NMDA receptor hypofunction has been implicated in the pathophysiology of schizophrenia, and pharmacological and genetic approaches have been used to model such dysfunction. We previously have described two mouse lines carrying point mutations in the NMDA receptor glycine binding site,Grin1 D481N andGrin1K483Q , which exhibit 5- and 86-fold reductions in receptor glycine affinity, respectively.Grin1D481N animals exhibit a relatively mild phenotype compatible with a moderate reduction in NMDA receptor function, whereas Grin1K483Q animals die shortly after birth. In this study we have characterized compound heterozygote Grin1D481N/K483Q mice, which are viable and exhibited biphasic NMDA receptor glycine affinities compatible with the presence of each of the two mutated alleles. Grin1D481N/K483Q mice exhibited a marked NMDA receptor hypofunction revealed by deficits in hippocampal long-term potentiation, which were rescued by the glycine site agonist d-serine, which also facilitated NMDA synaptic currents in mutant, but not in wild-type, mice. Analysis of striatal monoamine levels revealed an apparent dopaminergic and serotonergic hyperfunction. Behaviorally,Grin1D481N/K483Q mice were insensitive to acute dizocilpine pretreatment and exhibited increased startle response but normal prepulse inhibition. Most strikingly, mutant mice exhibited a sustained, nonhabituating hyperactivity and increased stereotyped behavior that were resistant to suppression by antipsychotics and the benzodiazepine site agonist Zolpidem. They also displayed a disruption of nest building behavior and were unable to perform a cued learning paradigm in the Morris water maze. We speculate that the severity of NMDA receptor hypofunction in these mice may account for their profound behavioral phenotype and insensitivity to antipsychotics.
The ionotropic glutamate receptors comprise the NMDA, AMPA, and kainate receptor families. NMDA receptors are heteromers composed of an NMDAR1 subunit and one or more of the four NR2 subunits (NR2A–NR2D) (Kutsuwada et al., 1992; Monyer et al., 1992). NMDA receptors are unique among ligand-gated ion channels in their requirement for an obligatory coagonist, glycine (Johnson and Ascher, 1987; Kleckner and Dingledine, 1988), in addition to the synaptically released neurotransmitter glutamate for receptor activation. Electrophysiological studies have demonstrated that NMDA receptor activation requires the occupation of two independent glycine sites and two independent glutamate sites (Benveniste and Mayer, 1991;Clements and Westbrook, 1991). Thus the minimal requirement for a functional NMDA receptor is likely to be a tetramer (Laube et al., 1998) composed of two NMDAR1 and two NR2 subunits that contain the glycine (Kuryatov et al., 1994; Wafford et al., 1995; Hirai et al., 1996; Kew et al., 2000) and glutamate (Laube et al., 1997; Anson et al., 1998) binding sites, respectively.
Pharmacological studies have demonstrated a critical role for NMDA receptor activation in both the induction of certain forms of long-term potentiation (LTP) and learning and memory (Morris et al., 1986; Larson and Lynch, 1988). In addition, the blockade of NMDA receptors in vivo by uncompetitive ion channel blockers such as phencyclidine (PCP) and dizocilpine induces behaviors in rats and mice that include hyperlocomotion, stereotypy, and reduction of prepulse inhibition and that are accompanied by the disruption of neuronal monoaminergic systems (Hiramatsu et al., 1989; Corbett et al., 1995; Geyer et al., 2001). PCP elicits a behavioral syndrome in man that resembles schizophrenia, incorporating positive (e.g., hallucinations, delusions, paranoia), negative (e.g., anhedonia, social isolation, flat affect), and cognitive symptoms (Javitt and Zukin, 1991). These observations provided the basis for the glutamate dysfunction hypothesis for the pathophysiology of schizophrenia; accordingly, PCP- or dizocilpine-treated animals have been adopted as models of the disorder.
Mutant mice exhibiting reduced NMDA receptor expression have been shown to exhibit diverse phenotypes, including developmental abnormalities and postnatal lethality (Forrest et al., 1994; Li et al., 1994;Kutsuwada et al., 1996), impaired synaptic plasticity, and learning and memory (Sakimura et al., 1995; Tsien et al., 1996) and behavioral changes that might model schizophrenia (Mohn et al., 1999; Miyamoto et al., 2001). Mutant mice expressing reduced levels of NMDAR1 exhibited hyperlocomotion, increased stereotypy, and abnormalities in social and sexual interactions that were ameliorated by antipsychotic pharmacotherapy (Mohn et al., 1999). Mutant mice lacking the NR2A subunit also exhibited a neuroleptic-sensitive hyperlocomotion together with an apparent hyperfunction of monoaminergic activity in the striatum and frontal cortex (Miyamoto et al., 2001).
We previously have described two mouse lines generated by site-directed mutagenesis carrying point mutations in the glycine binding site of the NMDAR1 subunit (Grin1, according to the Mouse Genome Database) (Kew et al., 2000). Homozygous mutantGrin1D481N animals, which exhibited a fivefold reduction in receptor glycine affinity, displayed mild deficits in LTP induction and spatial learning and increased startle reactivity but normal locomotor activity and prepulse inhibition. Homozygous mutant Grin1K483Q animals, which exhibited an 86-fold reduction in receptor glycine affinity, did not feed and died within a few days of birth. Here we describe compound heterozygote Grin1D481N/K483Q mice. These animals are viable and exhibited biphasic NMDA receptor glycine affinities compatible with the presence of each of the mutantGrin1 subunits, although receptor glutamate affinity was unchanged.
