Written informed consent was obtained from the parent of the patient reported. This study was approved by the Medical Ethics Committee and the Institutional Review Boards of Seattle Children’s Hospital, University of Washington (IRB: #13291). All clinical data of this study were analyzed anonymously. All functional studies were performed according to the guidelines of Emory University and The University of Texas Health Science Center.

We evaluated a female patient with a new clinical phenotype with a novel de novo deleterious variant in GRIN1 (c.1595C>A, p.Pro532His), who is from a large Caucasian family. A three generation pedigree was taken including proband and two full siblings of proband, biological parents and six uncles and aunts, four grandparents and nine grand uncles and aunts, and eight great grandparents. Her parents are nonconsanguineous. One younger sister and three brothers are healthy. One of her brothers has healthy children. There is a distant maternal relative with Down syndrome who died. There is a maternal family history of hemochromatosis. The patient’s father is 44 years old and in good health. He is 6 ft 1 in. tall. He has a healthy brother and sister, and his brother has a daughter with some hearing loss and minor anomalies but otherwise normal cognitive function. A brother of her father died in infancy of unknown cause. Clinical data was collected via electronic medical record review. Diagnostic workup, including karyotype, a chromosome microarray, metabolic testing (plasma amino and urine organic acids, serum lactate, serum pyruvate, oligosaccharides, acylcarnitine profile, and glycosylation), and a skin biopsy study for electron microscopy, were performed. Trio-whole exome sequencing (proband and biological parents) was performed at GeneDx. The Web resource gnomAD was used to search for GRIN1gene variants.

We utilized cDNA for wild-type human NMDA receptor GluN1-1a (hereafter GluN1; NCBI NM_007327/NP_015566), GluN2A (NCBI NM_000833/NP_000824), and GluN2B (NCBI NM_000834/NP_000825) subunits subcloned into the plasmid vector pCI-neo (Promega, Madison, WI, United States). The mutant GluN1 cDNA was generated by site-directed mutagenesis using the QuikChange protocol (Stratagene, La Jolla, CA, United States) and cRNA was synthesized in vitro from a linearized cDNA template according to manufacturer instructions (mMessage mMachine; Ambion, Austin, TX, United States) (Chen et al., 2017b). Xenopus laevis oocytes were prepared (XiangWei et al., 2019) from commercially available ovaries (Xenopus One Inc., Dexter, MI, United States); 5–10 ng of total cRNA in 50 nl of RNase-free water was injected into each oocyte using a microinjector (Drummond Nanoject II) with a ratio of 1:2 for GluN1:GluN2. Injected oocytes were stored at 15–17°C in Barth’s solution as previously described (Chen et al., 2017b).

Two-electrode voltage-clamp (TEVC) recordings from Xenopus oocytes were performed as previously described (Chen et al., 2017b). The recording solution contained 90 mM NaCl, 1 mM KCl, 10 mM HEPES, 0.5 mM BaCl₂, and 0.01 mM EDTA (23°C, pH 7.4 unless otherwise stated). The membrane potential was held at −40 or −60 mV. The amplitudes at each agonist concentration were fitted with

where EC₅₀ is the concentration that produces a half maximal response, [agonist] is the concentration of glutamate, glycine, or other agonists, and N is the Hill slope. Concentration–response curves were recorded for NMDAR current response as activated by maximally effective concentrations of glutamate and glycine, with coapplication of variable concentrations of negative allosteric modulators Mg²⁺ or Zn²⁺ (Zn²⁺ buffered with tricine) (Traynelis et al., 1998). The current response amplitudes were fitted with:

where minimum is the residual response in saturating concentration of the negative modulator (Mg²⁺ or Zn²⁺), IC₅₀ is the concentration of the modulator that produces a half-maximal inhibition, and N is the Hill slope.

The channel open probability (P_OPEN) was calculated from the degree of potentiation produced when maximally effective glutamate and glycine are coapplied with 200 μM MTSEA (Toronto Research Chemicals) (Yuan et al., 2005) according to

where γ is the channel chord conductance for GluN1/GluN2A before and after MTSEA modification (Yuan et al., 2005) and potentiation is the ratio of current after MTSEA to that observed before MTSEA.

All recording solutions for 24(S)-hydroxycholesterol, pregnenolone sulfate, and tobramycin were supplemented with 0.02% cremophor.

