Altered Actions of Memantine and NMDA-Induced Currents in a New Grid2-Deleted Mouse Line

Memantine is a non-competitive antagonist of the N-methyl-d-aspartate (NMDA) receptor, and is an approved drug for the treatment of moderate-to-severe Alzheimer’s disease. We identified a mouse strain with a naturally occurring mutation and an ataxic phenotype that presents with severe leg cramps. To investigate the phenotypes of these mutant mice, we screened several phenotype-modulating drugs and found that memantine (10 mg/kg) disrupted the sense of balance in the mutants. Moreover, the mutant mice showed an attenuated optokinetic response (OKR) and impaired OKR learning, which was also observed in wild-type mice treated with memantine. Microsatellite analyses indicated that the Grid2 gene-deletion is responsible for these phenotypes. Patch-clamp analysis showed a relatively small change in NMDA-dependent current in cultured granule cells from Grid2 gene-deleted mice, suggesting that GRID2 is important for correct NMDA receptor function. In general, NMDA receptors are activated after the activation of non-NMDA receptors, such as AMPA receptors, and AMPA receptor dysregulation also occurs in Grid2 mutant mice. Indeed, the AMPA treatment enhanced memantine susceptibility in wild-type mice, which was indicated by balance sense and OKR impairments. The present study explores a new role for GRID2 and highlights the adverse effects of memantine in different genetic backgrounds.


Introduction
Memantine (3,5-dimethyl-1-adamantanamine) is an approved drug in the treatment of moderate-to-severe Alzheimer's disease (AD). Ca 2+ -mediated excitotoxicity in neurons is one proposed mechanism of AD, and the N-methyl-D-aspartate (NMDA) receptor is one of the major receptor channels responsible for glutamate-induced Ca 2+ influx. Memantine binds and non-competitively inhibits NMDA receptors, which subsequently protects neurons from glutamate-induced excitotoxicity [1,2].
Glutamate is essential for excitatory synaptic transmission mediated by ionotropic glutamate receptors, which include NMDA receptors and non-NMDA receptors, Į-amino-3-hydroxy-5-methyl-4isoxazolepropionic acid (AMPA) and kainate receptors [3]. NMDA receptor activation is dependent on the membrane potential that is evoked by non-NMDA receptor function [4]. Under pathological conditions, the voltage-dependent regulation of NMDA receptors is believed to be impaired, and memantine is thought to alleviate excitotoxicity by the unregulated NMDA receptor.
Unlike memantine, other NMDA receptor antagonists, such as MK-801 [5], are considered neurotoxins because they inhibit normal excitotoxic neurotransmission as well as pathological physiological neurotransmission. Although memantine is considered a relatively safe drug, some adverse effects, such as dizziness [6], have been reported. Such adverse effects are likely to result from individual differences among patients, especially in their genetic backgrounds. However, the causes of adverse effects to memantine have not yet been clarified.
The glutamate receptor ionotropic delta 2 (GRID2, also known as GluRį2) is abundantly expressed in cerebellar Purkinje cells, and shares sequence homology with other glutamate receptors. Despite its name and these homologies, GRID2 does not bind glutamate or other glutamate analogs [7][8][9]. Mice with an impaired Grid2 gene exhibit a broad range of phenotypes, such as cerebellar ataxia, poor motor learning, and memory dysfunction. In addition to the known phenotypes in mice, new phenotypes presumably involving NMDA receptor dysfunction or memantine effects, such as nystagmus (in frogs) [10], oculomotor apraxia (in cats) [11,12], dementia (in humans) [13,14], have been observed in human patients with GRID2 gene deletions [15][16][17][18]. However, there is not sufficient evidence to ascribe these complex symptoms in human patients to GRID2 gene deletions.
