Enhanced stability of hippocampal place representation caused by reduced magnesium block of NMDA receptors in the dentate gyrus

We selected GluN2A for the dentate GC-specific abrogation of Mg²⁺ blockade of NMDARs because it is abundantly expressed in the hippocampus of adults, but not during early development [19]. To replace the asparagine residue 595 (N595) with glutamine (N595Q) in the GluN2A subunit, we constructed the targeting vector shown in Figure 1A and Additional file 1: Figure S1. We performed a two-step selection with G418 and Cre recombinase to obtain the desired recombinant ES clones (Figure 1A), and these were used to generate the flox homozygous (Figure 1C) mice. The mutually exclusive pre-mRNA splicing of adjacent exons is determined by the proximity of the downstream exon branch point to the upstream exon splice donor site, and a proximity of less than 51 nucleotides completely prevents the adjacent exons from being spliced together [20]. Therefore, the length of the spacer elements between the splice donor site of the wild-type (WT) exon 10 (Figure 1B, orange box) and the branching point of the mutant exon 10 (Figure 1B, green box) was designed to be 48 nucleotides. If the WT exon 10 between two loxP sequences were deleted by the action of Cre recombinase, the mutant exon 10 would then be spliced into mRNA (Figure 1C). GluN2Aflox/flox mice were first crossed with Nestin-Cre mice to confirm that the desired recombination event takes place, and their offspring (GluN2A+/flox-NestinCre) were subjected to analysis since the majority of homozygous GluN2A flox/flox-NestinCre mice died within 2 weeks after birth. In agreement with the distribution of Cre protein (data not shown), Cre-loxP recombination products were observed in the brain and spinal cord (Figure 2A). As expected, there was no significant difference in the size and expression levels of GluN2A mRNA and GluN2A subunits between genotypes (Figure 2B). We next confirmed the exclusive recruitment of WT exon 10 in the GluN2A flox/flox mice prior to Cre-loxP recombination and the substitution of WT mRNA (N595) with mutant mRNA (595Q) after Cre-loxP recombination (Figure 2C). These results clearly demonstrate that our method allows the introduction of a desired mutation into any gene of interest in a tissue-restricted manner without altering the level and pattern of target gene expression.

Under control of the Purkinje Cell Protein 2 (PCP2) promoter [21], one of the transgenic (Tag) lines obtained were mice (TDGCre mice) expressing Cre recombinase intensively in the dentate GCs of the hippocampus (Figure 3A) and in Purkinje cells of the cerebellum, with slight expression in the thalamus (Additional file 2: Figure S2). These expression patterns became detectable postnatally after approximately 3 weeks and were sustained into adulthood. We crossed GluN2Aflox/flox mice with heterozygous TDGCre mice to generate GluN2A+/flox-TDGCre and GluN2Aflox/flox-TDGCre mice. These mice are useful for analyzing the role of GluN2A subunits in the dentate GCs, because (i) the GluN2A subunit is reported to be expressed predominantly in the adult hippocampus and granular layer of the cerebellum, but not in the Purkinje cells, and (ii) NMDAR function in the thalamus is largely dependent on GluN1/GluN2B, GluN2C, and GluN2D complexes [19].

As expected, the Cre-loxP recombination of genomic DNA was selectively induced in the brain, but not in the liver, kidney, muscle, or testis (data not shown). To confirm the switching from wild-type mRNA (N595) to mutant mRNA (595Q) in the DG, brain RNA preparations were analyzed by RT-PCR amplification, followed by digestion of amplified DNA with Nco I (common to both mRNAs) and Eco RI (specific to mutant mRNA) (Figure 3B). Digestion with Eco RI resulted in two bands in the PCR products from the DG of mutant mice, but not from other brain regions (Figure 3B), indicating dentate-specific induction of the N/Q mutation. Consistent with the fact that Cre-recombinase was expressed in nearly half of the dentate GCs in TDG-Cre mice (Figure 3A), approximately half of RT-PCR products from the DG of homozygous mice (GluN2Aflox/flox-TDGCre) were successfully digested with Eco RI (Figure 3B). Notably, we also histologically confirmed that the brains from mutant mice did not show any obvious morphological abnormalities (data not shown).

