Positive allosteric modulators that target NMDA receptors rectify loss-of-function GRIN variants associated with neurological and neuropsychiatric disorders

Abstract

N-methyl-D-aspartate receptors (NMDARs) mediate a slow component of excitatory synaptic transmission that plays important roles in normal brain function and development. A large number of disease-associated variants in the GRIN gene family encoding NMDAR GluN subunits have been identified in patients with various neurological and neuropsychiatric disorders. Many of these variants reduce the function of NMDARs by a range of different mechanisms, including reduced glutamate potency, reduced glycine potency, accelerated deactivation time course, decreased surface expression, and/or reduced open probability. We have evaluated whether three positive allosteric modulators of NMDAR receptor function (24(S)-hydroxycholesterol, pregnenolone sulfate, tobramycin) and three co-agonists (D-serine, L-serine, and D-cycloserine) can mitigate the diminished function of NMDARs harboring GRIN variants. We examined the effects of these modulators on NMDARs that contained 21 different loss-of-function variants in GRIN1, GRIN2A, or GRIN2B, identified in patients with epilepsy, intellectual disability, autism, and/or movement disorders. For all variants, some aspect of the reduced function was partially restored. Moreover, some variants showed enhanced sensitivity to positive allosteric modulators compared to wild type receptors. These results raise the possibility that enhancement of NMDAR function by positive allosteric modulators may be a useful therapeutic strategy.

1. Introduction

N-methyl-D-aspartate receptors (NMDARs) are ionotropic glutamatergic receptors that contribute a slow Ca²⁺-permeable component to excitatory synaptic currents in the brain and spinal cord. NMDARs have important roles in many brain functions, including learning, memory, and development. NMDARs are a tetrameric complex that contains two glycine-binding GluN1 subunits that are encoded by the GRIN1 gene and two glutamate-binding GluN2 subunits, which are encoded by GRIN2A-D (Traynelis et al., 2010). The GluN1 subunit can be found in virtually all brain regions at all stages of development (Akazawa et al., 1994). By contrast, the four different GRIN2 genes show distinct temporal and spatial expression profiles. The NMDAR subunits comprise four semiautonomous domains, including an amino terminal domain (ATD), an agonist binding domain (ABD), a transmembrane domain (TMD) that contains three transmembrane helices (M1,M3,M4) and a reentrant pore-lining loop (M2), and an intracellular carboxyl terminal domain (CTD) (Hansen et al., 2018). Short linkers connect the extracellularly located ATD and ABD to each other as well as the transmembrane domain, which forms the ion-conducting pore of the channel. The binding of glutamate and glycine to the NMDAR ABDs triggers a sequence of conformational changes that leads to opening of a cationselective pore (Hansen et al., 2018).

Each of these different subunit domains is under distinct selective pressure (Li et al., 2019; Ogden et al., 2017; Swanger et al., 2016; XiangWei et al., 2019), with the regions that mediate key functions such as agonist binding, gating, and channel pore opening being highly intolerant to variation. Genetic variation in the GRIN genes that encode the GluN subunits have been associated with a wide range of neurologic and neuropsychiatric disorders that include autism, epileptic encephalopathy, seizure disorders, developmental delay, intellectual disability, motor function impairment, and schizophrenia (Burnashev and Szepetowski, 2015; Camp and Yuan, 2020; Hu et al., 2016; Myers et al., 2019; XiangWei et al., 2018; Yuan et al., 2015). The majority of disease-associated variants appear to reside in regions that are intolerant to variation in the general population (Li et al., 2019; Ogden et al., 2017; Swanger et al., 2016; Traynelis et al., 2017; XiangWei et al., 2019).

Missense genetic variants identified in patients with neurological disease change the protein composition of the gene product by specifying the incorporation of one different residue into the protein. The change in amino acid composition has the potential to change the secondary or tertiary protein structure and function, depending on where it resides and how different the side chain is. Functional and biochemical analysis of mutant recombinant NMDARs expressed in vitro can be used to assess overall consequences of the variation, which can often (but not always) be defined as gain- or loss-of-function (Swanger et al., 2016). If the altered residue resides in a critical region of the protein, it could make the protein more effective (referred to as a gain-of-function), less effective (referred to as a loss-of-function), or otherwise modify the temporal response profile, cellular localization, or posttranslational modification. While various algorithms estimate the probability that a change in amino acid composition will have “deleterious” effects on protein function, reliable information on consequences of a variant can only be determined by biological measurements of protein function. Moreover, once variants are categorized as gain- or loss-of-function, additional work is needed to understand the mechanisms leading to these actions, as well as therapeutic strategies to rectify the functional deficit.

The loss-of-function variants studied here reduce overall charge transfer associated with NMDAR activation, and may drive some initial aspects of clinical phenotypes, raising the possibility that rectification of the functional deficits might mitigate consequences of the mutation. As a first step towards assessing this possibility, we evaluated the effect of three positive allosteric modulators and three co-agonists on 21 loss-of-function variants in GRIN1, GRIN2A, and GRIN2B. These variants cause a range of actions, including reduced agonist potency, reduced surface expression, reduced synaptic-like response time course, and reduced channel open probability. Our results suggest that these positive allosteric modulators can rectify some aspects of the loss of function, which has intriguing therapeutic ramifications.