Here, we report a new deletion in the mouse Grid2 gene that is accompanied by ataxia. Administration of memantine led to impaired balance in ataxic mice, and the mutant mice showed deficits in the optokinetic response (OKR) and its learning. These optokinetic impairments were also sensitive to memantine. In addition, only a small population of cultured granule cells isolated from the mutant mice showed memantine-sensitive NMDA-induced currents. These phenomena were mimicked in wild-type (WT) mice following co-treatment with memantine and AMPA.

The Hereditary Ataxic Mouse is Sensitive to Memantine
Two ataxic (male and female) mice appeared spontaneously in the same litter from a mating pair of hetero mice with the knockout (KO) allele for Salt-Inducible Kinase 3 (Sik3: Sik3 tm1Htake/+ ) [19] on a C57BL/6J (B6) genetic background. The male was Sik3 +/í , but the female was Sik3 +/+ . Because these ataxic mice failed to produce offspring by natural mating, we performed in vitro fertilization (IVF) using these ataxic mice and confirmed the heredity of the phenotype. To expound the mouse population, the original sperm of the ataxic male was also used for other IVF with the oocyte of normal B6 female, and offspring with the ataxic phenotype were obtained at the F2 generation (numbers of normal male, ataxic male, normal female, and ataxic female were 4, 1, 9, and 2, respectively), suggesting a recessively-inherited phenotype. Despite showing normal forelimb movements, the mutant offspring were characterized by a short-stepped gait and frequent falls due to suspected leg cramps ( Figure 1A).
To examine whether the phenotype was caused by a neurogenic disorder, neuropharmacological compounds were tested in the normal and mutant mice. Because anesthesia (1% isoflurane) [20] alleviated leg cramps in the mutant mice, we first selected anticonvulsant drugs including gamma-aminobutyric acid (GABA) receptor activators and NMDA receptor antagonists ( Table 1). The GABA-A receptor activators felbamate and nitrazepam failed to modulate the phenotypes in the mutant mice. Whereas, memantine (10 mg/kg), an NMDA and 5HT3 receptor antagonist which is also reported to exert agonistic actions for the dopamine D2 receptor [21,22], impaired balance in the mutant mice, but not normal mice ( Figure 1B and Supplementary Movie 1). Movement traces of the mice ( Figure 1C) confirmed that 10 mg/kg memantine had no significant effect on walking in normal mice. Figure 1. Isolation of ataxic mice with memantine susceptibility. (A) A mouse (6-month-old female) that exhibited an ataxic phenotype with rigid hind limbs. The phenotype became more severe with age; (B) Footprint analyses 10 min after memantine treatment (10 mg/kg, 12-week-old male). The soles of the hind limbs were labeled with India ink for a mouse that walked freely on paper. Rollover by the mutant mouse after memantine treatment is shown (lower right); (C) Monitoring mice walking after memantine treatment (10 mg/kg, 12-week-old male). The position of the mouse's head was tracked (left), and the walking distance was recorded for 5 min (right; expressed as the mean and SD; n = 6). * p < 0.05, ** p < 0.01.
Next, we tested other NMDA receptor modulators. Similar to memantine, the NMDA receptor antagonist MK-801 (10 mg/kg) led to balance disturbances in the mutant mice. However, the normal mice also showed balance disturbances at this dose (Table 1). (R)-CPP also produced balance disturbances only in the mutant mice (Supplementary Movie 2), whereas ifenprodil, Ro25-6981, or the AMPA receptor antagonist DNQX did not produce balance disturbances in the mutant or normal mice. However, slower movements were observed in mutant mice following administration of DL-AP7.
Because serotonin reuptake inhibitors are used to treat vertigo [23], and the 5HT3 antagonist ondansetron causes headache and dizziness [24], we examined whether ondansetron affected the mutant mice. Both mutant and normal mice showed decreased activity with ondansetron (10 mg/kg; Supplementary Movie 3), but no effect that was specific to mutant mice was observed. Moreover, enhanced cholinergic signaling following donepezil treatment [21] and activation of dopamine signals by L-Dopa [21,22] had no significant effect on behavior in both normal and mutant mice (Table 1). These results suggest that, in addition to the ataxic phenotype, the mutant mice were characterized by enhanced memantine susceptibility, which was probably due to impaired NMDA receptor functions. The inconsistency in the efficacy of NMDA receptor antagonists on balance disturbance might result from differences in NMDA receptor subunit subtype-specificity and/or effective dose among the antagonists; 10 mg/kg and 10-20 mg/kg might be the threshold doses to affect balance in the mutant mice for memantine and (R)-CPP, respectively.

