Animals for molecular and histological experiments were killed under the protocol MPI/T-6/06; 15/08; 20/9; and 28/11, according to the guidelines of the Max Planck Institute that adhere to ethical guidelines set by the local governing body and were performed according to the German Animal Welfare Act: Regulation for the Protection of Animals Used for Experimental or Other Scientific Purposes (Animal Welfare Regulation Governing Experimental Animals (TierSchVersV). Electrophysiological experiments were conducted according to the Norwegian Animal Welfare Act and the European Union’s Directive 86/609/EEC. All behavioral experiments were conducted in accordance with the United Kingdom Animals Scientific Procedures Act (1986), under the project license number PPLs 30/1505 and 30/1989 of the UK Home Office and by the license of the regional council in Karlsruhe, Germany (35-9185.81/G-71/10). Genetic manipulations of mice were performed under the license of the regional council in Karlsruhe, Germany (35-9185.81/G-4/02). Efforts were made to minimize the number of animals used.

Gria1^R/R (Gria1^tm1Erk MGI:2178080) and Gria1^−/− mice (Gria1^tm1rsp; MGI:2178057) were generated by analogous targeting vectors and the same targeting strategy (Zamanillo et al., 1999; Vekovischeva et al., 2001) using R1 mouse embryonic stem cells (Nagy et al., 1993). Both targeting vectors for ES gene replacement of Gria1^R/R and Gria1^−/− just differed in the amino acid codon R600 and Q600 in the floxed exon 11, respectively. For the generation of Gria1^−/− mice exon 11 was removed together with the NEO selection marker by Cre-recombinase mediated deletion. For the generation of Gria1^R/R mice the floxed exon 11, encoding (Q600R) exon 11 is still present and only the floxed selection marker in intron 11 was removed by Cre. Similarly, in Gria1^f/f mice (MGI:3798452; Jackson labs: 019012-B6N129-Gria1^tm2Rsp/J) the floxed exon 11, encoding (Q600) is still present and expressed at regular levels (Fuchs et al., 2007). Cohorts of aged-matched Gria1^R/R and Gria1^+/+ (WT) littermates were produced by heterozygous matings, yielding approximately 25% WT and 25% Gria1^R/R mice. The GluA1(Q600R) mutation was previously called GluA1(Q582R) and is now renumbered according to the full-length GluA1 precursor (Genbank: NM_001113325). Gria1^−/− mice (MGI:2178057; Zamanillo et al., 1999) are available from the Jackson Laboratories (Gria1^tm1rsp, stock number: 019011). All behavioral experiments were performed during the light phase.

Anti-GluA1 (RRID:AB_390157), anti-GluA2 (RRID: AB_2336198), anti-GluA2/3 (RRID:AB_11213931), anti-GluA3 (RRID:AB_11152621), anti-GluN1 (RRID:AB_390129), anti-CaMKII (RRID:AB_2067919), anti-ß-actin (RRID: AB_476692), anti-mouse (RRID:AB_2336176), anti-rabbit (RRID:AB_2313567), gold-conjugated secondary antibodies (RRID:AB_106260) were used in these studies.

Hippocampi of mice were homogenized in 25 mM HEPES + Protease-Inhibitor Cocktail (Roche). Lysates were centrifuged for 5 min at 2000 rpm. Protein concentrations in supernatants were determined in triplets with BCA-Kit (Pierce). From each lysate 8 μg protein was separated by 7% SDS-PAGE and proteins were transferred to nitrocellulose membranes. The blotted proteins were probed with polyclonal antibodies against GluA1 (Merck Millipore 0.1 μg/ml), anti-GluA2 (Merck Millipore 0.16 μg/ml), anti-GluN1 (Merck Millipore, 0.25 μg/ml, monoclonal anti-GluA3 (ThermoFisher, 1:250), anti-CaMKII (MAB 8699 Merck Millipore 0.1 μg/ml) and anti-ß-actin (AC-15 Ascites, Sigma, 1:25,000), followed by peroxidase-linked anti-rabbit or -mouse secondary antibodies (Jackson Immuno Res., 1:20,000). ECL+ (ThermoFisher) was used to visualize immuno-labeled proteins.

