Abstract
N-methyl-D-aspartate receptors (NMDARs) are tetrameric assemblies of two glutamate N-methyl-D-aspartate receptor subunits, GluN1 and two GluN2, that mediate excitatory synaptic transmission in the central nervous system. Four genes (GRIN2A-D) encode four distinct GluN2 subunits (GluN2A-D). Thus, NMDARs can be diheteromeric assemblies of two GluN1 plus two identical GluN2 subunits, or triheteromeric assemblies of two GluN1 subunits plus two different GluN2 subunits. An increasing number of de novo GRIN variants have been identified in patients with neurologic conditions and with GRIN2A and GRIN2B harboring the vast majority (> 80%) of variants in these cases. These variants produce a wide range of effects on NMDAR function depending upon its subunit subdomain location and additionally on the subunit composition of diheteromeric versus triheteromeric NMDARs. Increasing evidence implicates triheteromeric GluN1/GluN2A/GluN2B receptors as a major component of the NMDAR pool in the adult cortex and hippocampus. Here, we explore the ability of GluN2A- and GluN2B-selective inhibitors to reduce excess current flow through triheteromeric GluN1/GluN2A/GluN2B receptors that contain one copy of GRIN2A or GRIN2B gain-of-function variants. Our data reveal a broad range of sensitivities for variant-containing triheteromeric receptors to subunit-selective inhibitors, with some variants still showing strong sensitivity to inhibitors, whereas others are relatively insensitive. Most variants, however, retain sensitivity to non-selective channel blockers and the competitive antagonist D-(-)-2-amino-5-phosphonopentanoic acid. These results suggest that with comprehensive analysis, certain disease-related GRIN2A and GRIN2B variants can be identified as potential targets for subunit-selective modulation and potential therapeutic gain.
SIGNIFICANCE STATEMENT Triheteromeric NMDA receptors that contain one copy each of the GluN2A and GluN2B subunits show intermediate sensitivity to GluN2A- and GluN2B-selective inhibitors, making these compounds candidates for attenuating overactive, GRIN variant-containing NMDA receptors associated with neurological conditions. We show that functional evaluation of variant properties with inhibitor pharmacology can support selection of a subset of variants for which GluN2 subunit-selective agents remain effective inhibitors of variant-containing triheteromeric NMDA receptors.
Introduction
N-methyl-d-aspartate receptors (NMDAR) are ligand-gated ion channels that mediate a slow, Ca2+-permeable component of excitatory synaptic transmission (MacDermott et al., 1986; Hansen et al., 2021). NMDARs can be blocked by extracellular Mg2+ in a voltage-dependent manner (Mayer et al., 1984; Nowak et al., 1984) and play a key role in many processes in the central nervous system, including learning, memory, and neuronal development (Hansen et al., 2021). NMDAR dysfunction has been suggested to be involved in various neurologic and neuropsychiatric disorders such as epilepsy, intellectual disability, autism spectrum disorder, neuropathic pain, depression, Parkinson’s disease, and schizophrenia (Lau and Zukin, 2007; Mony et al., 2009; Traynelis et al., 2010; Hansen et al., 2021).
The NMDAR receptors are encoded by the GRIN gene family, which includes GRIN1, GRIN2A-D, and GRIN3A-B genes. Whereas in vitro studies on recombinant receptors have largely been conducted with diheteromeric NMDARs that contain two copies of the same subunit of the glutamate N-methyl-D-aspartate receptor (GluN2) subunit co-assembled with two copies of GluN1 expressed in either Xenopus oocytes or transfected mammalian cell lines (Traynelis et al., 2010; Paoletti et al., 2013), a large fraction of NMDARs expressed in vivo are triheteromeric assemblies of two GluN1 subunits and two different GluN2 subunits in the tetrameric complex. For example, GluN1/GluN2A/GluN2B comprise over half of the NMDARs in the adult cortex and hippocampus (Sheng et al., 1994; Luo et al., 1997; Al-Hallaq et al., 2007; Rauner and Kohr, 2011; Tovar et al., 2013). Moreover, different combinations of GluN2 subunits assemble into the receptor complex in distinct brain regions (Rauner and Köhr, 2011; Tovar et al., 2013; Hansen et al., 2014; Hansen et al., 2021). Recent evidence has shown that the kinetic and pharmacological properties of GluN1/GluN2A/GluN2B triheteromeric receptors are distinct from the GluN1/GluN2A or GluN1/GluN2B diheteromeric receptors (Hansen et al., 2014; Stroebel et al., 2014; Bhattacharya et al., 2018; Yi et al., 2019). However, the biophysical and pharmacological properties of triheteromeric GluN2A- and GluN2B-containing NMDARs are strongly influenced by GluN2A (Hansen et al., 2014), whereas neither GluN2B nor GluN2D dominate the properties of GluN1/GluN2B/GluN2D triheteromeric NMDA receptors (Yi et al., 2019).