MATERIALS AND METHODS
Generation of Grin1 compound heterozygote mice. Grin1D481N/D481N andGrin1K483Q/+ mice were generated as described previously (Kew et al., 2000).Grin1D481N/D481N were mated withGrin1K483Q/+ , resulting in offspring carrying one of the mutations on each allele (Grin1D481N/K483Q ) or the D481N mutation on just one allele (Grin1D481N/+ ). Mice were genotyped by PCR on genomic tail DNA by using the primer described previously (Kew et al., 2000) and an additional primer specific for the K483Q mutation (5′-CCG CTC CTG TGT GCC AAA CTG-3′).Grin1D481N/K483Q can be distinguished fromGrin1D481N/+ by an additional amplicon of ∼200 bp size. Grin1D481N/K483Q were fed with wet food starting at approximately day 15 because they were smaller and weaker than heterozygous littermates and had difficulties in reaching normal food pellets and water supplied on top of the cage. C57BL/6J × 129/Ola F1 hybrids were used as wild-type controls.
Whole-cell voltage-clamp recordings from acutely dissociated hippocampal neurons. Brain slices (400 μm) from 5- to 12-d-old wild-type or Grin1D481N/K483Q mice were cut with a vibratome in an ice-cold solution that contained (in mm): 125 NaCl, 2.5 KCl, 2 CaCl2, 1 MgCl2, 1.25 NaH2PO4, 26 NaHCO3, and 25 d-glucose pH-adjusted to 7.4 with oxycarbon (95% O2/5% CO2); subsequently, the slices were incubated at 20°C in the same solution. When needed for electrophysiological experiments, the hippocampus was dissected out of each slice, and neurons were dissociated as described previously (Kew et al., 1998). Whole-cell voltage-clamp recordings were performed as described previously (Kew et al., 1998).
Equilibrium concentration–response curves. Best-fit lines were computed for equilibrium concentration–response data by using a two-equivalent binding site model for monophasic fits: Equation 1where mK D is the microscopic dissociation constant and [A] is the agonist concentration. Data were fit with a 2× two-equivalent binding site model for biphasic fits: Equation 2where I max(H) andI max(L) are the current amplitudes of the high- and low-affinity components of the concentration–response curve and mK D(H) andmK D(L) are the microscopic dissociation constants for the high- and low-affinity components of the curve.
Long-term potentiation. Hippocampal slices (400 μm) were cut from 6- to 7-month-old mice with a Sorvall tissue chopper. Slices were perfused at 35°C with a simple salt solution containing (in mm): 124 NaCl, 2.5 KCl, 2 MgSO4, 2.5 CaCl2, 1.25 KH2PO4, 26 NaHCO3, 10 glucose, and 4 sucrose gassed with 95% O2/5% CO2, pH 7.4. The CA1 stratum radiatum was stimulated (0.05 Hz, 100 μsec) at a stimulus strength adjusted to evoke field EPSPs equal to 30% of the relative maximum amplitude without superimposed population spike. Field EPSPs were recorded from the CA1 stratum radiatum with a glass micropipette (1–3 MΩ) containing 2 m NaCl. After stable baseline recordings the LTP was induced by using a theta burst stimulation (TBS) paradigm consisting of two stimulus patterns spaced by 8 sec. Each pattern consisted of 10 of the 50 msec stimulus trains at 100 Hz, each separated by 150 msec. The duration of the stimulation pulses was doubled during the tetanus. Results are expressed as means ± SE of the field EPSP slope as a percentage of the baseline values recorded 10 min before TBS.
Cortical wedge experiments. Experiments with cortical wedges were performed with the greased gap technique as described previously (Kew et al., 2000). Coronal slices (500 μm) were cut from a 3- to 4-mm-thick block of cerebral cortex/striatum with a vibratome. The tissue was submerged at all times in a simple salt solution containing (in mm): 124 NaCl, 2.5 KCl, 2 MgSO4, 2.5 CaCl2, 1.25 KH2PO4, 26 NaHCO3, 10 glucose, and 4 sucrose gassed with 95% O2/5% CO2, pH 7.4. Tissue wedges ∼1 mm wide consisting of frontoparietal motor cortex, corpus callosum, and underlying striatal matter were dissected from the cortical slices. The wedges were mounted in a Perspex perfusion chamber and perfused continuously with a modified salt solution containing 300 nm tetrodotoxin and 1.75 mmCaCl2 and lacking MgSO4. Population depolarizations of the cortical tissue were evoked by 1-min-duration applications of 20 μm NMDA or 6 μm AMPA, were recorded by using Agar/Ag/AgCl electrodes connected to a direct current amplifier, and were acquired by using MacLab8 software. AMPA applications were made at the beginning and end of the experiments to control for the stability of the preparation. All experiments were performed at room temperature.
Whole-cell hippocampal slice recordings. Hippocampal slices from 26- to 50-d-old mice were cut in a simple salt solution [containing (in mm): 124 NaCl, 2.5 KCl, 2.5 CaCl2, 1.25 KH2PO4, 6 MgCl2, 2 MgSO4, 26 NaHCO3, and 15 glucose gassed with 95% O2/5% CO2, pH 7.4] with a vibratome. Slices were maintained at room temperature in a storage chamber for a least 1 hr. Then single slices were placed in a recording chamber perfused at room temperature with simple salt solution without Mg2+ salts and with 10 μmNBQX and 50 μm picrotoxin. Patch pipettes had resistances of ∼2–4 MΩ when filled with patch pipette solution [containing (in mm): 120 CsF, 10 CsCl, 11 EGTA, 0.5 CaCl2, 10 HEPES, and 5 QX314 pH-adjusted to 7.25 with CsOH and osmolarity-adjusted to 290 mOsm]. Whole-cell recordings were made from CA1 pyramidal cell soma visualized via a 40× water immersion objective and infrared differential interference contrast optics (Olympus BX50WI, Schwerzenbach, Switzerland), using an Axopatch 1D amplifier (Axon Instruments, Foster City, CA). The CA1 stratum radiatum was stimulated (0.03 Hz, 100 μsec) at a stimulus strength adjusted to evoke NMDA field EPSCs of ∼500 pA. At the end of each recording d-AP5 (50 μm) was applied to confirm that the currents were NMDA receptor-mediated. Synaptic events were recorded onto digital audio tape (DTR-1204; BioLogic, Claiz, France). Data acquisition and analysis were performed with pClamp7 (Axon Instruments).