HEK293 cells (ATCC CRL-1573) were transiently cotransfected as previously described (Chen et al., 2017b) with plasmid cDNAs encoding wild-type human GluN1/GluN2A/GFP, GluN1/GluN2B/GFP, or the variant GluN1-P532H/GluN2A/GFP, or GluN1-P532H/GluN2B/GFP. The ratio of transfected cDNAs was 1:1:5 when co-expressed with GluN2A and 1:1:1 when co-expressed with GluN2B. Following an 18–24-h transfection, whole-cell voltage-clamp current recordings were made (Chen et al., 2017b; Ogden et al., 2017) with an extracellular solution containing 150 mM NaCl, 3 mM KCl, 10 mM HEPES, 0.01 mM EDTA, 0.5 mM CaCl₂, and 11 mM D-mannitol, with the pH adjusted to 7.4 by addition of NaOH (23°C). The recording electrodes were filled with an internal solution that contained 110 mM D-gluconic acid, 110 mM CsOH, 30 mM CsCl, 5 mM HEPES, 4 mM NaCl, 0.5 mM CaCl₂, 2 mM MgCl₂, 5 mM BAPTA, 2 mM Na-ATP, and 0.3 mM Na-GTP adjusted to pH 7.35 with CsOH; the osmolality was adjusted to 300–310 mOsmol kg^–1 using CsCl or water. The current responses to external application of glutamate (1,000 μM) and glycine (1 to 100 μM) were recorded using an Axopatch 200B patch-clamp amplifier (Molecular Devices) with the holding potential of −60 mV. The current responses were filtered at 8 kHz (−3 dB, 8 pole Bessel filter, frequency devices) and digitized at 20 kHz using a Digidata 1440A and Axon Instruments software (Molecular Devices, Sunnyvale, CA, United States).

HEK293 cells were plated in 96-well plates and transiently transfected with cDNA encoding WT β-lactamase (β-lac)-GluN1 or β-lac-GluN1-P532H with either WT GluN2A or WT GluN2B using Fugene6 (Promega) (Swanger et al., 2016). The background absorbance was determined by the cells treated with Fugene6 only. A negative control for surface β-lac activity was determined in cells that were not transfected with GluN2A and GluN2B cDNA. Eight wells were transfected for each condition, and the levels of surface and total β-lactamase were measured in four wells each. After 24 h transfection, the cells were washed with Hank’s balanced salt solution (HBSS) supplemented with 10 mM HEPES, and then 100 μl of a 100-μM nitrocefin (Millipore, Burlington, MA, United States) solution in HBSS with HEPES was added to each well to allow the measurement of surface activity (Swanger et al., 2016). In separate wells, the cells were lysed by a 30-min incubation in 50 μl H₂O prior to the addition of 50 μl of 200 μM nitrocefin to determine total activity. The absorbance was read on a microplate reader (SpectraMax M2) at 486 nm once every minute for 30 min (30°C). The rate of increase in absorbance was determined from the slope of the linear regression.

The dimer structure of GluN1/GluN2A bound to glycine and glutamate (PBD ID 5H8Q) was taken from the Protein Data Bank for MD simulation. Two simulation systems (wild-type and P532H variant) were prepared. The PSFGEN module of the Visual Molecular Dynamics Simulation software (VMD) was used for building missing residues and hydrogen atoms to the protein structure followed by protein solvation and ionization (Humphrey et al., 1996; Gullingsrud et al., 2006). First, the protein was placed in the water box of size 97 × 90 × 105 Å, with around 24,372 TIP3P water molecules and six chloride ions for charge neutralization (Humphrey et al., 1996; Gullingsrud et al., 2006). Total atoms in the simulation box ranged from 82,006 to 82,009. Each system was subjected to conjugate gradient energy minimization for 5,000 steps by applying restraint force of spring constant 4 kcal/mol/Å2 on the heavy atoms of protein and on the residues interacting with the ligands glutamate and glycine. The system was then equilibrated for 5 ns using 1 fs time step and gradually the restraint force constant was decreased to zero. Following system equilibration, 300 ns production was performed using 2 fs time step under constant NTP ensemble. NTP ensemble refers to constant number of particles (N), constant pressure of 1 bar controlled by Nose-Hoover Langevin piston and constant temperature at 310 K controlled by the Langevin thermostat. Particle mesh Ewald (PME) (York et al., 1993) method was applied for calculating long-range electrostatic interactions and SHAKE restraint was used on covalent bonds involving hydrogen atoms. Smooth switching of small-range nonbonded interaction was done between 10 and 12 Å with pair list cut-off updates on 14 Å. NAMD 2.12 (Phillips et al., 2005) program was used for performing the simulation. CHARMM36 force field (Best et al., 2012), with cMAP dihedral correction was used for the protein. The CHARMM general force field (CGENFF) was used for ligands (Vanommeslaeghe et al., 2010). Simulations were submitted at Stampede2 cluster of Texas Advancing Computing Center (TACC). MD trajectories were analyzed using TCL scripts, VMD software and python Matplotlib (Hunter, 2007).

