Classification of missense variants in the N-methyl-d-aspartate receptor GRIN gene family as gain- or loss-of-function

Abstract Advances in sequencing technology have generated a large amount of genetic data from patients with neurological conditions. These data have provided diagnosis of many rare diseases, including a number of pathogenic de novo missense variants in GRIN genes encoding N-methyl-d-aspartate receptors (NMDARs). To understand the ramifications for neurons and brain circuits affected by rare patient variants, functional analysis of the variant receptor is necessary in model systems. For NMDARs, this functional analysis needs to assess multiple properties in order to understand how variants could impact receptor function in neurons. One can then use these data to determine whether the overall actions will increase or decrease NMDAR-mediated charge transfer. Here, we describe an analytical and comprehensive framework by which to categorize GRIN variants as either gain-of-function (GoF) or loss-of-function (LoF) and apply this approach to GRIN2B variants identified in patients and the general population. This framework draws on results from six different assays that assess the impact of the variant on NMDAR sensitivity to agonists and endogenous modulators, trafficking to the plasma membrane, response time course and channel open probability. We propose to integrate data from multiple in vitro assays to arrive at a variant classification, and suggest threshold levels that guide confidence. The data supporting GoF and LoF determination are essential to assessing pathogenicity and patient stratification for clinical trials as personalized pharmacological and genetic agents that can enhance or reduce receptor function are advanced. This approach to functional variant classification can generalize to other disorders associated with missense variants.


Additional Assay Protocols that Allow Future Analysis of Variant Properties
We have previously published analysis of pH and Zn 2+ inhibition of NMDA receptors (see https://grinportal.broadinstitute.org/)using standard protocols.We suggest that it would be helpful to determine these parameters for variant NMDA receptors to obtain additional data that could be used in the future to further stratify patients and explore correlations in receptor function and clinical phenotype.We suggest that future experiments follow the methods previously reported to ensure data is harmonized with that already published (e.g.Yuan et al., 2014;Ogden et al., 2017;Chen et al., 2017;Xie et al., 2023).Below are proposed assay methods for these additional parameters.

Zn 2+ Assay
Xenopus laevis oocytes were prepared, maintained, and injected with cRNA encoding human NMDA receptor subunits as described in the Methods.Oocytes expressing recombinant human NMDA receptors were perfused with solution containing (in mM) 90 NaCl, 1 KCl, 0.5 BaCl2, 10 Tricine, and 10 HEPES adjusted to pH 7.3 (at 23 o C) with NaOH.The oocyte membrane potential was held under voltage clamp at -20 mV.After a steady baseline was obtained, oocytes were maximally activated with 50 µM L-glutamate and 50 µM glycine, and then in the continuous presence of maximal L-glutamate and glycine were superfused with increasing concentrations of Zn 2+ ; the concentration varied depending on the specific receptor and variant tested.If the variant reduced agonist potency, the concentration of L-glutamate and/or glycine were increased to be at least 10 times the EC50 values.To achieve low concentrations of extracellular Zn 2+ , ZnCl2 was added to the extracellular recording solution, which contained 10 mM of the Zn 2+ buffer tricine.We used a pKa for tricine of 8.15 to calculate the concentration of ionized tricine at pH 7.3 capable of binding Zn 2+ .The following buffered free Zn 2+ concentrations were achieved by adding nominal concentrations of ZnCl2 to tricine-containing external oocyte recording solution: for 1 nM free Zn 2+ 0.14 M ZnCl2 was added, for 3 nM free Zn 2+ 0.42 M ZnCl2 was added, for 10 nM free Zn 2+ 1.4 M ZnCl2 was added, for 30 nM free Zn 2+ 4.2 M ZnCl2 was added, for 100 nM free Zn 2+ 14 M ZnCl2 was added, for 300 nM free Zn 2+ 42 M ZnCl2 was added (see Traynelis et al., 1998 for equations).If the variant reduced Zn 2+ sensitivity, higher concentrations were tested.ZnCl2 was made up fresh immediately before addition to tricine-containing solution.Response amplitudes at each buffered Zn 2+ concentration should be normalized to the maximum receptor activation levels without Zn 2+ (defined as 100%) and IC50 values obtained by fitting concentration-inhibition data with Equation 2 in the Methods, which was where minimum is the residual percent response in saturating concentration (constrained to be > 0) of Zn 2+ , IC50 is the concentration of Zn 2+ that causes half maximal inhibition, and nH is the Hill slope.Statistical comparison between WT and variant receptor results should be made using a two-tailed unpaired t-test on the log of the fitted IC50 values or directly on the fitted minimum current in saturating concentrations of Zn 2+ .
pH Assay Xenopus laevis oocytes were prepared, maintained, and injected with cRNA encoding human NMDA receptor subunits as described in the Methods.Xenopus oocytes expressing recombinant human NMDA receptors were perfused with solution containing (in mM) 90 NaCl, 1 KCl, 0.5 BaCl2, 10 HEPES, and 0.01 EDTA, adjusted to pH 7.6 with NaOH, and then a portion of this solution adjusted to pH 6.8 with HCl.The order of pH adjustment is important to ensure the same concentration of the permeant ion Na + is present at both pH values.The oocyte membrane potential should be held under voltage clamp at -40 mV.After a steady baseline was obtained in pH 7.6 recording buffer, the oocytes should be activated with maximally effective concentration of co-agonists (100 µM L-glutamate and 100 µM glycine) in pH 7.6 buffer.If the variant reduced agonist potency, the concentration of agonists should be increased so that their concentration is at least 10 times the EC50 value.Following a washout period, the oocytes should be washed in pH 6.8 buffer, and then maximally activated with 100 µM L-glutamate and 100 µM glycine in pH 6.8 buffer.The current at pH 6.8 is then compared to the current at pH 7.6 (defined as 100%).All recordings should be made at 23 o C. Statistical comparison between wild type and variant receptor results should be made using a two-tailed unpaired t-test.S2).For some variants (Supplemental Table S4), we added new experiments for one parameter.