Impaired OKR and Learning in the Ataxic Mice
Gait disorders are a common pathology in patients with spinocerebellar degeneration, and are sometimes accompanied by vertigo resulting from oculomotor dysfunction [25,26]. Memantine also affects oculomotor functions in patients with cerebellar ataxia [27]. To examine whether our mutant mice also had oculomotor impairments, we measured the OKR and OKR adaptation, a form of cerebellum-dependent learning [28,29], and used the results from these assays to quantify memantine susceptibility (Supplementary Figure S1).
Normal and mutant mice were subjected to the OKR assay and horizontal visual pattern oscillations were given (Figure 2A). The normal mice could track the stimulus screen, whereas the mutant mice could not. Memantine (10 mg/kg) significantly lowered the overall OKR gain throughout a 1 h session in normal mice while the OKR in mutant mice was completely abolished ( Figure 2B).
To evaluate OKR adaptation, the mean OKR gain was plotted for each 10-min interval ( Figure 2C). Saline-injected normal mice (control) exhibited a time-dependent increase in the OKR gain, but no gain increase was observed in the mutant or the memantine-treated normal mice.

Figure 2.
The optokinetic response (OKR) and its susceptibility to memantine. (A) The OKR of normal and mutant mice after intraperitoneal injection of saline (control) or memantine-containing saline (10 mg/kg). Measurements commenced 10 min after injection. Representative OKRs from normal and mutant mice are shown. The relative pupil azimuth is plotted against time. Each trace indicates the average over each 10-min period of a 1-h measurement session. Screen, movement of the stimulus screen; (B) Overall OKR gain throughout the 1-h session with or without memantine. Dots and error bars, mean ± SD; n = 5 for each data point. * p < 0.05, ** p < 0.01, and **** p < 0.0001, multivariate ANOVA. The overall OKR gain was calculated by averaging the OKR gains for all 10-min periods in a single 1-h session (see panel C); (C) Time course of the mean OKR gain for normal (n = 5) and mutant (n = 5) mice. Each dot indicates the average over a 10-min period of a 1-h measurement session. Error bars indicate ±SD. **** p < 0.0001 vs. the control for time × drug interaction, repeated measures ANOVA.

Microsatellite Analysis in the Ataxic Mice
Because the pharmacological analyses were unable to predict the gene responsible for the ataxic phenotype, we performed a microsatellite analysis to identify the gene. Sperm were isolated from an ataxic mouse (N0: C57BL/6J[B6], B6/B6) and used to perform IVF with oocytes from C3H/HeN (C3) mice (normal: C3/C3). Oocytes were prepared from the N1 female (B6/C3 hetero-mice) and used for a second round of IVF using the original sperm from the N0 (B6/B6) ataxic mice, which produced 137 N2 mice (75 non-ataxic (the responsible genomic-region should be B6/C3) and 62 ataxic (the responsible genomic-region should be B6/B6)). Using 60 microsatellite markers and DNA from eight N2 non-ataxic mice, we screened heterogenic-regions composed of both B6 and C3 chromosomes. We observed that D6Mit149 (Chromosome 6, 48.93 cM) was amplified as the heterogenic-type in all eight mice. Next, we analyzed other markers near D6Mit149 using DNA from 70 mice ( Figure 3A) and narrowed the responsible region to 27.3-32.2 cM (D6Mit384-D6Mit243), which was also associated with the phenotype of enhanced memantine susceptibility. When we searched for genes that have known associations with ataxic phenotypes, we noticed the Grid2 gene at 29.77 cM [30]. To determine whether this gene was responsible for the observed phenotype, the Grid2 open reading frame (ORF) was amplified from cDNA prepared from the cerebellum. As shown in Figure 3B, the Grid2 ORF was approximately 1 kb shorter in the mutant mice than in the WT mice. In addition, western blotting for GRID2 revealed that this protein was absent from the mutant cerebellum ( Figure 3C).