Whole exome sequencing has identified a large number of rare do novo GRIN variants associated with neurologic disorders (Burnashev and Szepetowski, 2015; Yuan et al., 2015; Hu et al., 2016; Li et al., 2016; Swanger et al., 2016; XiangWei et al., 2018, 2019; Myers et al., 2019; Strehlow et al., 2019; Benke et al., 2021), with GRIN1, GRIN2A, GRIN2B, and GRIN2D genes being highly intolerant to genetic variation. As a result, a large number of missense GRIN2A and GRIN2B variants have been identified in patients with various neurologic conditions. These variants cluster in select regions of the subunit, such as the amino terminal domain (NTD), the agonist binding domain (ABD), the transmembrane domains (TMD), and the linker regions connecting these subdomains (Yuan et al., 2014; Chen et al., 2017; Ogden et al., 2017; Perszyk et al., 2020, 2021; Hansen et al., 2021). Variants localized in these subdomains can drive unique electrophysiological effects and thus distinct neurologic phenotypes in patients (XiangWei et al., 2018; Strehlow et al., 2019). Here, we investigate the utility of potential therapeutic agents for treatment of rare GRIN variants. We focused on gain-of-function (GoF) variants that increase current flow in GRIN2A and GRIN2B in the context of triheteromeric NMDAR complexes that include a single subunit that harbors the damaging rare variant, since this more likely reflects the NMDAR context in vivo. Prior studies showed that the GluN2B allosteric inhibitor radiprodil, which can reduce seizure burden in patients with infantile spasms (Auvin et al., 2020), inhibits diheteromeric GluN2B receptors harboring two copies of rare GluN2B de novo GoF variants (Mullier et al., 2017). Other GluN2B inhibitors that occupy the same binding site as radiprodil inhibit triheteromeric GluN1/GluN2A/GluN2B receptors with lower potency and diminished maximal inhibition at saturating inhibitor concentrations (Hansen et al., 2014; Yi et al., 2019). This raises the question of how sensitive triheteromeric GluN1/GluN2A/GluN2B receptors harboring GoF variants are to GluN2B-selective inhibitors. A similar question arises concerning the sensitivity of variant-containing triheteromeric receptors to the GluN2A-selective inhibitor, TCN-201 (Bettini et al., 2010).
Here, we evaluated the actions of three classes of NMDAR inhibitors on triheteromeric GluN1/GluN2A/GluN2B receptors that contain a single damaging variant: subunit-selective negative allosteric modulators (NAMs), channel blockers, and the competitive antagonist, D-(-)-2-amino-5-phosphonopentanoic acid (D-APV). Because most GoF variants are localized in the linker and TMDs of GluN2A and GluN2B NMDAR subunits, we selected a representative set of variants from these regions for study. Our results show that distinct subdomain variants have divergent responses to NMDAR inhibitors. Among these drug classes, channel blockers and D-APV were capable of attenuating the subset of variants we studied in triheteromeric receptors that contain one copy of the mutant allele, whereas the subunit selective NAMs showed variable effects that were dependent on the specific genetic variant.
Materials and Methods
Drugs and Reagents
Radiprodil, ifenprodil, CP-101,606, Ro 25-6981, and TCN-201 were purchased from Axon MedChem (Reston, VA USA), memantine and dextromethorphan from Sigma Aldrich (USA), and ketamine from Hospira, Inc. (Lake Forest, IL). Stock concentrations of each compound were prepared in deionized water (memantine and ketamine), DMSO (dextromethorphan, radiprodil, ifenprodil, CP-101,606, Ro 25-6981, and TCN-201); the 100 mg/ml ketamine stock solution contained 0.1 mg/ml of benzethonium chloride. After dilution into oocyte perfusion media, the final DMSO concentration was 0.1% (v/v).
cDNAs and Molecular Biology
Wild-type (WT) rat cDNAs for GluN1-1a (GenBank accession numbers U08261 for GluN1 lacking exon5 but containing exons 21 and 22), GluN2A (GenBank accession number D13211) , and GluN2B (GenBank accession number U11419) were subcloned into the pCI-neo mammalian expression vector. The constructs of triheteromeric receptors were generated using rat GluN1 and GluN2A with modified C-terminal peptide tags as previously described (Hansen et al., 2014). Briefly, C-terminal peptide tags were generated from the leucine zipper motifs found in GABAB1 (referred to as C1) and GABAB2 (referred to as C2). These tags were placed downstream of a synthetic helical linker and upstream of a KKTN endoplasmic reticulum retention signal (Jackson et al., 1990, 1993; Zerangue et al., 2001). The C1 and C2 leucine zipper motifs can enter into a coiled-coil interaction that will mask the KKTN retention motif and allow expression of mature receptors on the cell surface. The tag was introduced in frame and in place of the stop codon at the GluN2A C-terminal tail (to make 2AC1 and 2AC2). A chimeric GluN2B subunit was constructed in which the 2B carboxyl tail after residue 838 was replaced by the GluN2A carboxyl tail and C-terminal-linker-C1 or -C2-ER retention motifs (to make 2BAC1 and 2BAC2, hereafter called 2BC1 and 2BC2, respectively), as described by Hansen et al. (2014). The efficiency of the peptide tags for controlling surface expression could be assessed through the measurement of the “escape” currents estimated using a pair of mutations that rendered the agonist binding domain incapable of binding glutamate and thus unable to pass current. These mutations were GluN2A-R518K,T690I and GluN2B-R519K,T691I.