[3 H]-dizocilpine binding. Wild-type and mutant mice forebrains were homogenized individually at 4°C in 25 volumes of 50 mm Tris-HCl and 10 mm EDTA, pH 7.1, buffer with a polytron (10,000 rpm, 30 sec). The homogenate was centrifuged at 48,000 × g for 10 min, and the pellet was rehomogenized as above and incubated for 10 min at 37°C. After centrifugation the pellet was homogenized as above, and the homogenate was frozen at −80°C. [3H]-dizocilpine saturation isotherms were obtained by incubating various amounts of the radioligand (0.3–100 nm, final concentration) in the presence of 10 mg of brain membranes for 2 hr at room temperature in a 5 mm Tris-HCl, 3 mmglycine, and 100 μm glutamate, pH 7.4, binding buffer. The nonspecific binding was measured in the presence of 100 μm 1-[1-(2-thienyl)cyclohexyl]piperidine (TCP). After incubation the membranes were filtered on GF/B glass fiber filters preincubated for 1 hr in a polyethyleneimine 0.1% solution. The filters were washed three times with 3 ml of cold binding buffer, and the radioactivity bound to the membranes was measured by liquid scintillation counting. The binding parametersK D andB max were obtained from the fit to the data of the equation of a rectangular hyperbola (one-site model) by nonlinear regression.
Tissue levels of monoamines. Mice were decapitated. The brains were removed quickly from the skull, briefly washed in ice-cold saline, and laid down on a cooled (4°C) metal plate where they were dissected rapidly to remove the striatum and frontal cortex. The dissected brain regions were frozen, weighed, and stored at –80°C until analysis. The contents of monoamines and their metabolites were determined by using an HPLC system equipped with an electrochemical detector (Coulochem II detector, model 5200; ESA, Chelmsford, MA) essentially as described by Da Prada et al. (1989). Briefly, the striata and frontal cortices were homogenized with an ultrasonic processor in 0.1 m perchloric acid containing 0.5 mm EDTA disodium and 50 ng/ml of 3,4-dihydroxybezylamine as an internal standard. The homogenates were centrifuged at 51,000 × g, and 5 μl of the supernatant was injected in the HPLC. Monoamines and their metabolites were separated on a reverse phase ODS column (YMC-Pack, S-3 μm, 120 A; Stagroma, Switzerland). The column temperature was maintained at 33°C. The mobile phase was 34% citric acid 0.1m, 48% Na2HPO4 0.1m, 18% methanol, 50 mg/l EDTA, and 45 mg/l sodium octylsulfate, pH 4.5; the flow rate was set at 0.45 ml/min. The potential settings of the analytical cell (model 5011; ESA) were +0.45 V (first electrode) and –0.3 V (second electrode). Monoamines and their metabolites were read as the second electrode output signal.
Neurological assessment. The mice were assessed in a number of neurological tests, including flexion reflex, grip strength (g), and time (sec) spent on a rotarod at 16 and 32 rpm (methods as described byHiggins et al., 2001); for the horizontal wire test the mice were held by the tail and required to grip and hang from a 1.5 mm in diameter bar fixed in a horizontal position at a height of 30 cm above the surface for a maximum period of 1 min. The latency for the mice to fall was measured, and a cutoff of either 60 sec or the highest fall latency score from three attempts was used. Body weight (gm) was checked weekly in a second group of mice from an age of 9–24 weeks. Spontaneous locomotor activity was measured in activity chambers (36 × 24 × 19 cm, L × W × H; Benwick Electronics, UK) containing sawdust bedding for a 1 hr period. Mobile counts were measured by the interruption of vertically and horizontally located photo beams placed around the chamber.
Nest building. It was noted while testing the animals that wild-type mice consistently made nests by shredding a tissue that was provided in the cage, whereas most of theGrin1D481N/K483Q mice did not build nests. To quantify this, we placed a folded piece of tissue paper into each cage, and 24 hr later we assessed the nests.
Twenty-four hour locomotor activity. The mice were placed into a novel test chamber that consisted of a Plexiglas box (20 × 20 × 27 cm) with sawdust bedding on the floor. The animal's movement was recorded by using an electronic monitoring system (Omnitech Electronics, Columbus, OH). Movement of the animal resulted in interruption of an array of photo beams from vertically and horizontally located infrared sources placed around the test chamber. Total distance traveled (in centimeters) and stereotypy counts were measured. Animals were placed into the boxes at 6:00 A.M. for a 24 hr period. The room light was on from 6:00 A.M. to 6:00 P.M. and was switched off automatically from 6:00 P.M. to 6:00 A.M.; this corresponds exactly with the animals' normal light/dark cycle in the holding rooms. Food pellets (45 mg) were scattered over the floor, and water was available ad libitum from a drinking bottle that did not interfere with the activity system.
Locomotor activity: drug studies. The mice were tested via a Latin squares design twice weekly with at least a 2 d interval between test sessions. Dizocilpine (0.1, 0.3 mg/kg, i.p.) was administered to wild-type (n = 12) andGrin1D481N/K483Q (n = 7) mice immediately before testing for a 90 min period. Amphetamine (1, 3 mg/kg, i.p.) was administered to wild-type (n = 11) andGrin1D481N/K483Q (n = 5) mice 10 min before testing for a 1 hr period. Clozapine (0.1, 0.3, 1 mg/kg, i.p.) was administered to wild-type (n = 13) andGrin1D481N/K483Q (n = 12) mice 30 min before testing for a 1 hr period. Haloperidol (0.03, 0.1, 0.3 mg/kg, i.p.) was administered to wild-type (n = 10) and Grin1D481N/K483Q (n = 6) mice 30 min before testing for a 1 hr period. M100907 (0.003, 0.03, 0.3 mg/kg, i.p.) was administered to wild-type (n = 12) and Grin1D481N/K483Q (n = 11) mice 30 min before testing for a 1 hr period. Zolpidem (3, 10 mg/kg, i.p.) was administered to wild-type (n = 11) andGrin1D481N/K483Q (n = 5) mice immediately before testing for a 1 hr period.