Based on the above results, we analyzed the physical and neurological features, along with the behavior, of GluN2A+/flox-TDGCre and GluN2Aflox/flox-TDGCre mice. Homozygous and heterozygous mutant mice were viable, fertile, and appeared grossly normal under standard conditions. No significant effects on physical characteristics, neurological reflexes, and pain sensitivity were observed. Because slight abnormalities in motor function (wire hanging and rotarod tests) were observed in homozygous, but not in heterozygous, mice (Table 1 and Additional file 3: Figure S3), we analyzed the physiological and behavioral functions of heterozygous mice (GluN2A+/flox-TDGCre for the mutant and GluN2A+/flox for the control) in detail.

To verify the incorporation of the mutant GluN2A subunit into the functional NMDAR in the DG, we recorded excitatory postsynaptic currents (EPSCs) mediated by NMDAR in the dentate GCs. The medial perforant path (MPP) input was activated, and the current–voltage relationship of pharmacologically isolated NMDA EPSCs was examined. While NMDA EPSCs showed the typical J-shaped voltage dependence in the control and mutant mice, the relative EPSC amplitudes at −60 mV and −80 mV were significantly larger in the mutant mice compared to the control mice (Figure 4A and B). Thus, as expected, the voltage-dependent Mg²⁺ block of synaptic NMDARs was reduced in the GCs of the mutant mice. The reduction of the Mg²⁺ block was apparent in approximately half of the recordings (Figure 4C), which correlates with the expression ratio of Cre-recombinase in the GCs shown in Figure 3A.

Next, we recorded field excitatory postsynaptic potentials (fEPSPs) to examine synaptic plasticity at the MPP input in standard physiological saline. The input–output relationship of fEPSPs was indistinguishable between control and mutant mice (Figure 4D). Paired-pulse stimulation induced the depression of fEPSPs, and the depression ratio was indistinguishable between these mice (0.659 ± 0.016, n = 15 in controls; 0.659 ± 0.019, n = 15 in mutants). Thus, basal synaptic transmission, largely mediated by the AMPA receptor, is apparently not affected in mutant mice. LTP in the DG requires activation of NMDARs and is often difficult to induce in in vitro slices with intact synaptic inhibition [22, 23]. In our experimental conditions, tetanic stimulation at the test intensity failed to induce lasting potentiation and caused reversible short-term potentiation (STP) in both control and mutant mice. This STP was significantly larger in the mutant mice (Figure 4E). Whereas the tetanic stimulation at the increased stimulus intensity failed to induce LTP in the control mice, it consistently induced LTP in the mutant mice (Figure 4F). These results indicate that the induction of both STP and LTP is facilitated in the DG of the mutants.

To test the role of the Mg²⁺ block of DG NMDAR in hippocampal computational function, we analyzed place cell activity in the hippocampus. Place cells are neurons whose firing rate increases when an animal is in a specific location in a given space [14]. Their activity is thought to be involved in spatial navigation. We recorded hippocampal CA1 neurons, the major output of the hippocampal circuit. Mice were allowed to move freely in an open, white, round-cornered square box for a 15-min habituation session once daily (Figure 5A). After at least 7 days of habituation, CA1 place cell activity was recorded in the same box for 15 min (Figure 5B). We recorded the activity of 93 CA1 pyramidal cells and seven interneurons from control mice (n = 8) and 31 CA1 pyramidal cells and six interneurons from mutant mice (n = 5) under freely moving conditions. We first measured the basic firing properties of pyramidal cells and interneurons during exploratory behavior. As shown in Table 2, the mean firing rate, firing rate variability (measured by the Fano factor), and spike shape of the cell types were not significantly different. These data show that CA1 cells from mutants have similar firing properties to those from control mice.