Rare human GRIN2A (E551K, P552R, S644G, L649V, L812M, M817V) or GRIN2B (G543R, A639V, M818T, A819T) NMDAR variants were introduced into their respective rat cDNAs using a QuikChange mutagenesis protocol (Agilent, USA) according to the manufacturer’s instructions. After introduction of the genetic variant into the cDNA, the full open reading frame was sequence-verified (Sanger method, Eurofins, KY USA) to ensure only the intended genetic variant was introduced into the cDNA. The cDNA was then linearized with the Not I restriction enzyme and ribonucleic acid made from cDNA (cRNA) synthesized according to the manufacturer’s instructions using T7 RNA polymerase with mMessage mMachine (Invitrogen, USA).
Triheteromeric Three-Dimensional Missense Tolerance Determination
To determine three-dimensional missense tolerance (3DMTR, Perszyk et al., 2021) of the GluN1/GluN2A/GluN2B, we used the recently published cryo-EM structure of the triheteromeric NMDAR (PDB: 5UOW, Lü et al., 2017). This cryo-EM structure was determined using the Xenopus laevis analog subunits, so, in using the 3DMTR application (https://github.com/riley-perszyk/3DMTR), the Xenopus laevis grin1.L, grin2a.L, and grin2b gene sequences were used with the human GRIN1, GRIN2A, and GRIN2B sequences (NM_001185090.2_, NM_000833.5, and NM_000834.5, respectively) to align the gnomAD (version 2.1.1) missense and samesense variant entries for each gene.
Oocyte Preparation and Injections
Stage V-VI Xenopus laevis oocytes were prepared from commercially available ovaries (Xenopus 1 Inc., Dexter MI, USA). Briefly, the ovary was digested with Collagenase Type 4 (Worthington-Biochem) solution in Ca2+-free modified Barth’s solution, which contained (in mM) 88 NaCl, 2.4 NaHCO3, 1 KCl, 0.82 MgSO4, and 10 (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) (pH 7.4 with NaOH) and was supplemented with 1 U/ml of penicillin-streptomycin (Gibco #15140122). The ovary was incubated with gentle shaking at room temperature (23°C) for 2 hours, rinsed ten times (10 minutes each) with fresh Ca2+-free Barth’s solution, and then rinsed four times (5 minutes 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 (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid), adjusted to pH 7.4 with NaOH and supplemented with 100 μg/ml of gentamycin and 1 U/ml of penicillin-streptomycin. The prepared and sorted oocytes were then incubated in normal Barth’s solution prior to, and after, injections. cRNAs encoding GluN1, as well as C1- and C2-tagged GluN2A or GluN2B were expressed as previously described (Strong et al., 2021). Briefly, GluN1:GluN2C1:GluN2C2 cRNAs were injected at a 1:2:2 ratio with ∼10 ng of total cRNA in a total volume of 50 nl to achieve the following combinations of subunit RNA: GluN1:GluN2AC1:GluN2AC2 or GluN1:GluN2BC1:GluN2BC2 to generate diheteromeric receptors, or GluN1:GluN2AC1:GluN2BC2, GluN1:GluN2AC1-(variant):GluN2BC2, or GluN1:GluN2AC1:GluN2BC2-(variant) to make WT or variant-containing triheteromeric NMDARs. Injected oocytes were incubated at 15–19°C in Barth’s solution for 1 to 3 days before recording.
Two-Electrode Voltage Clamp Recordings
Two-electrode voltage clamp recordings were made at 22–23°C 1–-3 days after injection using OC725C amplifiers (Warner Instruments, Hamden, CT) similarly to that previously described (Strong et al., 2021). The recording solution contained (in mM) 90 NaCl, 1 KCl, 10 (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid), 0.5 BaCl2, and 0.01 EDTA and was adjusted to pH 7.4 with NaOH. Recording electrodes were filled with 0.3 − 3.0 M of KCl. Oocytes were perfused in recording solution supplemented with agonists or drugs as indicated, and the membrane was clamped at -40 mV. For single concentration drug inhibition experiments, test currents were recorded in response to maximal glutamate and glycine (100 μM each) and then also with the inhibitor at 3 μM (ifenprodil, radiprodil, Ro 25-6981, CP-101,606, TCN-201) for 3–-6 minutes. Concentration-response curves were determined similarly but with increasing concentrations of the GluN2B negative allosteric modulators; TCN-201 was co-applied with 100 μM glutamate and 3 μM glycine. Drug concentration-response curves were obtained until steady state inhibition was obtained at each concentration, typically 3–6 minutes. Drug concentration-response curves with the uncompetitive channel blockers memantine, ketamine, and dextromethorphan were conducted similarly, except each drug concentration was applied for 1 minute to reach steady-state.