2. Materials and Methods

2.1. GRIN variants

Table 1 summarizes the 21 missense variants in GRIN1, GRIN2A, and GRIN2B for which functional data were available in the peer-reviewed literature, or for which the variant was described in ClinVar. None of these variants were present in the gnomAD browser (https://gnomad.broadinstitute.org/; evaluated on October 20^th, 2019), suggesting they were absent from the general population.

Table 1:

Summary of GRIN1, GRIN2A, GRIN2B Variants

#	Gene	Genotype	Protein	Location	Phenotype	Functional Consequences	References
1	GRIN1	c.1656C>A, or >G	P.D552E	S1-M1 link	Epi/Sz, ID, DD, MD, CVI, hyPotonia	↓glu/gly Potency, ↓pA/pF	Ohba 2015, Ogden 2017, Lemke 2016
2	GRIN1	c.1858G>A, or >C	P.G620R	M2	ID, hyPotonia	↓glu/gly Potency, ↓Mg2+ block, ↓pA/pF, ↓expression	Lemke 2016, Chen 2017, Li 2019
3	GRIN1	c.1984G>A	P.E662K	M3-S2 link	ID	↑Ba2+ current	Hamdan 2011, Lemke 2016
4	GRIN1	c.2443G>A, or >C	P.G815R	M4	EPi/Sz, ID, DD, MD, CVI, hyPotonia	↓glu potency, ↓τw, ↓pA/pF, ↓POPEN, ↓Pca/PNa	Farwell 2015, Ohba 2015, Helbig 2016, Lemke 2016, Amin 2018
5	GRIN2A	c.1306T>C	P.C436R	ABD (S1)	EPi/Sz, DD, sPeech Problem	↑glu potency, ↓gly potency, ↓pA/pF, ↓charge, ↓expression	Lemke 2013, Swanger 2016, Addis 2017
6	GRIN2A	c.1447G>A	P.G483R	ABD (S1)	EPi/Sz , DD, ID, dysPhasia	↓glu potency, ↓τw, ↓pA/pF, ↓charge, ↓expression	Lesca 2013, Swanger 2016, Addis 2017
7	GRIN2A	c.1642G>C	P.A548T	Pre-M1	EPi/Sz , ID	↓glu potency, ↓gly potency, ↓pA/pF	Lesca 2013, Ogden 2017
8	GRIN2A	c.1928C>A	P.A643D	M3	ID, DD, MD	↑glu potency, ↑gly potency, ↓pA/pF	Fernández-Marmiesse 2018
9	GRIN2A	c.2054T>G	P.V685G	ABD (S2)	EPi/Sz , DD, ID	↓glu potency, ↓τw, ↓pA/pF, ↓charge, ↓expression	Swanger 2016
10	GRIN2A	c.2095C>T	P.P699S	ABD (S2)	EPi/Sz	↑glu potency, ↓POPEN, ↓charge, ↓expression	Lemke 2013, Swanger 2016
11	GRIN2A	c.2113A>G	P.M705V	ABD (S2)	EPi/Sz , DD, sPeech delay	↓glu potency, ↓POPEN, ↓charge, ↓expression	Lemke 2013, Swanger 2016, Addis 2017
12	GRIN2A	c.2146G>A	P.A716T	ABD (S2)	EPi/Sz, DD, verbal dysPraxia, hyPotonia	↓glu potency	Lesca 2013, Fainberg 2016, Swanger 2016
13	GRIN2A	c.2179G>A	P.A727T	ABD (S2)	EPi/Sz , ID, DD, language Problems	↓glu potency, ↓POPEN, ↓charge, ↓expression	Lemke 2013, Swanger 2016, Addis 2017, Von Stülpnagel 2019
14	GRIN2A	c.2191G>A	p.D731N	ABD (S2)	EPi/Sz , ID, DD, verbal dysPraxia	↓glu potency, ↓τw, ↓pA/pF, ↓charge	Lesca 2013, Swanger 2016, Addis 2017, Gao 2017
15	GRIN2A	c.2200G>C	P.V734L	ABD (S2)	EPi/Sz	↓glu potency, ↓τw, ↓charge trans	Lemke 2013, Swanger 2016
16	GRIN2A	c.2314A>G	P.K772E	ABD (S2)	EPi/Sz, learning/reading Problems	↓glu potency, ↓pA/pF, ↓POPEN, ↓charge, ↓expression	Lemke 2013, Swanger 2016
17	GRIN2B	c.1238A>G	P.E413G	ABD (S1)	ID, DD, hyPotonia	↓glu potency, ↓tw, ↓pA/pF, ↓charge, ↓expression	Adams 2014, Swanger 2016, Wells 2018, Bell 2018
18	GRIN2B	c.1306T>C	P.C436R	ABD (S1)	EPi/Sz , ID, ASD	↓pA/pF, ↓expression	Swanger 2016, Platzer 2017
19	GRIN2B	c.1367G>A	P.C456Y	ABD (S1)	ASD, ID	↑glu potency, ↓gly potency, ↓charge, ↓pA/pF, ↓expression	O'Roak 2012, Swanger 2016, Stessman 2017, Fedele 2018
20	GRIN2B	c.1382G>T	P.C461F	ABD (S1)	EPi/Sz , ID, ASD	↓glu potency, ↓gly potency, ↓τw, ↓pA/pF, ↓charge	Swanger 2016, Fedele 2018, Allen 2014
21	GRIN2B	c.1623C>G	P.S541R	S1-M1 link	EPi/Sz , ID, ASD, MD	↓glu/gly potency	Platzer 2017