Prepulse inhibition. Testing was conducted in eight startle devices (SRLAB; San Diego Instruments, San Diego, CA), each consisting of a Plexiglas cylinder (d = 8.8 cm) mounted on a Plexiglas platform in a ventilated sound-attenuated cubicle with a high-frequency loudspeaker (28 cm above the cylinder) producing all acoustic stimuli. Movements within the cylinder were detected and transduced by a piezoelectric accelerometer attached to the Plexiglas base, digitized, and stored by the computer. Beginning at the stimulus onset, 150 of the 1 msec readings were recorded to obtain the startle amplitude (Ouagazzal et al., 2001). The first protocol was used to look at the effect of different prepulse intensities (74, 82, and 90 dB) on the startle pulse (110 dB; prepulse–pulse interval, 100 msec). Then the mice were tested in a second prepulse inhibition (PPI) protocol, using a variable prepulse–pulse interval (interstimulus interval: 30, 100, and 300 msec); prepulse intensity was fixed at 90 dB.
Water maze: cued acquisition. The ability of the wild-type (n = 10) and theGrin1D481N/K483Q (n = 9) mice to learn to swim to and climb onto a visible, flagged platform (7 cm in diameter) in a water maze (1 m in diameter) was assessed. The platform was moved to a different position in the maze on each trial. Cued learning consisted of four sessions of training over 2 consecutive days; each session consisted of three trials (maximum duration of 60 sec) separated by a 10 min intertrial interval. All data were captured and analyzed by video tracking software (HVS Systems, UK).
Statistics. Behavioral observations were recorded among the following: mean values ± SE and analyzed with an unpairedt test, median values with interquartile ranges and analyzed with a Mann–Whitney U test, or proportion of each group and analyzed with a χ2 test. Locomotor activity data (total distance and total stereotypy counts) were analyzed with a two-factor (Genotype and Dose) ANOVA with repeated measures. Comparisons of dose effects in each genotype were undertaken with a repeated measures ANOVA, followed in significant cases by pairedt tests. PPI data were analyzed with an ANOVA, followed in significant cases by a post hoc Newman–Keuls test. Percentage of PPI was calculated according to the formula: [100 × (ST110-PP74, PP82, PP90/ST110] (variable prepulse intensity experiment) or [100 × (ST110-ISI30, ISI100, ISI300/ST110] (variable interstimulus interval experiment). Water maze data were analyzed with a two-factor ANOVA (Genotype and Session) with repeated measures. A p value of <0.05 was accepted as statistically significant.
Severe reduction in NMDA receptor glycine affinity inGrin1D481N/K483Q mice
Glycine and glutamate concentration–response analysis was performed by using whole-cell patch-clamp recordings from acutely dissociated hippocampal neurons from 7-to 10-d-oldGrin1D481N/K483Q mice. Glycine concentration–response curves were generated by rapidly jumping from a control solution to one containing 100 μm NMDA in the presence of increasing concentrations of glycine in both the control and NMDA-containing solutions. Glycine concentration–response curves were biphasic and were better fit by a 2× two-equivalent binding site model than a monophasic two-equivalent binding site model (p < 0.1; F test). The fitted curve through the mean data yielded high- and low-affinitymK D values of 0.22 and 3.41 μm, respectively, with amplitudes of the high- and low-affinity components of 16 and 84%, respectively (Fig.1 A). These glycine affinities are in good agreement with those we have reported previously for homozygous Grin1D481N (mK D = 0.19 μm) and Grin1K483Q (mK D = 3.26 μm) mice (Kew et al., 2000).
Glutamate concentration–response curves were constructed by jumping rapidly from a control solution into one containing increasing concentrations of glutamate in the continuous presence of 100 μm glycine and 10 μm NBQX in the control and glutamate-containing solutions. Glutamate concentration–response curves were monophasic, and a fitted curve through the mean data yielded an mK Dof 3.2 μm (Fig. 1 B). The meanpmK D value of 5.49 ± 0.03 (n = 3) was not significantly different from that we have described previously (Kew et al., 2000) for control mice, 5.70 ± 0.08 (n = 9; two-tailed t test;p > 0.1).
To examine the relative level of NMDA receptor glycine site occupancy in brain slices from control andGrin1D481N/K483Q mice, we used a greased gap cortical wedge technique. In cortical wedges from wild-type animals the level of occupancy of the glycine site is such that the application of NMDA (20 μm) alone results in a robust depolarization. The addition of the NMDA glycine site agonistd-serine, which is not taken up by the GLYT1 glycine transporter (Supplisson and Bergman, 1997), at up to 300 μm resulted in a small, nonsignificant potentiation of control NMDA responses with mean depolarizations of 1.67 ± 0.17 and 1.91 ± 0.23 mV (n = 22) under control conditions and in the presence of 300 μm d-serine, respectively (Fig. 2 A). The addition of d-serine to cortical wedges fromGrin1D481N/K483Q mice resulted in a marked concentration-dependent increase in response amplitude with mean depolarizations of 1.19 ± 0.11 and 2.70 ± 0.33 mV (n = 7–14) under control conditions and in the presence of 300 μm d-serine, respectively. Thus, whereas in cortical wedges from wild-type animals the NMDA receptor glycine site apparently is close to saturated, the d-serine-mediated potentiation of response amplitude in slices from mutant mice reveals a population of nonliganded receptors.
We next assayed the effect of d-serine on synaptically evoked NMDA receptor-mediated currents. NMDA receptor-mediated EPSCs were recorded from identified CA1 pyramidal neurons by using the whole-cell patch-clamp technique in the absence of Mg2+ and in the presence of 10 μm NBQX and 50 μm picrotoxin. Stimulation strength was adjusted to elicit EPSCs of ∼500 pA amplitude. The addition of 100 μm d-serine did not affect the amplitude of EPSCs elicited in hippocampal slices from wild-type mice (97 ± 3%; n = 7) but resulted in a significant increase (263 ± 13%; n = 8; two-tailed t test; p < 0.0001) of EPSC amplitude in hippocampal slices fromGrin1D481N/K483Q mice (Fig.2 B).