For drug inhibition studies, oocyte recordings were made from 2 or more independent experiments (i.e., oocyte injection cycles) when testing inhibitors against variant-containing triheteromeric NMDARs and WT controls. Drug potency was quantified as the concentration that inhibits 50% of the current for each oocyte by fitting Eq. (1) to the concentration-response data: where minimum is the residual response in saturating concentration of the inhibitor (constrained to be >0), IC50 is the concentration of inhibitor that causes half-maximal inhibition, and nH is the Hill slope. For channel blockers, minimum was constrained to be equal to or greater than 0.
In experiments testing the effects of the competitive antagonist D-APV, the glutamate concentrations were adjusted to the EC50 values as determined for the variant triheteromeric receptors as shown in Supplemental Table S2. Here, glutamate agonist potency was quantified as the concentration that elicited half-maximal activation of the current response for each oocyte by fitting Eq. (2) to the concentration-response data: where concentration is the concentration of the agonist, EC50 is the concentration of agonist that causes half-maximal activation, and nH is the Hill slope.
The uncompetitive channel blockers memantine (100 μM), ketamine (100 μM), and dextromethorphan (300 μM), and the NTD-site negative allosteric modulator radiprodil (3 μM) were also tested for inhibition at different membrane potentials by stepping the oocyte membrane potential in 10 mV increments from -90 mV to +30 mV in the absence and presence of drug under maximal receptor activation by glutamate and glycine (100 μM each).
Statistical Analysis
Data are expressed as mean ± SEM, or mean ± 95% confidence interval, and the number of replicates (n) reported for each measure. One-way ANOVA with Dunnett’s Multiple Comparison Test (GraphPad Prism 9.0) was used to assess significance of potential changes in the mean current in response to a single concentration of inhibitor. For potency shifts of agonists or modulators, statistical comparisons were made and 95% confidence intervals were calculated from the log IC50 or log EC50 values for comparison among conditions.
Results
GRIN Variants Selected for Rescue Pharmacology in Triheteromeric NMDARs
We selected a set of rare GRIN variants identified in patients with a neurologic phenotype, a de novo heterozygous genotype (Table 1), and an overall GoF designation based on electrophysiological endpoints measured in vitro by recombinant expression in Xenopus laevis oocytes (Supplemental Table S2; Fig. 1). We report here GoF effects for three variants reported in ClinVar (Table 1) by measuring both glutamate and glycine potency of diheteromeric variant containing NMDA receptors expressed in vitro in Xenopus oocytes. In this receptor context, the 2A-E551K variant (Fig. 1A,B) exhibits several-fold enhanced potency for both glutamate of 0.47 μM [0.41–0.52 μM 95% confidence interval (CI)] versus 2.9 μM (2.2–3.5 μM 95% CI) for WT 2A and glycine of 0.24 μM (0.19–0.29 μM 95% CI) versus 1.2 μM (1.0–1.4 μM 95% CI) for WT 2A EC50 values that were significant (P < 0.001, t-test performed on log EC50’s). Similarly, the 2B-G543R and 2B-A639V variants (Fig. 1C,D) also exhibited several-fold enhanced potency for both glutamate of 0.22 μM (0.19–0.26 μM 95% CI) for 2B-G543R and 0.28 μM (0.21–0.37 μM 95% CI) for 2B-A639V versus 1.0 μM (0.90–1.2 μM 95% CI) for WT 2B and glycine of 0.09 μM (0.08–0.12 μM 95% CI) for 2B-G543R and 0.07 μM (0.05–0.09 μM 95% CI) for 2B-A639V versus 0.33 μM (0.26–0.41 μM 95% CI) for WT 2B EC50 values that were significant (P < 0.001, one-way ANOVA, Dunnett’s, performed on log EC50s).