Direct sequencing of the mutant ORF suggested that the region from exon 3 to exon 8 might be deleted, which is the longest deletion reported in Grid2 mutant mice [31]. Chromosome walking ( Figure  3D) identified the break points 110 kb upstream from exon 3 and 30 kb downstream from exon 8, which yielded a truncated GRID2 protein similar to that in the Grid2 trp/trp , Grid2 ho8J/ho8J , and Grid2 ho13J/ho13J mutants, but with different flanking peptides ( Figure 3E). Our laboratory code for mutant mouse lines is Htake, thus the mutant allele is named Grid2 Htake .

Altered Sensitivity to NMDA in Cultured Granule Cells of Grid2 Htake/Htake Mice
GRID2 is highly expressed in Purkinje cells, and disruption of GRID2 signaling impairs Purkinje cell functions. However, little or no expression of functional NMDA receptors in Purkinje cells of the adult cerebellum has been reported [32]. Thus, it is possible that Grid2 gene deletion influences cerebellar function by affecting NMDA receptors on granule cells. We examined whether Grid2 Htake/Htake mice possessed memantine-sensitive NMDA receptors on cerebellar granule cells, using granule cell-enriched (Purkinje cell-free) primary cultures from P2 mice. Quantitative PCR ( Figure 4A) and western blot analyses ( Figure 4B) showed that cultured granule cells expressed low, but detectable, levels of Grid2 mRNA and GRID2 protein. No significant difference in NMDAR1 protein level was observed between the cultures derived from normal and Grid2 Htake/Htake mice, which was consistent with the result of mRNA (microarray) analyses (Supplementary Table S1).
We have to note, however, that no significant interaction between GRID2 protein and NMDAR1 protein in the total cerebella or cultured granule cell lysate was observed (Supplementary Figure S2), despite the presence of PKC gamma binding to GRID2 protein [33].
Next, we monitored NMDA-induced currents in cultured granule cells using a whole-cell voltage-clamp technique. At cell densities as low as 1.25 million cells/mL and a holding potential of í90 mV, granule cells from both WT and mutant mice showed no baseline activity ( Figure 4C). In the absence of memantine (before memantine), NMDA (20 μM, 2 s) induced inward currents that were larger in WT cells than in mutant cells. Furthermore, NMDA-induced currents were reduced in the presence of memantine (after memantine: 10 μM, 30 s). However, the magnitude of the memantinedependent suppression of the NMDA-induced currents varied in the individual cells.
To accurately categorize the memantine-sensitive NMDA-induced currents, we divided the cells into two categories based on the ability of memantine to reduce the charge density of NMDA-induced currents by more than 75%. According to this classification, 51.9% of WT cells and only 21.4% of mutant cells were categorized as memantine-sensitive ( Figure 4D). Moreover, the classification (only open circles) revealed that the values of charge density from the mutant cells fell into a lower range, which produced a significant difference in the variance of the charge density between the WT and mutant groups ( Figure 4E). These results suggest that mutant mice might possess a poor variation of NMDA receptor with reduced function on their granule cell population. The raw data for quantitative PCR analyses and normalized levels (inset: n = 4) of Gird2 mRNA are indicated. Total RNA was prepared from total cerebellum, 2-week-cultured cerebellar granule cells from neonates (P2), B16F10 melanoma cells (non-neuronal negative control); (B) Western blotting was performed using WT and Grid2 Htake/Htake (Mut) total cerebella or cultured granule cells. There is no significant difference in NMDAR1 protein level; (C) Whole-cell current responses of cultured granule cells derived from WT and Mut mice to local application of NMDA (20 μM, 2 s) before and after treatment with memantine (10 μM, 30 s). Representative currents recorded from single WT and Mut cells. Arrowheads indicate artifacts due to the opening and closure of the electromagnetic valve controlling delivery of NMDA-containing saline; (D) The examined cells were categorized into two groups by memantine susceptibility, and their populations are represented as % fractions. When more than 75% of NMDA-induced inward current was suppressed by memantine, the cells were categorized into the memantine-sensitive group (white area). Cells that experienced reduced inward current suppression after memantine treatment, or no significant inward current were categorized into the memantine-insensitive group (black area). * p = 0.0178, likelihood ratio test (WT, n = 27; Mut, n = 28); (E) Distribution of the total charge density of NMDA-induced inward currents. Open circles (white) and closed triangles (black) indicate the data from cells categorized as D. The magnitude of the total data (white and black) was not significantly different between the WT and mutant cells (medians, 1.01 and 0.64 pC/pF, respectively). However, in a comparison of the data from the memantine-sensitive cells (white), there was a significant difference in variance between the WT and mutant cells (* p = 0.0120, Brown-Forsythe test).

