for spine density analysis and electrophysiology experiments), or 1 µg pKvenus-Scribble1 (for surface expression and recycling assays, as indicated).
Immunofluorescence microscopy
NMDAR trafficking was analyzed using immunocytochemistry. We expressed Myc-tagged GluN2A (WT or mutant) in DIV13 rat hippocampal neurons, using Lipofectamine 2000 transfection. At DIV17, we analyzed receptor surface expression, as previously described (Suh et al., 2010; Sanz-Clemente et al., 2013; Liu et al., 2017). Briefly, we stained surface Myc-tagged GluN2A using a Myc antibody for 15 min at RT, followed by three washes with PBS. We immediately fixed the cells with 4% PFA, for 7 min at RT, and labeled surface receptors with Alexa Fluor 555-conjugated secondary antibody. After permeabilization with PBS 0.25% Triton X-100, we stained the intracellular pool of receptors using a Myc antibody followed by labeling with Alexa Fluor 647-conjugated secondary antibody. For colocalization analysis, we co-expressed the PSD-95.FingR-GFP plasmid and labeled the intracellular GFP expression using a rabbit GFP antibody, followed by the Alexa Fluor 488-conjugated secondary antibody.
To analyze spine density, we co-transfected pCAG-GFP with the GluN2A constructs, and, at DIV 17, we fixed the cells and mounted the coverslips. All immunofluorescence imaging was carried out in a Zeiss LSM800 confocal microscope, and serial optical sections collected at 0.1 μm intervals were used to create maximum projection images.
Recycling assays were performed as previously described (Suh et al., 2010; Mota Vieira et al., 2020). Briefly, we transfected the neurons at DIV 13 with the indicated constructs (Myc-GluN2A WT or P1199Rfs*32 co-transfected with pCAG-GFP or pKvenus-Scribble) and assayed them at DIV 17. We incubated the transfected neurons with a Myc antibody at RT for 15 min. We then returned the cells to 37°C for 30 min to allow for NMDAR internalization. After this incubation, we added an excess of anti-mouse antibody to neurons for 20 min, at RT, to block surface-labeled receptors. The cells were then returned to 37°C for 1 h to induce NMDAR recycling. We washed and fixed neurons in 4% PFA in PBS for 7 min at RT and then labeled surface Myc (recycled receptors) with Alexa Fluor 555 antibody. Cells were then permeabilized and blocked with 10% NGS. Internalized Myc and GFP or Venus-Scribble1 were labeled with Alexa Fluor 647 or 488 secondary antibodies, respectively. Finally, we washed the coverslips again and mounted them on glass slides. We performed the analysis of recycled Myc-GluN2A using a Zeiss LSM800 confocal microscope and the ImageJ software (integrated density) and presented the data as the index of recycled receptors. This index was calculated as a ratio of the integrated density values of the recycled receptor (surface) over the integrated density values of the internalized receptor (intracellular); the ratio of each experimental condition was normalized to that of the control condition (GluN2A WT).
Quantification and analysis of immunocytochemistry
For surface expression of GluN2A in neurons, three dendritic regions were randomly selected per cell (secondary or tertiary dendrites). The experimenter was blinded to the conditions during image acquisition and quantification. We used ImageJ to measure surface and intracellular integrated intensity. For the evaluation of spine density and morphology, we analyzed neuronal dendrites using Neurolucida 360 (Dickstein et al., 2016). Data are presented as ratios of intensity (mean ± SEM) or as the number of spines/dendritic area.
For colocalization quantification, we calculated Mander's coefficients using the JACoP plugin in ImageJ. We used the Puncta tool in the Neurolucida 360 software, version 2020.2.1, to quantify the number of puncta.
Electrophysiological recordings in hippocampal neurons
We co-transfected cultured rat hippocampal neurons at DIV 13 with pCAG-GFP and GluN2A constructs. At DIV 15–18, we performed whole-cell voltage clamp recordings of NMDA mEPSCs in cultured rat hippocampal neurons, at 20–25°C, using glass patch electrodes filled with an internal solution consisting of 135 mM CsMeSO4, 8 mM NaCl, 10 mM HEPES, 0.3 mM Na-GTP, 4 mM Mg-ATP, 0.3 mM EGTA, 5 mM QX-314, and 0.1 mM spermine, as previously described (Mota Vieira et al., 2020). The external solutions contained 119 mM NaCl, 2.5 mM KCl, 26 mM NaHCO3, 1 mM Na2PO4, 11 mM glucose, 2.5 mM CaCl2, and 1.3 mM MgCl2; 10 μM bicuculline and 0.5 μM TTX were added to the external solutions for recording AMPA mEPSCs and 10 μM NBQX, 0.1 mM Picrotoxin and 0.5 μM TTX were added to the external solutions for recording NMDA mEPSCs. We visualized transfected cells with fluorescence, AMPA mEPSCs were measured at a holding potential of -70 mV, and NMDA mEPSCs were measured at +40 mV. Series resistance was monitored and not compensated, and cells in which series resistance was more than 25 MΩ or varied by 25% during a recording session were discarded. Synaptic responses were collected with a Multiclamp 700B amplifier (Axon Instruments), filtered at 2 kHz, and digitized at 10 kHz. The analysis of the mEPSCs was performed with pCLAMP 11 software.
