Differential responses of disease-related GRIN variants located in pore-forming M2 domain of N-methyl-D-aspartate receptor to FDA-approved inhibitors

Abstract

N-methyl-D-aspartate receptors (NMDAR), ionotropic glutamate receptors, mediate a slow component of excitatory synaptic transmission in the central nervous system and play a key role in normal brain function and development. Genetic variations in GRIN genes encoding NMDAR subunits that alter the receptor’s functional characteristics are associated with a wide range of neurological and neuropsychiatric conditions. Pathological GRIN variants located in the M2 re-entrant loop lining the channel pore cause significant functional changes, the most consequential alteration being a reduction in voltage-dependent Mg²⁺ inhibition. Voltage-dependent Mg²⁺ block is unique feature of NMDAR biology whereby channel activation requires both ligand binding and postsynaptic membrane depolarization. Thus, loss of NMDAR Mg²⁺ block will have profound impact on synaptic function and plasticity. Here, we choose eleven missense variants within the GRIN1, GRIN2A and GRIN2B genes that alter residues located in the M2 loop and significantly reduce Mg²⁺ inhibition. Each variant was evaluated for tolerance to genetic variation using the 3-dimensional structure and assessed for functional rescue pharmacology via electrophysiological recordings. Three FDA-approved NMDAR drugs – memantine, dextromethorphan, and ketamine – were chosen based on their ability to bind near the M2 re-entrant loop, potentially rectifying dysregulated NMDAR function by supplementing the reduced voltage-dependent Mg²⁺ block. These results provide insight of structural determinants of FDA-approved NMDAR drugs at their binding sites in the channel pore and may further define conditions necessary for use of such agents as potential rescue pharmacology.

Introduction

N-methyl-D-aspartate receptors (NMDARs) are ligand-gated glutamatergic ion channels that mediate a slow component of excitatory synaptic neurotransmission in the central nervous system. NMDARs play critical roles in normal brain function, including neurodevelopment, learning, memory, motor, and sensory function (Hansen et al., 2021; Traynelis et al., 2010). The hetero-tetrameric NMDARs are composed of two glycine-binding GluN1 subunits and two glutamate-binding GluN2 subunits (Hansen et al., 2021; Traynelis et al., 2010). GluN1 is the product of a single GRIN1 gene with eight splice variants and is expressed throughout the brain The GluN2 subunits (GluN2A-2D), are encoded by four genes (GRIN2A-2D), show unique spatio-temporal expression profiles as well as distinct pharmacological and biophysical properties (Akazawa, Shigemoto, Bessho, Nakanishi, & Mizuno, 1994; Hansen et al., 2021; Jantzie et al., 2015; Law et al., 2003; Monyer, Burnashev, Laurie, Sakmann, & Seeburg, 1994; Paoletti, Bellone, & Zhou, 2013; Traynelis et al., 2010; Watanabe, Inoue, Sakimura, & Mishina, 1993; Wyllie, Livesey, & Hardingham, 2013). All NMDAR subunits share a similar architecture, containing four semi-autonomous domains: an extracellular amino-terminal domain (ATD, also known as NTD), a bi-lobed agonist binding domain (ABD, also known as LBD, ligand binding domain), a pore-forming transmembrane domain comprising three transmembrane helices (TMD; M1, M3, M4), a re-entrant loop (M2), and an intracellular carboxy-terminal domain (CTD) (Hansen et al., 2021; Traynelis et al., 2010). Activation of NMDARs requires both agonists (glutamate and glycine) to bind and simultaneous postsynaptic membrane depolarization to dispel Mg²⁺ ions lodged in the pore, resulting in current flow through the open cation-selective pore (Hansen et al., 2021; Traynelis et al., 2010).

Next-generation whole exome sequencing have identified a large number of de novo missense variants in GRIN genes in a wide range of pathological conditions (see (Burnashev & Szepetowski, 2015; Hansen et al., 2021; Hu, Chen, Myers, Yuan, & Traynelis, 2016; Soto, Altafaj, Sindreu, & Bayes, 2014; XiangWei, Jiang, & Yuan, 2018; Yuan, Low, Moody, Jenkins, & Traynelis, 2015). These rare missense variants were identified in all NMDAR GluN subunits and located across all NMDAR domains, with significant numbers in ABD and TMD (Hansen et al., 2021; XiangWei et al., 2018). Among them, the disease-associated missense variants in the pore-forming re-entrant loop M2 have been associated with a number of neurological and neuropsychiatric disorders (Chen et al., 2017; Endele et al., 2010; Farwell et al., 2015; Hamdan et al., 2014; Lemke et al., 2016; Lemke et al., 2014; Li et al., 2019; Moller et al., 2016; Platzer et al., 2017; Retterer et al., 2016; Strehlow et al., 2019; von Stulpnagel et al., 2017). Moreover, the majority of missense variants found within the M2 re-entrant loop are associated with various levels of developmental delay, intellectual disability, and epilepsy which may be due in part to their diminished levels of voltage-dependent Mg²⁺ block (see Supplemental Table S1). Within a synapse, Mg²⁺ block of NMDARs prevents current flow unless the postsynaptic membrane is sufficiently depolarized, regardless of whether agonist is bound (Jahr & Stevens, 1990; Mayer, Westbrook, & Guthrie, 1984; Nowak, Bregestovski, Ascher, Herbet, & Prochiantz, 1984; Premkumar & Auerbach, 1996). This feature of NMDAR biology enables coincidence detection, which is important for modulating bidirectional synaptic plasticity (Calabresi, Pisani, Mercuri, & Bernardi, 1992; Coan, Irving, & Collingridge, 1989; Kampa, Clements, Jonas, & Stuart, 2004; Yuste, Majewska, Cash, & Denk, 1999).