ATD: amino-terminal domain, ABD: agonist-binding domain (composed by S1 and S2), TMD-link: transmembrane domains (M1, M2, M3, M4) and linker regions (S1-M1 link, M1-M2 link, M2-M3 link, M3-S2 link, and S2-M4 link)

ASD: autism spectrum disorder, CVI: cortical visual impairment, DD: developmental delay (including mental retardation), Epi/Sz: epilepsy/seizures, ID: intellectual disability, MD: movement disorders ↑: increased, ↓: decreased, charge: charge transfer, expression: receptor cell surface expression, glu: glutamate, gly: glycine, pA/pF: current response amplitude (current density), P_Ca/P_Na: relative calcium permeability, P_OPEN: channel open probability, τ_W: weighted deactivation time course tau

2.2. Molecular biology and expression of NMDA receptors in Xenopus laevis

We used cDNA for human NMDA receptor subunit GluN1-1a (hereafter GluN1, NM_007327/ NP_015566), GluN2A (NM_000833/NP_000824) and GluN2B (NM_000834/ NP_000825) subunits subcloned into the pCI-neo plasmid (Promega, Madison, WI). Variants were introduced into cDNA using site-directed mutagenesis via the QuikChange protocol (Stratagene, La Jolla, CA). We subsequently synthesized cRNA from linearized cDNA as described by the manufacturer (mMessage mMachine, Ambion, Austin, TX) (Chen et al., 2017b). Xenopus laevis oocytes were obtained from commercially available ovaries (Xenopus One Inc, Dexter, MI), as previously described (XiangWei et al., 2019). We microinjected 5-10 ng of cRNA (diluted in 50 nl of RNase-free water) into oocytes at a ratio of GluN1:GluN2 that was 1:2. Oocytes were incubated at 15-17°C in Barth’s solution for 1-3 days, as previously described (Chen et al., 2017b; Yuan et al., 2014).

2.3. Two electrode voltage-clamp current recordings from Xenopus oocytes

NMDAR-mediated currents were recorded under two-electrode voltage-clamp (TEVC) from Xenopus laevis oocytes (Chen et al., 2017b; Yuan et al., 2014). The recording solution contained 90 mM NaCl, 1 mM KCl, 10 mM HEPES, 0.5 mM BaCl₂, 0.01 mM EDTA (pH 7.4). Unless stated otherwise, the membrane potential was held at −40 mV and temperature was 23 °C. D-serine, L-serine, and D-cycloserine concentration-response curves were recorded by varying these agonists in the presence of a maximal concentration of glutamate (1000 μM). The current response amplitudes at each concentration were fitted by

$$Response(\%)=100∕(1+(E{C}_{50}∕[agonist]{)}^N)$$ (1)

where EC₅₀ is the concentration that produces a half maximal response, [agonist] is the concentration of agonist, and N is the Hill slope. The concentration-response curves for the positive allosteric modulators were recorded for NMDAR current responses activated by maximally effective concentration of glutamate and glycine, with co-application of variable concentration of positive allosteric modulator. The current response amplitudes were fitted with:

$$Response(\%)=(Maximum-100)∕(1+(E{C}_{50}∕[modulator]{)}^{{}_N})+100$$ (2)

where maximum is the response in saturating concentration of the positive allosteric modulator, EC₅₀ is the concentration of the modulator that produces a half-maximal potentiation, and N is the Hill slope.

The stock concentration of 24(S)-hydroxycholesterol was 10 mM in DMSO, which was aliquoted and stored at room temperature up to one week. Pregnenolone sulfate was made as a 100 mM stock in DMSO, and stored as aliquots at −20°C. Tobramycin, D-serine, L-serine, and D-cycloserine were made as 100 mM stock solutions in deionized water and stored at −20°C. All recording solutions for 24(S)-hydroxycholesterol, pregnenolone, and tobramycin supplemented with 0.02% cremaphor, unless indicated otherwise.

2.4. Whole cell voltage-clamp current recordings from mammalian cells

HEK293 cells (CRL 1573, ATCC; hereafter HEK cells) were maintained in DMEM + GlutaMAX (4.5 g/L D-Glucose, 110 mg/L Sodium Pyruvate. Thermo Fisher, catalogue #10569-010) supplemented with 10% fetal bovine serum, 10 U/ml penicillin, and 10 μg/ml streptomycin. HEK cells were maintained in 5% CO₂ at 37°C and plated on poly-D-lysine coated glass coverslips (0.1 mg/mL) and placed in a 24 well plate for 24-48 hr. HEK cells in a 24-well plate were transiently transfected using calcium phosphate method with 500 ng of DNA at a ratio of 1:1:5 (GluN1:GluN2A:GFP) or 1:1:1 (GluN1:GluN2B:GFP), and four hours after transfection, the NMDAR antagonists D,L-2amino-5-phosphonovalerate (200 μM, DL-APV) and 7-chlorokynurenic acid (200 μM) were added to the culture medium to reduce NMDAR-mediated toxicity (Chen et al., 2017b).