Cysteine-light constructs of human GluN1 wild type, GluN1-P532H variant, and GluN2 were generated by mutating nondisulphide-bonded extracellular cysteines (Cys15, Cys22 in GluN1 and Cys231, Cys399, and Cys460 in GluN2) to serines, and cysteines were introduced at positions 502 and 701 in the GluN2A cysteine-light construct to measure conformational changes across ABD of GluN2. All the mutations were confirmed by Sanger sequencing (Genewiz). HEK293 cells were transiently cotransfected with the plasmids harboring cDNAs encoding GluN1/GluN2A (1:3; 10 μg of total cDNA per 10 cm dish) or the variant GluN1-P532H/GluN2A with the FRET mutations mentioned above. Transfections were performed the day before the smFRET experiment using JetPRIME transfection reagent (polyplus). On the day of experiment, i.e., 24 h post-transfection, the cells were labeled with donor and acceptor fluorophore and the sample was prepared as described previously (Dolino et al., 2017; Litwin et al., 2019; Durham et al., 2020).

For smFRET slide preparation, we used the same protocol as described previously (Durham et al., 2020). To selectively pull down the NMDA receptors onto the slide, we used biotinylated anti-mouse antibody bound to the streptavidin-coated slide followed by anti-NMDAR1 antibody (ab64572, Abcam). After antibody treatment, the slide was treated with bovine serum albumin and then with lysate of HEK293 cells expressing the modified NMDA FRET construct. The slide was then flushed with oxygen scavenging buffer containing (3.3% (w/w) glucose, 3 units/ml pyranose oxidase, 0.001% (w/w) catalase, 1 mM ascorbic acid, and 1 mM methyl viologen, in 1× imaging buffer (Durham et al., 2020), containing 1 mM glutamate and 1 mM glycine. smFRET data was acquired using a custom-built Pico-Quant MicroTime 200 Fluorescence Lifetime Microscope (Picoquant). Pulsed interleaved excitation was used and donor excitation was at 532 nm (LDH-D-TA-530) and acceptor at 637 nm (LDH-D-C-640, Picoquant). SPAD photodiodes (SPCM CD3516H; Excelitas Technologies) were used to detect the photons emitted from the sample and emission filters [550 nm (FF01-582/64; AHF) and 650 nm (2XH690/70; AHF)] were also used. For data analysis, we used molecules showing single acceptor and donor photobleaching step in addition to anticorrelation between acceptor and donor to ensure that the data was from a single-molecule showing FRET. FRET efficiencies were calculated using donor and acceptor intensities using the Forster equation (detailed description provided in Cooper et al., 2015; Litwin et al., 2019; Durham et al., 2020). Numbers of conformational states were estimated using step transition and state identification (STaSI).

The relative change in synaptic and non-synaptic charge transfer was evaluated as a ratio to the WT receptors by the following equations (Swanger et al., 2016; Li et al., 2019):

where [glutamate] is 1 × 10^–3 M for R_{GLU, Synaptic} and 1 × 10^–7 M for R_{GLU, Non–synaptic}, [glycine] is 3 × 10^–6 M, and N is the Hill slope. t_w is the weighted constant for deactivation time course, P is the channel open probability, Surf is cell surface protein levels, Mg is percentage inhibition by 1 mM Mg²⁺ (V_HOLD −60 mV), R_GLY and R_GLU are relative response to a given concentration of glutamate or glycine. The τ_w, P, Surf, and Mg for the variant were calculated as a ratio to the WT receptors.