Supplemental Figure S1:
The relationship between log(fold shift in glutamate EC50 calculated as variant/WT) and log (fold shift in weighted tau deactivation calculated as variant/WT).The log(fold shift in glutamate EC50) was determined in oocytes and log (fold shift in tau deactivation) was determined in HEK cells.The relationship can be empirically approximated either by a linear (GRIN1, GRIN2B) or an exponential function (GRIN2A), which allows estimation of fold changes in tau deactivation from fold changes in EC50 for (A) GRIN1 variants (co-expressed with WT GRIN2A/GluN2A), (B) GRIN2A variants, and (C) GRIN2B variants.For GRIN2B, GluN2B-V821F was omitted because the small current amplitude complicated determination of tau and GluN2B-G689S was omitted because the large reduction in potency made determination of EC50 ambiguous.Black symbols are published variants, red symbols are patientderived variants, and blue symbols are variants in gnomAD.There is not enough data for GRIN2D variants (D) to determine the relationship.These relationships are an approximation, and not as reliable as direct measure of tau deactivation, given that other factors can influence the time course for deactivation.The fold change in weighted tau deactivation for GRIN1 and GRIN2B variants can be empirically estimated by Equation S1: Fold change in tau = 10 (slope x log(fold change in EC50) + intercept) where slope and intercept were determined by linear regression.For GRIN1, the slope was -0.5319 and intercept was 0.1137 (R 2 =0.91).For GRIN2B, the slope was -0.6226 and intercept was 0.01787 (R 2 =0.81).The fold change in weighted tau deactivation for GRIN2A variants can be empirically estimated by Equation S2: Fold change in tau = 10 (-0.4629 + 0.4749 exp (-log(fold change in EC50) / 0.8370))

Table S2 .
Predicted synaptic and non-synaptic charge transfer changes relative to WT NMDARs

GRIN Variants Parameters of Variant Actions on Receptor Function Non-Synaptic Function Source
Xie et al. 2023).The threshold for promoting classification to Possible GoF or Possible LoF was a relative fold change in synaptic or non-synaptic charge transfer that was greater 2.5 or less than 0.4, respectively.Potential promotion (if needed, see Table5) of GoF fold values are blue and LoF fold values are red.Supplemental

Table S3 .
Published variant-mediated changes in parameters supporting GoF and LoF status and blue is GoF.The number of high (H) and moderate (M) changes are given when there are no conflicts.Conflicting and subthreshold variants were promoted to Possible GoF or Possible LoF if the fold change in the synaptic (left number) or non-synaptic (right number) charge transfer change was >2.5-fold or <0.4-fold (Supplemental Table