Tissue selective deletion of GluR2 in c-KO mice
To study the effects of Ca2+
-permeable AMPARs on plasticity and memory we generated a line of mice in which GluR2 was deleted from pyramidal cells of the CA1 region of the hippocampus as described in Text S1
. In order to identify the loss of GluR2 mRNA we performed in situ hybridization on brain slices obtained from floxed mice (controls), GluR2-cKO mice and global GluR2-KO mice. Representative sagittal sections are shown in . In the GluR2-cKO mice, there was significant loss of GluR2 in the dorsal hippocampus and in cortex layer III (). When compared to control mice, the largest loss of GluR2 from the dorsal hippocampus of GluR2-cKO mice was seen in the CA1 region while minor loss of GluR2 was observed in CA3. In contrast to this conditional deletion, we observed a total loss of GluR2 mRNA from GluR2-KO mice, as expected (). Expression of GluR2 in the ventral hippocampus (CA1, CA3 and dentate gyrus) was completely normal in GluR2-cKO mice (). We also noted no difference in expression of GluR2 in the amygdala between GluR2-cKO mice and controls.
Conditional deletion of GluR2 in cKO mice.
In order to quantify the loss of GluR2 mRNA, we analyzed the percentage of neurons that contain GluR2 mRNA at 6 and 8 weeks of age in control and GluR2-cKO mice in these brain regions (). As expected, we observed a significant loss of GluR2 expression in dorsal CA1 pyramidal neurons in the GluR2-cKO mice at 6 weeks and 8 weeks of age compared to control mice of the same age (57.7±3.0% vs 91.8±1.4% at 6 weeks and 52.8±2.2% vs 95.2±0.4% at 8 weeks; p values<.05). Similarly, when cortex layer III neurons were counted, we observed a significant loss of GluR2 in the GluR2-cKO mice at 6 weeks and 8 weeks of age when compared to control mice of the same age (39.8±3.5% vs 70.6±5.5% at 6 weeks, and 40.0±2.2% vs 76.2±1.0% at 8 weeks; p values<.05). Meanwhile, in the dorsal CA3 region a modest but significant loss was seen at both time points when comparing GluR2-cKO mice to controls (78.5±2.6% vs 91.4±2.1% and 79.7±1.6% vs 91.0±1.3%; p values<.05). As expected, the dorsal dentate gyrus (DG) showed no loss in percentage expression (92.8±1.1% vs 91.3±2.6% and 92.7±0.9% vs 95.7±0.5%; p values>.05). We also found no statistically significant loss of GluR2 from pyramidal neurons in the CA1 region of the ventral hippocampus of the GluR2-cKO mice at 6 and 8 weeks of age when compared to controls of the same ages (83.9±1.9% vs 80.9±2.2% at 6 weeks and 86.7±4.1% vs 90.5±1.6% at 8 weeks; p values>.05).
In all subsequent experiments, we only used mice that were between 6 and 8 weeks of age. In order to assess if there was any change in the percentage of cells expressing GluR2 during this time period in mice within the same genotype, we compared percentage of cells that contained GluR2 mRNA at these two time points within each genotype, at all five anatomic regions analyzed. In the controls we observed no loss of GluR2 in dorsal hippocampus CA1 at 8 weeks compared to 6 weeks of age (p>.05). Importantly, it was also apparent that in the GluR2-cKO mice there was no further loss of GluR2 from dorsal CA1 pyramidal neurons between 6 and 8 weeks of age (p>.05). Likewise, the percentage of cortical layer III cells positive for GluR2 mRNA did not change between 6 and 8 weeks of age in the controls (p>.05), or, importantly, in the GluR2-cKO mice (p>.05). Nor was there any significant loss of cells positive for GluR2 in the GluR2-cKO mice between 6 weeks and 8 weeks of age in pyramidal neurons in the CA1 region of the ventral hippocampus, or in the DG, or CA3 regions of the dorsal hippocampus (p values >.05).
Based on these results we conclude that the majority of GluR2 loss is restricted to the dorsal hippocampus CA1 region and cortex layer III in the GluR2-cKO mice, with lesser loss in the dorsal CA3, and that expression levels are stable between 6 and 8 weeks of age in both genotypes in all five anatomical regions investigated.
Altered GluR2 gene expression does not affect cell survival
We next confirmed that loss of hippocampal GluR2 mRNA results in loss of GluR2 protein in GluR2-cKO mice. Immunohistochemical staining () with an anti-GluR2 antibody revealed that, when compared with the control mice (), the GluR2-cKO mice showed profound loss of GluR2 protein from cells in the CA1 region (), little loss of GluR2 from the CA3 region () and no evident loss of GluR2 protein from the DG (). Meanwhile GluR2 was completely absent from the hippocampus of GluR2 global KO mice ().