Hippocampi of mice (>P60) were dissected and prepared as described above. Extracts were diluted 1:1 in buffer2 (25 mM HEPES pH7.4; 300 mM NaCl; 2% TritonX-100), incubated for 60 min at 4°C and subjected to centrifugation (10,000× g for 10 min at 4°C). Solubilized proteins of the supernatant (4 μg/μl) were mixed with buffer3 (25 mM HEPES pH7.4; 150 mM NaCl) in a 1:5 ratio. About 80 μg of solubilized protein was used for Immunoprecipitations (IPs) using protein A-Sepharose (ThermoFisher) washed in buffer4 (buffer2 and buffer3 in 1:5 ratio), and subsequently incubated for 1 h with the 80 μg of solubilized protein. About 15% of the supernatants were kept as “Input”. The rest of the supernatants were incubated overnight with anti-GluA1 or anti-GluA2/3 (1.5 μg; Merck Millipore), and then incubated for 1 h with 20 μl of the washed and pre-saturated protein A-sepharose. The A-sepharose was pelleted and kept as IP-pellet1. The supernatants were subjected to a second precipitation. The washed and pre-saturated protein A-sepharose beads were incubated with the specific antibodies for 1 h, washed with buffer4 after incubation with the supernatant for 4 h, pelleted (2 min; 3000× g), and pooled with pellet1. The pooled pellets were washed twice in buffer4 and buffer3 (800 μl each) and resuspended in 60 μl 0.5% SDS; 1% β-mercaptoethanol. After incubation at 95°C for 5 min the sample was briefly centrifuged and half of the sample was used for immunoblot analysis.

At P14 and P42 mice were prepared for immunohistochemical analysis as described previously (Jensen et al., 2003). Fixed, agarose embedded brains were cut coronally into 50–75 μm vibratome sections (VT1000S, Leica Microsystems). For DAB-staining, sections were incubated for 10 min in 0.5% H₂O₂ in PBS, washed twice with PBS (10 min each) and blocked in 1% BSA, 0.3% TritonX-100, 2% NGS in PBS for 30–60 min. The sections were incubated at 4°C overnight in the same buffer containing the following primary antibodies: polyclonal anti-GluA1 (1 μg/ml, Merck Millipore), polyclonal anti-GluA2 (4 μg/ml, Merck Millipore), polyclonal anti-GluA2/3 (1 μg/ml, Merck Millipore,). After the incubation, sections were washed three times in buffer1 (0.3% BSA, 0.1% Triton X-100 in PBS) and exposed to horseradish peroxidase coupled secondary antibody (1:1000 anti-rabbit or -mouse, Jackson Immuno Research) in buffer1 for 2 h. Finally, sections were washed twice (10 min) with buffer1 followed by two washes with PBS. Sections were developed by incubation in 0.05% DAB solution (Sigma Aldrich) and 0.01% H₂O₂ in 20 mM Tris·Cl pH7.6. The reaction was stopped with 20 mM Tris·Cl pH7.6. The sections were mounted on glass slides, dried overnight and embedded in vitro-Clud (Langenbrinck Labor- und Medizintechnik).

Bright-field images of immunostainings were captured with the Axio-Imager (Zeiss) and digitized with an AxioCam CCD camera using the AxioVision™ software. Brightness and color adjustments were performed in Photoshop CS5 (Adobe, RRID:SCR_014199). Confocal images of coronal brain slices immuno-stained by anti-GluA1 or GluA2 and Cy3 or FITC labeled secondary antibodies were recorded using a Leica TCS SP8 microscope (Leica).

Image data analysis was performed in ImageJ (V2.0.0.). Bright-field: the mean average signal intensity was measured for the corresponding ROIs in str. pyramidale, radiatum and oriens and was used to calculate the relative signal intensity in the str. pyramidale relative to the str. oriens + str. radiatum (Somatic Accumulation index SAi). For the confocal image analysis the mean signal intensity of big regions (at least 200 μm in length) covering the pyramidal cell layers and the neuropil in the str. radiatum were measured and used to calculate the somatic accumulation ratio.