We then assessed the place cell properties of CA1 pyramidal cells in mutant mice. As shown in Figure 5C and Additional file 4: Figure S4, CA1 pyramidal cells in the mutant mice showed robust place-specific discharge. Therefore, the DG-restricted NMDAR mutation did not interfere with place field formation by CA1 cells. To quantitatively compare the specificity of the place fields, we measured the size of the place fields. As shown in Figure 5E, CA1 cells in the mutant mice exhibited no significant change in the size of the firing field. However, we found a difference in the properties of the place fields. In both controls and mutants, many CA1 cells have more than one firing field (Figure 5C and Additional file 4: Figure S4). To explore the properties of the place fields in more detail, we compared the size of the largest firing field (main place field) and the number of firing fields. We found that mutants have a significantly larger main firing field (Figure 5F) and the total number of firing fields was smaller (3.6 ± 0.18 in controls; 2.4 ± 0.20 in mutants, p = 0.00052, Wilcoxon). Measurement of the mean running speed revealed no differences between genotypes (5.89 ± 0.91 cm/s for controls, 5.82 ± 0.02 cm/s for mutants, p = 0.84, Wilcoxon). Therefore, although running speed can influence place cell activity [24], it fails to explain the observed differences in place cell properties.

Because NMDARs have been shown to regulate place field stability [25], one possible mechanism for the larger main place field and the smaller number of firing fields in the mutants is that the temporal fluctuations in the place fields are diminished by the mutation. To test this, we used the Pearson product-moment correlation coefficient for two place fields recorded successively in the same environment (Figure 5D, left). Place fields are known to exhibit fluctuations even when the animal is in the same environment [26], and in particular, larger fluctuations are observed in the freely-moving condition than in food-seeking [27]. Place fields of the mutant mice showed higher correlations than control mice (p = 0.00063, Wilcoxon) (Figure 5G), indicating that the mutants have more stable place fields. In contrast, the shift in the place fields between the two recordings (Figure 5D, right), another measure of similarity between two place fields, exhibited no detectable differences (Figure 5H). These two measures have different properties: place field shift (Figure 5H) is only affected by changes in the high-firing-rate field, whereas the correlations for the place fields (Figure 5G) evaluate the differences between entire place field maps. Therefore, the latter is more sensitive to small changes in place fields. Taking into account the different characteristic of these two measures, the mutation does not affect the persistence of place receptive fields, but reduces small fluctuations in the place fields.

The DG has long been proposed to perform pattern separation, an ability to distinguish between similar input patterns. Therefore, we tested the pattern separation of hippocampal place representation by comparing the place fields of two visually similar but distinct environments. The two testing environments shared the same recording room and box, but the local cue (a paper towel) was rotated 90 degrees clockwise (Figure 6). Place fields were first recorded in the ‘original cue’ condition and then recorded in the ‘rotated cue’ condition. During cue translocation, mice were allowed to rest in their home cage (~1 min). In control mice, most of the cells changed their place fields after local cue rotation (Figure 6, upper and Additional file 5: Figure S5A, upper), which is consistent with previous results in mice [27, 28]. In contrast, as shown in cross correlation histograms (Figure 6, lower-left and Additional file 5: Figure S5 A, lower) and place field shift histograms (Figure 6, lower-right), more than half of the mutant cells were insensitive to local cue rotation. Statistical analyses confirmed that the place fields of the mutant mice in the two environments were more similar than those of the control mice (cross correlation; p = 0.000077, Wilcoxon, and place field shift; p = 0.0072, Wilcoxon). We used a distal cue-centric coordinate frame for this analysis, because correlation coefficients for local cue-centric coordinate frames were significantly lower than those for distal cue (recording room, door, etc)-centric coordinates in both controls and mutants (Additional file 5: Figure S5B and Additional file 6: Figure S6). These results indicate that place representation in the CA1 of the mutant mice is less sensitive to environmental differences. Another possible interpretation is that attention to local cues was only decreased in the mutants. However, this is unlikely because no significant difference was detected in the correlation coefficient for local cue-centric coordinates between controls and mutants (Additional file 6: Figure S6).

We showed that place representations in the mutants were more stable and less sensitive to environmental differences compared with those of the controls (Figures 5 and 6). To test whether the mutation affects hippocampal-related behavior, we examined the spatial working memory of the mutants. In the eight-arm radial maze test, mutants showed spatial learning deficits in the initial phase of training, but they acquired normal performance in the late phase (Figure 7). This result indicates that the mutants have a slight but significant defect in the performance of a task requiring spatial working memory.