Mice Treated with Memantine and AMPA Were Unable to Walk Smoothly
GRID2 deficiency results in dysregulation of AMPA receptors [34,35]. To examine whether impaired AMPA receptor functions affected memantine susceptibility, mice were treated with memantine simultaneously with the AMPA receptor agonist AMPA or the antagonist DNQX [36], and the movements of these mice were monitored ( Figure 5A). The mice treated with AMPA (20 mg/kg) walked slowly and sometimes crouched on the floor. However, mice treated with both memantine (10 mg/kg) and AMPA had increased activity and did not stop walking. In addition to these abnormal behaviors, the mice walked with a mild staggering gait and sometimes slipped (roll-over: Figure 5B, Supplementary Movie 4). These combined effects of memantine on mouse behavior were not observed when memantine was administered with DNQX (10 mg/kg), although the mice treated with DNQX were also sometimes crouched. Moreover, in the Grid2 Htake/Htake mice, co-treatment with AMPA and a low dose of memantine (5 mg/kg) caused more evident balance impairment than memantine treatment alone (Supplementary Figure S3), suggesting that GRID2 deficiency may augment the synergistic action of AMPA and memantine. The effect of AMPA co-treatment could not be evaluated with a higher dose of memantine (10 mg/kg) because the maximal effect was induced by this dose of memantine.  (12-week-old WT male, n = 6) after memantine treatment (10 mg/kg) combined with AMPA (20 mg/kg) or the AMPA receptor antagonist DNQX (10 mg/kg). Ten minutes after the treatments, the walking distance for 5 min was expressed as the mean and SD, * p < 0.05 vs. the other conditions (unpaired t-test; p < 0.0001 for overall differences, one-way ANOVA); (B) Number of rollovers in 5 min was counted. NO indicates that rollover was not observed. n = 6 for Control and AMPA + memantine and n = 3 for the other conditions. The occurrence of rollover was significantly different between AMPA + memantine and the other conditions (p < 0.05, Fisher's exact test); (C) OKRs measured in female WT mice with sequential injections of the control saline, AMPA (10 mg/kg), or DNQX (5 mg/kg) and memantine (5 mg/kg). A set of traces indicates representative responses of individual mice. Dotted line, the movement of the stimulus screen; (D) Graph shows the mean and ±SD of OKR gain (n = 3 for each condition). p value (without a bracket, one-tailed unpaired t-test; with brackets, paired t-test) is indicated when significance was observed. Finally, the OKR was monitored after co-treatment with memantine and AMPA or DNQX ( Figure 5C). Mice were first treated with AMPA, DNQX, or saline, and the OKR was subsequently monitored for 5 min (because some mice closed their eyes after treatment with higher doses, AMPA (10 mg/kg) and DNQX (5 m/kg) were used for OKR measurement). To evaluate synergies between memantine and AMPA receptor modulators, the mice were further treated with a low dose of memantine (5 mg/kg), and again subjected to OKR measurements for another 5 min. AMPA significantly impaired the OKR, and the combined treatment with memantine further impaired the OKR ( Figure 5D). These effects were not observed in mice co-treated with memantine and DNQX.