Cells were maintained for 18–24 h in DMEM as described above, after which the coverslip was placed in the recording chamber for whole-cell voltage-clamp recordings (Chen et al., 2017b; Ogden et al., 2017). Recordings were made in an extracellular solution containing 150 mM NaCl, 3 mM KCl, 10 mM HEPES, 0.01 mM EDTA, 0.5 mM CaCl₂ and 11 mM Dmannitol, with the pH adjusted to 7.4 by addition of NaOH (23 °C). The recording electrodes consisted of thin-walled filamented borosilicate glass (TW150F-4; World Precision Instruments, Sarasota, FL, USA), were fabricated using a vertical puller (Narishige P-10, Tokyo, Japan), and 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, 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 (1000 μM) and glycine (30 μM) supplemented with vehicle (DMSO) or allosteric modulator were recorded using an Axopatch 200B patch-clamp amplifier (Molecular Devices, Sunnyvale, CA) at −60 mV. The current responses were filtered at 8 kHz (−3dB, 8 pole Bessel filter, Frequency Devices, Ottawa, IL, USA) and digitized at 20 kHz using a Digidata 1440A and Axon Instruments software (Molecular Devices, Sunnyvale, CA).

The NMDAR deactivation time course following rapid glutamate removal was fitted by a two-component exponential function

$$\text{Response}={\text{Amp}}_{\text{FAST}}\text{exp}(\text{-time}∕{\tau }_{\text{FAST}})+{\text{Amp}}_{\text{SLOW}}\text{exp}(\text{-time}∕{\tau }_{\text{SLOW}})$$ (3)

where τ_FAST is the fast deactivation time constant, τ_SLOW is the slow deactivation time constant, AmpFAST is the current amplitude of the fast deactivation component, and Amp_SLOW is the current amplitude of the slow deactivation component. The weighted deactivation time constant (τ_W) was calculated using the following equation:

$${\tau }_{\mathrm{W}}=[{\text{Amp}}_{\text{FAST}}∕({\text{Amp}}_{(\text{FAST}}+{\text{Amp}}_{\text{SLOW}})]{\tau }_{\text{FAST}}+[{\text{Amp}}_{\text{SLOW}}∕({\text{Amp}}_{\text{FAST}}+{\text{Amp}}_{\text{SLOW}})]{\tau }_{\text{SLOW}}.$$ (4)

2.5. Statistical Analysis

Data are reported as mean ± SEM, and significance tested using unpaired t-test (p < 0.05), or by comparison of overlapping confidence intervals. For comparison of the EC₅₀ value, the confidence interval was determined from the logEC₅₀. For all statistical tests, sample size was adjusted so that the power to detect an effect size of 1 was > 0.9 (GPower 3.0).

3. Results

3.1. Loss-of-function GRIN variants

Here we have selected 21 disease-associated missense variants in the GRIN1, GRIN2A, and GRIN2B genes that have been functionally shown to produce a loss-of-function for the NMDA receptors that these genes encode. These variants are associated with neurological and neuropsychiatric symptoms that include autism (19%, 4/21), epileptic encephalopathy/seizure disorders (76%, 16/21), , intellectual disability/developmental delay (including mental retardation; 86%, 18/21), movement disorders (14%, 3/21), hypotonia (24%, 5/21), cortical vision impairment (9.5%, 2/21), and/or various language/speech problems (29%, 6/21) (Adams et al., 2014; Chen et al., 2017a; Epi et al., 2013; Fainberg et al., 2016; Farwell et al., 2015; Fernandez-Marmiesse et al., 2018; Gao et al., 2017; Hamdan et al., 2011; Helbig et al., 2016; Lemke et al., 2016; Lemke et al., 2013; Lesca et al., 2013; O'Roak et al., 2012; Ohba et al., 2015; Platzer et al., 2017; Stessman et al., 2017; Swanger et al., 2016; von Stulpnagel et al., 2017). Figure 1A shows the location of these variants, which all fall in key domains that either control agonist binding or opening of the ion-conducting pore. Table 1 summarizes the functional effect of the 21 variants that we have selected to study, and shows that the variants reduce function by a variety of mechanisms (Adams et al., 2014; Addis et al., 2017; Amin et al., 2018; Bell et al., 2018; Chen et al., 2017a; Fedele et al., 2018; Fernandez-Marmiesse et al., 2018; Lemke et al., 2016; Li et al., 2019a; Ogden et al., 2017; Platzer et al., 2017; Swanger et al., 2016; Wells et al., 2018), including decreases in agonist potency, decreases in channel opening, decreases in receptor cell surface expression, decrease in calcium permeability, and/or acceleration of the synaptic-like deactivation following removal of agonist (Table 1), which reduces the total current passed by brief activation, as occurs at synapses. The goal of this study is to utilize this functional information to test whether allosteric modulators known to enhance NMDA receptor-mediated charge transfer during activation of the receptor can also enhance the function of variants that diminish the ability of the NMDARs to pass current. That is, we are testing whether positive allosteric modulators can partially rescue functional defects for a range of missense variants.