Theta burst-induced potentiation in hippocampal slices fromGrin1D481N/K483Q mice was attenuated significantly, relative to wild-type controls, throughout the post-tetanus period (ANOVA; 4–12 min, p < 0.001; 90–120 min, p < 0.01) (Fig. 2 C). The deficit in LTP was rescued by the addition of 100 μm d-serine before the tetanus, although post-tetanic potentiation (PTP) remained somewhat attenuated in slices from mutant mice (ANOVA; 4–12 min, p < 0.05; 90–120 min,p > 0.1). In slices from wild-type mice the addition of d-serine had no effect on either PTP or LTP.
Saturation-binding analysis with the uncompetitive NMDA receptor antagonist [3H]-dizocilpine that used mouse forebrain membranes in the presence of saturating concentrations of glutamate and glycine revealed no significant differences inpK D: wild-type = 8.40 ± 0.07 and Grin1D481N/K483Q = 8.46 ± 0.06 (n = 4 and 3, respectively) or inB max values: (pmol/mg tissue) wild-type = 0.066 ± 0.003 andGrin1D481N/K483Q = 0.070 ± 0.04.
We assayed the tissue content of the neurotransmitters dopamine, noradrenaline, and serotonin in the striatum and frontal cortex of mutant and wild-type mice. Levels of dopamine and serotonin and their metabolites were elevated in the striatum, but not frontal cortex, of mutant relative to wild-type mice (Table1).
Grin1D481N/K483Q mice were impaired significantly in their performance of motor tests, such as grip strength (df = 26; t = 9.7; p < 0.0001) and rotarod at both 16 rpm (z = –4.2; p< 0.0001) and 32 rpm (z = –4.1; p < 0.0001) compared with wild-type mice (Table 2). However, this impairment also may be a consequence of their hyper-reactivity to situations, because it was observed that a proportion of the Grin1D481N/K483Q group would jump off the rotarod repeatedly. In addition, theGrin1D481N/K483Q mice showed increased reactivity to noise and were startled easily. The body posture of 5 of 14 Grin1D481N/K483Q mice differed from wild-type in that they had a raised abdomen. Eight of 14Grin1D481N/K483Q mice also were observed to have self-inflicted wounds, such as scratched and swollen eyes and ears. As a consequence, some of theGrin1D481N/K483Q mice were killed over the course of this study; accordingly, the group size differs in the following experiments. A second group ofGrin1D481N/K483Q mice was assessed weekly from the age of 9 to 24 weeks; it was noted that, although their weight increased over this period (F (11,11) = 59.5; p < 0.0001), it was always significantly less than that of the wild-type group (F (1,31) = 37.4; p < 0.0001) (Fig. 3 A). Furthermore, the Grin1D481N/K483Q mice also were impaired on grip strength and rotarod throughout the testing period (assessed every 2 weeks) compared with wild-type mice (data not shown).
Spontaneous locomotor activity
Grin1D481N/K483Q mice were significantly more active than wild-type mice (F (1,26) = 24.2; p < 0.0001) and did not habituate to the chambers during the test period (Group × Time bin interaction:F (11,286) = 10.1; p < 0.0001) (Fig. 3 B). Total mobile counts (t = –4.9; p < 0.0001) in the 1 hr period were significantly higher in theGrin1D481N/K483Q group compared with the wild-type group (Fig. 3 B, inset). This hyperactivity persisted across repeated testing. A second group ofGrin1D481N/K483Q mice was tested in the same activity chambers every 2 weeks for a 1 hr period from the age of 9 to 24 weeks (data not shown), and the total distance was increased significantly compared with the wild-type group on every test day.
Figure 3 C shows two photographs of a representative mouse from each group: a wild-type mouse with a complete nest and aGrin1D481N/K483Q mouse without a nest. At 24 hr after the introduction of nesting material all of the wild-type mice had built an adequate nest characterized by the mice sitting within a mound of shredded tissue, whereas only one of eightGrin1D481N/K483Q mice had built a nest (Fig. 3 C). In the remainder of cases the tissue paper mainly was untouched by the mutant mice.
Twenty-four hour locomotor activity
The Grin1D481N/K483Q mice had a normal circadian rhythm, i.e., their activity reduced during the light phase and then markedly increased at the beginning of the dark phase (Fig. 3 D). However,Grin1D481N/K483Q mice were more active than wild-type mice throughout the 24 hr test period in both the distance traveled (F (23,23) = 6.9;p < 0.0001) and the number of stereotypy counts (F (23,23) = 6.9; p < 0.0001). The Grin1D481N/K483Q mice did not habituate to their surroundings completely, as indicated by the distance traveled (Group × Time bin interaction:F (23,506) = 1.6; p < 0.05).
Locomotor activity: drug studies
Because the Grin1D481N/K483Q mice were significantly hyperactive, even on repeated testing, compounds from different pharmacological classes were tested to determine whether this behavior could be modified. The compounds that were selected included two psychostimulant drugs, dizocilpine and amphetamine; the “atypical” antipsychotic, clozapine; the “typical” antipsychotic, haloperidol; the 5-HT2A receptor antagonist, M100907 (Kehne et al., 1996); and the benzodiazepine α1 subunit-selective agonist, Zolpidem (Graham et al., 1996). M100907 was selected given accumulating evidence that 5-HT2A receptor antagonists can normalize certain behaviors related to reduced NMDA function (Maurel-Remy et al., 1995; Varty and Higgins, 1995; Carlsson et al., 1999) and Zolpidem because its potent sedative properties are not mediated directly via dopaminergic, serotonergic, or glutamatergic mechanisms (Crestani et al., 2000).