In this present study, we focus on 11 de novo missense variants located in the M2 re-entrant loop within the GRIN1, GRIN2A, and GRIN2B genes. These variants were identified in patients with epilepsy syndrome, developmental delay, intellectual disability, hypotonia, autism, and/or speech disorder. Additionally, each of these variants display a significantly reduced voltage-dependent Mg²⁺ block (Chen et al., 2017; Fedele et al., 2018; Li et al., 2019; Marwick, Skehel, Hardingham, & Wyllie, 2019; Vyklicky et al., 2018) (see Supplemental Table S1). Using 3-dimensional structurally derived missense tolerance ratio (3DMTR) (Perszyk, Kristensen, Lyuboslavsky, & Traynelis, 2021), we show that the M2 re-entrant loop variants are found within highly intolerant regions of the receptor. We also investigated whether a set of FDA-approved NMDAR inhibitors – memantine, dextromethorphan, and ketamine – could rectify the reduced Mg²⁺ inhibition caused by these variants as a first step to evaluate their potential use as rescue pharmacological agents.

Materials and Methods

Three-dimensional missense tolerance ratio analysis

Analysis of genetic variation on the M2 domain in GRIN1, GRIN2A, and GRIN2B were performed using three-dimensional tolerance ratio analysis (3DMTR) (Perszyk et al., 2021). 3DMTR analysis was performed using a GUI based analysis application (https://github.com/riley-perszyk/3DMTR) which was generated using MATLAB (Mathworks, Natick, MA). 3DMTR calculates the missense tolerance ratio for each residue in the protein structure. To convert the sporadic data into a less noisy MTR score, a smoothing window is applied in the 3DMTR calculation that includes the 30 nearest neighboring residues foreach resisdue. This 31-residue (central residue plus the surrounding 30 residues) score provides a metric indicating how tolerated variation is in the general vicinity of a specific residue. When a residue has a value of 0, it means there are no missense variants in that specific residue or in the 30 nearest neighboring residues. Variance data is obtained from the gnomAD database (https://gnomad.broadinstitute.org/, Version 2.1.1 non-neuro subset, accessed 5–1-2023) that is a repository containing hundreds of thousands of whole genome/exome data from healthy individuals. Two tetrameric receptors were analyzed, GluN1/GluN2A and GluN1/GluN2B, using models (Perszyk et al., 2021) based on the non-active GluN1/GluN2B structure (pdb: 6WHS, (Chou, Tajima, Romero-Hernandez, & Furukawa, 2020). Depictions of the receptor structure were created with pymol after generating (using the MATLAB application) pdb files with the 3DMTR score loaded into the b-factor data column.

Site-directed Mutagenesis

Human plasmids cDNAs encoding GluN1–1a (hereafter GluN1; GenBank accession numbers NP_015566), GluN2A (NP_000824), and GluN2B (NP_000825) were used for experiments involving di-heteromeric variants or WT GluN1/GluN2 NMDARs. The cDNA fragments of the full open reading frames for human NMDAR subunits were obtained from the I.M.A.G.E. Consortium (Carlsbad, CA) and Origene (Rockville, MD) and were assembled and subcloned into the mammalian expression vector pCI-neo (U47120) (Hedegaard, Hansen, Andersen, Brauner-Osborne, & Traynelis, 2012). The GenBank accession numbers for rat cDNAs were U08261 for GluN1–1a (hereafter GluN1), D13211 for GluN2A, and U11419 for GluN2B. Missense mutations were introduced into NMDAR cDNAs using the Quikchange protocol from Stratagene (Chen et al., 2017). For expression in Xenopus oocytes, cDNA constructs were linearized by enzyme Not I (NEB, catalog# R0189S, Ipswich, MA) and used as templates for cRNA synthetization in vitro (Ambion, catalog# AM1626, Austin, TX) (Chen et al., 2017).

Two electrode voltage clamp (TEVC) current recordings from Xenopus oocytes

TEVC current recordings were performed on unfertilized, defolliculated stage V-VI Xenopus laevis oocytes acquired directly from EcoCyte Bioscience (Austin, TX), which were injected with cRNA for WT or variant receptors. Following injection, oocytes were maintained at 15–19 °C in Barth’s culture medium containing (in mM) 88 NaCl, 2.4 NaHCO₃, 1 KCl, 0.33 Ca(NO₃)₂, 0.41 CaCl₂, 0.82 MgSO₄, and 5 Tris-HCl (pH 7.4 with NaOH) supplemented with antibiotics (0.1 mg/mL gentamicin sulfate, 1 μg/mL streptomycin, and 1 U/mL penicillin). After 2–4 days post-injection, oocytes were continuously perfused with extracellular oocyte recording solution containing (in mM) 90 NaCl, 1 KCl, 10 HEPES, 0.5 BaCl₂, and 0.01 EDTA (except for the experiments of Mg²⁺ sensitivity) (pH 7.4 with NaOH). Solution exchange was accomplished by a computer-controlled 8-modular valve positioner (Digital MVP Valve, Hamilton, CT). Electrodes were prepared from borosilicate glass (World Precision Instruments, catalog #TW150F-4) by a dual-stage glass micropipette puller (PC-10, Narishige, Tokyo, Japan). Voltage clamp was achieved, and current responses were recorded at a holding potential of −40 mV (unless otherwise stated) with voltage and current electrodes filled with 0.3 M and 3.0 M KCl, respectively, and a two-electrode voltage clamp amplifier (Warner Instruments model OC-725C, Hamden, CT). Currents were filtered with a low-pass filter at 10 Hz and digitized at 20 kHz by LabWindows/CVI (National Instruments, Austin, TX). Maximal concentrations of glutamate (100 μM) and glycine (100 μM) were used in all oocyte experiments unless otherwise stated. Potency (IC₅₀ values) for FDA-approved NMDAR-targeted channel blockers (memantine, Sigma-Aldrich, catalog# M9292; dextromethorphan, Sigma-Aldrich, catalog# 1180503; ketamine, Sigma-Aldrich, catalog# K2753) were obtained by fitting the concentration-response curves data according with

$$\text{Response}\left(\%\right)=\left(100-\text{minimum}\right)/\left(1-{\left(\left[\text{modulator}\right]/{\text{IC}}_{50}\right)}^{\text{nH}}\right)+\text{minimum}$$ Equation 1

where IC₅₀ is the concentration that produces a half-maximal effect and minimum is the degree of residual inhibition at a saturating concentration of Mg²⁺ or FDA-approved NMDAR channel blockers.