All data are given as mean with the 95% confidence intervals. Statistical significance was set to p < 0.05 and assessed by an unpaired Student’s t-test. The number of independent experiments was represented by n. Samples sizes were determined a priori from power analysis for an effect size of 1–2 (Gpower) and were 6–17 to ensure at least 0.8 power with α = 0.05. All compounds and agents were purchased from Sigma-Aldrich (St. Louis, MO, United States) unless stated otherwise.

The patient was born at full term by C-section due to breech presentation with maternal age of 26 years and paternal age of 34 years. The pregnancy was complicated by decreased fetal movement at 6 months of gestation. Parents are nonconsanguineous Caucasian. Birth weight was 7 lbs, and length was 20 in. Birth occipital frontal circumference (OFC) was not available but was 52.7 cm at the age of 10 years (65 percentile). There is no family history of developmental impairment in the three-generation pedigree. The patient had some tongue thrusting at birth but was able to breast feed, fix, and follow. She had head lag which was noted when she was 3 months old. However, by the age of 3 months, the patient lost these previously acquired skills, developed difficulties with feeding, increased irritability in infancy, and required G-tube placement at the age of 4 years. She could hold her head up at age 12 years in sitting position. Her weight ranged from 25% to 32% from age 3 months to 4 years, then increased after the G-tube placement to 50% from age 6 to 8 years, then gradually decreased to less than 3% at the age of 13 years. Her height ranged from 50% to 80% from age 3 months to 8 years, then gradually drop to less than 3% by the age of 13 years.

The patient had severe gastroesophageal reflux and scoliosis, as well as progressive contractures, mostly of her legs. She has mild to moderate equinus contractures of the feet. Knee contractures were at 45° to 90°. Her gross motor function was classified as (GMFCS) V. She has a left thoracolumbar scoliosis at 80° from T8 to L4. The patient had spasticity in all extremities. The patient has not had confirmed seizures. EEG at the age of 5 years showed diffuse slowing. The ophthalmological evaluation at 3 months of age showed normal visual evoked potential and normal dilated fundi examination. The patient has profound developmental delay and intellectual disability. She is nonverbal, does not make eye contact, and has no sign or other forms of language. She has remained non-mobile and her gross motor function classification system (GMFCS) Level V. She is G-tube dependent at the age of 16–18 years. The patient has marked brachycephaly, midface (maxillary) hypoplasia, and mildly protuberant ears. The patient has low normal muscle bulk and mildly increased tone in the arms and moderate to severe increased tone in the legs. The patient has very frequent stimulus-sensitive myoclonic jerks. The patient did not have a muscle biopsy. The EEG monitoring captured the myoclonus with no EEG correlation, suggesting subcortical origin. Brain MRI at 3 months and about 5 years of age showed enlarged extra axial space, best seen over the frontal lobes with some mildly enlarged sulci (Figure 1A). Her corpus callosum was mildly thin. The brainstem, cerebellum, and posterior fossa appeared normal, and there was no polymicrogyria detected (Figure 1A).

Karyotype and a chromosome microarray were normal. Metabolic testing was normal, including plasma amino and urine organic acids, serum lactate, serum pyruvate, oligosaccharides, acylcarnitine profile, and glycosylation. The patient had a skin biopsy study for electron microscopy with normal results. The whole-exome sequencing was performed by GeneDx, which identified a de novo, heterozygous variant p.Pro532His, c.1595C>A, in exon 11 in the GRIN1 gene (NM_007327) (Figure 1B). The GluN1-P532H variant has not been reported previously. This variant was interpreted as likely pathogenic by GeneDx. The variant was not observed in healthy populations in the gnomAD database¹ (evaluated on March 3, 2021) and in about 6,400 individuals of European and African American ancestry in the NHLBI Exome Sequencing Project.