The preparation of the mouse brains was as described previously (Jensen et al., 2003). The fixed brains were cut on a vibratome into 300 μm sections. The CA1 area of the hippocampus was excised and the tissue blocks were cryoprotected in glycerol, cryofixed in nitrogen-cooled propane, and substituted in methanol-containing 1.5% uranyl acetate. After embedding in Lowicryl HM20 (Addivant Germany GmbH) ultrathin sections were processed for post embedding immunocytochemistry, employing anti-GluA1 and anti-GluA2 antibodies (5 μg protein/ml, Merck Millipore). Immunolabeling was visualized by 10 nm gold-coupled secondary antibodies (1:50, GE-Healthcare). Data analysis was performed on 12 grids per animal (six incubated with anti-GluA1; six with anti-GluA2). Sections from WT and Gria1^R/R mice were processed simultaneously. On each grid, 300 randomly selected synapses were counted in CA1 str. radiatum. The number of labeled PSDs at contacts and the number of gold grains per individual PSD were evaluated. Only synapses labeled with at least two gold particles were counted. Data were fitted by Poisson distributions, and numbers of labeled PSDs and the mean numbers of gold grains per PSD were determined from the fitted distributions. Differences were assessed by Student’s t-test for paired samples.

Since in heterologous systems the GluA1(Q600R) point mutation is trafficking compromised (Greger et al., 2003), and since hippocampal GluA2 expression levels are reduced in absence of GluA1 (Jensen et al., 2003), we first analyzed the expression and the channel assembly of AMPARs in Gria1^R/R mice by immunoblots and co-IPs. Our analysis of hippocampal protein extracts isolated at different postnatal days (P2 – P90) from controls and Gria1^R/R mice showed that the dominant partner GluA2 and GluA1(Q600R) itself were reduced in Gria1^R/R animals (Figure 1A). GluA1(Q600R) was reduced at all developmental stages. The reduction of GluA2 expression became obvious from P42 onwards. No changes were observed for the protein levels of GluA3, the NMDAR subunit GluN1 and for the alpha subunit of the Ca²⁺/Calmodulin dependent protein kinase 2 (α-CaMKII). Thus, the half-life of AMPARs containing GluA1(Q600R) seems to be reduced in adult Gria1^R/R mice. The observed reduction of GluA2 was very similar to what we have described previously in Gria1^−/− mice (Jensen et al., 2003).

For the analysis of the GluA1(Q600R) assemblies, we IP-purified GluA1- or GluA2/3-containing AMPARs from solubilized membrane preparations of hippocampi from WT controls and Gria1^R/R mice. Comparable amounts of GluA2 could be co-purified with anti-GluA1 antibody, whereas very little GluA3 associated with GluA1 in WT mice and GluA1(Q600R) in Gria1^R/R mice (Figure 1B). This is in line with the observation that heteromeric GluA1/A2 assemblies provide the majority of AMPAR assemblies in the hippocampus (Wenthold et al., 1996; Lu et al., 2009), and indicates that GluA1(Q600R) does form assemblies with GluA2. Similarly, anti-GluA2/3 antibodies co-precipitated GluA1 and GluA1(Q600R) from WT controls and Gria1^R/R mice, respectively. The specificity of the IP was controlled by the crude membrane preparation of Gria1^−/− mice, and IP input by visualizing immunoglobulin of the primary antibody. These results show that the GluA1(Q600R) subunit can form dimeric or tetrameric assemblies with GluA2 and GluA3 subunits indicating that the reduced levels of GluA1(Q600R) and GluA2 in Gria1^R/R mice were not caused by a channel assembly impairment, but most likely reflect a shorter half-life of the low conductance AMPAR channels (Swanson et al., 1997) containing four subunits with R at the tip of the channel pore: GluA1(Q600R) and GluA2 homomeric and GluA1(Q600R)/GluA2 heteromeric channels (Figure 1C).