Figure 1. Effects of positive allosteric modulators on loss-of-function GRIN variants.

A, The locations of the GRIN1 and GRIN2 variants studied are shown in red on a model developed from the GluN1/GluN2 crystallographic data (Karakas and Furukawa, 2014). B-F, Representative current recordings of 24(S)-hydroxycholesterol (24(S)HC, 10 μM), pregnenolone sulfate (PS, 100 μM), and tobramycin (300 μM) on WT GluN1/GluN2A and GluN1/GluN2A-D731N (B,D), and WT GluN1/GluN2B, GluN1/GluN2B-E413G (C,E,F) using two electrode voltage clamp (TEVC) recordings from Xenopus oocytes. G/G is 1-3 mM glutamate/100 μM glycine. G, Concentration-response curves for 24(S)HC on WT and two GluN2B variants. The values for fitted EC₅₀, maximal potentiation, and Hill slope for 24(S)HC were: 0.46 μM, 149%, 0.69 (WT GluN2B); 2.0 μM, 187%, 0.50 (GluN2B-E413G); 0.15 μM, 156%, 0.71 (GluN2B-S541R). H, Concentration-response curves for glutamate on GluN2B-E413G in the absence (open square) and presence (filled circle) of 10 μM 24(S)HC were determined from TEVC from oocytes. Left panel: all data points were normalized to the current response at 3000 μM glutamate in the absence of 24(S)HC. Right panel: data points were normalized to the current response at 3000 μM glutamate in the absence of and in the presence of 24(S)HC, respectively. The values for fitted glutamate EC₅₀ were: 119 μM of control (in the absence of 24(S)HC), 121 μM of in the presence of 24(S)HC. Error bars are SEM.

3.2. Positive allosteric modulators potentiate current responses of loss-of-function GRIN variant receptors

We have tested the effect of three positive allosteric modulators, 24(S)-hydroxycholesterol, pregnenolone sulfate, and tobramycin on NMDA receptor function. 24(S)-hydroxycholesterol is a cholesterol metabolite produced by the cytochrome P450 CYP46A1, the expression of which is largely restricted to the brain, leading to high concentration of 24(S)-hydroxycholesterol in the central nervous system as compared to the rest of the body (Lund et al., 1999). 24(S)-hydroxycholesterol can enhance the maximal response of NMDARs, as well as increase the agonist potency and prolong the deactivation time course (Linsenbardt et al., 2014), with structural determinants associated with the transmembrane domains (Wilding et al., 2016). Pregnenolone sulfate can potentiate neuronal NMDA receptor function (Wu et al., 1991), showing subunit selectivity for GluN2A- and GluN2B-containing NMDARs over GluN2C- and GluN2D-containing NMDARs (Ceccon et al., 2001; Horak et al., 2004; Malayev et al., 2002). Tobramycin is a GluN2B-selective positive allosteric modulator, presumably acting at the polyamine binding site (Masuko et al., 1999). We also test whether D-serine, L-serine, and D-cycloserine can replace glycine to enhance NMDAR function, raising the possibility that therapeutic administration of these agents might allow full occupancy of the glycine site and thereby enhance NMDAR-mediated currents. L-serine can also be converted to D-serine by serine racemase (Wolosker, 2018).

Figure 1B-F shows the effects of these three allosteric modulators and co-agonists on two representative loss-of-function variants, GluN1/GluN2A-D731N (Gao et al., 2017) and GluN1/GluN2B-E413G (Adams et al., 2014; Swanger et al., 2016; Wells et al., 2018). In each case there is substantial potentiation by the positive allosteric modulator of the current response to maximally effective glutamate and glycine for both wild type and mutant NMDARs (Fig 2). 24(S)-hydroxycholesterol potentiated GluN1/GluN2B-E413G and GluN1/GluN2B-S541R activated by maximally effective glutamate plus glycine with similar EC₅₀ values (2.0 μM and 0.15 μM, respectively) as WT GluN1/GluN2B (0.46 μM; Fig 1G, Supplemental Table S1). The effects of 24(S)-hydroxycholesterol on agonist potency were also evaluated. The current amplitude was increased in the presence of 24(S)-hydroxycholesterol (e.g. 241% of the control at 3,000 μM glutamate; Fig 1H, left panel), while the GluN2B-E413G variant receptors showed similar glutamate potency (EC₅₀ values) in both the absence and presence of 24(S)-hydroxycholesterol (119 μM vs. 121 μM, respectively; Fig 1H, right panel). These data suggest that the potentiation of the current responses of the GluN2B-E413G variant by 24(S)-hydroxycholesterol occurs without any detectable change in the glutamate potency.

Figure-2. Summary of percentage changes of the response current of WT and GluN1, GluN2A or GluN2B variant receptors.

Current responses to 1 mM glutamate and 100 μM glycine were recorded under TEVC in oocytes at holding potential −40 mV. A, The ratio of the current response to maximally active glutamate and glycine plus 10 μM 24(S)-hydroxycholesterol (24(S)HC) and vehicle is shown. B, The ratio of the current response to 100 μM pregnenolone sulfate, PS and vehicle is shown. C, The ratio of the current response to 300 μM tobramycin and vehicle is shown. Error bars are 95% CI. Data are from 7-34 oocytes. Filled bars show non-overlapping confidence intervals.