The current-voltage curves were generated by applying voltage steps (−90 mV to +30 mV with 15 mV step) in Xenopus oocytes expressing WT or variant receptors. The relative response at each holding potential was normalized to the current at +30 mV. The voltage dependence and blocker affinity at 0 mV can be assessed by fitting the current-voltage curve obtained in the presence of 1 mM Mg²⁺ with the Woodhull equation (Li et al., 2019; Woodhull, 1973)

$$I_{\mathit{\text{UNBLOCKED}}}\left(V\right)=I\left(V-V_{rev}\right)/\left(1+{\left[{\text{Mg}}^{2+}\right]}_{\text{o}}/K_{D,0mV}\text{\hspace{0.17em}exp}\left(z\delta VF/RT\right)\right)$$ Equation 2

where I_UNBLOCKED is the current response at a given holding potential in the presence of 1 mM Mg²⁺, I is the current response in the absence of Mg²⁺ at a given holding potential, [Mg²⁺]_o is the extracellular Mg²⁺ concentration (1000 μM), K_{D,0 mV} is the K_D in the absence of an applied electric field, z is the valence for Mg²⁺ (z = 2), δ is the effective fraction of the electric field at the binding site, V is the voltage, Vrev is the reversal potential, F is Faraday constant (96,485 C/mol), R is gas constant (8.31 J·K⁻¹·mol⁻¹), and T is temperature. The potential permeation by the blocking ion for this equation was not accounted for. In the presence of 1 mM Mg²⁺ plus channel blockers ketamine or memantine, we determined an apparent K_{D,0 mV} and zδ assuming both Mg²⁺ and channel blocker follow to a first approximation the assumptions necessary for equation 2. In this case, the K_{D,0 mV} and zδ will reflect a weighted apparent value for the combined block by both agents and provide insight into how the channel blockers supplement voltage-dependent Mg²⁺ block.

All in vitro studies were conducted according to the guidelines of Emory University Health and Safety Office and the Institutional Review Board. All reagents were purchased from Sigma (unless otherwise stated). For all experiments, data were expressed as mean ± SEM or mean [± 95 confidence interval] and analyzed statistically using unpaired t-test (two tailed) or one-way ANOVA with Dunnett’s multiple comparison. Significance for all tests was set at p < 0.05. Error bars in all figures represent SEM. The number of independent experiments is represented by n. The number of samples was calculated to have a power to detect a 50% change greater than 0.8. Samples sizes were determined by a priori power analysis for effect size = 2, power = 0.8, and α = 0.05.

Results

Intolerance to variation of M2 re-entrant loop in GRIN1, GRIN2A, and GRIN2B

The M2 re-entrant loop is a small but critical component of the NMDAR (Figure 1A–C). The M2 is comprised of 21 residues in GluN2 whereas the GluN1 M2 is comprised of 22 residues. In a NMDAR tetramer, the M2 loops line a portion of the channel pore and control permeation and gating properties (Hansen et al., 2018). The 3DMTR score (see Methods) uses a protein structure and the gnomAD database (Version 2.1.1 accessed 5–1-2023) to compute how tolerant specific residues are to missense variation in the general human population. Using the “non-neuro” gnomAD dataset and a smoothing function of 31 residues, we see that the M2 loop, and in particular the apex of the M2 loop, are highly intolerant across all NMDAR subunits. The average 3DMTR score (Perszyk et al., 2021) for all residues represented in the re-entrant loop were 0.07 for GluN1, 0.36 for GluN2A, 0.23 for GluN2B, and 0.33 for GluN2D, with GluN2C being the exception with a tolerant 3DMTR score of 0.65. In all cases the M2 segment is more intolerant than the overall average 3DMTR score for all residues in each subunit (GluN1, 0.46; GluN2A, 0.53; GluN2B, 0.34; GluN2C, 0.80; GluN2D, 0.63), suggesting that the M2 plays a critical role in receptor function that is necessary for normal brain function. In each NMDAR subunit there are several residues that have a 3DMTR score equal to 0, meaning that there are no missense variants reported in each specific residue or in their 30 nearest neighboring residues. In GluN1 there are 15 residues with a 3DMTR score of 0, in GluN2A and GluN2B there are 3 each, and in GluN2D there are 5 (there are no residues in the GluN2C M2 that have a 3DMTR score of 0). The deleterious de novo variants described here (Li et al., 2019) are found in these highly intolerant portions of the M2 segment (Supplementary Table S1; Figure 1 C–E).

Figure 1. Genetic tolerance of the GRIN variants located in the pore-forming M2 domain.