In the absence of GluA1, the GluA2 subunits accumulate in the cellular layers of hippocampal principle cells in young and in adult mice (Zamanillo et al., 1999; Jensen et al., 2003). To determine the expression pattern of GluA1(Q600R) and GluA2 in Gria1^R/R mice we generated and evaluated immunostaining of dorsal hippocampi in coronal brain sections from young and adult GluA1(Q600R) expressing mice by using GluA1, GluA2 and GluA2/3 specific antibodies. As shown in Figure 2A, all three antibodies revealed an expression in all hippocampal layers in WT mice at P14 and P42. Hippocampi of Gria1^R/R mice, however, exhibited strong accumulation of GluA1 and GluA2 immunosignal in principle cell layers of the hippocampus of Gria1^R/R mice at P14 and P42 similar to the GluA2 accumulation, which can be observed in Gria1^−/− mice (Jensen et al., 2003). To quantify the subcellular distribution of the AMPAR subunits we calculated somatic accumulation indices (SAi) in the CA1 region showing the highest AMPAR expression in the hippocampus. In Gria1^R/R mice ratios of mean signal intensities in the principle cell (str. pyramidale) vs. dendritic layers (str. oriens, str. radiatum) above one (SAi = 1.39–2.14) indicated strong somatic accumulation of the GluA1, A2 and A2/A3 imunosignal at P14 and at P42 whereas AMPAR subunits in WT mice showed no clear hippocampal layer specific preference, with a SAi of approximately one. Confocal fluorescent images of anti-GluA1 and anti-GluA2 immunostainings confirmed the somatic accumulation of GluA1 and GluA2 in the hippocampus of Gria1^R/R mice (Supplementary Figure S1). To analyze the effect of the GluA1(Q600R) mutation on synaptic receptors directly we performed a quantitative electron microscopic immunogold analysis for GluA1 and GluA2 subunits at CA3-to-CA1 PSD containing synapses (Figure 2B). As shown in Figure 2C, there was no significant difference in the number of postsynaptic anti-GluA1, GluA1(Q600R) and anti-GluA2 immunosignals between WT controls and Gria1^R/R mice. Furthermore, we were unable to detect any statistically significant difference between Gria1^R/R and WT mice in the number of immunoreactive GluA1 and GluA2 subunits per synapse. A slight but significant reduction in the number of synapses expressing GluA1(Q600R) could, however, be detected in Gria1^R/R mice when compared to the GluA1 expression in WT littermates (Figure 2C). The possibility of a direct loss of extrasynaptic AMPARs could not be analyzed by our current immunogold analysis and would require cryosection immunogold analysis.

Thus, the replacement of GluA1 by the mutated GluA1(Q600R) had no substantial effect on the synaptic presence of GluA1(Q600R) and GluA2 subunits but induced a retention of GluA1(Q600R) and GluA2 at somata of principal neurons of Gria1^R/R mice.

In GluA1 deficient mice somatic redistribution of functional AMPARs in CA1 cells could be documented by loss of somatic and dendritic AMPAR responses, slightly reduced synaptic efficiency and impaired LTP (Zamanillo et al., 1999; Hoffman et al., 2002; Andrásfalvy et al., 2003; Jensen et al., 2003). To provide electrophysiological evidence that the expression of the GluA1(Q600R) mutation has a similar electrophysiological signature, we first monitored the somatic AMPAR/NMDAR ratios by analyzing glutamate-induced currents in nucleated patches. We found that, similar to GluA1 deficient mice, the glutamate-induced AMPAR currents were strongly reduced in nucleated patches of Gria1^R/R mice at P42, and reached, at most, only 20% of the WT control values (AMPAR/NMDAR ratios; 1.87 ± 0.68 in Gria1^R/R, n = 7 cells from three mice; 5.43 ± 0.97 in WT, n = 9 cells from 4 mice; p < 0.01; Figure 3A). To assess changes in excitatory synaptic transmission and synaptic excitability, we recorded simultaneously in CA1 apical dendritic (middle part of the str. radiatum) and soma layers of hippocampal slices from Gria1^R/R and WT littermates. We measured the fiber volley, the fEPSP, and the population spike as a function of different stimulation strengths. The stimulation strengths necessary to elicit fiber volleys of given amplitudes (0.5 and 1.0 mV) were numerically lower in Gria1^R/R (4.4 ± 0.5 nC, n = 44; 7.5 ± 0.9 nC, n = 44) compared to control WT mice (6.2 ± 0.4 nC, n = 82; 10.4 ± 0.7 nC, n = 80), although this didn’t reach statistical significance p = 0.08 and 0.06, respectively (Figure 3B). The evoked fEPSPs for presynaptic fiber volleys of 0.5 and 1.0 mV in Gria1^R/R mice (1.4 ± 0.1 mV, n = 44 and 2.3 ± 0.1 mV, n = 44), were not significantly different from the fEPSPs elicited in brain slices of control littermates (1.5 ± 0.1 mV, n = 82; 2.5 ± 0.1 mV, n = 80, p = 0.70 and 0.41, respectively), indicating that basal, fast glutamatergic transmission was unaltered in Gria1^R/R mice, at least across this range of stimulation intensities. Furthermore, postsynaptic excitability measured as the fEPSP threshold for generating a just detectable population spike was unchanged in Gria1^R/R mice (2.3 ± 0.1 mV, n = 44), when compared to controls (2.4 ± 0.1 mV, n = 83) p = 0.58 (Figure 3B). In a similar manner, the fEPSP size necessary to elicit a population spike of 2 mV was also of the same magnitude in the two genotypes (Gria1^R/R 3.1 ± 0.1 mV, n = 40; WT 3.4 ± 0.1 mV, n = 73 p = 0.24). A comparison of paired-pulse facilitation (PPF ratio) also failed to reveal a significant difference between Gria1^R/R (1.41 ± 0.02 mV, n = 48) and WT littermates (1.43 ± 0.02 mV, n = 44 p = 0.50; Figure 3B). Thus similar to GluA1-deficient mice, the somatic AMPAR currents could hardly be detected despite regular synaptic transmission in field recordings in hippocampal CA1 cells of Gria1^R/R mice.