We focused on GoF variants given the availability of several NMDAR inhibitors that are either FDA-approved for use in man (memantine, ketamine, dextromethorphan), have been tested in phase 2 clinical studies (radiprodil, CP-101,606; e.g., Preskorn et al., 2008, Auvin et al., 2020), or are tool compounds that are known to be highly selective for GluN2B (ifenprodil, Ro 25-6981) or GluN2A (TCN-201; Hansen et al., 2021) diheteromeric receptors. We also evaluated a competitive antagonist at the glutamate site (D-APV) (Fig. 2). Six GRIN2A GoF variants were evaluated, including E551K and P552R in the pre-M1 linker region, S644G and L649V in the M3 transmembrane domain, L812M in the pre-M4 linker, and M817V in the M4 transmembrane domain (Table 1; Fig. 3). Four GRIN2B variants were also selected, including G543R in the pre-M1 linker region, A639V in the M3 transmembrane domain, and the M818T and A819T in the M4 transmembrane domain (Table 1; Fig. 3). All ten variant residues are located in critical receptor subunit subdomains that are intolerant to genetic variation based on absence of variants reported in the gnomAD database (v2.1.1 accessed 10-19-2021) (Fig. 3). Furthermore, these ten amino acid residues are highly conserved both phylogenetically and across the four NMDAR GluN2 subunits (GluN2A, 2B, 2C, 2D). These de novo genetic variants were identified in patients with epilepsy and/or developmental delay (Table 1). However, there is no information available about the pharmacological effects of different NMDAR inhibitors on these variants in the context of missense variant-containing triheteromeric NMDA receptors and only limited information when in the context of diheteromeric receptors (Supplemental Table S1).
To selectively express triheteromeric NMDA receptors at the cell surface, we used an expression paradigm with two engineered C-terminal peptide tags to control subunit stoichiometry within the tetrameric NMDAR, as described previously (Hansen et al., 2014; see Methods). An additional control with a glutamate-insensitive mutation that only produced measurable agonist-dependent currents when diheteromeric receptor “escape” from ER retention was used to verify the efficiency of the peptide tags for controlling surface expression (see Methods). The percent of escape diheteromeric currents measured were between 1 and 6% of the recorded triheteromeric response, suggesting that greater than 94% of the current was from surface triheteromeric NMDARs (Supplemental Figure S1).
We evaluated the noncompetitive GluN2A-selective inhibitor TCN-201 and noncompetitive GluN2B-selective inhibitors ifenprodil, radiprodil, Ro 25-6981, and CP-101,606 for actions on triheteromeric NMDARs containing GoF variants. We compared the effects of these subunit-selective compounds to the FDA-approved channel blockers memantine, ketamine, and dextromethorphan in addition to the competitive antagonist D-APV, none of which show any appreciable subunit-selectivity. We first established efficiency of inhibition of GluN2A- and GluN2B-selective inhibitors on triheteromeric receptors by testing saturating concentrations (e.g., 3 μM) of these compounds using two-electrode voltage clamp current recording from Xenopus oocytes expressing WT and variant-containing triheteromeric NMDARs. For inhibitors with appreciable activity, we evaluated the concentration-response relationship to determine IC50s.
Inhibition of Triheteromeric NMDARs by a GluN2A-Selective Negative Allosteric Modulator
The GluN2A-selective inhibitor TCN-201 showed strong inhibitory effects on diheteromeric GluN1/GluN2AC1/GluN2AC2 receptors and no inhibition on GluN1/GluN2BC1/GluN2BC2 NMDARs, as expected (Table 2; Fig. 4A). TCN-201 also showed relatively strong inhibition on GluN1/GluN2AC1/GluN2BC2 WT triheteromeric receptors that approached 59% (Fig. 4; Table 2), as previously reported (Hansen et al., 2014). By contrast, TCN-201 displayed weak or no inhibition on pre-M1 and M3 GluN2A GoF variants within triheteromeric receptors (E551K, P552R, S644G, L649V) with maximal inhibition ranging from 0–17%. This may in part reflect the higher potency of glycine at these variants, given that the mechanism of action of TCN-201 involves an allosteric reduction in glycine affinity (Hansen et al., 2012). TCN-201 was more effective at inhibiting pre-M4/M4 GRIN2A variants L812M and M817V (24–31%; Table 2), yet was still less effective than its inhibition at WT GluN1/GluN2AC1/GluN2BC2 NMDARs (59%; Table 2).
We also evaluated the ability of TCN-201 to inhibit triheteromeric NMDARs that contained a GRIN2B variant. TCN-201 produced relatively strong inhibition at saturating concentrations on the GRIN2B pre-M1 variant G543R (46%; Table 2) but exhibited modest actions on GRIN2B M3 and M4 variants A639V, M818T, and A819T (6–15% inhibition; Table 2).
Because the GRIN2B-G543R variant showed the largest % inhibition at saturating TCN-201 concentrations, we also determined an IC50 for this variant by TCN-201. In a GluN1/GluN2AC1/GluN2B-G543RC2 triheteromeric NMDAR, the IC50 was 1.5 μM, which compares well with the IC50 of 1.2 μM at WT GluN1/GluN2AC1/GluN2BC2 NMDARs (Table 3, Fig. 4B).