In addition, three glycine site agonists, D-serine, L-serine, and D-cycloserine, could be used to increase NMDAR function when glycine levels are not saturating in brain tissue. These two agonists thus could be potentially useful therapeutic agents to enhance NMDAR function, given that they show similar potency for wild-type and mutant GluN1/GluN2B NMDARs (Fig 3, Table 2, Supplemental Table S2). These results raise the possibility that supplementation of these co-agonists might increase the current response if the glycine binding site is not fully occupied during normal brain function by endogenous glycine.

Figure 3. Effects of co-agonists on loss-of-function GRIN variants.

Concentration-response curves for co-agonists D-serine (A) and D-cycloserine (B) on WT and variant GluN1/GluN2 receptors determined by TEVC recordings from 8-49 oocytes. Glutamate was present at 1000 μM for WT NMDARs, GluN1/GluN2A-D731N, and GluN1/GluN2B-E413G.

Table 2:

Rescue pharmacology for loss-of-function GRIN variants

	I(24(S)HC)/ I(control), %	I(pregnen.sulf.)/ I(control), %	I(tobramycin)/ I(control), %	D-serine EC50, uM	D-cycloserine EC50, uM
WT GluN1/GluN2A	124 ± 4.5% (21)	331 ± 66% (8)	NA	1.3 ± 0.08 (51)	29 ± 0.98 (45)
GluN1-D552E/GluN2A	175 ± 5.7% (11)	219 ± 11% (9)	NA	3.0 ± 0.14 (18)	65 ± 3.7 (18)
GluN1-G620R/GluN2A	158 ± 6.9% (10)	250 ± 10% (15)	NA	2.5 ± 0.06 (10)	35 ±1.2 (10)
GluN1-E662K/GluN2A	155 ± 4.2% (10)	298 ± 43% (9)	NA	1.7 ± 0.06 (13)	41 ±1.3 (16)
GluN1-G815R/GluN2A	158 ± 3.0% (12)	519 ± 19% (12)	NA	3.3 ± 0.16 (13)	55 ± 2.6 (15)
GluN1/GluN2A-C436R	126 ± 3.4% (13)	619 ± 40% (18)	NA	1.8 ± 0.05 (17)	24 ± 1.0 (16)
GluN1/GluN2A-G483R	107 ± 2.7% (11)	126 ± 6.1% (15)	NA	0.89 ± 0.04 (15)	12 ± 1.4 (16)
GluN1/GluN2A-A548T	238 ± 5.1% (11)	391 ± 20% (12)	NA	4.2 ± 0.10 (16)	53 ± 1.2 (15)
GluN1/GluN2A-A643D	101 ± 1.4% (13)	160 ± 9.1% (16)	NA	0.26 ± 0.03 (11)	5.1 ± 0.51 (11)
GluN1/GluN2A-V685G	185 ± 6.2% (12)	225 ± 25% (14)	NA	1.9 ± 0.12 (18)	27 ± 1.1 (13)
GluN1/GluN2A-P699S	133 ± 6.6% (8)	353 ± 50% (9)	NA	1.6 ± 0.07 (17)	28 ± 1.8 (15)
GluN1/GluN2A-M705V	143 ± 2.8% (13)	357 ± 46% (17)	NA	1.6 ± 0.06 (15)	29 ± 0.82 (15)
GluN1/GluN2A-A716T	155 ± 4.7% (13)	302 ± 35% (11)	NA	1.3 ± 0.05 (14)	22 ± 0.68 (14)
GluN1/GluN2A-A727T	147± 5.4% (10)	368 ± 61% (14)	NA	1.4 ± 0.05 (12)	23 ± 0.7 (12)
GluN1/GluN2A-D731N	160± 5.3% (10)	579 ± 43% (11)	NA	1.2 ± 0.07 (10)	11 ± 0.58 (10)
GluN1/GluN2A-V734L	109 ± 2.8% (11)	147 ± 5% (12)	NA	1.2 ± 0.08 (10)	29 ± 0.12 (9)
GluN1/GluN2A-K772E	116 ± 4.3% (12)	151 ± 9% (15)	NA	1.3 ± 0.13 (13)	22 ± 1.2 (10)
WT GluN1/GluN2B	175 ± 4.4% (34)	391 ± 44% (15)	190 ± 7.6% (10)	0.54 ± 0.02 (18)	9.1 ± 0.38 (17)
GluN1-D552E/GluN2B	227 ± 9.2% (10)	534 ± 33% (9)	361 ± 17% (10)	0.68 ± 0.01 (9)	12 ± 0.18 (9)
GluN1-G620R/GluN2B	184 ± 6.3% (10)	601 ± 39% (9)	464 ± 14% (10)	0.61 ± 0.02 (11)	10 ± 0.23 (11)
GluN1-E662K/GluN2B	180 ± 3.6% (10)	415 ± 51% (10)	267 ± 8.3% (11)	0.58 ± 0.03 (10)	9.5 ± 0.22 (8)
GluN1-G815R/GluN2B	197 ± 7.4% (12)	892 ± 188% (6)	392 ± 12% (14)	0.43 ± 0.02 (10)	8.3 ± 0.10 (10)
GluN1/GluN2B-E413G	179 ± 5.8% (15)	387 ± 31% (17)	185 ± 7.0% (19)	0.55 ± 0.02 (13)	10 ± 0.29 (13)
GluN1/GluN2B-C436R	199 ± 4.1% (10)	299 ± 24% (8)	154 ± 2.1% (10)	NA	NA
GluN1/GluN2B-C456Y	139 ± 3.4% (14)	422 ± 40% (27)	97 ± 4.1% (13)	1.5 ± 0.02 (12)	28 ± 0.86 (12)
GluN1/GluN2B-C461F	181 ± 6.1% (9)	465 ± 42% (11)	162 ± 7.7% (7)	0.24 ± 0.02 (12)	5.8 ± 0.46 (14)
GluN1/GluN2B-S541R	184 ± 7.7% (13)	517 ± 66% (18)	334 ± 22% (11)	1.3 ± 0.07 (12)	22 ± 1.00 (11)