Dizocilpine significantly increased the total distance traveled (F (2,22) = 47.8; p < 0.0001) in the wild-type group at 0.3 mg/kg (p< 0.001), whereas there was not a significant increase in activity in the Grin1D481N/K483Q mice (F (2,12) = 0.4; p = 0.7) (Fig. 4 A). This was supported by a significant Genotype × Dose interaction (F (2,34) = 20.1;p < 0.0001). Dizocilpine significantly increased stereotypy counts (F (2,22) = 24.4;p < 0.0001) in the wild-type group at 0.3 mg/kg (p < 0.001) but did not increase activity in the Grin1D481N/K483Q mice, as shown by a significant Genotype × Dose interaction (F (2,12) = 16.6; p < 0.0001) (Fig. 4 B).
Amphetamine significantly increased the distance traveled (F (3,30) = 44.4; p < 0.0001) in the wild-type group at 1 mg/kg (p < 0.01) and 3 mg/kg (p < 0.0001). There was also a significant increase in total distance in theGrin1D481N/K483Q mice (F (3,12) = 10.4; p < 0.01) after 3 mg/kg amphetamine (p < 0.001) (Fig. 4 C). This was supported by a main effect of dose (F (3,3) = 32.6; p < 0.0001). Amphetamine also significantly increased stereotypy counts (F (3,30) = 66.9; p < 0.0001) in the wild-type group at 1 mg/kg (p < 0.01) and 3 mg/kg (p < 0.0001). There was a significant effect (F (3,12) = 7.7; p < 0.01) of amphetamine on stereotypy counts in theGrin1D481N/K483Q mice at 3 mg/kg (p < 0.05) (Fig. 4 D). This is supported by a main effect of dose (F (3,3) = 46.4; p < 0.0001).
Clozapine significantly reduced the distance traveled (F (3,36) = 19.3; p < 0.0001) in the wild-type group at 0.3 mg/kg (p< 0.01) and 1 mg/kg (p < 0.0001). There was not a significant effect of clozapine on total distance in theGrin1D481N/K483Q mice (F (3,27) = 1.9; p = 0.2) (Fig. 5 A). Clozapine also significantly decreased stereotypy counts (F (3,36) = 24.9; p < 0.0001) in the wild-type group at 0.1 mg/kg (p< 0.01), 0.3 mg/kg (p < 0.001), and 1 mg/kg (p < 0.0001). There was also a significant effect (F (3,27) = 5.01;p < 0.01) of clozapine on stereotypy counts in theGrin1D481N/K483Q mice at 1 mg/kg (p < 0.01) (Fig. 5 B).
Haloperidol significantly reduced the distance traveled (F (3,27) = 11.4; p < 0.0001) in the wild-type group at 0.1 mg/kg (p< 0.05) and 0.3 mg/kg (p < 0.01). There was not a significant effect of haloperidol on total distance in theGrin1D481N/K483Q mice (F (3,15) = 1.4; p = 0.3) (Fig. 5 C). Haloperidol significantly decreased stereotypy counts (F (3,27) = 12.5;p < 0.0001) in the wild-type group at 0.1 mg/kg (p < 0.05) and 0.3 mg/kg (p < 0.01). There was also a significant effect (F (3,15) = 6.3; p < 0.01) of haloperidol on stereotypy counts in theGrin1D481N/K483Q mice at 0.3 mg/kg (p < 0.05) (Fig. 5 D).
M100907 significantly reduced the distance traveled (F (3,33) = 15.5; p < 0.0001) in the wild-type group at 0.3 mg/kg (p< 0.0001). There was not a significant effect of M100907 on total distance in the Grin1D481N/K483Q mice (F (3,30) = 0.7; p = 0.6) (Fig. 5 E). M100907 significantly decreased stereotypy counts (F (3,33) = 13.2;p < 0.0001) in the wild-type group at 0.03 mg/kg (p < 0.05) and 0.3 mg/kg (p < 0.0001). There was not a significant effect (F (3,30) = 1.2;p = 0.3) of M100907 on stereotypy counts in theGrin1D481N/K483Q mice (Fig.5 F).
Zolpidem significantly reduced the distance traveled (F (2,20) = 26.4; p < 0.0001) in the wild-type group at 3 mg/kg (p < 0.01) and 10 mg/kg (p < 0.001). There was not a significant effect of Zolpidem on total distance in theGrin1D481N/K483Q mice (F (2,8) = 0.5; p = 0.6) (Fig. 5 G). Zolpidem significantly decreased stereotypy counts (F (2,20) = 69.6;p < 0.0001) in the wild-type group at 3 mg/kg (p < 0.0001) and 10 mg/kg (p < 0.0001). There was not a significant effect (F (2,8) = 0.08;p = 0.9) of Zolpidem on stereotypy counts in theGrin1D481N/K483Q mice (Fig.5 H).
The Grin1D481N/K483Q mice had exaggerated startle reactivity at 82, 90, and 110 dB (F (4,104) = 14.9; p < 0.0001) (Fig. 6 A) but normal PPI (F (2,52) = 1.1;p = 0.3) compared with the wild-type mice (Fig.6 B). Under a second protocol the startle reactivity again was increased significantly inGrin1D481N/K483Q mice compared with wild-type mice (data not shown). There was a small reduction in PPI in the Grin1D481N/K483Q group compared with the wild-type group, with an interstimulus interval of 30 msec (Fig.6 C). However, there was not an interaction between genotype and interval (F (2,52) = 0.6;p = 0.6), and this effect may be attributable to interference from the increased startle response by theGrin1D481N/K483Q mice to the prepulse with this short prepulse–pulse interval.