GluR2-cKO mice exhibit loss of GluR2 protein in CA1 and unchanged neuronal numbers.
It is possible that the loss of GluR2 protein could lead to increased calcium influx into cells through AMPARs which might lead to cell death. To assess whether there was loss of hippocampal neurons in 8 week old mice, we stained sections using an antibody to NeuN, which is a marker of mature neurons (), and we employed stereology 
to count the number of neurons in dorsal hippocampus between bregma AP -1.34 mm and bregma AP -2.06 mm in 9 control and 6 GluR2-cKO mice. There was no difference in the number of neurons in GluR2-cKO mice compared with controls (). These combined results demonstrate that while GluR2 expression is significantly reduced from the expected regions of the dorsal hippocampus of mutant mice this loss does not lead to cell death in mice at eight weeks of age. In addition, knockout of GluR2 did not affect the expression of other glutamate receptor subunits (Fig. S2
; for primer sequences see Table S1
Loss of GluR2 leads to enhanced LTP at synapses
A number of studies have characterized the effects of GluR2 deletion on synaptic transmission in the hippocampus 
. Therefore, we first confirmed that our GluR2-cKO mice showed similar changes (see Text S1
). As predicted from previous work, synaptic transmission was sensitive to blockers of GluR2-lacking AMPARs (Fig. S3
), input/output functions were right-shifted (Fig. S3B
) and there was little change in paired-pulse facilitation, (Fig. S3C
The induction of LTP by a conventional high-frequency stimulation (HFS) protocol was significantly enhanced in the CA1 region of GluR2-cKO mice (60 minutes post-HFS, fEPSPs were potentiated to 255±19% of baseline in GluR2-cKO mice, n
7, compared to 188±9% of baseline in control littermates, n
8, p<.05) (). Consistent with the notion that the larger LTP observed in GluR2-cKOs is associated with Ca2+
influx via GluR2-lacking AMPARs, HFS induced a significant LTP in slices from GluR2-cKO mice bathed in ACSF containing the NMDAR antagonist D-APV (100 µM; 60 minutes post-HFS fEPSPs were potentiated to 134±4% of baseline in cKO slices, n
4, compared to 110±3% of baseline in control slices, n
4, p<05). Nonetheless, the amount of potentiation was still reduced by D-APV in both groups (p<.05) suggesting that NMDARs contribute to LTP in GluR2-cKO mice and controls. While the induction of LTP by a less robust pattern of synaptic stimulation (30 seconds of 5 Hz stimulation) was also significantly enhanced in GluR2-cKO mice (), a shorter train of 5 Hz stimulation, near to the threshold for LTP induction in control mice, produced similar levels of LTP in slices from GluR2-cKOs and controls (). When LTD was induced by a long train of low-frequency stimulation (1 Hz for 15 min) similar levels of decreased responding were observed in GluR2-cKO and control slices (). Together, these findings are consistent with previous studies showing that the absence of GluR2 AMPAR subunits enhances the induction of LTP by some patterns of synaptic stimulation with little or no effect on LTD 
, however see 
. It is notable that other studies also provide data that is consistent with our findings, in that they demonstrate that Ca2+
-permeable AMPARs alter NMDAR dependency of HFS-induced CA1 LTP 
LTP is enhanced in GluR2-cKO mice.
GluR2-cKO mice express a unique form of synaptic plasticity
Hebbian plasticity refers to strengthening that is contingent upon the concurrent release of neurotransmitter and post-synaptic membrane depolarization (see Text S1
). The current-voltage (I/V) relations for NMDARs are such that channel conductance is low at negative membrane potentials and increases with membrane depolarization as the voltage-dependent Mg2+
ion block of the channel is relieved. Consequently, LTP is induced by NMDARs when the post-synaptic membrane is depolarized, coincident with activation of the receptor by glutamate. Hence LTP that is mediated by the NMDAR is referred to as Hebbian.