Two-electrode voltage-clamp current recordings from Xenopus laevis oocytes
Two-electrode voltage clamp (TEVC) current recordings 2–3 d after injections were performed at RT (23°C) as previously described (Chen et al., 2017). Current and voltage electrodes were filled with 0.3 M KCl and used for TEVC recordings (OC-725C; Warner Instruments). Electrodes were pulled (PC-10) using borosilicate glass (#TW150F-4; World Precision Instruments). Oocytes were transferred to a dual track recording chamber and perfused with extracellular recording solution that contained (in mM) 90 NaCl, 1 KCl, 0.5 BaCl2, 10 HEPES, and 0.01 EDTA (pH 7.4 with NaOH unless otherwise stated; no EDTA was added for experiments measuring Mg2+ or Zn2+ sensitivity). The extracellular recording solution was supplemented with 10 mM tricine for experiments evaluating Zn2+ sensitivity according to Erreger and Traynelis (2008) and recorded with 50 mM glutamate and 50 mM glycine for maximal activation (Erreger and Traynelis, 2008). Solution exchange was computer-controlled through an 8-valve positioner (Digital MVP Valve). Current responses to agonist application were recorded under a voltage clamp at a holding potential of −40 mV unless otherwise stated. All solutions for concentration–response experiments were made in the extracellular recording solution. Maximal concentrations of agonists (100 μM glutamate and 100 μM glycine) were used in all oocyte recordings unless otherwise stated.
Xenopus laevis data and statistical analysis
Statistical analyses were performed in GraphPad Prism 9.4.0 and performed on the Log IC50 or Log EC50 values where appropriate. Statistical significance was assessed with an unpaired t test with p < 0.05 considered significant. Data are presented as mean ± SEM.
EC50 values and IC50 values were obtained by fitting the concentration–response curve with Equation 1 and Equation 2, respectively:$$mathtex$${\rm Response}\lpar {\rm \% } \rpar {\rm}\equals {\rm 100\sol }\lpar { 1\plus {\lpar {{\rm E}{\rm C}_{ 50}{\rm \sol }\[ {{\rm agonist}} \] } \rpar }^N} \rpar \;$$mathtex$$ (1)where N is the Hill slope and EC50 is the concentration of the agonist that produces a half-maximal effect,$$mathtex$${\rm Response}\equals \lpar {{\rm 100\% }- {\rm minimum}} \rpar {\rm \sol }\lpar { 1\plus {\lpar {\[ {{\rm inhibitor}} \] {\rm \sol I}{\rm C}_{ 50}} \rpar }^N} \rpar {\rm}\plus {\rm minimum \;\;$$mathtex$$ (2)where N is the Hill slope, IC50 is the concentration of the inhibitor that produces a half-maximal effect, and minimum is the degree of residual inhibition at a saturating concentration of the antagonist.
Statistical analysis
We performed the statistical analysis of the data on the GraphPad Prism9 software. Normality tests (D'Agostino–Pearson omnibus or Shapiro–Wilk test) were performed. We used parametric tests (Student's t tests when comparing between two groups; one-way ANOVA for comparisons between three or more groups) on normally distributed data. For data that did not pass the normality test, we used nonparametric tests, or data were log-transformed, and parametric tests were used. Information on the test used for each figure is indicated in the corresponding figure legend. For all assays, 3–5 independent experiments were done. Each n corresponds to one animal or one independent cell culture. Statistical significance was established for p < 0.05. Data are presented as means ± SEM.
Results
Case study
This pediatric patient was found to have a de novo heterozygous variant GRIN2A c.3596delC, encoding the frameshift mutation GluN2A p.P1199Rfs*32, which is absent from the unaffected population (gnomAD database, verified on March 13, 2023). The patient's early development was notable for mild developmental delay, particularly with speech and fine motor delays. Then, at 5 years of age, he developed behavioral changes and the onset of bilateral tonic–clonic seizures, followed by brief myoclonic seizures and eyelid myoclonia and possible absence seizures with only partial response to treatment with multiple medications including valproic acid, clonazepam, levetiracetam, a trial of ketamine, two courses of treatment with prednisolone, and several months on the ketogenic diet. The parents noted a possibly more significant response to ethosuximide and sulthiame. A trial of memantine, an NMDAR antagonist, was also apparently ineffective. Around the time of seizure onset, the patient also experienced severe progressive cognitive and behavioral impairment with loss of expressive speech, impaired comprehension, inattention, incoordination, significant anxiety and depression, and tic disorder. EEG initially showed multifocal predominantly centrotemporal interictal epileptiform discharges activated during sleep, with progression to CSWS by age 10. MRI was reportedly unremarkable. Of note, the patient also has a family history of two siblings diagnosed with autism but no family history of seizures or other seizure risk factors. The patient is now 13 years old, and over the past year, the parents have noted significant improvement in his overall status, with better-controlled seizures requiring less medication, improvement in his language function, improved coordination, and still reduced but improved social engagement. Resolution of seizure activity also occurs near adolescence for patients with GRIN2A null variants (Camp et al., 2021).