Water maze: cued acquisition
Previous work has demonstrated that homozygousGrin1D481N mice are mildly impaired on a spatial learning task (Kew et al., 2000); therefore, we decided to assess the Grin1D481N/K483Q mice in a water maze task. Because of the neurological deficits noted with these mice, they were trained initially on a cued task (visible platform) to ensure that they were able to locate and climb onto a platform before the spatial learning task. The wild-type group learned to find a visible platform across the testing sessions, as indicated by a significant reduction in path length (F (3,3) = 11.0; p < 0.0001). However, the Grin1D481N/K483Q mice failed to learn to find a visible platform in the water maze across the sessions (Genotype × Session:F (3,48) = 6.0; p < 0.01) (Fig. 7 A). Figure7 B represents the swim paths of a mouse from the wild-type group, showing that by the third session the mouse swam directly toward the platform. The Grin1D481N/K483Q mice never found the platform during testing. When placed into the maze, the mice swam quickly and erratically, often in circles with no clear search strategy (Fig. 7 C). Most of the mice had difficulty in swimming (one mouse had to be removed because it was unable to swim), and, when placed onto the platform, the mice would jump off repeatedly, similar to the behavior noted with some mice on the rotarod. Because the mice were unable to learn this cued task, no further water maze testing was undertaken.
We have reported previously that homozygousGrin1D481N mice, exhibiting a fivefold reduction in NMDA receptor glycine affinity, survive to adulthood, whereas homozygous Grin1K483Q animals do not feed and die within a few days of birth (Kew et al., 2000). In this study we have characterized compound heterozygousGrin1D481N/K483Q mice. These mice are viable and exhibit biphasic NMDA receptor glycine concentration–response curves compatible with the presence of each of the two mutated NMDAR1 subunits, although receptor glutamate affinity is unchanged.
The high- and low-affinity components of the biphasic glycine concentration–response curves obtained from theGrin1D481N/K483Q mice are in good agreement with those of the monophasic curves previously determined in homozygous Grin1D481N andGrin1K483Q animals, respectively (Kew et al., 2000). Interestingly, no intermediate affinities were observed. The proportions of the high- and low-affinity components that were observed (16 and 84%, respectively) are compatible with a tetrameric receptor stoichiometry of two NMDAR1 and two NR2 subunits (Laube et al., 1998), subject to a number of assumptions. These include the equal expression and random assembly of the two mutated NMDAR1 subunits, the independence of the two glycine binding sites (which is supported by the absence of any discernible intermediate glycine affinities), and the determination of the net receptor glycine affinity by the lower affinity of the two obligatory glycine binding sites. With these assumptions in place, any receptor containing aGrin1K483Q NMDAR1 subunit must exhibit a low affinity; thus only receptors containing twoGrin1D481N subunits can exhibit a high affinity so that for an NMDA receptor population of tetramers containing two NMDAR1 and two NR2 subunits, a 25% high-affinity, 75% low-affinity distribution would be predicted. This prediction is similar to the ratios that have been determined experimentally. Saturation binding analysis with [3H]-dizocilpine in the presence of saturating glutamate and glycine revealed no differences inK D orB max between wild-type andGrin1D481N/K483Q mice, demonstrating that the number of functional membrane receptors does not differ between wild-type and mutant animals.
The behavioral changes we have reported previously inGrin1D481N homozygous animals were apparent even with a relatively small (fivefold) reduction in receptor glycine affinity, suggesting that the ambient glycine affinity in wild-type animals is not far above threshold (Kew et al., 2000). Because the large majority of NMDA receptors inGrin1D481N/K483Q mice appears to exhibit an ∼90-fold reduction in glycine affinity as revealed by concentration–response analysis, it is likely that, in the absence of any compensatory mechanism, a far larger proportion of NMDA receptors will not be functional in these compound heterozygote animals. Both population NMDA receptor responses assayed by using a greased gap cortical wedge preparation and NMDA receptor EPSCs were potentiated significantly by d-serine in slices from mutant, but not wild-type, animals, illustrating that a significant proportion of NMDA receptors is nonfunctional because of the presence of limiting concentrations of endogenous glycine site ligands. The impact of this reduction in functional NMDA receptors on receptor-dependent signaling was illustrated by the large deficit in LTP induction observed in hippocampal slices from Grin1D481N/K483Q mice, which was rescued by the addition ofd-serine (100 μm). PTP remained somewhat attenuated in mutant animals in the presence ofd-serine, which is a full agonist at the NMDA receptor glycine site with a similar affinity to glycine (Priestley et al., 1995). Thus the PTP and LTP phases of potentiation might be differentially dependent on the extent of NMDA receptor activation. NR2A-containing NMDA receptors, which are expressed together with NR2B in hippocampal slices from mature mice (Kew et al., 1998), exhibit an ∼10-fold reduction in glycine affinity relative to those containing NR2B as the sole NR2 subunit (Kutsuwada et al., 1992; Priestley et al., 1995; Kew et al., 1998). Thus in adultGrin1D481N/K483Q animals the saturating glycine concentration is likely to be in the millimolar range, as suggested by our cortical wedge data, somewhat higher than that we observed in our concentration–response analysis, using neurons from 7 to 12 animals in which NR2B is likely to be the predominant NR2 subunit. The absence of the facilitatory effect ofd-serine on NMDA receptor-mediated activity in all brain slices from wild-type animals also suggests that under our experimental conditions the glycine site agonist concentration at NMDA receptors in these slices is at, or close to, saturating levels.
The most striking behavioral phenotype of theGrin1D481N/K483Q mice was hyperactivity, which persisted over repeated testing and was evident over a 24 hr period, although some circadian pattern was evident. There was also a significant increase in stereotypy counts recorded in these experiments, which through visual inspection suggested a relationship with excessive grooming and scratching in the mutants. Indeed, the severity of this behavior was such that over the course of these studies many of the Grin1D481N/K483Q mice developed self-inflicted facial lesions, resulting in the loss of some animals because of health considerations.Grin1D481N/K483Q mice also were impaired in nest building, suggesting further disruption to normal home cage behaviors.
During the neurological assessment it was noted that theGrin1D481N/K483Q mice were hyper-reactive to noise. This was assessed formally in PPI tests. TheGrin1D481N/K483Q mice had a significantly increased startle response compared with controls, with a startle threshold at 82 dB. Homozygous mutantGrin1D481N mice also show increased startle reactivity, with a threshold startle at 90 dB (Kew et al., 2000). This indicates that the increased startle response is dependent on the degree of reduction in receptor glycine affinity. There was no PPI difference between the wild-type andGrin1D481N/K483Q group, which is also consistent with the homozygous Grin1D481N mice (Kew et al., 2000).