The residue Pro532 is located in the S1 region of the ABD (Figures 1C–E), which harbors the binding pocket for the coagonist glycine. This proline is conserved in GluN1 through all vertebral species and across all human GluN2 subunits (Figure 1D), suggesting a potential important role in channel function. NMDAR complexes with wild-type (WT) GluN1 subunit or variant GluN1-Pro532His were coexpressed with either wild-type GluN2A or GluN2B subunits. The electrophysiological properties of these receptor complexes were compared by determining the potency of endogenous agonists and modulators. Since the residue at which this missense variation occurred is located in the GluN1 ABD, which fully encodes the glycine-binding pocket, we first assessed the effects on glycine potency. The concentration of agonist that produces a half-maximal current response (EC₅₀) was determined by measuring the response to a range of glycine concentrations coapplied with a maximally effective concentration of glutamate (1,000 μM) onto NMDARs expressed in oocyte using TEVC (V_HOLD: −40 mV). The GluN1-P532H receptors showed only minimal effects on glycine potency, with an EC₅₀ value for GluN1-P532H/GluN2A of 2.0 μM compared with 1.8 μM for WT GluN1/GluN2A. The EC₅₀ value for GluN1-P532H/GluN2B was 0.34 μM, slightly lower than 0.45 μM found for WT GluN1/GluN2B (Figures 2A,B and Table 1), indicating this variant may have a modest effect on glycine potency. We subsequently evaluated glutamate potency on GluN1-P532H-containing NMDARs in the presence of maximally effective concentration of glycine (100 μM). Unexpectedly, GluN1-P532H-containing receptors significantly decreased glutamate potency, increasing EC₅₀ values by 16-fold for GluN1-P532H/GluN2A (EC₅₀: 105 μM) compared with GluN1/GluN2A (6.4 μM) and 43-fold for GluN1-P532H/GluN2B (EC₅₀: 82 μM) compared with WT GluN1/GluN2B) (1.9 μM; p < 0.01, unpaired t-test; Figures 2C,D and Table 1). These data suggest that activation of variant receptors requires higher concentrations of glutamate.

It is well known that NMDAR function can be regulated by a set of endogenous negative allosteric modulators, including Mg²⁺, protons, and Zn²⁺ (Traynelis et al., 2010). We therefore evaluated the effects of GluN1-P532H on the sensitivity of the NMDAR to these modulators. The concentration–response curves for Mg²⁺ inhibition at −60 mV showed a comparable potency for GluN1-P532H/GluN2A with an IC₅₀ value of 22 μM compared with 19 μM of WT GluN1/GluN2A. The IC₅₀ value for GluN1-P532H/GluN2B was 35 μM, which was similar to 38 μM for WT GluN1/GluN2B (p = 0.99 and 0.88, unpaired t-test; Figures 2E,F and Table 1). Determination of the concentration–response relationship of Zn²⁺ revealed a modest increase in Zn²⁺ sensitivity in GluN1-P532H/GluN2A by almost twofold, with IC₅₀ values decreasing from 8.5 nM for WT GluN1/GluN2A to 4.1 nM for GluN1-P532H/GluN2A (p < 0.05, unpaired t-test; Figure 2G and Table 1). The GluN1-P532H receptor showed no change in proton sensitivity, assessed by comparison of NMDAR-mediated current amplitude recorded at pH 6.8 with that recorded at pH 7.6 when coexpressed with either GluN2A or GluN2B (Figure 2H and Table 1).

Taken together, these data suggested that GluN1-P532H may reduce excitatory drive as a result of the decreased activation of glutamate (reduced potency) and the enhanced inhibition by endogenous zinc. They also reveal the unexpected effect of a variant located in the GluN1 glycine-binding domain on glutamate EC₅₀, the binding site for which resides in the GluN2 subunit.

The deactivation response time course following rapid removal of agonist (e.g., glutamate) from NMDARs has been proposed to determine the time-course of the NMDAR component of the excitatory postsynaptic current (EPSC) (Lester et al., 1990). To evaluate the influence of GluN1-P532H on the deactivation time course, we measured the current response time course following glutamate removal using a rapid solution exchange system. Current responses to prolonged application of glutamate (1.5 s, 1,000 μM) were recorded under whole-cell voltage clamp from transiently transfected HEK293 cells expressing WT GluN1/GluN2A, GluN1-P532H/GluN2A, WT GluN1/GluN2B, or GluN1-P532H/GluN2B. GluN1-P532H significantly decreased glutamate deactivation time course, which could be described by two exponential components with a weighted time constant (τw) of 26 ms for GluN1-P532H/GluN2A compared with 66 ms for WT GluN1/GluN2A. Weighted time constant for deactivation was 37 ms for GluN1-P532H compared with 821 ms for WT GluN1/GluN2B (p < 0.01, unpaired t-test; Figures 3A,B left panels and Table 1). To mimic synaptic events, we also recorded the current response time course while briefly moving the transfected HEK cell into the agonist solution for 5 ms (brief application). Similar to the prolonged (1.5 s) application of glutamate, GluN1-P532H/GluN2A had a faster deactivation time course with a τ_W of 18 ms compared with 61 ms for WT GluN1/GluN2A. Similarly, GluN1-P532H/GluN2B had a faster deactivation time course with τ_W of 28 ms compared with 704 ms for WT GluN1/GluN2B (Figures 3A,B right panels). These data suggest that the variant GluN1-P532H-containing NMDARs have a faster deactivation response time course, and therefore will likely produce brief synaptic currents.