The P1199Rfs*32 variant of GluN2A displays increased extra-synaptic localization
The CTD of NMDAR subunits contributes to the fine modulation of receptor trafficking (Vieira et al., 2020), thereby impacting the number and localization of NMDARs at the PSD. To determine if the frameshift variant GluN2A P1199Rfs*32 (Fig. 1) leads to differential surface expression of GluN2A-containing NMDARs, we labeled surface and intracellular receptors in primary cultures of rat hippocampal neurons transfected with the WT or mutant subunit (Fig. 2A,B). In neurons, we observed that the Myc-tagged mutant GluN2A had increased surface expression relative to the WT subunit (P1199Rfs*32: 1.306 ± 0.113), indicating that the mutant GluN2A subunit has increased expression at the neuronal membrane, relative to WT GluN2A.
GluN2A P1199Rfs*32 has altered surface expression. A,B, DIV13 rat hippocampal neurons were transfected with Myc-tagged GluN2A subunits (WT or P1199Rfs*32), as indicated. At DIV17, surface (red) and intracellular (blue) receptors were labeled. Coverslips were analyzed in a Zeiss LSM800 confocal microscope, and quantifications were performed using the ImageJ software. The data were plotted as a ratio of surface/intracellular receptors (relative to WT) and represented as means ± SEMs (n = 3 independent experiments; Mann–Whitney U test; p = 0.0182). Scale bar, 10 μm. C–E, DIV13 rat hippocampal neurons were co-transfected with Myc-tagged GluN2A subunits (WT or P1199Rfs*32) and PSD-95.FingR-GFP. At DIV17, surface GluN2A (red) and the PSD-95 intrabody (blue) were visualized. Coverslips were analyzed in a Zeiss LSM800 confocal microscope, and Mander's colocalization coefficients were calculated using the ImageJ software (JACoP plugin). The Mander's coefficients are presented as the fraction of PSD-95 colocalizing GluN2A (F) or the fraction of GluN2A colocalizing PSD-95 (G) [n = 3 independent experiments; unpaired Student's t test; p = 0.5344 (F), p < 0.0001 (G)]. Scale bar, 10 μm. F,G, The images acquired in (E) for colocalization analysis were used in Neurolucida 360 to detect puncta in the GluN2A (red) and the PSD-95 (green) channels. The density of GluN2A (H) and PSD-95 (I) puncta were calculated. Data are presented as means ± SEMs [n = 4 independent experiments; Student's t test; p = 0.7181 (H), p = 0.0006 (I)].
We next assessed whether the localization of the receptors was altered due to the frameshift and truncation of the subunit (Fig. 2C–G). To do this, we co-transfected hippocampal cultures with the Myc-tagged GluN2A (WT or mutant) and a plasmid encoding an intrabody for PSD-95 [PSD-95.FingR-GFP (Gross et al., 2013)] that allowed us to label the PSD while avoiding morphological changes arising from PSD-95 overexpression (El-Husseini et al., 2000). We analyzed the colocalization coefficient between GluN2A and PSD-95 using ImageJ (Fig. 2C–E). We observed that GluN2A P1199Rfs*32 was targeted to the PSD to a similar extent as the WT (Fig. 2D; Mander's M2—a fraction of PSD-95 colocalized with GluN2A; WT, 0.518 ± 0.024; P1199Rfs*32, 0.543 ± 0.031). Thus, the fraction of PSD-95 that colocalizes with the mutant GluN2A does not differ from the WT. However, we observed an increase in the extra-synaptic localization of GluN2A P1199Rfs*32, relative to GluN2A WT (Fig. 2E; Mander's M1—a fraction of GluN2A colocalized with PSD-95; WT, 0.766 ± 0.026; P1199Rfs*32, 0.594 ± 0.026), indicated by a bigger fraction of WT subunit colocalizing with PSD-95. We then used Neurolucida 360 to identify puncta corresponding to surface GluN2A or PSD-95 and calculated the number of puncta in both channels (Fig. 2F,G). In accordance with the surface expression data and the ImageJ colocalization analysis, we observed an increased number of GluN2A puncta in neurons expressing the P1199Rfs*32 mutant, relative to WT GluN2A (Fig. 2F; WT, 0.815 ± 0.059; P1199Rfs*32, 1.180 ± 0.077). The number of PSD-95 puncta was not changed (Fig. 2G; WT, 0.951 ± 0.063; P1199Rfs*32, 0.979 ± 0.046). Taken together, these results indicate that the truncation of the CTD of GluN2A in the P1199Rfs*32 variant does not prevent the targeting of the subunit to the synapse but leads to increased surface expression at extra-synaptic sites.