(A) A linear representation of the semi-autonomous domains of a single GluN subunit (CTD not shown). (B) Image of a model of the GluN1/GluN2 extracellular and transmembrane domains (the semiautonomous domains are labeled matching the linear map shown in A) and a view of the TMD (showing only two subunits, one GluN1 and one GluN2) with the M2 segment highlighted. (C) Sequence alignment of the M2 segment of GluN1 and GluN2 subunits. Residues marked by a box below the amino acid letter ID denote instances of synonymous (green) or missense (orange) variants in the gnomAD database (v2.1.1, non-Neuro). The 3DMTR score (closest 31 residues, non-Neuro) is illustrated for each residue using the color scale, shown in the left of panel A. A dark gray box in the 3DMTR row is used when the residue is not present in the receptor model. (D) Side view (same orientation as in panel B) of the TMD showing the M2 variants tested (residue side chains are shown as spheres, D₁, GluN2A; D₂, GluN2B), the two closer GluN1 and GluN2 subunits have been removed to highlight the pore lining sides of M2 segments. The left images show the subunit coloring used in panel B, and the right images show the 3DMTR score using the same color scale as in panel C. (E) Bottom view (view depicted by the eye cartoon shown in panel B) of the TMD showing the M2 variants tested (residue side chains are shown as spheres, E₁, GluN2A; E₂, GluN2B). The left images show the subunit coloring used in panel B, and the right images show the 3DMTR score using the same color scale as in panel C.

Pharmacology on M2 rare variants

All eleven M2 variants evaluated showed a reduction in voltage-dependent Mg²⁺ block, which should strongly drive increased charge transfer when NMDARs are activated. In addition, because patients’ seizures associated with several M2 variants (e.g. GluN2A-L611Q) were not fully controlled by conventional anti-epileptic drugs, a number of FDA-approved NMDAR channel blockers (memantine, dextromethorphan, and ketamine) were evaluated for their ability to offset the reduced Mg²⁺ block of these variant receptors, both in the absence and presence of physiological concentrations of Mg²⁺ (i.e., 1.0 mM). In vitro analyses by TEVC recordings from Xenopus oocytes at a holding potential of −40 mV revealed that the M2 variants have differential sensitivity (potency, IC₅₀) to these channel blockers. In the absence of Mg²⁺, memantine showed enhanced potency on GluN1-G620R- and GluN2B-N615I-containing NMDARs (2.7-fold and 4.3-fold, respectively), but decreased potency by 2- to 13-fold on GluN2A-L611Q, GluN2A-N614S, GluN2B-W607C, GluN2B-N616K, and GluN2B-V618G (Figure 2A,B, upper panels; Table 1). The antitussive dextromethorphan showed an enhanced potency on GluN1-G620R-, GluN2A-N615K, and GluN2B-V620M-containing NMDARs in comparison to WT NMDARs (2.5-fold, 2.7-fold, and 2.7-fold, respectively). The potency of dextromethorphan was reduced by 2- to 9-fold on GluN2A-L611Q, GluN2A-N614S, GluN2B-N615I, and GluN2B-V618G (Figure 2C,D, upper panels; Table 1). The anesthetic ketamine displayed enhanced potency on GluN1-G620R-, GluN2B-N615I-, and GluN2B-N616K-containing NMDARs in comparison to WT NMDARs (2.2-fold, 8.1-fold, and 1.8-fold, respectively). Ketamine was less potent by ~3-fold on GluN2A-N614S, GluN2B-W607C, and GluN2B-V618G (Figure 2E,F, upper panels; Table 1). The GluN2B-N615K variant almost abolishes the sensitivity to all three channel blockers evaluated, with 24-fold or larger reduced potency. This led to only 14–22% of percentage inhibition by saturating 100–300 μM concentrations of blockers compared to 95–97% of WT receptors (Figure 4; Table 2; Supplemental Figure S1).

Figure 2. The M2 domain variants influence memantine sensitivity.

A,B,C, Representative current traces were generated by TEVC recording (V_HOLD: −40 mV) from Xenopus oocytes to evaluate the effects of different concentrations of memantine (0.3 to 100 μM) on agonist-evoked currents (GG: 100 μM glutamate and 100 μM glycine) of WT GluN1/GluN2B (A), GluN1/GluN2B-N615I (B), and GluN1/GluN2B-N615K (C) in the absence (left panels) and presence (right panels) of I mM Mg²⁺. D,E, Composite concentration-response curves of memantine on WT GluN1/GluN2A, GluN1 and GluN2A variants (D) and on WT GluN1/GluN2B and GluN2B variants (E) in the absence (left panels) and presence (right panels) of 1.0 mM Mg²⁺. Number of oocytes recorded: n = 7–30. Fitted values are provided in Table 1.

Table 1.