Inhibition of Triheteromeric NMDARs by GluN2B-Selective Negative Allosteric Modulators
We next evaluated four GluN2B-selective inhibitors (ifenprodil, Ro 25-6891, radiprodil, CP-101,606) at saturating concentrations for actions on triheteromeric receptors that contained one copy of either a GRIN2A or GRIN2B variant. As reported previously, all GluN2B-selective inhibitors blocked GluN1/GluN2BC1/GluN2BC2 WT NMDARs with similar maximum efficacy as GluN1/GluN2B WT NMDARs (Table 2, Fig. 4C,F,G,H; Hansen et al., 2014; Mullier et al., 2017), with little effect on GluN1/GluN2AC1/GluN2AC2 NMDARs (Fig. 4C,F,G,H; Table 2). GluN2B-selective inhibitors have been reported to have reduced potency and a diminished maximal inhibition of triheteromeric GluN1/GluN2AC1/GluN2BC2 WT NMDARs (e.g., Hatton and Paoletti, 2005; Hansen et al., 2014), and we also report diminished maximal inhibition ranging from 18–31% for four GluN2B-selective NAMs (Fig. 4C-H, Table 2).
To better understand the potential effects of the GluN2B-selective NAMs on GRIN2A variants that exist within triheteromeric GluN1/GluN2AC1/GluN2BC2 NMDARs, we first measured the ability of these GluN2B-selective inhibitors to maximally inhibit triheteromeric NMDARs that contain a single copy of a GRIN2A variant (Fig. 3C-H; Table 2). These four inhibitors showed variable effects. No GluN2B-selective NAM was particularly effective at maximally inhibiting GluN1/GluN2AC1/GluN2BC2 triheteromeric NMDARs that contained GRIN2A M3, pre-M4, or M4 variants, with 0.6–14% maximal inhibition observed at saturating concentrations for each inhibitor. However, the GluN2B NAMs were more effective at maximally reducing current responses for GRIN2A pre-M1 variants E551K or P552R (19–54%; Fig. 4C,E,F,G,H; Table 2).
We also tested the GluN2B-selective inhibitors on triheteromeric GluN1/GluN2AC1/GluN2BC2 receptors that contained a single copy of a damaging GRIN2B variant. Here, maximal inhibition of triheteromeric receptors that contained GRIN2B variants G543R and M818T (24–46% inhibition) (Fig. 4C,F,G,H) was similar across the four inhibitors, whereas for the GRIN2B variant A819T, maximal inhibition was more variable across the four inhibitors (13–41% inhibition). All four inhibitors were only modestly effective at blocking responses of the GRIN2B M3 variant A639V (5.6–12% inhibition; Table 2, Fig. 4C,F,G,H).
We subsequently measured the potency of ifenprodil to inhibit five variants in the context of triheteromeric NMDARs. We selected ifenprodil as a representative GluN2B-selective inhibitor given its widespread use in the literature. In these studies, ifenprodil potency was reduced approximately 1.2- to 2.4-fold for the GRIN2A E551K and P552R variants, and from 3.2- to 3.6-fold for the GRIN2B G543R, M818T, and A819T variants when tested in the context of a triheteromeric receptor (Table 3; Fig. 4D; Supplemental Table S3), suggesting that while they retain effectiveness on inhibiting variant receptors, somewhat higher concentrations are required. For example, radiprodil reduced the current response of GluN1/GluN2AC1/GluN2BC2 WT NMDARs with an IC50 of 50 nM and maximal inhibition of 29%, and inhibited GluN1/GluN2A-P552RC1/GluN2BC2 NMDARs with an IC50 of 135 nM and maximal inhibition of 54% (Supplemental Figure S2).
Inhibition of Triheteromeric NMDARs by Channel Blockers
Three US Food and Drug Administration-approved NMDAR channel blockers (memantine, ketamine, dextromethorphan) were also evaluated at variant-containing triheteromeric NMDARs. All channel blockers tested inhibited both triheteromeric GluN1/GluN2AC1/GluN2BC2 WT and variant NMDARs. We first determined the degree of inhibition by a maximally effective concentration of channel blocker (100 μM for memantine, ketamine; 300 μM for dextromethorphan). Maximally effective concentrations of memantine inhibited WT GluN1/GluN2AC1/GluN2BC2 by 95%, triheteromeric GRIN2A GoF variants by 87–95%, and triheteromeric GRIN2B GoF variants by 83–95% (Table 2). We subsequently recorded the concentration-response curve for NMDAR channel blockers dextromethorphan, memantine, and ketamine on triheteromeric GluN1/GluN2AC1/GluN2BC2 NMDARs harboring either GoF GRIN2A or GRIN2B variants (Fig. 5; Table 3). Triheteromeric GluN1/GluN2AC1/GluN2BC2 NMDARs showed intermediate sensitivity to memantine (IC50 3.4 μM) compared with diheteromeric GluN1/GluN2AC1/GluN2AC2 (IC50 6.4 μM) or GluN1/GluN2BC1/GluN2BC2 (IC50 2.3 μM). Ketamine similarly showed an intermediate potency at triheteromeric receptors compared with diheteromeric receptors, whereas dextromethorphan had similar potency at triheteromeric NMDARs (Table 3). This is consistent with the idea that the sites of interaction within the pore blockers are not identical between different channel blockers (e.g., LePage et al., 2005). The IC50 was determined for all ten GoF GRIN2A and GRIN2B variants when expressed in triheteromeric receptor contexts. Here, memantine showed similar potency for pre-M1 GRIN2A variants E551K and P552R when expressed in GluN1/GluN2AC1/GluN2BC2 triheteromeric