Scribble1 interacts with GluN2A P1199Rfs*32
NMDAR subunits possess, at their extreme C terminus, a well-characterized protein–protein interaction motif, named the PDZ-binding domain. This ESDV, four-amino acid PDZ-binding motif, mediates the interaction with a variety of synaptic proteins that contain PDZ domains, most prominently proteins of the MAGUK family—PSD-93, PSD-95, SAP97, and SAP102. These interactions and others occurring via the CTD of NMDARs are important for receptor trafficking, stabilization, and coupling to intracellular signaling pathways (Vieira et al., 2020). Due to the truncation and frameshift in our variant of interest, the expectation is that interactions established via the PDZ-binding domain are abolished in the mutant subunit, since the stop codon occurs upstream of the PDZ-binding motif (Fig. 1). We confirmed such an effect by performing both in silico (Table 1) as well as in vitro (Fig. 3A–D) analyses of the interactions established by WT and mutant GluN2A. Our in silico analysis to predict PDZ domain-containing interactors was performed in the MoDPepInt server, using the PDZPepInt tool (Kundu et al., 2014). As expected, the server predicted interactions of the GluN2A subunit with a variety of proteins with PDZ domains such as all the MAGUKs (DLG1-4), GIPC (Yi et al., 2007), and MAGI-2 (Hirao et al., 1998), among others (Table 1). On the other hand, these PDZ-mediated interactions were predicted to be absent for the frameshift mutant subunit, as anticipated. However, the in silico analysis predicted an interaction between the mutant subunit and the PDZ3 domain of Scribble1. Whereas Scribble1 interacts with WT GluN2A via the sequence IESDV, at positions 1460–1464 (PDZ-binding motif), it is predicted to interact with the mutant subunit via the sequence QATSP, at positions 1225–1229, at the end of the frameshift region of GluN2A P1199Rfs*32 (Table 1, Fig. 1B). To test the predictions from the in silico analysis, we performed co-IPs from HEK293T cells expressing both GluN1/GluN2A (WT or mutant) and PSD-95 (Fig. 3A), SAP102 (Fig. 3B), or Scribble1 (Fig. 3C). We were able to detect the interaction between WT GluN2A and the MAGUKs but observed that the interaction between GluN2A P1199Rfs*32 and both PSD-95 and SAP102 was completely ablated (Fig. 3D; PSD-95, P1199Rfs*32, 0.047 ± 0.022; SAP102, P1199Rfs*32, 0.056 ± 0.003). Conversely, we observed that both GluN2A WT and P1199Rfs*32 were able to interact with Scribble1 (Fig. 3D; Scribble1, P1199Rfs*32, 1.498 ± 0.5131), thus validating the in silico prediction from the MoDPepInt server (Table 1). These data indicate that the trafficking of the truncated GluN2A subunit to the neuronal membrane (Fig. 2) is likely due to the preserved interaction between the subunit and Scribble1, although we cannot exclude interactions with other synaptic or endocytic proteins that may contribute to receptor trafficking.
GluN2A P1199Rfs*32 has altered protein–protein interactions. HEK293T cells were transfected with Myc-tagged GluN2A subunits, GluN1, and GFP-PSD-95 (A), GFP-SAP102 (B), or Venus-Scribble1 (C), as indicated. In addition, 44–48 h after transfection, lysates were prepared, and co-IPs were performed, using a GFP antibody. Immunoblots were performed to visualize protein levels in the total lysate (inputs) and immunoprecipitated fractions. D, Data were plotted as a ratio of immunoprecipitated Myc/GFP (relative to WT) and represented as means ± SEMs (n = 4, PSD-95 and SAP102; n = 6, Scribble1; unpaired Student's t test of log-transformed data; p = 0.0018, PSD-95; p < 0.0001, SAP102; p = 0.6937, Scribble1).
In silico predictions of PDZ domains interacting with GluN2A WT or P1199Rfs*32
Scribble1 mediates the trafficking of the GluN2A P1199Rfs*32 variant
Scribble1 has been shown to contribute to GluN2A trafficking, specifically the recycling of GluN2A-containing receptors, in an activity-mediated manner (Piguel et al., 2014). To examine the functional role of the interaction between Scribble1 and the frameshift GluN2A variant, we performed surface biotinylation assays in HEK293T cells (Fig. 4A,B) and recycling assays in neurons co-expressing WT or P1199Rfs*32 GluN2A and GFP or Scribble1 (Fig. 4C,D). In the surface biotinylation assay, we observed that even without overexpressing Scribble1, the mutant receptor had an increased expression in the cell surface (Fig. 4B; P1199Rfs*32 − Scribble1, 2.087 ± 0.181), relative to the WT subunit. These data are consistent with our observations shown in Figure 2. When we overexpressed Scribble1, we observed that Scribble1 overexpression led to similar surface expression levels for WT and P1199Rfs*32 subunits (Fig. 4B; WT + Scribble1, 1.807 ± 0.288; P1199Rfs*32 + Scribble1, 2.485 ± 0.414). This suggests that endogenous levels of Scribble1 in HEK293T cells could lead to the efficient transport of the mutant receptor to the cell surface [HEK293 expression of Scribble1 available from v22.0.proteinatlas.org/ENSG00000180900-SCRIB/cell + line (Uhlen et al., 2015)]. However, we cannot exclude the potential effect of other endogenously expressed proteins in HEK293T cells, which may be involved in this NMDAR trafficking mechanism. In accordance with these observations, our recycling assay demonstrated that recycling of GluN2A P1199Rfs*32 with endogenous levels of Scribble1 in neurons is higher than that for WT GluN2A (Fig. 4D; P1199Rfs*32, 1.402 ± 0.123). In conditions with overexpressed Scribble1, relative to the GFP-expressing conditions, recycling of both WT and mutant subunit was increased to similar levels (Fig. 4D; WT, 2.388 ± 0.240; P1199Rfs*32, 2.331 ± 0.396). This finding shows that Scribble1 interacts with both WT and mutant GluN2A, to promote the receptor's trafficking.