Discussion
Here, we report cross-talk between GRID2 signaling and memantine in mice, which may, in part, account for the adverse effects of memantine in patients with individual differences in congenital or acquired genetic factors, such as GRID2.
Major phenotypes have been identified in GRID2 mutant mice, including impaired motor coordination, learning, and memory. GRID2 is located on the postsynaptic membrane of Purkinje cells and binds to cerebellin precursor protein 1 (CBLN1) and neurexin 1 beta (NRXN1b) [7,8] on the parallel fibers of granule cells [7,37]. On the other hand, the memantine target, NMDA receptors, are also expressed on the developing [38] and adult [39] cerebellar granule cells. Furthermore, contrary to the neurotoxic effects of glutamate, NMDA receptors were found to promote the survival of cultured Purkinje [40] and granule cells [41,42]. However, double-KO mice with disrupted Nr2A and Nr2C genes (encoding two major NMDA receptor subunits in the adult mice cerebellum) demonstrate a mild impairment in motor coordination, but they do not exhibit an ataxic phenotype [43]. This observation suggests that NMDA receptors and GRID2 functions may not be directly linked [44].
Thus, in the present study, pharmaco-behavioral approaches failed to identify Grid2 as a candidate for the cause of the observed phenotypes. Unexpectedly, our efforts resulted in the observation of new phenotypes in our Grid2 deficient mice, which appeared as enhanced memantine susceptibility, which was likely mediated by dysfunctional NMDA receptors [1,2,45]. OKR measurements in Grid2 Htake/Htake mice revealed impaired basal cerebellar functions in eye movement, which was also mimicked by memantine treatment. Dizziness has been reported as a major adverse effect of memantine treatment in humans [6,46]. The cerebellar flocculus is thought to be responsible for the early stage formation of OKR adaptation and its memory [47,48].
The Grid2 gene is located in a hot spot of genomic deletions [31], and a number of mutant lines with defects in this gene have been identified. In addition to naturally occurring mutants, targeted disruption and knock-in mutations of Grid2 have been reported [30,49,50]. In contrast to these loss-of-function mutations, Grid2 Lc (Lurcher) [51] was identified as a spontaneous dominant mutation characterized by cerebellar ataxia and atrophy of Purkinje and granule cells [30]. Physiological studies of Grid2 Lc/+ mice have shown that the Grid2 Lc mutation produces constitutive inward Ca 2+ /Na + currents that induce cell death [52].
The importance of the genetic background on Grid2 Lc/+ mice phenotypes was also reported [47,53]. In congenic Grid2 Lc/+ mice, almost 99.99% of those on a C57BL/6 genetic background lost Purkinje cells, whereas no Purkinje cell loss was observed in Grid2 Lc/+ mice on a 93% C57BL/6 genetic background, indicating that phenotypes in Grid2 Lc/+ mice are highly dependent on their genetic backgrounds. Interestingly, abnormal eye-movement and impaired motor-coordination were only observed in 93% of C57BL/6 background-Grid2 Lc/+ mice possessing Purkinje cells, but not in 99.99% of C57BL/6 background-Grid2 Lc/+ mice without Purkinje cells [47], suggesting that gain of GRID2 signaling is also a cause of motor deficits in the presence of Purkinje cells. In contrast to Grid2 Lc/+ mice, we noticed during gene mapping that memantine-induced balance impairment was observed in Grid2 Htake/Htake mice irrespective of their genetic backgrounds (mixed B6 and C3 backgrounds). This is also the case for random eye movements commonly observed in different Grid2-deleted mice with different genetic backgrounds, Grid2-KO in C57BL/6 and Grid2 ho-15J/15J on a C3HJ background [17], suggested that memantine-induced balance impairment may occur in other Grid2 deficient mice irrespective of their genetic backgrounds.