Adult neurogenesis in the DG [29] has been shown to contribute to pattern separation [30‐32], and NMDARs are involved in the survival of newborn neurons in the DG [33]. This raised the possibility that the NMDAR mutation might affect the number of adult-born dentate granule cells. To test this, we counted the number of neurons newly produced in the subgranular zone (SGZ) of the DG. There was no detectable difference in numbers of BrdU-positive cells in the SGZ and hilus between the control and mutant mice at 24 h (Additional file 7: Figure S7) and 28 days (data not shown) after BrdU administration. Therefore, impaired pattern separation in our mutant mice was unlikely due to changes in adult neurogenesis.

Because N595 is located in exon 10 of the GluN2A gene, [57], we generated a mouse line (GluN2Aflox) harboring a tandem array of the wild-type (WT) exon 10 encoding N595 and a mutant exon 10 encoding Q595 separated by a 70-nucleotide (nt) artificial intron. This artificial intron was comprised of a 5′-donor sequence (9 nt) and a 3′-acceptor sequence, including the branch point (27 nt) from the mouse β-globin intron and loxP sequences (34 nt) that were inserted between the donor and acceptor sequences (Figure 1B). The length of the spacer elements between the splice donor site of the WT exon 10 (Figure 1B, orange box) and the branch point of the mutant exon 10 (Figure 1B, green box) was designed to be 48 nt. However, it is possible that on occasions, the 5′ normal exon will be skipped due to unforeseen mechanisms. In order to ensure that the normal exon 10 is exclusively spliced into the mRNA, the splicing acceptor element of normal exon 10 was substituted with a 73-nt Sxl-binding splicing acceptor sequence that is preferentially selected among mutually spliced exons in the absence of Sxl protein (notably, Sxl-like protein has not been found in mice) [58] (Figure 1B, violet). The Neo-TK cassette, consisting of neomycin-resistant and thymidine-kinase genes flanked by loxP sequences, was located on the 5′ side of the Sxl-binding splicing acceptor sequences (Figure 1A and B). Two-step selection was performed to obtain the desired recombinant ES clones. The targeting vector was electroporated into TT2 ES cells [59]. After G418 selection, one homologous recombinant clone (out of 399 G418-resistant clones) was identified. This clone was expanded, transfected with a Cre recombinase-expression plasmid (pIC-Cre), and further selected with gancyclovir. We confirmed that one clone out of 216 gancyclovir-resistant clones exhibited precise excision of the loxP-Neo-TK-loxP cassette by Cre-loxP recombination. This secondary recombinant ES clone was then microinjected into 8-cell embryos to generate flox homozygous (GluN2Aflox/flox) mice using standard procedures, and was confirmed by Southern blot analyses.

After the fidelity of PCR genotyping for GluN2A was confirmed, subsequent genotyping performed by PCR using primers Pr137: 5′GTGGTAAAATCCAGTTAGATAG3′ and E1R: 5′GGGTTATAGAATGGATGGTTA3′. PCR conditions were as follows; denaturation at 94°C for 1 min, followed by 30 cycles of 1 min at 94°C, 1 min at 60°C, 1 min at 72°C, a final extension at 72°C for 5 min, and storage at 4°C. The floxed, wild-type and Cre-mediated recombined GluN2A genes correspond to PCR products of 807, 599 and 454 bp, respectively. For analysis of Cre-loxP recombination in various tissues by PCR, genomic DNA samples from various tissue were subjected to the same PCR condition as above, using primers CRE10: 5′CAACGAGTGATGAGGTTCGCAA3′ and CRE13: 5′CCCCAGAAATGCCAGATTACGT3′.

The GluN2A cDNA fragments were amplified using total RNA by RT-PCR with primers mE1F: 5′ATCAACGAGCAGTTATGGCC3′ and mE1R: 5′GTCATGAGGTCTCTGGAACT3′. The amplified DNA fragment of 521 bp was digested by EcoRI or NcoI.