By contrast, the current-voltage (I/V) relations for GluR2-lacking (Ca2+
-permeable) AMPAR channels is inwardly rectifying, so that channel conductance and Ca2+
influx through GluR2-lacking AMPARs is most profound at negative (i.e. hyperpolarized) membrane potentials. Consequently, LTP mediated by GluR2-lacking (Ca2+
-permeable) AMPARs can be generated when presynaptic fiber stimulation is paired with hyperpolarization of the postsynaptic membrane 
In the next experiment we examined whether the loss of GluR2 in CA1 pyramidal neurons enables the expression of LTP at synapses onto these cells when they are hyperpolarized. In these experiments whole-cell current-clamp recordings were used to pair short trains of presynaptic fiber stimulation (100 pulses at 10 Hz) with either depolarization or hyperpolarization of the postsynaptic cells to approximately −20 or −120 mV, respectively. As shown in , a “Hebbian” protocol (i.e. pairing a train of 10 Hz synaptic stimulation with postsynaptic depolarization) induced robust LTP in control cells (30 minutes post-pairing EPSPs were potentiated to 221±12% of baseline, n
5). In pyramidal cells from GluR2-cKO mice this same pairing protocol also induced robust LTP that was significantly larger than that seen in cells from control slices (EPSPs were potentiated to 283±5% of baseline, n
4, p<.05 compared to controls). In contrast, pairing 10 Hz stimulation with postsynaptic hyperpolarization to approximately −120 mV () never induced significant LTP in cells from control mice (0 out of 12 cells, 30 minutes post-pairing EPSPs were 106±3% of baseline, n
6), however in approximately half of the cells (5 out of 11 cells) we recorded from in slices from GluR2-cKO mice this pairing protocol did induce significant LTP (across all cells EPSPs were potentiated to 150±15% of baseline, n
9, p<0.05 compared to control). Stimulation in the absence of postsynaptic hyperpolarization did not produce potentiation in GluR2-cKO mice (). This demonstrates that LTP in these animals can be induced via a second, separate mechanism from classical Hebbian LTP induction.
LTP can be induced in GluR2-cKO mice at both hyperpolarized and depolarized membrane potentials.
In summary, the experiments above demonstrate that LTP in the CA1 region of our mutants is dependent on both NMDARs and Ca2+-permeable GluR2-lacking AMPARs. The next experiments examined the functional effects of adding AMPAR-dependent plasticity to CA1 pyramidal cells.
Contextual fear conditioning is impaired in GluR2-cKO mice
We first examined contextual fear conditioning, a form of learning known to depend on the hippocampus and NMDAR-mediated plasticity in CA1 
. Mice were placed in a novel environment and allowed to explore for two minutes before receiving either 1 or 5 footshocks. Control mice and GluR2-cKOs showed significant and equivalent increases in freezing immediately after shock relative to the baseline period (no effect of genotype F (1,55)
1.48, p>.05, effect of shock number F (1, 55)
67.93, p<.05, no genotype x shock number interaction F<1, effect of period (baseline vs. shock) F (1, 55)
113.09, p<0.05), no period x genotype interaction F (1, 55)
2.125, p>.05, period x shock number interaction F (1, 55)
49.85, p<.05, no period x genotype x shock number interaction F<1) (). There are substantial data indicating that freezing during this period is entirely a conditional response to contextual stimuli that have become associated with shock. Therefore, similar levels of post-shock freezing suggest that short-term memory is intact in GluR2-cKO animals 
(see also Text S1
). The same mice were brought back to the context 24 hours later to test for long-term memory (). During this test, GluR2-cKOs froze significantly less than control mice (main effect of genotype F (1, 55)
24.69, p<0.05) both in the 1 and 5 shock groups (no genotype x shock number interaction F<1). Baseline freezing levels from the training session (prior to shock delivery) are shown for comparison. These data demonstrate that deletion of GluR2 in the CA1 region impairs long-term memory for context fear.
Deletion of GluR2 in CA1 impairs context fear.
In the next experiment, we examined the time course of memory loss by testing animals immediately, 2 hours or 24 hours after training (). Based on the results above, we predicted that memory in GluR2-cKO mice would be normal immediately after training but impaired at longer delays. To test this idea we conducted a set of planned contrasts (Fisher's PLSD) between controls and GluR2-cKOs that revealed memory was intact immediately after training (p>.05) but significantly impaired 2 hours and 24 hours later (p values <0.05). The fact that we observed normal post-shock freezing in our first two experiments and in our subsequent c-fos experiment strongly suggests that short-term memory is intact in GluR2-cKO mice. These data are in line with reports that place field stability is significantly impaired in global GluR2 knockout mice 
To ensure that controls and GluR2-cKO mice were using the dorsal hippocampus to store context fear memories we selectively removed this structure 1 day after training (histology in Fig. S4D
). Previous studies have shown that post-training lesions of the dorsal hippocampus severely impair memory for context fear in both rats and mice 
. Consistent with these results, we found that controls and GluR2-cKOs with dorsal hippocampus lesions showed significantly less context fear than sham-operated animals (main effect of lesion, (F (1,53)
49.49, p<.05), main effect of genotype, (F (1, 53)
35.63. p<.05), no genotype x lesion interaction, F<1) (). This demonstrates that the dorsal hippocampus is used to store and retrieve context fear memory in our experiments.