NMDARs, but a 3- to 7-fold loss in potency for inhibition of the other four GRIN2A variants in M3, pre-M4, and M4 domains (Fig. 5A; Table 3; Supplemental Table S3). A similar pattern was observed for ketamine and dextromethorphan, where pre-M1 variants had only modest effects on IC50 values for these channel blockers (Fig. 5B,C; Table 3; Supplemental Table S3). We also evaluated the actions of channel blockers on the four GRIN2B variants expressed as triheteromeric GluN1/GluN2AC1/GluN2BC2 NMDARs. Here, memantine and ketamine were slightly more potent at GRIN2B variant G543R, slightly less potent at M818T, and 3- to 5-fold less potent but still effective inhibitors of GRIN2B variants A639V and A819T (Fig. 5; Table 3; Supplemental Table S3). We additionally confirmed that all channel blockers still show expected voltage dependence with block at negative potentials and minimal actions at positive potentials, with the exception of GluN2A-P552R, which showed memantine block of inward and outward currents (Fig. 6A-F). These results confirm that variants did not markedly alter the voltage-dependent mechanism of these channel blockers in triheteromeric receptors harboring most variants. In addition, the representative GluN2B-selective inhibitor radiprodil produced a voltage-independent inhibition for all receptor subunit combinations tested, as expected given its binding site in the NTD, which is at considerable distance from the transmembrane potential (Fig. 6G-I).
Inhibition of Triheteromeric NMDARs by a Competitive Antagonist
Lastly, we evaluated the effect of these GoF variants on inhibition produced by a competitive antagonist that bound to the glutamate binding site on the GluN2 subunit. To determine comparable IC50 values for D-APV, we first determined the glutamate EC50 value for all ten variants (Supplemental Table S2). This allowed us to produce concentration-response curves for D-APV inhibition of NMDARs activated by appropriate EC50 levels of glutamate per variant. In this situation, the IC50 value that we determine for D-APV approximates its KB value at each variant receptor, allowing us to determine whether the variants altered the sensitivity of the receptor to binding of D-APV. Determination of the IC50 values for D-APV inhibition of NMDAR responses with EC50-levels of glutamate showed less than 2.1-fold variation across all variants, (Fig. 7; Table 3; Supplemental Table S3). These data are consistent with the location of the variants, which in all cases were in the transmembrane or linker regions and outside of the agonist binding domain that contains the agonist and D-APV binding site (Fig. 2).
Discussion
The mammalian brain contains a wide range of NMDA receptors of varying subunit composition. Many of these are triheteromeric receptors that contain two different GluN2 subunits (Sheng et al., 1994; Luo et al., 1997; Al-Hallaq et al., 2007; Rauner and Kohr, 2011; Tovar et al., 2013; Hansen et al., 2021). We have studied the pharmacology of triheteromeric NMDARs that contain a single copy of a GRIN2A or a GRIN2B variant identified in patients with neurologic conditions. Several important insights are apparent from the results presented here. First, we found that the prototypical GluN2A-selective inhibitor TCN-201 is capable of attenuating current for triheteromeric receptors that contain some (but not all) GRIN2B variants tested. Conversely, GluN2B-selective inhibitors, including ifenprodil and its analogs (CP-101,606, Ro 25-6981, radiprodil), can attenuate triheteromeric receptor responses that contain certain GRIN2A variants. These results suggest that if such subunit-selective inhibitors become available for clinical use in the future, their utility in the context of NMDAR variants could extend beyond their specified subunit selectivity. That is, since GluN2B-selective inhibitors are still modestly active at GluN2A- and GluN2B-containing triheteromeric NMDA receptors, they might also be considered for use against GRIN2A GoF variants. While this idea is raised by the results here, it will require more thorough evaluation in animal models. One animal model has shown a modest potential effect of radiprodil on a GRIN2A variant, S644G (Amador et al., 2020). Second, the variable actions of these inhibitors underscore the necessity of functional evaluation of drug action at variant-containing receptors prior to any consideration of clinical use, as it is presently impossible to know a priori whether a given compound will be effective. This idea is consistent with results for a number of channel blockers that have been tested against NMDAR variants and have shown a range of potencies, some lower, some higher than wild-type receptors (Supplemental Tables S2; S3; Pierson et al., 2014; Marwick et al., 2019; Amador et al., 2020; Xu et al., 2021). Third, these results support the intermediate pharmacological phenotype observed for many triheteromeric receptors, which emphasizes the importance of studying these receptor subtypes in addition to diheteromeric receptors.