Nestin-Cre and TDG-Cre transgenic lines were generated. The Nestin-Cre mice were produced carrying tandem arrays of the Cre-recombinase gene driven by the rat nestin promoter, the internal ribosome entry sites (IRES) sequence, lacZ gene, and a neuron specific enhancer [60]. In TDG-Cre we used a Purkinje Cell Protein 2 (PCP2) promoter to drive the expression of Cre-recombinase. We used a transgenic strain, CAG-CAT-Z, as a reporter line [61].

A neurological screen was conducted as previously described [62]. The righting, whisker touch, and ear twitch reflexes were evaluated. A number of physical features were also recorded, including the presence of whiskers or bald patches in the hair.

Neuromuscular strength was tested with the wire hanging test. The mouse was placed on a wire mesh and the mesh was then inverted, to force the animals to grip the wire. Latency to fall was recorded, with a 60 sec cut-off time.

Motor coordination and balance were tested using an accelerating rotarod (UGO Basile Accelerating Rotarod). The test was performed by placing a mouse on a rotating drum (3 cm diameter) and the time each animal was able to maintain its balance on the rod was measured. The speed of the rotarod accelerated from 4 to 40 rpm over a 5-min period.

The hot plate test was used to evaluate sensitivity to a painful stimulus. Mice were placed on a 55.0 (±0.3) °C hot plate (Columbus Instruments, Columbus, Ohio), and latency to the first hind-paw response was recorded. The hind-paw response was either a foot shake or a paw lick.

Adult male mutant and control mice (14–17 weeks old) were used. Mice were decapitated under halothane anesthesia and both hippocampi were isolated. Transverse hippocampal slices (370–400 μm) were obtained using a tissue slicer (VT1000S; Leica Microsystems, Nussloch, Germany or Vibratome 3000 plus; Vibratome Company, St. Louis, MO, USA) in sucrose-containing saline composed of (in mM): sucrose, 72; NaCl, 80; KCl, 2.5; NaH₂PO₄, 1.0; NaHCO₃, 26.2; glucose, 20; CaCl₂, 0.5; and MgCl₂, 7 (equilibrated with 95% O₂/5% CO₂) for whole-cell recordings or in standard saline (see below) for field potential recordings. Slices were incubated for 30 min at 30°C in a standard saline solution composed of (in mM): NaCl, 125; KCl, 2.5; NaH₂PO₄, 1.0; NaHCO₃, 26.2; glucose, 11; CaCl₂, 2.5; and MgCl₂, 1.3 (equilibrated with 95% O₂/5% CO₂) and maintained in a humidified interface holding chamber at room temperature (24–27°C) before use. Electrophysiological recordings were made from slices placed in a submersion-type chamber superfused at 2 mL/min. The bath temperature was maintained at 26.5–27.5°C using an automated temperature controller.

Whole-cell recordings were made from GCs in the DG using a blind whole-cell patch-clamp technique. Voltage-clamp recordings were made with a pipette filled with a solution composed of (in mM) CsCl, 140; HEPES, 20; NaCl, 8; MgATP, 2; Na₂GTP, 0.3; EGTA, 0.5; and CaCl₂, 0.05 (pH adjusted to 7.3 with CsOH). The recording pipette was placed in the middle third of the GC layer. Synaptic currents were evoked at 0.2 Hz by bipolar tungsten stimulating electrodes that were placed in the middle third of the dentate molecular layer to activate the input from the medial perforant path (MPP). Activation of MPP input was verified by the depression of synaptic currents in response to paired stimuli. EPSCs mediated by NMDARs were recorded in standard saline solution supplemented with 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX, 20 μM), picrotoxin (100 μM), and SR95531 (2 μM). EPSC amplitudes were measured on analysis. Series resistance was monitored during the recordings, and data deviating by more than 15% were excluded. fEPSPs at the MPP synapse were recorded in the middle third of the molecular layer using a glass pipette filled with 2 M NaCl. The MPP was stimulated at 0.05 Hz in standard saline. The initial slopes of fEPSPs were measured.

All recordings were done using a Multiclamp 700B amplifier (Axon Instruments, Union City, CA, USA), filtered at 2–5 kHz, and stored in a personal computer via an interface (digitized at 10 kHz). CNQX and d-2-amino-5-phosphonovaleric acid (d-APV) were purchased from Tocris (Bristol, UK). Picrotoxin was from Wako Pure Chemical Industries, Ltd. (Osaka, Japan). SR95531 was from Sigma (St. Louis, MO, USA). All values are expressed as the mean ± S.E.M. Statistical significance was evaluated using two-tailed Mann–Whitney tests, Student’s t-tests, or ANOVA.