Summary inhibitor potency for potential rescue pharmacology

	Memantine IC50, μM		Dextromethorphan IC50, μM		Ketamine IC50, μM
	w/o Mg2+	w/ 1 mM Mg2+	w/o Mg2+	w/ 1 mM Mg2+	w/o Mg2+	w/ 1 mM Mg2+
WT GluN1/WT 2A	4.3 [3.3, 5.4] (15)	11 [8.1, 14] (30)$	11 [7.3, 14] (19)	25 [12, 37] (10)	5.5 [4.1, 6.9] (25)	17 [10, 23] (24)$
1-G620R/2A	1.6 [1.2, 2.0] (8)&	0.66 [0.5, 0.8] (8)$&	4.4 [3.2, 5.6] (14)&	4.4 [2.1, 6.7] (8)&	2.5 [1.8, 3.3] (12)&	1.0 [0.8, 1.1] (7)$&
2A-L611Q	9.3 [6.8, 12] (12)&	3.0 [2.2, 3.8] (9)$&	27 [22, 31] (16)&	36 [27, 45] (8)	7.2 [5.4, 8.9] (17)	4.0 [3.0, 5.0] (10)$&
2A-N614S	40 [18, 61] (9)&	>100 (10)$&	22 [12, 32] (10)	32 [24, 40] (11)	29 [19, 39] (12)&	31 [16, 47] (9)
2A-N615K	12 [6.6, 17] (7)&	22 [17, 28] (8)$&	3.8 [2.9, 4.6] (8)&	3.0 [1.9, 4.1] (12)&	4.3 [2.7, 5.8] (18)	1.6 [1.1, 2.1] (7)$&
WT GluN1/WT 2B	1.5 [1.1, 1.8] (18)	4.1 [3.4, 4.9] (26)$	4.8 [2.7, 6.9] (15)	22 [16, 28] (18)$	4.2 [2.7, 5.6] (13)	14 [12, 17] (12)$
2B-W607C	6.9 [4.7, 9.1] (10)&	7.1 [1.4, 13] (11)	3.7 [2.5, 4.9] (11)	11 [6.3, 16] (12)$&	16 [8.8, 23] (12)&	11 [4.8. 16] (9)
2B-G611V	1.9 [1.3, 2.6] (10)	3.1 [1.2, 5.0] (14)	3.6 [1.7, 5.4] (10)	6.2 [4.2, 8.2] (10)&	3.1 [1.2, 5.0] (7)	8.0 [6.2, 9.8] (16)$&
2B-N615I	0.35 [0.3, 0.4] (11)&	0.57 [0.3, 0.8] (13)&	15 [13, 16] (14)&	17 [14, 20] (11)	0.51 [0.2, 0.8] (10)&	0.76 [0.2, 1.4] (8)&
2B-N615K	>100 (14)&	>100 (21)&	>300 (8)&	>300 (18)&	>100 (17)&	>100 (16)&
2B-N616K	6.4 [3.7, 9.1] (7)&	13 [8.0, 17] (24)&	3.2 [2.2, 4.2] (10)	4.5 [3.0, 6.1] (7)&	2.3 [1.4, 3.1] (8)	2.5 [1.9, 3.1] (13)&
2B-V618G	19 [16, 21] (16)&	13 [8.1, 17] (14)&	45 [38, 52] (11)&	71 [43, 99] (10)&	18 [12, 24] (8)&	24 [18, 29] (8)
2B-V620M	4.0 [3.0, 5.0] (8)&	4.4 [3.5, 5.3] (7)	1.8 [1.1, 2.5] (8)&	3.5 [2.0, 5.1] (8)&	3.4 [1.9, 5.0] (10)	7.0 [5.2, 8.7] (8)$&

Data were generated from TEVC recordings (holding at −40 mV) on Xenopus oocytes expressing WT and variant NMDARs that were activated by maximally effective concentrations of glutamate and glycine (100 μM). IC₅₀ values are expressed as mean [± 95% CI] (n). CI: confidence interval. n: number of oocytes recorded. Data are from oocytes isolated from 2 or more frogs.

^$indicates 95% confidence intervals in the presence of 1 mM Mg²⁺ that are non-overlapping with the same variant (or WT) in the absence of 1 mM Mg²⁺.

^&indicates 95% confidence intervals that are non-overlapping with the corresponding WT receptors

Figure 4. Percentage inhibition of maximal concentration of FDA-approved NMDAR blockers in the presence of Mg²⁺.

The plots of percent inhibition are the ratio of the current response at a maximally effective concentration of memantine (100 μM; A,B), dextromethorphan (300 μM; C,D), and ketamine (100 μM; E,F) to the agonist-evoked current (100 μM glutamate and 100 μM glycine) in the presence of I mM Mg²⁺ at holding potential of −40 mV. Number of oocytes recorded: 7–30. The data are expressed as mean ± SEM. The percentage inhibition values are provided in Table 2.

Table 2.

Summary of % inhibition observed for potential rescue pharmacology

	% inhibition by 100 μM memantine		% inhibition by 300 μM dextromethorphan		% inhibition by 100 μM ketamine
	w/o Mg2+	w/ 1 mM Mg2+	w/o Mg2+	w/ 1 mM Mg2+	w/o Mg2+	w/ 1 mM Mg2+
WT GluN1/WT 2A	94 ± 1.2% (15)	89 ± 1.3% (30)*	95 ± 1.6% (19)	88 ± 3.3% (10)*	92 ± 1.1% (25)	79 ± 2.4% (24)*
1-G620R/2A	96 ± 1.3% (8)	98 ± 0.4 (8)	97 ± 1.4% (14)	96 ± 1.5% (8)¥	95 ± 1.8 (12)	97 ± 0.6 (7)¥
2A-L611Q	87 ± 2.1% (12)	94 ± 0.6% (9)*	91 ± 1.4% (16)	89 ± 2.3% (8)	87 ± 1.5% (17)	93 ± 2.1% (10)*¥
2A-N614S	60 ± 4.1% (9)¥	25 ± 6.1% (10)*¥	85 ± 3.0% (10)¥	85 ± 2.7% (11)	78 ± 3.0% (12)¥	77 ± 4.1% (9)
2A-N615K	90 ± 2.5% (7)	81 ± 1.6% (8)*¥	98 ± 1.4% (8)	97 ± 0.9% (12)¥	93 ± 1.5% (18)	92 ± 3.3% (7)¥
WT GluN1/WT 2B	97 ± 1.1% (18)	93 ± 0.8% (26)*	95 ± 1.2% (15)	90 ± 1.9% (18)*	95 ± 1.0% (13)	86 ± 1.0% (12)*
2B-W607C	91 ± 2.0% (10)	92 ± 2.8% (11)	90 ± 2.8% (11)	95 ± 1.4% (12)	80 ± 3.6% (12)¥	89 ± 2.8% (9)
2B-G611V	95 ± 1.4% (10)	97 ± 1.2% (14)	94 ± 2.3% (10)	97 ± 1.0% (10)	92 ± 2.6% (7)	88 ± 1.6% (16)
2B-N615I	96 ± 1.1% (11)	98 ± 0.7% (13)	95 ± 1.4% (14)	94 ± 1.5% (11)	89 ± 1.9 (10)	94 ± 1.8% (8)¥
2B-N615K	22 ± 2.5% (14)¥	12 ± 1.9% (21)*¥	14 ± 2.7% (8)¥	33 ± 3.7% (18)*¥	22 ± 3.8 (15)¥	6.5 ± 1.5% (16)*¥
2B-N616K	93 ± 1.8% (7)	88 ± 1.6% (24)¥	94 ± 1.7% (10)	93 ± 1.6% (7)	91 ± 2.7% (8)	93 ± 1.3% (13)¥
2B-V618G	85 ± 1.5% (16)¥	87 ± 2.1% (14)¥	85 ± 1.7% (11)¥	83 ± 4.0% (10)	77 ± 3.0% (8)¥	81 ± 2.1% (8)
2B-V620M	96 ± 0.7% (8)	95 ± 0.7% (7)	93 ± 3.2% (8)	87 ± 3.2% (8)	93 ± 1.0% (10)	91 ± 2.8% (8)