Given the robust hyperactivity seen in theGrin1D481N/K483Q mice, we focused drug interaction studies on this aspect of behavior. Dizocilpine elicited a significant increase in locomotor and stereotypy counts in wild-type, but not Grin1D481N/K483Q , mice, thus providing in vivo confirmation of reduced NMDA receptor tone in the Grin1D481N/K483Q mice. Interestingly, even if we take into account the baseline phenotypic differences, the magnitude of amphetamine effect on activity and stereotypy measures was similar. This indicates that the null effect of dizocilpine pretreatment on motor function inGrin1D481N/K483Q mice is not attributable to a ceiling effect.
Clozapine did not affect the activity of theGrin1D481N/K483Q mice across a dose range that significantly reduced locomotion and stereotypy counts in the wild-type mice. Only at the highest dose that was tested (1 mg/kg) was there a slight, nonsignificant, reduction in the distance traveled and a significant reduction in stereotypy counts in theGrin1D481N/K483Q mice. Haloperidol, M100907, and Zolpidem, in particular, also exerted relatively little effect on activity or stereotypy measures ofGrin1D481N/K483Q compared with the wild-type mice. Taken together, the results clearly show that the hyperlocomotor phenotype ofGrin1D481N/K483Q mice is resistant to pharmacological suppression.
It is of interest to note that, unlike channel blockers (e.g., PCP and dizocilpine), NMDA receptor glycine site antagonists such as L-701,324 produce little or no stimulation of locomotor activity at anti-convulsant doses in mice (Bristow et al., 1996), with pronounced ataxia evident at higher doses. However, L-687,414, a low-efficacy partial agonist, exhibited stimulant activity at high doses (Tricklebank et al., 1994). The reason for these discrepancies is unclear but might relate to the rate of block of NMDA receptors. Full glycine site antagonists of the L-701,324 class penetrate the CNS quite poorly when compared with PCP, dizocilpine, or L-687,414, and studies with competitive glutamate site antagonists indicate that rapid intravenous dosing induces more severe side effects than drug administration by routes with slower rates of adsorption (Lowe et al., 1994).
Nevertheless, in multiple studies with rodents the reduction of NMDA receptor activity either pharmacologically or genetically has been reported to result in behaviors including hyperactivity, increased stereotypy, and deficits in social interaction together with the alteration of monoaminergic systems, and many of these changes can be attenuated by antipsychotic drugs (Hiramatsu et al., 1989; Corbett et al., 1995; Mohn et al., 1999; Miyamoto et al., 2001). Accordingly, these strategies have been adopted to model schizophrenia.Grin1D481N/K483Q mice exhibited a phenotype with similarities to other mice with genetically reduced NMDA receptor function. However, in contrast to the previous studies the hyperactivity exhibited byGrin1D481N/K483Q mice was not attenuated by antipsychotic drugs at concentrations that were without effect in wild-type animals (Mohn et al., 1999; Miyamoto et al., 2001). Moreover, even exposure to higher levels of antipsychotic drugs that were markedly sedative in wild-type animals did not reduce the hyperlocomotion and only partially attenuated the increased stereotypy of Grin1D481N/K483Q mice.Grin1D481N/K483Q mice also were distinguished by the sustained, nonhabituating nature of the observed hyperactivity. It is possible that the apparently more severe phenotype observed with Grin1D481N/K483Q mice reflects a relatively larger deficit in NMDA receptor function compared with the NR2A knock-out (Miyamoto et al., 2001) and NR1 “knock-down” mice (Mohn et al., 1999). In agreement, the severity of the phenotype prevented the inclusion of cognitive tests such as spatial water maze learning, previously performed with NR2A knock-outs (Sakimura et al., 1995). A further distinction is thatGrin1D481N/K483Q mice exhibit reduced NMDA receptor function while retaining a full receptor complement, whereas the NR2A knock-out and NR1 knock-down mice both exhibit significant reductions in the absolute number of receptors (Mohn et al., 1999;Miyamoto et al., 2001). Such a marked loss of a major component of postsynaptic density may result in compensatory changes that could contribute to their phenotype.
The lack of effect of the benzodiazepine agonist Zolpidem on the hyperactivity of the Grin1D481N/K483Q mice is of potential interest. This may reflect a reduction of GABAergic tone in these mice, i.e., a disinhibition, which would be compatible with the observed hyperactivity and excessive stereotypy. Indeed, GABAergic dysfunction has been proposed as a contributing factor to the pathophysiology of schizophrenia (Lewis and Lieberman, 2000; Carlsson et al., 2001). Disinhibition as a result of NMDA receptor hypofunction has emerged as a core principle of the NMDA receptor model of schizophrenia proposed by Olney et al. (1999). The severity of NMDA receptor hypofunction in Grin1D481N/K483Q mice may render these mice of limited value as models of this disease, at least in terms of response to pharmacological challenge. Nonetheless, the potential for a more subtle disruption of NMDA function via the generation of heterozygousGrin1K483Q or heterozygous and homozygousGrin1D481N mice might yield improved models for studying diseases related to NMDA receptor hypofunction such as schizophrenia.
↵* T.M.B. and M.P.-E. contributed equally to this work.
We thank Marie-Claire Pflimlin, Urs Humbel, Sylvie Chaboz, Hans Ehrsam, Patricia Glaentzlin, Caroline Kuhn, Bernard Morand, Stefanie Saenger, and Yeter Kolb for expert technical assistance.
Correspondence should be addressed to James N. C. Kew at his present address: Psychiatry Centre for Excellence, GlaxoSmithKline, New Frontiers Science Park, Third Avenue, Harlow, Essex CM19 5AW, UK. E-mail:.
Abdel-Mouttalib Ouagazzal's present address: Institut de Génétique et de Biologie Moléculaire et Cellulaire, 64704 Illkirch, France.