All procedures were approved by our Institutional Animal Care and Use Committee. Adult male mutant and control mice (14–21 weeks old) were used. The control mice consisted of six WT mice (GluN2A+/+) and two floxed mice (GluN2A+/flox). We did not detect any significant differences in firing properties and place field measurements between the WT and floxed animals.

Animals were chronically implanted with a miniature microdrive equipped with four electrodes (Neuralynx, Tucson, AZ). Each electrode was composed of four individually insulated tungsten wires (12.7 μm diameter) that were twisted together. The impedance of the electrodes was approximately 400 kΩ at 1 kHz. Animals were anesthetized with a mixture of 0.9% ketamine and 0.2% xylazine. The skull was exposed, and four stainless screws were inserted into the bone for structural support. A small hole (<0.5 mm) was made over the left hemisphere (1.5 mm lateral to the midline, 2.0 mm posterior to the bregma), and the electrodes were inserted through the hole. The electrode tips were positioned 0.9 mm deep from the brain surface. The space between the bone and the microdrive was filled with petroleum jelly, and the microdrive and the bone were glued using dental acrylic.

The electrodes were connected to a unity gain buffer consisting of an operational amplifier (TL074; Texas Instruments, Dallas, TX). Signals from the buffer amplifier were amplified (×3000), and filtered between 300 Hz and 5000 Hz. Amplified signals were sampled by a data acquisition card (PCI-6259; National Instruments, Austin, TX) at 20 kHz. The location of the animal was tracked using an overhead video camera, which tracks a pair of infrared LEDs placed on the top of the microdrive. Neural waveforms and video images were processed and stored using custom software written in LABVIEW7 (National Instruments, Austin, TX).

At least one week before surgery, animals were handled daily and allowed to explore the testing platform (a white, round-cornered square box, 40 × 40 cm, 30 cm high) for 15 min each day. The box was housed in a 180 × 120 cm soundproof room. A brown paper towel was hung on the wall of the box as a local cue. Four incandescent bulbs in 15-cm hemispherical reflectors were mounted in the corners of the room (Figure 5A). One week after surgery, electrodes were screened daily for complex spike cells. Complex spike cells were identified as those that had both a wide peak-to-trough width (>0.3 ms) and complex spike bursts. If no complex spike cell was found, electrodes were lowered by 40 μm and were screened the next day until complex spike cells were found. The recording session began only if complex spike cells persisted for at least 1 day. The animals were exposed to both cue configurations (‘orig’ and ‘rot’) for 15 min each during the cell screening period (11–19 days, depending on individuals) and recording sessions (5 days). The complex spiking cells recorded in the CA1 appear to be pyramidal cells [63, 64]; thus, we designated the complex spiking cells in the CA1 as pyramidal cells. The data were excluded if the mouse crossed less than 90% of the platform in a single session or the mean firing rates were below 0.1 Hz. At the end of the recordings, marker lesions were made at the electrode tips by passing a current (20 μA, 10 s). The location of a recording site was determined by estimating the distance along the electrode track associated with the microelectrode position at the time of recording.

Recorded neural waveforms were digitally high-pass filtered above 800 Hz. Spikes with amplitudes 3.5 times greater than noise were extracted for further analyses. The spikes were segregated into multiple single units using automatic clustering software Klustakwik [65] and manual cluster cutting software Klusters [66]. The presence of a refractory period in isolated units was inspected visually by using Klusters. Units with a clear refractory period (>2 ms) in their autocorrelograms were included in the present analysis. The firing rate maps of the complex spike cells were constructed by finding the spike number at each location bin (2 cm × 2 cm) divided by the dwell time in the bin. To eliminate nonspecific firing at rest, spikes occurring when the animal's velocity was less than 1 cm/s were filtered out [28]. The maps were smoothed using a Gaussian spatial and temporal filter with a standard deviation of 1 pixel (2 cm) and 2.5 s, respectively. A place field was defined as a group of at least one contiguous pixel in which the firing rates exceeded the overall mean firing rate plus one standard deviation. “Main place field” was defined as the largest place field for each cell. Signal processing and statistical analyses were performed using MATLAB 6.5 (Mathworks, Natick, MA) and GraphPad Prism5 (GraphPad Software, La Jolla, CA).