The loss of synaptic plasticity at hippocampal CA3-to-CA1 synapses in the presence of synaptic AMPAR mediated transmission was the most significant result provided by the Gria1 knock-out mice and is an additional signature for the loss of GluA1. Therefore we examined LTP at hippocampal CA3-to-CA1 synapses in brain slices from adult mice. Forty to 45 min after a single tetanization (100 Hz, 1 s) of the afferent fibers in the str. radiatum or str. oriens in slices from control mice, the average slope of the field EPSP was 1.44 ± 0.05 (mean ± SEM; n = 26) of the pre-tetanic value, whereas the untetanized control pathway was unchanged (1.02 ± 0.02). We found a significant reduction in LTP in slices from Gria1^R/R mice (1.17 ± 0.03, n = 25, when compared to WT; p < 0.01; Figure 4A), although this reduction in the magnitude of LTP was not as dramatic as observed previously in Gria1^−/− mice (Zamanillo et al., 1999). To ascertain whether the reduced LTP in the Gria1^R/R mice was caused by an induction failure, such as a change in the LTP induction threshold, we conducted a further series of experiments in the presence of the GABA_A-receptor blocker (–)-bicuculline methochloride (10 μM), which facilitates the induction of LTP without significantly changing its magnitude (Wigström and Gustafsson, 1983, 1985). Similar to the previous results with Gria1^−/− mice (Zamanillo et al., 1999), the evoked LTP under these circumstances was not significantly different from the values obtained in the control solution, either in WT mice (1.33 ± 0.04, n = 37, p = 0.10) or in Gria1^R/R mice (1.07 ± 0.04, n = 26, p = 0.08), suggesting that the tetanization procedure used previously was sufficient. LTP was also reduced in Gria1^R/R mice when a repeated tetanization paradigm was applied (100 Hz for 1 s, repeated 4× with 5 min interval). LTP induced by this procedure in CA1 slices from WT controls was 1.51 ± 0.06 (n = 17), whereas in Gria1^R/R mice LTP measured 1.22 ± 0.06 (n = 15; Figure 4B; genotype comparison p < 0.01). The residual LTP in Gria1^R/R mice was NMDAR-mediated. It was completely blocked by the presence of 50 μM AP5 (data not shown), as previously also observed in Gria1^−/− mice (Jensen et al., 2003). Notably, despite this pronounced deficit in tetanus-induced LTP (Zamanillo et al., 1999), more recent studies have shown that a theta-burst pairing paradigm (Larson and Lynch, 1986) can elicit robust LTP in Gria1^−/− mice (Hoffman et al., 2002; Romberg et al., 2009). We therefore adopted a similar theta-burst pairing paradigm, but used extra-cellular antidromic activation instead of depolarizing soma injection in order to elicit a single postsynaptic spike (Figure 4C). In control mice, the magnitude of LTP induced with the theta-burst pairing paradigm was not significantly different from the magnitude obtained by the tetanization procedures (1.43 ± 0.06, n = 30). Furthermore, although reduced in magnitude, the amount of LTP induced using the theta-burst stimulation paradigm in both full knockouts and point mutants was not significantly different from that observed in the WT control mice (Gria1^R/R: 1.23 ± 0.05, n = 14; Gria1^−/−: 1.30 ± 0.06, n = 17; p = 0.15 and p = 0.21, respectively, when compared to WT). However, in agreement with our previous observations in Gria1^−/− mice (Zamanillo et al., 1999; Hoffman et al., 2002; Jensen et al., 2003), the potentiated fEPSP responses during the first few minutes after tetanization were reduced in Gria1^R/R mice. In WT animals this immediate increase in synaptic transmission comprises post-tetanic potentiation (PTP), short-term potentiation (STP) and LTP (Bliss and Lomo, 1973). This immediate component of synaptic plasticity was significantly reduced (p < 0.05) during the first 7–10 min following LTP induction in both Gria1^−/− and Gria1^R/R mice and thus might be dependent on the presence of Ca²⁺-permeable GluA1-containing AMPA receptors that are absent in Gria1^−/− and Gria1^R/R mice (Rozov et al., 2012).