Data were generated from TEVC recordings (holding at −40 mV) on Xenopus oocytes expressing WT and variant NMDARs that were activated by maximally effective concentrations of glutamate and glycine (100 μM). Percentage inhibition by maximal concentrations of FDA-approved NMDAR inhibitor (100 μM memantine, 300 μM dextromethorphan, 100 μM ketamine) is expressed as mean ± SEM (n). n: number of oocytes recorded. Data are from oocytes isolated from 2 or more frogs.

^*p < 0.05, unpaired Student t-test (two tailed), in the presence of 1 mM Mg²⁺ that are non-overlapping with the same variant (or WT) in the absence of 1 mM Mg²⁺.

^¥p < 0.05, one way ANOVA with Dunnett’s multiple comparison, compared with the corresponding WT in the presence or absence of 1 mM Mg²⁺. The full statistical reports have been provided as supplemental materials (Supplemental Table-S2,S3)

We subsequently repeated the experiments in the presence of 1 mM extracellular Mg²⁺. For all three channel blockers evaluated, the WT receptors showed 2- to 5-fold reduced potency in the presence of 1 mM Mg²⁺ (Kotermanski & Johnson, 2009). GluN1-G620R-, GluN2A-L611Q-, and GluN2B-N615I-containing NMDARs showed an enhanced potency to memantine (by 17-fold, 3.7-fold, and 7.2-fold, respectively) in the presence of 1 mM Mg²⁺ compared to the WT receptors, whereas GluN2A-N614S, GluN2A-N615K, GluN2B-N615K, GluN2B-N616K, and GluN2B-V618G reduced memantine potency by 2- to 24-fold (Figure 2A,B, lower panels; Table 1). In the presence of 1 mM Mg²⁺, dextromethorphan showed an enhanced potency on GluN1-G620R-, GluN2A-N615K-, GluN2B-G611V-, GluN2B-N616K-, and GluN2B-V620M-containing NMDARs by 5.7-fold, 8.3-fold, 3.5-fold, 4.9-fold, and 6.3-fold, respectively. Dextromethorphan showed reduced potency at GluN2B-N616K (14-fold) and GluN2B-V618G (3.2-fold; Figure 2C,D, lower panels; Table 1). Ketamine showed 4- to 17-fold enhanced potency on GluN1-G620R-, GluN2A-L611Q-, GluN2A-N615K-, GluN2B-N615I-, and GluN2B-N616K-containing NMDARs in the presence of 1 mM Mg²⁺ in comparison to WT NMDARs. Ketamine displayed over 5.9-fold reduced potency on GluN2B-W607C, and GluN2B-N615K (Figure 2E,F, lower panels; Table 1). In the presence of 1 mM Mg²⁺, the GluN2A-N614S variant significantly reduced the sensitivity to memantine by 9-fold (Figure 2D; Tables 1). Memantine (100 μM) produced 25% percentage inhibition on GluN2A-N614S variant-containing NMDARs compared to 89% of WT receptors (Tables 2). The GluN2B-N615K variant largely abolished the sensitivity to all three channel blockers, with over 7- to 24-fold reduced potency and 6.5–33% of percentage inhibition at 100–300 μM compared to 86–93% of WT receptors (Figure 3; Tables 1,2). In addition, the physiological concentration of Mg²⁺ makes GluN1-G620R- and GluN2A-L611Q-containing NMDARs more sensitive to both memantine (increased potency by 2.4-fold and 3.1-fold, respectively) and ketamine (increased potency by 2.5-fold and 1.8-fold, respectively) compared to the same variant in the absence of Mg²⁺. GluN2A-N616K was more sensitive to ketamine only (increased potency by 2.7-fold) in Mg²⁺ (Table 1).

Figure 3. The M2 domain variants influence dextromethorphan and ketamine sensitivity.

Composite concentration-response curves of dextromethorphan (A,B), and ketamine (C,D) in the absence (upper panels) and presence (lower panels) of 1.0 mM Mg²⁺ were obtained from TEVC recordings from Xenopus oocytes in the presence of 100 μM glutamate and 100 μM glycine at holding potential of −40 mV. Number of oocytes recorded: 7–25. Fitted values are provided in Table 1.

Interestingly, all seven GRIN2B/GluN2B variants showed a comparable sensitivity to the three channel blockers in the absence and presence of 1 mM Mg²⁺, except for GluN2B-W607C (reduced potency to dextromethorphan by 3-fold in 1 mM Mg²⁺), GluN2B-G611V (reduced potency to ketamine by 2.6-fold in 1 mM Mg²⁺), and GluN2B-V620M (reduced potency to ketamine by 2.1-fold in 1 mM Mg²⁺) (Table 1, Figure 3).