All procedures were approved by our Institutional Animal Care and Use Committee. Adult male mutant and control mice (10–13 or 17–21 weeks old) were used. Mice were housed in a room with a 12-h light/dark cycle (lights on at 7:00 a.m.) with free access to food and water. Behavioral testing was performed between 9:00 a.m. and 6:00 p.m. After the tests, all apparatus were cleaned with super hypochlorous water to remove the scent of mice. The eight-arm radial maze test was conducted in a manner similar to that described previously [62]. The floor of the maze consisted of white plexiglass, and the wall (25 cm high) consisted of transparent plexiglass. Each arm (9 × 40 cm) radiated from an octagonal central starting platform (perimeter 12 × 8 cm) like the spokes of a wheel. Identical food wells (1.4 cm deep and 1.4 cm in diameter) with pellet sensors were placed at the distal end of each arm. The pellet sensors were able to automatically record pellet intake by the mice. The maze was elevated 75 cm above the floor and placed in a dimly lit room with several extra-maze cues. During the experiment, the maze was maintained in a constant orientation. One week before pre-training, animals were deprived of food until their body weight was reduced to 80–85% of the initial level. Pre-training started on day 8. Each mouse was placed in the central starting platform and allowed to explore and consume food pellets scattered on the whole maze for a 5-min period (one session per mouse). After completion of the initial pre-training, mice received another pre-training to take a pellet from each food well after being placed at the distal end of each arm. A trial was finished after the subject consumed the pellet. This series of experiments was repeated eight times, using eight different arms, for each mouse. After these pre-training trials, the actual maze acquisition trials were performed. All eight arms were baited with food pellets. Mice were placed on the central platform and allowed to find all eight pellets within 25 min. A trial was terminated immediately after all eight pellets were consumed or 25 min had elapsed. An “arm visit” was defined as traveling more than 5 cm from the central platform. The mice were confined in the center platform for 5 s after each arm choice. The animals went through one trial per day (34 trials in total). For each trial, choices of arms, latency to get all pellets, distance traveled, number of different arms chosen within the first eight choices, the number of revisiting, and omission errors were automatically recorded. Data acquisition, control of guillotine doors, and data analyses were performed using Image RM software written by Tsuyoshi Miyakawa (available through O’Hara & Co., Tokyo, Japan). The software is based on NIH Image (http://​rsb.​info.​nih.​gov/​nih-image/​). Statistical analyses were conducted using StatView (SAS Institute).

To assess the effects of GluN2A mutations on the number of BrdU-positive cells, mice (18 weeks old) were administered BrdU, 4 × 75 mg/kg i.p., dissolved in saline, every 2 h. Mice were sacrificed 24 h after the last BrdU injection. After anesthesia, mice were transcardially perfused with 4% paraformaldehyde and brains were collected for immunohistochemistry. All brains were post-fixed overnight in 4% paraformaldehyde at 4°C and soaked in a series of 10%, 20% and 30% sucrose at 4°C. Serial sections of the brains (30 μm sections) were cut through the entire hippocampus on a cryostat. DNA denaturation was conducted by incubation for 2 h in 50% formamide/2× SSC at 65°C, followed by several rinses. Sections were then incubated for 30 min in 2 N HCl and then for 10 min in boric acid. After washing with PBS, sections were incubated in 3% H₂O₂ for 30 min to eliminate endogenous peroxidases. After blocking with 3% normal goat serum in 0.01% Triton X-100, cells were incubated with anti-mouse BrdU (1:100; Becton Dickinson) overnight at 4°C. Sections were then incubated for 1 h with secondary antibody (1:1000; biotinylated goat anti-mouse; Vector Laboratories, Burlingame, CA) followed by amplification using an avidin-biotin complex (Vector Laboratories) and visualization with DAB (Vector Laboratories).