We next determined if GluR2-cKO mice were able form a long-term context memory in the absence of shock. To do this we used a context pre-exposure procedure that has previously been shown to depend on the hippocampus and NMDAR activation 
. In this procedure, animals learn about the context prior to fear conditioning and have to retain this information across a 24-hour period 
. On day 1, half of the mice were pre-exposed to the training context for 10 minutes (in the absence of shock) while the other animals remained in their homecages. The next day, all animals were trained with an immediate shock delivered 5 seconds after placement in the context. Without pre-exposure, this short interval does not provide enough time to learn about the context before shock is delivered. Consistent with this fact, pre-exposed control animals showed considerably more context fear than non-exposed mice (). In contrast, pre-exposed GluR2-cKOs did not benefit from this experience (main effect of genotype (F (1, 46)
16.07, p<.05, main effect of exposure (F (1, 46)
14.77, p<.05), significant genotype x exposure interaction (F (1, 46)
9.874, p<.05). Post-hoc tests revealed that pre-exposed control mice froze significantly more than pre-exposed GluR2-cKO animals (p<.05). These results demonstrate that GluR2-cKO mice have an impaired ability to form long-term memories of the context.
To determine if the GluR2-cKO deficit in fear conditioning was specific to context fear we also examined hippocampus-independent auditory fear conditioning 
. This type of conditioning depends on NMDAR activation in the amygdala 
. Mice were trained with 5 white noise-shock pairings and then received an auditory test in a novel environment 24 hours later (). During the baseline period of this test, GluR2-cKO mice exhibited less generalized fear to the novel environment (main effect of genotype F (1, 31)
4.6, p<0.05), a phenotype consistent with their overall reduction in context fear. Despite this fact, GluR2-cKO mice showed robust increases in freezing during white noise presentations that did not differ from controls (effect of period (baseline vs. tone) F (1, 31)
434.93, p<0.05, main effect of genotype (F (1, 31)
7.18, p<.05, no period x genotype interaction F<1). Post-hoc tests revealed that white noise freezing levels were similar in controls and GluR2-cKO mice (p>.05). The same knockout animals exhibited significantly less context fear than controls (main effect of genotype F (1, 31)
9.89, p<0.05) when tested in the training environment the next day (). These results demonstrate that the loss of GluR2 in CA1 selectively impairs context fear (see Text S1
for a more detailed discussion).
Lastly, we assessed motor function using rotarod and openfield tests (Fig. S4A
) and pain processing by analyzing shock reactivity (Fig. S4C
) and found that them to be normal in GluR2-cKO mice. This indicates that the reduced context fear observed in these animals does not result from hyperactivity or reduced pain sensation. These data are also consistent with the fact that GluR2-cKO mice show normal short-term memory for context fear (). Intact short-term memory suggests that context exploration and pain processing are normal 
Spatial learning and memory are impaired in GluR2-cKO mice
Spatial learning in the Morris watermaze is dependent on NMDAR activation in the hippocampus 
. We trained GluR2-cKO mice on a fixed-visible version of this task that produces robust spatial memory and facilitates procedural learning 
. Across training days, both controls and GluR2-cKO mice showed a reduction in the distance traveled to reach the platform (main effect of day F (4, 104)
22.299, p<0.05, no effect of genotype (F<1), no day x genotype interaction (F (4, 104)
1.119, p>.05) (). Spatial memory was assessed on day 6 by administering a 60 s probe test with the platform removed. Control mice showed selective searching in the target quadrant where the platform was located during training compared to the other quadrants (main effect of quadrant, F (1, 15)
6.102, p<.05) (). In contrast, GluR2-cKO mice searched equally in all quadrants suggesting they did not learn the spatial location of the platform (no effect of quadrant, F<1). These data suggest that deletion of GluR2 in the CA1 region of the hippocampus impairs the formation of spatial memory.
Long-term spatial memory is impaired in GluR2-cKO mice.