Scribble-1 interaction contributes to the trafficking of both WT and P1199Rfs*32 GluN2A-containing receptors. A,B, HEK293T cells were transfected with GluN2A (WT, Fs, frameshift), GluN1, and ± Scribble1. Biotinylation assays were performed 44–48 h after transfection, as described. Data are presented as means ± SEMs (n = 3; one-way ANOVA of log-transformed data, followed by Sidak's multiple comparison test; p values indicated in the graph). C,D, Hippocampal neurons were co-transfected with GluN2A constructs and GFP or Venus-Scribble-1. The pCAG-GFP plasmid was used as a co-transfection control for the pKvenus-Scribble conditions and to allow visualization of transfected neurons; these constructs, however, were in different backbones. Recycling assays were performed and recycled, and internalized Myc-tagged receptors were labeled. Cells were imaged using a Zeiss LSM800 microscope, and the recycling index was quantified using ImageJ. Data are expressed as a ratio of recycled/endocytosed receptors and represented as means ± SEM (n = 4; Kruskal–Wallis test followed by Dunn's multiple comparison test; p values indicated in the graph). Scale bar = 10 mm. End., endocytosed; Rec., recycled.
Most channel properties are unchanged by the variant GluN2A P1199Rfs*32
We next examined the functional and pharmacological properties of GluN2A P1199Rfs*32-containing NMDARs. The CTD of NMDARs is primarily involved in the modulation of receptor trafficking and downstream signaling. Therefore, it is less likely that mutations in the subunits’ CTD will affect the channel properties of NMDARs, since these properties are generally determined by the ligand binding and transmembrane domains and linker regions (Chen et al., 2008; Wrighton et al., 2008; Puddifoot et al., 2009). To examine any potential functional effects due to the frameshift, we measured the pharmacological properties of GluN2A P1199Rfs*32 using the Xenopus laevis oocyte expression system and TEVC recordings. We determined the concentrations of glutamate and glycine that can produce a half-maximal current response (EC50), as well as the sensitivity to endogenous negative allosteric modulators and channel blockers: Mg2+, protons, and Zn2+ (Fig. 5; Table 2). We observed that both glutamate and glycine agonist potency, sensitivity to the endogenous modulator Zn2+, and sensitivity to endogenous channel blocker Mg2+ were not affected by the frameshift mutation. The only property that was significantly changed in the mutant receptor was the inhibition by extracellular protons, assessed as a modestly reduced ratio of current at pH 6.8 to that at pH 7.6 [WT 38%, CI (36, 41), and for P1199Rfs*32 45%, CI (42%, 49%)]. This indicates that receptors containing the P1199Rfs*32 subunit may have less tonic inhibition at physiological pH and reduced sensitivity to pH changes that occur during normal neuronal activity or hypersynchronous firing.
Functional measures of GluN1/GluN2A WT (circle symbols, black traces in insets, WT 2A) and GluN1/GluN2A P1199Rfs*32 (triangle symbols, blue traces in insets, P1199Rfs*32) NMDARs expressed in Xenopus oocytes. In all panels, inset current traces shown are normalized to the maximum current generated in maximal ʟ-glutamate and glycine, defined as 100%. A–E, Panel A shows the concentration-effect curve for activation by ʟ-glutamate in the continuous presence of 100 μM glycine at Vhold of −40 mV. The inset maximal current for WT 2A and P1199Rfs*32 was 837 nA and 1539 nA, respectively. B, Concentration-effect curve for activation by glycine in the continuous presence of 100 μM ʟ-glutamate at Vhold of −40 mV. The inset maximal current for WT 2A and P1199Rfs*32 was 693 nA and 1423 nA, respectively. C, Concentration-effect curve for inhibition by Mg2+ in the continuous presence of 100 μM ʟ-glutamate and 100 μM glycine at Vhold of −60 mV. The inset maximal current for WT 2A and P1199Rfs*32 was 476 nA and 461 nA, respectively. D, Concentration-effect curve for inhibition by Zn2+ in the continuous presence of 50 μM ʟ-glutamate and 50 μM glycine at Vhold of −20 mV. The inset maximal current for WT 2A and P1199Rfs*32 was 663 nA and 1007 nA, respectively. E, Ratio of receptor response at pH 6.8 versus pH 7.6 when maximally activated with 100 μM ʟ-glutamate and 100 μM glycine at Vhold of −40 mV. The inset maximal current for WT 2A and P1199Rfs*32 was 624 nA and 2797 nA, respectively. All nonlinear fits of concentration–response data were performed in GraphPad Prism 9.4.0 with 100% maximum and 0% minimum (panels A–C) or 100% max and >0% minimum (panel D) values. In all panels, the mean ± SEM, n = 6 oocytes, for data points is shown.