There is a negative allosteric coupling between glycine and glutamate binding such that when glutamate first binds to the receptor, there is a decrease in affinity for glycine. If glycine is present at subsaturating levels, glutamate binding can produce a slowly decreasing current because glycine unbinds as the system relaxes to a new equilibrium. This glycine-dependent relaxation in current has been referred to as glycine-dependent desensitization (Mayer et al., 1989; Benveniste et al., 1990). To evaluate whether the variant influences glycine-dependent desensitization, we recorded current responses to prolonged application of two concentrations of glycine (1 and 30 μM; 1.5 s) in the presence of 1,000 μM glutamate under whole-cell voltage clamp from HEK cells transfected with WT GluN1/GluN2A and GluN1-P532H/GluN2A. GluN1-P532H variant-containing NMDA receptors showed a diminished desensitization to high concentration of glycine (p < 0.01, unpaired t-test; Figure 3C and Table 1), while the fast component to desensitization in a subsaturating concentration of glycine appears to be retained in the variant receptors. These data suggest that GluN1-P532H does not eliminate glycine-dependent desensitization.

To assess the effects of the GluN1-P532H variant on single-channel open probability, we measured the degree by which covalent modification of an M3 residue by MTSEA potentiates NMDARs with a cysteine mutation in GluN2A M3 SYTANLAAF gating region (GluN2A-A650C); covalent modification by MTSEA of this residue locks the channels open (Jones et al., 2002; Yuan et al., 2005). We calculated the channel open probability based on the degree of MTSEA potentiation of the NMDAR response to maximally effective agonists (1,000 μM glutamate and 100 μM glycine) in TEVC oocytes recordings (V_HOLD: −40 mV; Figure 3D). The MTSEA-mediated current increase is reciprocally related to the channel open probability prior to MTSEA application (see the section “Materials and Methods”). These MTSEA-derived estimations indicated that channel open probability for GluN1-P532H/GluN2A decreased by 2.6-fold from 0.26 (n = 41) for WT GluN1/GluN2A to 0.11 (n = 34) (p < 0.01, unpaired t-test; Figure 3D and Table 1).

To evaluate whether the variant influences receptor cell surface expression, we measured the cell surface protein level and total protein level using a reporter assay in which β-lac was fused to the extracellular NTD of WT GluN1 (β-lac-GluN1) and the variant (β-lac-GluN1-P532H). The β-lac-GluN1 fusion protein was coexpressed with WT GluN2A or WT GluN2B in HEK293 cells. The level of surface receptor expression was determined by the β-lac cleavage of the cell-impermeable chromogenic substrate nitrocefin in the extracellular solution (Lam et al., 2013; Swanger et al., 2016). When coexpressed with WT GluN2A or WT GluN2B, the GluN1-P532H variant showed a significant decrease of surface-to-total protein level, compared with WT GluN1 receptors (GluN1-P532H/GluN2A: 64% of WT, GluN1-P532H/GluN2B: 67% of WT; p < 0.05; Figures 3E,F and Table 1). However, the variant when coexpressed with GluN2B showed a tendency to have higher total protein levels (total: 193% of WT; p = 0.06; Figure 3F and Table 1). These data suggest that the GluN1-P532H variant has a significant influence on surface expression. This may involve alterations in receptor assembly, as have been noted for mutations at other interdomain GluN1 residues (Farina et al., 2011).