Given that the changes in AMPAR expression and AMPAR mediated synaptic plasticity are comparable between Gria1^−/− and Gria1^R/R mice, we hypothesized that Gria1^−/− and Gria1^R/R mice would show similar behavioral impairments. Most notably, Gria1^−/− mice exhibit locomotor hyperactivity in a novel environment, and a profound and selective spatial short-term memory deficit (Reisel et al., 2002; Schmitt et al., 2003; Bannerman et al., 2004). In agreement with this, the Gria1^R/R mice displayed a pronounced hyperactivity in photocell locomotor activity cages. ANOVA revealed a significant main effect of genotype (Table 1). There was no main effect of sex, or any interaction between sex and genotype (F < 1 for both).

Gria1^R/R mice also displayed a robust and selective spatial short-term memory deficit. In the discrete trial, spontaneous alternation, spatial working memory task the WT controls showed a high level of alternation performance across the 10 trials (80.7%), whereas the Gria1^R/R mice exhibited chance levels (54.0%; Table 1), thus resembling the result seen with Gria1^−/− mutants (Bannerman et al., 2004; Sanderson et al., 2007). An ANOVA revealed a significant effect of genotype (Table 1), and a significant effect of sex that was due to an overall lower level of alternation in the females (F_(1,25) = 7.68; p < 0.05). There was, however, no interaction between genotype and sex (F < 1; p > 0.20).

Gria1^R/R mice also demonstrated a profound spatial short-term memory impairment during the appetitively motivated, discrete trial, rewarded alternation task on the elevated T-maze (Figure 5A, left). Whereas the WT controls performed well on the task (74.2 ± 2.3% correct responses), the Gria1^R/R mice never differed from chance levels of performance (50.4 ± 2.3%). An ANOVA confirmed that there was a significant effect of genotype (F_(1,25) = 54.09; p < 0.01). There was no effect of sex, nor a genotype by sex interaction (F’s < 2.77; p’s > 0.10 for both).

In contrast to these spatial working memory tasks, Gria1^R/R mice performed normally on an appetitively motivated, spatial reference memory task on the elevated Y-maze (Figure 5A, right). Previous studies have demonstrated that Gria1^−/− mice acquire this task as well as WT controls (Reisel et al., 2002). In the present study, both WT and Gria1^R/R mice learned the appetitive, spatial reference memory task at a similar rate, and reached the same high level of asymptotic performance (Figure 5A, right). An ANOVA of the percent correct responses across the 10 days of testing confirmed these impressions. There was a significant effect of day (F_(9,117) = 41.68; p < 0.01), but no significant effect of genotype, nor any genotype by day interaction (F < 1 for both).

Similarly, Gria1^R/R mice also performed well on the spatial reference memory version of the Morris water-maze task (Figure 5B), again paralleling the results that have previously been obtained for Gria1^−/− mice (Zamanillo et al., 1999; Reisel et al., 2002). Both groups of mice became gradually more efficient at reaching the escape platform, as shown by the gradual reduction in escape latencies as training proceeded (main effect of day—F_(8,200) = 62.32; p < 0.01; Figure 5B, left). Although the mutant mice appeared to take slightly longer than their WT counterparts to find the escape platform throughout training, this was not reflected in a statistically significant main effect of genotype (F_(1,25) = 3.55; p > 0.05). Furthermore, there was no interaction between genotype and day (F < 1). In addition, the ANOVA revealed that there was no significant effect of sex (F_(1,25) = 1.77; p > 0.10), and sex did not interact with any of the other experimental variables (F < 1 for all interaction terms). Analysis of path lengths to the escape platform revealed a similar pattern of results. There was a highly significant main effect of day, reflecting the fact that the mice learned to swim more directly towards the location of the hidden platform (F_(8,200) = 53.08; p < 0.01). Again, although the Gria1^R/R mice traveled slightly further before finding the platform, this did not result in a significant main effect of genotype (F_(1,25) = 3.21; p > 0.05). Furthermore, there was no genotype by day interaction (F_(8,200) = 1.32; p > 0.20). As before, the ANOVA also showed no main effect of sex, nor any interaction between sex and the other experimental variables (p > 0.10 for all interaction terms).