How does the Grid2 Htake deletion enhance the actions of memantine? GRID2 regulates long-term depression (LTD) at synapses between immature parallel fibers and Purkinje cells by inducing AMPA receptor endocytosis [34,35]. D-Serine is an endogenous ligand for GRID2 and one of the factors that induce LTD [9]. Developing mice that express GRID2 with a disrupted D-serine binding site show impaired motor coordination and learning, suggesting the importance of LTD for motor regulation. On the other hand, NMDA receptor activation requires the removal of Mg 2+ block, which occurs when the membrane potential increases through activation of non-NMDA receptors including AMPA receptors [4].
Cooperative signal-transmission from cerebellar mossy fiber-granule cells to Purkinje cells mediated by NMDA and AMPA receptors has been shown in both Mg 2+ block dependent and independent manners [54]. We observed that co-treatment with memantine and AMPA impaired gait and OKR in wild type mice, suggesting that dysregulation of AMPA receptor function in Grid2 deficient mice may cause the enhanced memantine susceptibility. This may be implicated in the decreased number of memantine-sensitive NMDA-responsible granule cells in Grid2 Htake/Htake mice.
Whilst mapping the gene responsible for the ataxic phenotype, we observed new phenotypes in Grid2 deficient mice, which were latent balance defects, and OKR impairments. These deficits were also mimicked by memantine with AMPA in the impaired WT mice. Because, recently, the wide distributions of Grid2 mRNA and GRID2 protein were reported in the adult rodent brain [55], the phenotypes in Grid2 Htake/Htake mice could be ascribed to not only attenuated NMDA receptor responsiveness in a subset of granule cells but also other cellular mechanisms.
When granule cells are collectively activated in vivo, glutamate spilt over from parallel fiber-Purkinje cell synapses stimulate adjacent interneurons (volume transmission), which in turn exert inhibition of Purkinje cells [44,[56][57][58] crucial for normal motor coordination [59]. The GRID2 mutation may reduce NMDA receptor-mediated excitation of granule cells, and this might decrease the volume transmission and Purkinje cell inhibition via interneurons. Moreover, cerebellar interneurons (basket, stellate, and Golgi cells) also express functional NMDA receptors at their presynaptic membrane [44,[56][57][58] and these receptors could be affected by the change of GRID2 signaling. Taken as a whole, the present study using a naturally occurring Grid2 deleted mouse line may lead to a better understanding of NMDA, AMPA, and GRID2 receptors. Further studies are required to elucidate the precise mechanisms underlying GRID2 signaling.

Reagents
Microsatellite markers were used to identify the chromosomal region responsible for the ataxic phenotype, ( 48.93 cM; D6Mit200, 89.28 cM). To map the end points of the Grid2 deletion, we used following primers: mGrid2 intron 2H F 5'-GCT ACT TTG GTA CAA GTG GAC A and mGrid2 intron 2H R 5'-GAC AAG TTG CTC TCT GTA TCT; mGrid2 Intron 2I F 5'-CAT GCT CAC ATC AAA ATA CAT CAA and mGrid2 Intron 2I R 5'-TGT AAT TGA GGAA AAT ACA TAA T. Other primers used in Figure 3D were prepared with reference to [31]. To identify the Grid2 deletion in Grid2 Htake/Htake mice, we used the following primers for PCR amplification: mGrid2 Intron 2HIF 5'-TGG ATC CTT CTA CGT GCA AC, mGrid2 Intron 8YF2 5'-GGA CCA CAC TGA GGT TCG AAA GA, and mGrid2 Intron 8YR 5'-ATC TCT TGG CAT GCA TTA GAC. The mutant allele of Grid2 Htake produces a PCR band corresponding to a product of approximately 600 bp, and that for the WT allele is approximately 400 bp.