One important caveat to bear in mind when considering the results described here is that our method for controlling subunit stoichiometry involved both addition of the coiled-coil domains plus an ER retention signal onto GluN2 subunits downstream of the C-terminal. In addition, we replaced the GluN2B C-terminal with that for GluN2A. While the wild-type GluN2B C-terminal may be necessary for some aspects of inhibition by negative allosteric modulators, such as radiprodil that binds to the NTD, several observations suggest the data we obtained here is representative of that found for wild type triheteromeric receptors. For example, deletion of the C-terminal tails for both GluN2A and GluN2B prior to addition of coiled-coil domains yields the same results in terms of the actions of ifenprodil (Stroebel et al., 2014). Furthermore, use of a different strategy to inhibit diheteromeric receptors that does not alter the C-terminal also yields the same results for ifenprodil inhibition of triheteromeric receptors (Hatton and Paoletti, 2005), suggesting the data we obtain is representative of unmodified triheteromeric GluN1/GluN2A/GluN2B receptors. We also note that ifenprodil showed similar pharmacological properties at both wild-type GluN1/GluN2B receptors as receptors that contained our modified C-termini, GluN1/GluN2BC1/GluN2BC2 (Hansen et al., 2014).
An interesting result that emerged from the sensitivity of variants within triheteromeric receptors was the manner in which subunit-selective inhibitors showed activity for variants in some but not other regions of the protein. This suggests potential downstream actions of the allosteric modulators, in that variants within specific domains that are downstream of modulator binding may interrupt the activity of that modulator, rendering the inhibitor less effective. For example, GRIN2A pre-M1 variants were permissive for inhibition by GluN2B-selective inhibitors, but GRIN2A M3/Pre-M4/M4 variants were not. One hypothesis raised from this observation is that the actions of GluN2B inhibition impact molecular motions involving M3 and M4 transmembrane elements that are required for inhibition, such that variants in these regions interrupt actions of the GluN2B allosteric inhibitors by interfering with downstream conformational changes. However, this may be a special case of the triheteromeric context, perhaps due to the NTD-ABD cross-over which occurs in NMDA receptors. Thus, the mechanistic pathway connecting the NTD, agonist binding, and channel gating may not necessarily run through the GluN2B subunit alone and may be influenced by the alternate GluN2 subunit in the triheteromer context. For some variants, either in GluN2A or GluN2B, the available therapeutic options may be more broad given the triheteromeric receptor is the primary NMDA receptor subtype found in several key brain regions.
While these data provide a useful perspective on potential means to rectify aberrant NMDAR activity, it will be important and informative to evaluate drug actions at different time points during development in animal models harboring different variant NMDAR subunits to further guide precision medicine.
Acknowledgment
The authors thank Rebecca Bui, Courtney Ning, and Vincent Peterson for technical assistance in DNA preparation and oocyte injections and Jing Zhang and Phuong Le for excellent technical assistance.
Authorship Contributions:
Participated in research design: Han, Yuan, Traynelis, Myers.
Conducted experiments: Han, Allen, Shaulsky, Myers, Kim.
Performed data analysis: Han, Yuan, Allen, Perszyk, Kim, Traynelis, Myers.
Wrote or contributed to the writing of the manuscript: Han, Yuan, Allen, Kim, Perszyk, Traynelis, Myers.
Footnotes
- Received October 29, 2021.
- Accepted January 7, 2022.
↵1 Current affiliation: Department of Neurology, Children’s Hospital of Chongqing Medical University, Chongqing, China.
This work was supported by the National Institutes of Health National Institute of Neurologic Disorders and Stroke [Grant NS111619] (S.F.T.), the National Institute of Mental Health [Grant MH127404] (H.Y.), the Eunice Kennedy Shriver National Institute of Child Health and Human Development [Grant HD082373] (H.Y.), and the National Institute on Aging [Grant AG072142] S.J.M.
H.Y., S.J.M., and S.F.T. are co-inventors of Emory-owned intellectual property. S.F.T. is a PI on research grants from Biogen and Janssen to Emory, is a member of the SAB for Sage Therapeutics, Eumentis Therapeutics, the GRIN2B Foundation, and the CureGRIN Foundation. H.Y. is the PI on a research grant from Sage Therapeutics to Emory. S.F.T. and S.J.M. own stock in a company (NeurOp, Inc.) that is developing a GluN2B antagonist.
Abbreviations
- CI
- confidence interval
- cRNA
- ribonucleic acid made from cDNA
- D-APV
- D-(-)-2-amino-5-phosphonopentanoic acid
- GoF
- gain of function
- GluN1
- the 1 subunit of the glutamate N-methyl-D-aspartate receptor
- GluN
- the subunit of the glutamate N-methyl-D-aspartate receptor
- NAM
- negative allosteric modulator
- NMDA
- N-methyl-D-aspartate
- NMDAR
- N-methyl-D-aspartate receptor
- NTD
- amino terminal domain
- WT
- wild-type.
- Copyright © 2022 by The American Society for Pharmacology and Experimental Therapeutics