Results on NMDA receptor function
GluN2A P1199Rfs*32 reduces synaptic activity
NMDAR subunit CTDs have a role in the fine modulation of receptor trafficking and mobility in the neuronal membrane. As such, truncation of the CTD could impact receptor transport and localization. As shown in Figure 2, the variant subunit GluN2A P1199Rfs*32 is efficiently trafficked to the cell surface, likely due to the preserved interaction with Scribble1 (Fig. 3), which contributes to receptor recycling (Fig. 4). To assess whether receptors harboring this epilepsy-associated frameshift variant affect synapse number or function, we analyzed the spine density of neurons expressing WT or mutant GluN2A (Fig. 6). For this purpose, we co-transfected cultured hippocampal neurons with Myc-tagged GluN2A (WT or P1199Rfs*32) and pCAG-GFP. We then counted dendritic spines (three dendrites/neuron) and analyzed their morphology using Neurolucida 360 (Fig. 6A–C). We found a significant decrease in spine density in neurons expressing the GluN2A P1199Rfs*32 variant (Fig. 6B; WT, 14.11 ± 0.708; P1199Rfs*32, 12.18 ± 0.646), but did not detect any changes in spine morphology (Fig. 6C; thin, WT, 0.171 ± 0.014; P1199Rfs*32, 0.140 ± 0.012; stubby, WT, 0.309 ± 0.017; P1199Rfs*32, 0.297 ± 0.0.021; mushroom, WT, 0.141 ± 0.015; P1199Rfs*32, 0.112 ± 0.011) or dendritic branching in variant-expressing neurons (data not shown).
Neurons expressing the epilepsy variant GluN2A P1199Rfs*32 have decreased neuronal activity. A–C, Hippocampal neurons were co-transfected with GluN2A constructs (WT or P1199Rfs*32) and pCAG-GFP. Spine density was assessed using a Zeiss LSM800 microscope for image acquisition and Neurolucida 360 for analysis. B, Spine density data are expressed as the number of spines/20 μm and represented as mean ± SEM (n = 4; Mann–Whitney U test; p = 0.0197). C, Spine morphology data were obtained from Neurolucida 360 and are expressed as number of spines/μm and represented as mean ± SEM (n = 4; two-way ANOVA, followed by Sidak's multiple comparison test; p = 0.2521, thin; p = 0.9578, stubby; p = 0.3476). D, Representative AMPAR-mediated mEPSC traces recorded in cultured hippocampal neurons expressing either GluN2A WT or P1199Rfs*32. E, mEPSC mean amplitude and cumulative probability plots of mEPSC amplitude. The bar graph indicates mean ± SEM (n = 11 cells from 3 independent cultures; paired Student's t test; p = 0.0189). F, mEPSC mean frequency and cumulative probability plots of mEPSC interevent intervals (n = 11 cells from 3 independent cultures; paired Student's t test; p = 0.0007). G, Representative NMDAR-mediated mEPSC traces recorded in cultured hippocampal neurons expressing either GluN2A WT or P1199Rfs*32. H, NMDAR-mediated mEPSC mean amplitude. The bar graph indicates mean ± SEM (n = 12 cells from 3 independent cultures; paired Student's t test; p = 0.1108). I, NMDAR-mediated mEPSC mean frequency interevent intervals (n = 12 cells from 3 independent cultures; paired Student's t test; p < 0.0001).
Next, to evaluate any effect that our variant subunit has on synaptic transmission, as suggested by the decrease in spine density, we analyzed the AMPAR mEPSC (Fig. 6D–F) and NMDAR mEPSC (Fig. 6G–I) frequency and amplitude in neurons expressing either WT GluN2A or the P1199Rfs*32 mutant. We observed a significant decrease in AMPAR mEPSC frequency and amplitude in hippocampal neurons expressing the GluN2A P1199Rfs*32 subunit (Fig. 6E,F; amplitude, WT, 24.55 ± 2.590; P1199Rfs*32, 17.15 ± 1.493; frequency, WT, 2.164 ± 0.288 P1199Rfs*32, 0.908 ± 0.057). Additionally, we observed that the frequency of NMDAR-mediated mEPSC frequency (Fig. 6H,I; amplitude, WT, 19.22 ± 0.847; P1199Rfs*32, 17.41 ± 0.556; frequency, WT, 1.813 ± 0.066; P1199Rfs*32, 0.880 ± 0.049). These data indicate that even though the frameshift variant subunit is efficiently transported to the neuronal membrane, it likely exerts a dominant-negative effect, affecting NMDAR- and AMPAR-mediated synaptic transmission.
Discussion
NDDs are debilitating disorders whose etiology is still being uncovered. Research into the molecular mechanisms underlying these disorders is complicated by the high degree of comorbidity with other conditions, hindering the study of pathways specific to the different disorders (Satterstrom et al., 2020; Willsey et al., 2022). Over the past decade, however, advances in genetic sequencing speed and availability have facilitated the identification of many NDD-associated genes, and specific genetic patterns have started to emerge. Indeed, even though hundreds of genes have been identified, most seem to cluster in specific networks. One of those networks is the synaptic function group of genes that encodes a variety of proteins important for neurotransmission and synaptic plasticity, namely, scaffolding, transsynaptic, and receptor subunit proteins (Krumm et al., 2014). Among the genes found to be associated with diseases, such as epilepsy, ID, or ASDs, are the GRINs, or NMDAR subunit genes. The GRIN2A and GRIN2B genes are the most studied among NMDAR genes, but the GRIN1 and GRIN2D genes have also been found to be pathogenic in some cases of NDDs (Vieira et al., 2021).