Measurements of Walking Distance and the OKR
Male mice (12-14 weeks old) were placed in a rectangular box (25 cm × 40 cm), and their head position was tracked for 5 min using ANY-maze software (Brain Science Idea, Osaka, Japan). Each mouse was monitored three times, and the longest distance recorded was used in the data analysis.
The OKR was measured in adult mice using previously described methods [28], which are depicted in Supplementary Figure S1. Briefly, the mice were anesthetized with isoflurane (2%) and a stainless steel screw was glued to the skull. The mice were then habituated to the experimental conditions for 2 days before the measurement. Ten minutes before the measurement were made, mice were administered a saline (8 mL/kg) injection with or without memantine (10 mg/kg) intraperitoneally. The mouse was then mounted on a stereotaxic apparatus and exposed to continuous sinusoidal horizontal oscillations (17°, 0.25 Hz) of a cylindrical checkerboard-patterned screen (diameter, 65 cm; single square, 1.8 × 1.8 cm; brightness, ~30 lx). Right eye movement was captured at 30 Hz with an infrared camera. For each image frame, we used a machine vision system to estimate the pupil azimuth from the location of the pupil center. The OKR was expressed on a time plot of the relative pupil azimuth for each round of screen oscillation, with the pupil azimuth at the beginning of backward eye movement set to 0°. An OKR gain was defined as the ratio of the maximal relative pupil azimuth change to that of the screen (17°).

Cell Culture and Electrophysiology
Granule cell-enriched cultures were prepared as previously described [61]. Briefly, 2-day-old (P2) mutant or WT mice were anesthetized by cooling and then sacrificed by decapitation. The cerebella were dissociated with trypsin, plated on poly-L-ornithine-coated plastic dishes (Becton Dickinson, Franklin Lakes, NJ, USA) at 1.25 million cells/mL, and maintained in low-serum, nutrient-supplemented Dulbecco's Modified Eagle Medium/F-12 (Life Technologies; 5% CO2, 37 °C) for 12 days.
Ruptured-patch whole-cell recordings were performed on cultured granule cells. The pipette solution contained 134 mM potassium D-gluconic acid, 7.6 mM KCl, 9 mM KOH, 10 mM NaCl, 1.2 mM MgCl2, 4 mM ATP magnesium salt, 10 mM HEPES, and 0.5 mM EGTA (pH 7.3). The culture dish was perfused at a rate of 1.4 mL/min with 145 mM NaCl, 5 mM KCl, 2 mM CaCl2, 10 mM HEPES, 10 mM D-glucose, and 10 μM glycine (pH 7.4). Current signals were recorded using an EPC-8 amplifier (holding potential, í90 mV; cut-off frequency, 5 kHz; sampling rate, 20 kHz; HEKA, Lambrecht/Pfalz, Germany) controlled by Patchmaster software (version, 2.35; HEKA). The command potentials were corrected for a liquid junction potential between the pipette and bath solutions. Electronic capacitance cancellation and series resistance compensation were not used. The series resistance (33.2 ± 5.1 Mȍ, n = 55) and membrane capacitance were estimated from the amplitude and time constant of the capacitive current evoked by a 10 mV voltage jump. The bath solution containing 20 μM NMDA or 10 μM memantine was locally applied to the cell through a theta tube under the control of gravity and electromagnetic valves (VM8, ALA Scientific Instruments, Farmingdale, NY, USA). The magnitude of an NMDA-induced current was quantified as the inward current charge over a 2 s NMDA application normalized to the membrane capacitance (charge density). The background charge was estimated from a 0.5 s pre-application period and subtracted from the charge density.

Statistical Analyses
Data from each group were characterized by the mean ± SD, unless otherwise stated. Data from biochemical assays were examined with one-way ANOVA followed by unpaired two-tailed t-tests to detect statistically significant differences. All the statistical examinations were performed using JMP software (versions 9.0.2 and 10.0.1, SAS Institute, Cary, NC, USA).

Conflicts of Interest
No conflicts of interest, financial or otherwise, are declared by the author(s).