Data are expressed as mean ± SEM of the ratio of current responses to modulator and vehicle (10 μM 24(S)-hydroxycholesterol, 100 mM pregnenolone sulfate, and 300 μM tobramycin) or EC₅₀ values (D-serine, D-cycloserine). The number in parentheses is the number of oocytes recorded from. Cremophor was added to solutions for experiments with 24(S)-hydroxycholesterol, pregnenolone sulfate and tobramycin, but not experiments with D-serine and D-cycloserine. NA is not available, since tobramycin has no effect on GluN2A-containing NMDARs.

3.3. Effects of positive allosteric modulators and co-agonists on current-voltage curve relationships

We evaluated the potentiation observed by these three positive allosteric modulators as well as the effects of the co-agonists D-serine and D-cycloserine over range of holding potentials in the nominal absence of extracellular Mg²⁺. Figure 4 shows the current-voltage relationships for NMDAR current responses to a maximally effective concentration of glutamate and glycine co-applied with the indicated concentrations of each of the three positive allosteric modulators. The current-voltage curve was unaffected by 24(S)-hydroxycholesterol and pregnenolone sulfate for WT GluN1/GluN2A and or WT GluN1/GluN2B. We observed a similar result for co-application of these the neurosteroids 24(S)-hydroxycholesterol and pregnenolone sulfate for NMDARs that contained either GluN2A-D731N or GluN2B-E413G (Fig 4A-D). By contrast, we observed modest outward rectification for GluN1/GluN2B in the presence of tobramycin, which mimics that actions of polyamines and is positive charged at physiological pH (Masuko et al., 1999). We similarly noted that modest rectification was produced by tobramycin for variant NMDARs, which we interpret to reflect low affinity channel block by tobramycin, which tends to attenuate potentiation of the response to maximally effective glutamate and glycine at holding potentials more negative than −60 mV. Supplemental Tables S3 and S4 compare the potentiation for each NMDAR variant at −60 mV and +30 mV holding potential. Supplemental Figure S1 shows that, as expected, the current voltage curve (Supplemental Figure S1) for WT and mutant NMDAR currents activated by glutamate and D-serine or D-cycloserine lacked any rectification in the absence of extracellular Mg²⁺.

Figure 4. Positive allosteric modulators enhance NMDA current function without detectable voltage dependence.

Current-voltage relationships for WT GluN1/GluN2A and GluN2A-D731N, WT GluN1/GluN2B, GluN2B-E413G, and GluN1/GluN2B-C456Y are shown for the current response to maximally effective concentrations of the co-agonists (Glu/Gly: 1000 μM glutamate and 100 μM glycine; open circle) and in the agonists supplemented with 10 μM 24(S)-hydroxycholesterol (24(S)HC) (A,B), 100 μM pregnenolone sulfate (PS) (C,D), or 300 μM tobramycin (E). Data are the mean ± SEM from 7-13 oocytes. The measured current response ratios for modulator and vehicle at −60 mV and +30 mV is given in Supplemental Table S3.

3.4. Positive allosteric modulators alter response time course for some GRIN variants.

We subsequently tested the actions of these three positive allosteric modulators on the current response of NMDARs transiently expressed in mammalian HEK cells and recorded under whole cell voltage clamp. Figure 5 illustrates the actions of 24(S)-hydroxycholesterol on WT GluN1/GluN2B and GluN1/GluN2B-E413G. The actions of 24(S)-hydroxycholesterol was similar for both WT and the variant receptor, with strong potentiation of the maximum response to saturating agonists, slowing of deactivation time course, and increasing synaptic-like charge transfer (Fig 5A,B,C). Fitted time constants are given in Supplemental Table S5.

Figure 5. 24(S)-hydroxycholesterol prolongs the NMDAR response time course.

A,B, Representative current responses recorded using whole cell patch clamp configuration from HEK cells co-transfected with GluN1 plus WT GluN2B (A) or GluN2B-E413G (B) in response to 1 mM glutamate and 100 μM glycine with and without 10 μM 24(S)-hydroxycholesterol at a holding potential of −60 mV. Note the slow rise time of the response in the presence of 24(S)-hydroxycholesterol, which is indicative of a use-dependent mechanism. C, 10 μM 24(S)-hydroxycholesterol significantly (p < 0.05, paired student t-test) increased current amplitude (left panel), prolonged deactivation time course (τ_W: weighted tau) (middle panel), and increased charge transfer (right panel) in GluN2B-E413G-containing NMDA receptors. The mean fitted time constants and current response amplitudes are summarized in Supplemental Table S5.