Effect of NMDAR inhibitors on voltage-dependent channel block

We assessed whether FDA-approved drugs could supplement the reduced Mg²⁺ block and restore the current-voltage relationship for variant NMDARs closer to that observed for Mg²⁺ acting on WT NMDARs. We selected two variants that were associated with medically refractory epilepsy and showed similar or increased potency to therapeutic agents. We evaluated the current-voltage relationship for GluN2A-L611Q and GluN2B-V620M, both of which showed reduced Mg²⁺ sensitivity, enhanced glutamate potency, prolonged deactivation rate, and increased channel open time (Li et al., 2019). Current responses were recorded under different holding voltages for Mg²⁺ alone and for Mg²⁺ plus increasing concentrations of ketamine for GluN2A-L611Q and memantine for GluN2B-V620M (Figure 5A,B). These two drugs were selected for evaluation because they showed comparable or enhanced inhibition on the corresponding variants compared to the WT receptor in the presence of Mg²⁺. We evaluated the nature of the combined block by fitting the current-voltage curves with the Woodhull equation (see Methods), which incorporated voltage-dependent term for the blocker affinity and an estimation of the depth of the electric field sensed within the pore. This equation was developed for a single blocking species (Woodhull, 1973), however when two blockers are present, the results should be a weighted average of their combined actions. We assessed both the K_{D,0 mV} and fraction of the electric field sensed by the blocking ions (zδ) as an indication of whether these organic cation blockers could re-introduce similar block and similar voltage dependence as Mg²⁺ at WT receptors.

Figure 5. Ketamine and memantine can offset dismissed magnesium sensitivity caused by the two M2 domain variants.

The effects of increasing concentrations of ketamine (A,B) and memantine (C,D) on the current-voltage (I-V) curves in the presence of 1 mM Mg²⁺ on GluN1/GluN2A-L611Q (A,B) and GluN1/GluN2B-V620M (C,D) receptors, respectively. The data were expressed as mean ± SEM. The relative response at each holding potential was normalized to the current at +30 mV. Smooth lines are the fit of equation 2 to the data. Current-voltage curves are shown (A,C) in addition to the rectification ratio determined at −60 and +30 mV (B,D). Number of oocytes recorded: n = 7–36.

The fitted results from this analysis suggested that low concentrations of both of these organic cation channel blockers were effective at reducing the increased current observed for variant receptors near their resting potential (Figure 5C,D). However, the change in zδ, which is the valence times the fraction of the electric field felt by the combined blockers, clearly indicates that the nature of the channel block produced by Mg²⁺ and organic cations at variant receptors is different than the block that Mg²⁺ alone produces at WT receptors (Table 3). The electric field zδ sensed by the blocking ion Mg²⁺ alone was 1.97 for GluN1/GluN2A and 1.96 for GluN1/GluN2B WT NMDARs. The GluN2A-L611Q variant reduced this value to 1.21, and addition of increasing concentrations of the drug ketamine only reduced the value for zδ further (down to 0.80 above 10 μM ketamine; Table 3). The reduction of zδ by addition of organic cations may reflect increased proportion of block by a monovalent ion compared to divalent Mg²⁺. Similarly, the apparent affinity for the combined blockage increased (i.e., K_{D,0 mV} decreased) with increasing concentrations of ketamine, suggesting qualitatively dissimilar blocking by the combination of Mg²⁺ and ketamine at variant NMDARs. A similar result was observed for the GluN2B-V620M variant, which decreased zδ from 1.96 for WT GluN1/GluN2B receptors to 1.47 for GluN2B-V620M NMDARs. This value was reduced further to 1.28 when Mg²⁺ was supplemented with10 μM memantine. The value for K_{D,0 mV} was also decreased in presence of memantine, suggesting that memantine can minimize some of the excess current associated with loss of Mg²⁺ block, but cannot restore the qualitative features of the block for this GRIN variant (Table 3). This result is also apparent from the incongruent nature of the shape of the current-voltage curves for WT in Mg²⁺ compared to variant in Mg²⁺ plus channel blocker. The reversal potential for the oocyte recordings is from −3.0 mV to +3.3 mV (average: −2.53 +/− 0.28 mV).

Table 3.

Summary of voltage dependence and apparent blocker affinity

Constructs/Conditions	KD, 0 mV, mM	z δ
WT 2A	1.52	1.97
2A-L611Q + 100 μM ketamine	0.26	0.80
2A-L611Q + 30 μM ketamine	0.41	0.80*
2A-L611Q + 10 μM ketamine	0.53	0.84
2A-L611Q + 3.0 μM ketamine	1.73	1.01
2A-L611Q + 1.0 μM ketamine	4.01	1.02
2A-L611Q + 0.3 μM ketamine	8.68	1.14
2A-L611Q	13.0	1.21
WT 2B	1.83	1.96
2B-V620M + 10 μM memantine	0.53	1.28
2B-V620M + 3.0 μM memantine	1.08	1.38
2B-V620M + 1.0 μM memantine	3.66	1.52
2B-V620M + 0.3 μM memantine	4.90	1.54
2B-V620M	3.93	1.47

The current-voltage relationships were generated from TEVC recordings on Xenopus oocytes expressing NMDARs that were activated by maximally effective concentrations of glutamate and glycine (100 μM) in the presence of 1 mM Mg²⁺. Current was recorded at holding potential ranging from −90 to +30 mV. The voltage dependence and apparent blocker affinity at 0 mV were obtained by fitting the data using the Equation-2 (see Methods).

^*zδ was fixed to 0.80 for this condition.

Discussion

The most important finding of this study was that different variants that alter residues comprising the pore lining regions of the NMDAR have differential effects on the sensitivity for channel block. The differential response of these M2 variant-containing receptors to FDA-approved NMDAR inhibitors is consistent with previous reports that these channel blockers may engage different structural determinants at their binding site in the channel pore (Chou et al., 2022). This suggests that use of channel blockers with variants located in the pore are unlikely to be uniformly effective when evaluated indiscriminately for all patients. Rather, these data suggest that in vitro analysis of channel blocker potency should identify candidate blockers for potential testing for off-label use in patients or future clinical trials. In some cases, missense variants alter the residues within the pore to render channel blockers more potent at the variant (e.g. (Pierson et al., 2014; Xu et al., 2021)), and in these situations one might expect improvement in clinical phenotype since channel blocking drugs will reduce aberrant activity through variant receptors to a greater extent than activity through WT receptors.