The performance during the 90 s probe trial conducted at the end of water-maze training confirmed that both the WT and Gria1^R/R mice had learned about the spatial location of the escape platform, with both groups showing a strong preference for the training quadrant (Figure 5B, right top). ANOVA comparing the time spent in each of the four quadrants of the pool during the probe trial revealed a significant effect of quadrant (F_(2,75) = 47.79; p < 0.01), reflecting the fact that both groups of mice showed a strong preference towards the training quadrant (% time in the training quadrant during the 90 s probe trial—WT = 47.7; Gria1^R/R = 59.9). Note that for the analysis of the distribution of time spent searching the four quadrants, the numerator term in the degrees of freedom was reduced by one to control for the fact that the quadrant dwell times were not independent. There was no interaction between quadrant and genotype, quadrant and sex, nor quadrant by genotype by sex (p > 0.10 for all interactions). A second ANOVA was then conducted comparing only the amount of time spent in the training quadrant. For this analysis the 90 s probe trial was divided into three 30 s time bins (Figure 5B, right bottom). This ANOVA confirmed that the overall time spent in the training quadrant during the 90 s probe trial by the two groups did not differ; although the mutant mice spent slightly more time in the training quadrant, this tendency did not attain statistical significance (F_(1,25) = 2.92; p > 0.05). ANOVA revealed no significant effect of sex or time bin (p > 0.20 for both comparisons), no genotype by sex, sex by time bin, or genotype by sex by time bin interaction (p > 0.10 for all interactions). There was, however, a statistically significant genotype by time bin interaction (F_(2,50) = 10.20; p < 0.01), which reflected the fact that as the probe trial proceeded, the Gria1^R/R mice were spending more time in the training quadrant relative to the WT controls. This was confirmed by a subsequent analysis of simple main effects which showed that there were no significant group differences during either the 0–30 or 30–60 s time bins of the probe trial (p > 0.10 for both comparisons), but that there was a significant difference between the groups in terms of the amount of time spent in the training quadrant during the final 30 s of the transfer test (60–90 s F_(1,25) = 22.17; p < 0.01).

One possible explanation as to why the Gria1^R/R mice might have remained in the training quadrant to a greater extent than the WT mice as the probe test proceeded is that they exhibited increased floating behavior, having discovered that the platform was not present. However, only two mice showed any signs of floating behavior during the probe test. Both were male Gria1^R/R mice, and they each spent approximately 10 s floating during the 90 s test. Re-analysis of the probe test data for the time spent in the training quadrant only, excluding these two subjects, revealed essentially the same pattern of results, although now the overall main effect of genotype did reach statistical significance (F_(1,23) = 10.94; p < 0.01), with the point mutants spending significantly more time in the goal quadrant during the probe test. Importantly, there was still a genotype by time bin interaction (F_(2,46) = 10.20; p < 0.01), and subsequent analysis of simple main effects again showed that the stronger preference for the training quadrant in the point mutants only developed as the probe test proceeded (main effect of genotype for 0–30 s F < 1; 30–60 s F_(1,23) = 4.55; p < 0.05; 60–90 s F_(1,23) = 43.36; p < 0.01).

To ensure that the pattern of results that we obtained across the spatial memory tests did not reflect the order of testing, we re-tested the mice on the discrete trial, rewarded alternation test after they had completed the spatial reference memory tasks. As before, the WT animals again showed a good level of spatial short-term memory performance (85.6%), whereas the Gria1^R/R mice performed at near chance levels (45.0%; F_(1,13) = 62.98; p < 0.01). The results were therefore very similar to the earlier assessments of spatial short-term memory during both spontaneous and rewarded alternation, and demonstrate that the pattern of positive and negative results in the preceding tasks was not due to the order of testing. Thus, in summary, Gria1^R/R mice exhibited an impairment on spatial working/short-term memory tasks but preserved spatial reference memory acquisition, both in the appetitively motivated Y-maze task and in the Morris water-maze.