4. Discussion

In this study we report that positive allosteric modulators retain their ability to enhance the response to maximally effective glutamate and glycine for recombinant NMDARs that include subunits with missense variants. This demonstrates that in principle, positive allosteric modulators have the capacity to be considered for their ability to rescue NMDAR function, or at least mitigate some of the effects of the loss-of-function variant. The neurosteroids in particular are intriguing modulators since they act primarily on GluN2A and GluN2B-containing receptors, showing some potential specificity (Ceccon et al., 2001; Horak et al., 2004; Malayev et al., 2002). Even greater subunit selectivity is shown by the aminoglycoside tobramycin (Masuko et al., 1999), although the positive charge at physiological pH, relatively low potency, and potential for side effects makes it difficult to see a path forward for clinical use of this compound to enhance the function of NMDARs that contain GluN2B variants.

We selected a range of variants with missense changes in different domains from three GRIN genes (Table 1). These variants were associated with a subset of pediatric neurological and neuropsychiatric disorders. The data obtained demonstrate that the potentiating actions can occur for variants located in three different subunits, and present in both the transmembrane domain linkers as well as in the agonist binding domain. This suggests that strategies to enhance receptor function might be broadly applicable to loss-of-function receptors that result from missense variants. Similarly, neurosteroids were effective for variants in all three subunits tested (GluN1, GluN2A, GluN2B), raising the possibility that they may provide broader utility against loss of function variants. These subunits harbor the vast majority of GRIN variants described (XiangWei et al., 2018). This is an important point, in that it simplifies diagnostic testing of compounds at every individual variant, making it easier to potentially design clinical studies and use treatments. A recent study indicated that early D-cycloserine treatment of young mutant mice hosting a loss-of-function GluN2B-C456Y variant can correct NMDAR function and NMDAR-dependent synaptic long-term depression, and improve anxiolytic-like behaviors in adult mice (Shin et al., 2020), suggesting application of positive allosteric modulators or supplemental glycine site agonists in early temporal window across CNS development as a strategy for enhancing mutant NMDAR function could be beneficial.

Although none of the allosteric modulators tested are under consideration for FDA approval or approved for neurological conditions, they demonstrate that in principle it is possible to enhance NMDAR function. Tobramycin is a FDA-approved aminoglycoside antibiotic used to treat infections, however its potential ototoxicity and renal toxicity limits its usage in pediatric patients. D-Serine is a potent co-agonist at the glycine-binding site in NMDA receptor that is generated from L-serine by serine racemase (Wolosker 2018). D-serine has a significant modulatory effect in NMDA receptor-mediated neurotransmission and synaptic plasticity (Balu et al., 2012; Billard, 2012). Considerable preclinical and clinical studies suggest that D-serine may be effective in reducing cognitive dysfunction in patients with schizophrenia or major depressive disorder (MacKay et al., 2019). L-serine, D-serine’s stereoisomer and a natural nonessential amino acid, has been administered to a 5-year-old patient with GRIN2B-related Rett-like syndrome and severe encephalopathy as a dietary supplementation (Soto et al., 2019). Adding L-serine to the patient’s food or drink over one year showed significant improvements in all neurodevelopment assessments, including motor skill, cognitive performance, and communication skill, suggesting that this dietary supplementation might mitigate neurological conditions related to NMDA receptor deficiency.

The natural neurosteroids show more potent actions than the aminoglycoside tobramycin, which appears to act in a similar fashion as the polyamine spermine. Our previous study showed that pregnenolone sulfate prolonged the NMDAR deactivation time course, which should prolong synaptic activity and significantly enhance measured charge transfer (Swanger et al., 2016). Exogenous administration of endogenous neurosteroids or their synthetic analogs may be beneficial to treat NMDAR-associated neuropathological conditions. While more work needs to be done, especially testing effects of these positive allosteric modulators on transgenic animals hosting the loss-of-function GRIN variants, these results raise the possibility of pharmacological intervention to mitigate some of the effects of loss of function variants. Evaluation of the effectiveness of FDA-approved drugs and dietary supplements that enhance mutant NMDARs could promote our understanding of channel modification, lead to new ideas about how these drugs work, as well as suggest potential personalized therapies.

Highlights.

• Effects of PAMs and co-agonists on disease-associated GRIN variants were evaluated.

• PAMs can enhance function of NMDARs harboring loss-of-function GRIN variants.

• Co-agonists can augment glycine to enhance NMDAR function.

• Enhancement of NMDAR function by PAMs may be a useful therapeutic strategy.

Abbreviations

ABD

agonist binding domain

ATD

amino terminal domain

CTD

cytosolic carboxyl terminal domain

LoF

loss-of-function

NGS

next-generation sequencing

NMDAR

N-methyl-D-aspartate receptor

TEVC

two-electrode voltage clamp

TMD

transmembrane domain, which contains M1-M4