In this study, we investigated the functional implications of a frameshift variant of the GRIN2A subunit gene, identified in the patient described in this case report. GRIN2A mutations have been clinically associated with variable neurodevelopmental and epilepsy phenotypes, with more than 80% of cases expressing some form of epilepsy and speech disorder, with ID in 62.7% of patients, although it is often mild (Strehlow et al., 2019). Mild phenotypes can present with near-normal development or with speech delay/apraxia, and mild epilepsy often presents as benign or atypical epilepsy with centrotemporal spikes. More severe phenotypes include profound developmental and epileptic encephalopathy, often in the epilepsy–aphasia spectrum, including Landau–Kleffner syndrome (LKS) and CSWS. Several seizure types have been reported, including focal with impaired or retained awareness, bilateral tonic–clonic seizures, and tonic, atonic, myoclonic, and eyelid myoclonias, often incompletely controlled with seizure medications (von Stulpnagel et al., 2017; Strehlow et al., 2019).
In a large case series of GRIN2A disorders, additional neuropsychiatric comorbidities as seen in our patient were found in 24.3% of patients, including attention deficit hyperactivity disorder, ASD, schizophrenia, and anxiety disorder (Strehlow et al., 2019). Although our patient's presentation was on the severe end of this spectrum, it is important to note that his course is typical of that reported for CSWS overall, with seizures and neurocognitive status appearing around the time of emergence of the EEG findings, followed by eventual improvement after months to years of both the EEG and clinical symptoms. Although the EEG pattern nearly universally resolves, residual abnormalities can remain, and clinical recovery is variable and often incomplete, with the extent of recovery often appearing to be dependent on the severity and duration of the initial regression (Caraballo et al., 2019).
Rare variants associated with disease can have wide-ranging effects on receptor function. Indeed, the type of mutation, as well as its location, can influence the effects of the variant on protein function (Vieira et al., 2021). The genetic variant in our study corresponds to the single nucleotide deletion c.3596delC, which results in a codon reading frame shift, a premature stop codon, and protein truncation. This truncation only affects the CTD of GluN2A, which is encoded by the final exon. Therefore, we do not anticipate that the patient's mRNA undergoes mRNA decay. mRNA decay is a cellular quality-control mechanism that is activated when an aberrant mRNA, containing a premature stop codon, is recognized. The mRNA is targeted for degradation, and thus the protein is not expressed. This mechanism only recognizes premature stop codons when these occur prior to the last exon–exon junction (Kurosaki et al., 2019). The implication is that truncations that occur in the area corresponding to the last exon usually lead to regular mutant protein expression. Since our mutation occurs in the middle of the CTD, which is mostly encoded in the last exon, this mechanism is unlikely to be activated in the patient's cells, and, consequently, we predict that the mutant subunit is expressed at similar levels to the WT.
The CTD of NMDARs is known to have important modulatory functions in the trafficking, localization, and interactions of the receptor with its multimolecular complex (Vieira et al., 2020). As such, we hypothesized that, since our frameshift mutation leads to subunit truncation, it would impact receptor function by disturbing receptor expression at the neuronal membrane, changing its membrane localization, and/or impacting its ability to bind to protein interactors. Our data indicated, however, that the mutant subunit was efficiently expressed at the cell surface. In fact, in neurons expressing the mutant subunit, we even observed a small but significant increase in the expression levels of GluN2A P1199Rfs*32 relative to WT-expressing neurons. In addition, we found that the mutant subunit was expressed at synaptic sites, although we found it to be increased at extra-synaptic locations. These observations are supported by evidence in the literature that reported that ablation of the CTD of GluN2A does not lead to reduced surface expression of the receptor. Instead, deleting the CTD leads to a change in receptor distribution, which becomes more extra-synaptic (Steigerwald et al., 2000; Kohr et al., 2003). In these studies, a reduction in the synaptic levels of the truncated GluN2A subunit was reported, whereas we did not detect such a change. A reason for this difference could be that while our mutant subunit still has approximately half of the CTD, in those of previous works, the CTD is completely absent. Thus, mechanisms that are regulated by membrane-proximal regions of the CTD could still be functional in our frameshift subunit. Few studies have addressed the functional implications of GRIN2A variants residing in the CTD of the GluN2A subunit (Addis et al., 2017; Mota Vieira et al., 2020; Yong et al., 2021; Li et al., 2022). Of the studied variants, only GluN2A S1459G (Mota Vieira et al., 2020; Yong et al., 2021) and K879R (Li et al., 2022) were found to significantly affect receptor properties. GluN2A S1459G was found to inhibit important protein–protein interactions established by the subunit via the PDZ-binding domain, affecting the receptor's trafficking mechanisms. Conversely, GluN2A K879R enhanced receptor surface expression and postsynaptic currents, thereby resulting in NMDAR gain of function. Thus, the CTD variants in the literature have been found to not affect receptor function (Addis et al., 2017) or cause changes in receptor trafficking and synaptic currents. The resulting synaptic dysfunction can be either a gain or loss of function. Our findings for GluN2A P1199Rfs*32 in the current study reveal an unusual mistargeting of receptors, which still ultimately results in fewer spines and synaptic currents consistent with a loss-of-function outcome. Therefore, just as has been observed for other domains of NMDAR subunits, the consequences of disease variants in the CTD are variable and depend on the specific residues affected (Vieira et al., 2021). Thus, establishing parallels between different mutations is challenging, and each variant should be examined independently.