To evaluate the net impact of GluN1-P532H variant on receptor function, we estimated the functional consequence of all measured changes of multiple parameters that resulted from this variant on synaptic and nonsynaptic charge transfer to wild-type receptors (Swanger et al., 2016; Li et al., 2019). This calculation indicated that the GluN1-P532H variant reduces synaptic and nonsynaptic charge transfer by 4.5- and 250-fold, respectively (Table 1). Overall, these data suggested that GluN1-P532H is a LoF variant and may induce NMDAR hypofunction as a result of a shortened synaptic response time course, a decreased channel open probability, and a reduced receptor cell surface expression levels.

To investigate the structural effects of GluN1-P532H mutation, we performed MD simulation using the structure of the dimeric glycine and glutamate-binding domains of human NMDA (GluN1/GluN2A) receptor (PDB ID 5H8Q) (Figure 4A). Snap shots of the wild-type and mutant protein structure in the MD simulation trajectories are shown in Figure 4B. The structures show that the histidine in the mutant protein at site 532 in GluN1 is within 2.6 Å of the hydroxyl group of threonine 748 on the lower lobe of GluN2A. Consistent with the interaction across the dimer, there is a motion of the lower lobe of GluN2A ABD toward the GluN1 ABD and an opening of GluN2 ABD cleft as seen in the distance between the Cα atoms of sites Q503 and M701 (Figure 4C). No significant changes are observed in the glycine-binding cleft between wild-type and GluN1-P532H mutant, as seen in the Cα atom distance between sites Ser507 and Thr701 (Figure 4D). The changes in the ABD clefts are consistent with the functional measurements that show reduced glutamate potency but only slight changes in the glycine potency between the mutant and wild-type GluN1/GluN2A receptors.

To confirm the results from the MD simulations showing a more open GluN2 glutamate-binding domain cleft in the GluN1-P532H mutant, we investigated the distance across the ABD cleft in the GluN2 subunit using smFRET. Figures 4E,F depicts the normalized cumulative histograms generated from efficiency traces of 25–30 molecules for the wild-type and GluN1-P532H bound to glutamate and glycine. Concatenated FRET efficiency traces showing the population of the different FRET states shown in the smFRET histograms are provided in the Supplementary Figures 1–5. The smFRET histograms show a shift in the population to lower FRET states in the GluN1-P532H mutant as compared with wild-type receptor. The low FRET states correspond to longer distances across the agonist-binding cleft, and hence more open cleft states, confirming the MD simulations. The more open cleft state is consistent with the decrease in glutamate potency observed in receptors containing the GluN1-P532H mutant relative to the wild-type protein.

We tested whether the NMDAR hypofunction caused by GluN1-P532H variants can be rectified by a set of positive allosteric modulators (PAMs) and coagonists (Tang et al., 2020). 24(S)-Hydroxycholesterol [24(S)HC], an endogenous cholesterol metabolite in the brain, can enhance the current response of NMDARs (Linsenbardt et al., 2014; Wilding et al., 2016). Pregnenolone sulfate (PS) can enhance neuronal NMDAR with selectivity for GluN2A- and GluN2B (Wu et al., 1991; Ceccon et al., 2001; Malayev et al., 2002; Horak et al., 2004). Tobramycin, an aminoglycoside antibiotic, can selectively potentiate GluN2B-containing NMDAR function (Swanger et al., 2016). Figures 5A–C shows that all three PAMs significantly potentiate the current response to maximally effective agonists (1 mM glutamate and 100 μM glycine) for both wild-type and GluN1-P532H-containing NMDARs (Table 2). We evaluated the potentiation by these three PAMs over a range of holding potentials (−90 to 40 mV) in the absence of extracellular Mg²⁺. Figures 5D,E shows the current-voltage relationships for NMDAR current responses to a maximally effective concentration of agonists coapplied with the indicated concentrations of each PAMs. We observed similar potentiating effects for coapplication of these PAMs for NMDARs that contained either WT or GluN1-P532H at both negative and positive holding potentials (Figures 5D,E and Table 2). In addition, three coagonists at glycine site, D-serine, L-serine, and D-cycloserine, were evaluated. These three coagonists showed a similar potency for WT and variant GluN1-P532H-containing NMDARs, with the exception of L-serine at GluN1-P532H/GluN2B (Figures 5F–H and Table 2). This result suggests that these PAMs and coagonists might enhance the current response caused by the LoF variant.