There are multiple ways in which the loss or reduction in channel block by extracellular Mg²⁺ might impact neuronal and circuit function. These include an increase in potential current through synaptically-activated NMDARs at or near resting membrane potentials, as well as increased inward current produced by perisynaptic or extrasynaptic receptors that are activated by ambient levels of glutamate (or glutamate released by glia). Such an increase in current could drive strong excitation within the circuit, which could have multiple effects and also lead to aberrant homeostatic plasticity. Provided variants do not reduce Ca²⁺ permeation, this also could lead to increased Ca²⁺ influx, which might drive different second messenger-linked signaling pathways. However, GluN2A-L611Q reduces Ca²⁺ permeation threefold (Li et al., 2019), suggesting reduced Mg²⁺ block will drive strong excitation without an equally large increase in intracellular Ca²⁺. Similarly, GluN2B-V620M reduced Ca²⁺ permeability, although not as strongly as GluN2A-L611Q, and thus one would expect less Ca²⁺ influx for this variant in the context of strong excitatory drive.

In addition to increased current at resting levels, the nature of the voltage-dependent block is an important element used to trigger synaptic plasticity (Bliss & Collingridge, 1993; Bliss & Gardner-Medwin, 1973). In particular, the voltage range over which the negative slope conductance occurs within the current-voltage relationship designated the extent to which depolarization can diminish Mg²⁺ block. For example, there is a region of negative slope conductance over the voltage range of −30 to −70 mV for WT NMDARs. This means that a reduction in membrane potential from, say −65 mV to −45 mV will diminish voltage-dependent Mg²⁺ block, thus allowing NMDARs to mediate more current and provide a bolus of Ca²⁺ that can trigger various forms of plasticity. By contrast, the region of negative slope conductance for L611Q occurs over the range −90 to −70 mV. However, the curve is relatively flat over the range of −65 to −45 mV, so a depolarization will not drive much additional current through this variant NMDAR. The shallow nature of the Mg²⁺ block curve in this voltage range for variants (as well as shallow nature of supplemental organic monovalent cation blockers with Mg²⁺) means that modest depolarization from the resting membrane potential will not produce much relief of channel block, and will trigger a smaller influx of Ca²⁺ than observed for similar depolarization of WT receptors. Attempts to reestablish voltage-dependent block with organic cation channel blockers will not markedly change this situation since the voltage dependence of monovalent blockers will not be as strong. This could have profound effects on neurodevelopment and synaptic plasticity. This lack of robust depolarization-induced current over the relevant voltage range needed as a trigger for plasticity will be exacerbated by variant-associated reductions in Ca²⁺ permeability.

Analysis of the current-voltage curves revealed concentration-dependent inhibition on GluN2A-L611Q and GluN2B-V620M by ketamine and memantine, respectively. The reduced affinity and voltage-dependence of Mg²⁺ binding (the apparent affinity K_{D,0 mV} and the product zδ) caused by these two disease-related variants suggested that the loss of voltage-dependent Mg²⁺ inhibition involves both disruption of Mg²⁺ binding site by these M2 pore loop variants in addition to changes in intra-pore electrostatics. These parameters cannot be fully restored by the addition of increasing concentrations of the FDA-approved drugs ketamine and memantine. Thus, while these channel blockers could minimize excessive excitatory drive associated with reduced Mg²⁺ block (which is important), they will not restore the plasticity that is dependent on the steep region of negative slope conductance within the current-voltage curve that allows small depolarization to produce substantial increases in current.

These results also provide insight into the nature of the Mg²⁺ binding site. For all three channel blockers evaluated, the WT receptors showed reduced potency in the presence of physiological concentration Mg²⁺ (1 mM) compared to nominally Mg²⁺-free control. By contrast, most M2 variants evaluated showed a similar or enhanced potency to at least one channel blocker compared to the WT receptors in the presence of Mg²⁺ (e.g., Figures 2,3). In addition, in the presence of physiological concentrations of Mg²⁺, comparing to the GluN2B-W607C reduced potency to dextromethorphan by 3-fold, GluN2B-G611V reduced potency to ketamine by 2.6-fold, and GluN2B-V620M reduced potency to ketamine by 2.1-fold, respectively (Table 1, Figure 3). These observations may reflect changes in the nature of the interaction between Mg²⁺ and channel blockers (Glasgow, Wilcox, & Johnson, 2018; Kotermanski & Johnson, 2009). Interestingly, GluN2A-N614S reduced the potency for memantine and GluN2B-N615K reduced the potency for all three channel blockers. These two asparagines (position 614 for GluN2A, position 615 for GluN2B) are equivalent, indicating shared features among GluN2A and GluN2B variant actions.

Taken together, these in vitro analyses of missense GRIN variants revealed a wide range of effects on channel blocker potency and voltage dependence. Some variants showed a similar or modestly enhanced potency to FDA-approved NMDAR channel blockers compared to WT receptors, suggesting these drugs might help to reduce overexcitation produced by relief of extracellular Mg²⁺ block at resting membrane potentials. However, the addition of FDA-approved channel blockers will not fully restore NMDAR function, which could limit effectiveness in fully restoring circuit function, synaptic plasticity deficits, and aberrant neurodevelopment. Moreover, further testing using knock-in mouse models is needed to determine the extent to which data obtained for voltage dependent channel block of NMDARs expressed in heterologous expression systems faithfully matches properties of channel block in native receptors.

Abbreviations:

3DMTR

three-dimensional tolerance ratio analysis

ABD

agonist binding domain (also known as LBD, ligand binding domain)

CTD

carboxy-terminal domain

NMDAR

N-methyl-D-aspartate receptors

NTD

amino-terminal domain (also known as ATD)

TEVC

two electrode voltage clamp

TMD

transmembrane domain