Another important function of the CTD of NMDAR subunits is mediating a variety of protein–protein interactions, which can contribute to several outcomes, mediating the stabilization of the receptor at the PSD. One of the best-characterized interaction motifs in the NMDAR CTD is the PDZ-binding domain, encoded in the last four amino acids of the protein (ESDV). This motif mediates interactions with PDZ domain-containing proteins such as the MAGUK family members or SNX27 (Clairfeuille et al., 2016; Vieira et al., 2020). Since our frameshift GluN2A subunit is truncated, it does not possess the typical C-terminal PDZ-binding domains, and, therefore, those interactions are expected to be disrupted. Indeed, we observed that truncation of GluN2A blocked the interaction with PSD-95 and SAP102. However, we found that GluN2A was still able to interact with the PDZ protein Scribble1. This is an endocytic protein, known to contribute to the regulation of GluN2A-containing receptor's surface expression (Piguel et al., 2014). As predicted by an in silico analysis using the PDZPepInt server, GluN2A interacts with Scribble1 via the C-terminal amino acids resulting from the frameshift of the mutated subunit. We observed that the preserved interaction has functional implications, as Scribble1 enhanced GluN2A P1199Rfs*32 recycling. Thus, our data support a role for this interaction with Scribble1 in promoting the surface expression of our frameshift mutant subunit.
Despite reaching the neuronal surface and even localizing to the synapse, are the functional properties of GluN2A P1199Rfs*32 similar to the WT subunit's properties? We analyzed functional and pharmacological channel properties in oocytes and the spine density and mEPSCs in mutant-expressing neurons to answer this question. In oocytes, we observed that channel properties, including glutamate and glycine potencies, were preserved and the sensitivity to most endogenous inhibitors, namely, Mg2+ and Zn2+ inhibition, was unchanged. We did not anticipate that a mutation in the CTD would affect the channel properties of NMDARs since previous work has shown that deletion of the CTD of GluN2A does not lead to changes in agonist potency or sensitivity to endogenous modulators (Puddifoot et al., 2009). Furthermore, previous studies of disease variants affecting the CTD of NMDAR subunits, for the most part, did not affect the receptor's channel properties, relative to WT (Liu et al., 2017; XiangWei et al., 2019). The only parameter that we observed to be changed in the mutant receptor was pH sensitivity. In GluN2A P1199Rfs*32-expressing neurons, we observed a significant reduction in both dendritic spine density and in AMPAR-mediated mEPSC frequency and amplitude, and in NMDAR-mediated mEPSC frequency, consistent with an overall decrease in excitatory synapses. These observations suggest that even though the mutant subunit is expressed at the neuronal surface, it causes a synaptic deficit, resulting from reduced NMDAR transmission, which likely impacts AMPAR transmission, given the reported roles exerted by NMDAR activation on AMPAR insertion (Lu et al., 2001; Chater and Goda, 2022). These effects are likely compounded by the dominant-negative effect in these neurons by saturating PSDs with receptors that cannot engage the full repertoire of important intracellular signaling proteins and, consequently, resulting in reduced neuronal activity.
Overall, our observations suggest that the mutant subunit GluN2A P1199Rfs*32 perturbs synaptic NMDAR function by altering the subcellular distribution of the GluN2A-containing NMDARs. This NMDAR mistargeting reduces excitatory drive, likely secondary to changes in the molecular complex associated with NMDARs containing this subunit. The ramifications of this for the balance of inhibition and excitation are unknown. This observation is reinforced by the fact that dendritic spine density, and mEPSCs were reduced in neurons expressing the mutant subunit. Thus, we propose that the GluN2A mutation P1199Rfs*32 results in loss of function, due to altered synaptic currents; in the patient, the NMDAR mistargeting and altered NMDAR-mediated currents may contribute to the complex phenotype, which includes epilepsy and ID. Potential therapeutic strategies to mitigate the consequences of this variant remain to be explored.
Footnotes
We thank Carolyn Smith and the NINDS imaging facility for their assistance. This work was supported by the NIH Intramural Research Program (M.M.V., S.W., E.H., I.S., A.T., J.D.B., S.P., A.H.T., S.K.I., W.L., and K.W.R.), the Simon’s Foundation (S.F.T.), and the NIH grants NINDS NS111619 (S.F.T.) and AG072142 (S.J.M.).
S.F.T. is a member of the SAB for Eumentis Therapeutics, Sage Therapeutics, and Combined Brain, is a member of the Medical Advisory Board for the GRIN2B Foundation and the CureGRIN Foundation, is an advisor to GRIN Therapeutics, is co-founder of NeurOp Inc. and AgriThera Inc., and is a member of the Board of Directors of NeurOp Inc. S.J.M. is PI on a research grant from GRIN Therapeutics to Emory University School of Medicine.
- Correspondence should be addressed to Katherine W. Roche at rochek{at}ninds.nih.gov.