Long-term spatial memory was also assessed using the reference memory version of the radial maze 
. In this task, mice were required to remember the spatial location of 4 baited arms across a 24-hour period. The same arms were baited during each session and the mice were trained for 9 consecutive days. There was an increase in the percentage of correct choices across training days (main effect of day F (8, 208)
9.94, p<0.05) that was larger in control animals than GluR2-cKOs (significant day x genotype interaction F (8, 208)
2.31, p<0.05, no effect of genotype F (1, 26)
2.849, p>.05) (). Post-hoc tests (Fisher's PLSD) revealed that control mice performed significantly better than knockout animals on the last day of training (p<0.05). Control mice also showed a significant decrease in errors (i.e. visits to unbaited arms) across days (main effect of day F (8, 104)
2.52, p<0.05) while GluR2-cKO mice did not (no effect of day F<1) (). Post-hoc tests (Fisher's PLSD) showed that control mice made significantly fewer errors on days 8 and 9 than GluR2-cKO animals (p<0.05). These results are consistent with our watermaze data and demonstrate that the deletion of GluR2 in CA1 impairs the formation of long-term spatial memories.
In the next experiment we examined spatial working memory using the win-shift version of the radial maze 
. On this version of the task, animals were required to remember the spatial location of 4 baited arms across a 2-minute period. New arms were randomly chosen on each day to eliminate the contribution of long-term spatial memory. We found that controls and GluR2-cKO mice showed significant (main effect of day F (8, 160)
15.36, p<0.05) and equivalent (no effect of genotype, F<1, no day x genotype interaction F<1) increases in the percentage of correct choices across training days (). In addition, the number of errors (i.e. visits to unbaited arms) made by controls and cKO mice decreased across days (main effect of day F (8, 160)
7.65, p<0.05) and were not different between groups (no day x genotype interaction F<1) (). To demonstrate that the mice were not forming long-term spatial memories we increased the interval between phases (2–480 min) and found a systematic decrease in the performance of controls and cKO animals (main effect of time F (5, 100)
15.55, p<0.05) that did not differ between groups (no time x genotype interaction F<1) (data not shown). Animals returned to naive levels of performance at the longest interval and made the same number of errors as they did on the first day of training (no effect of session (first vs. last) F (1, 20)
1.08, p>0.05). These results suggest that selective deletion of GluR2 in CA1 does not impair spatial working memory. This finding is consistent with recent data showing that NMDAR-dependent plasticity in CA1 is not required for the retention of spatial information across short intervals 
Thus, in addition to showing deficits in contextual fear conditioning GluR2-cKO animals exhibit impaired long-term memory in spatial learning tasks that are known to require NMDAR activation in the CA1 region of the hippocampus. Therefore, our study is the first to demonstrate that hippocampus-dependent learning impairments in GluR2 knockout mice can be produced by targeted deletion of this subunit in the CA1 region. This deficit does not result from impaired NMDAR-dependent LTP as our electrophysiology experiments demonstrate that D-APV significantly reduces potentiation in GluR2-cKO mice.
Immediate early gene expression is normal in GluR2-cKO mice
It is possible that the learning impairments observed in our GluR2-cKO mice result from reduced synaptic transmission and/or excitability and not altered plasticity. We addressed this issue by determining if CA1 neurons are similarly engaged by fear conditioning in controls and GluR2-cKO mice. To do this we examined the expression of c-fos, an immediate early gene that is an indicator of neural activity and whose expression is significantly increased in the CA1 region of the hippocampus following context fear conditioning 
. Homecage controls were compared to mice that received 5 unsignaled shocks (identical to the training procedures above). Conditioned mice showed robust short-term memory at the end of training that did not differ between controls and GluR2-cKO animals (no effect of genotype F<1). Ninety minutes later we sacrificed the animals, froze their brains on dry ice and performed immunohistochemical staining to determine the level of c-fos expression in the CA1 region of the hippocampus (). Controls and GluR2-cKO mice that remained in their homecages showed similar levels of expression (F<1). Following fear conditioning, there was a robust increase in c-fos expression that also did not differ between control and GluR2-cKOs () (main effect of training F (1, 8)
4439.43, p<0.05, no training x genotype interaction F (1, 8)
1.17, p>0.05). These results demonstrate that context fear conditioning produces normal activation of CA1 neurons in GluR2-cKO animals. We also looked at maximal activation following kainate-induced seizures (). Once again there was no difference in the level of c-fos expression between controls and GluR2-cKO animals (no effect of genotype F<1). In addition, no difference in seizure susceptibility was observed (). Together, these results suggest that the learning deficits observed in our GluR2-cKO mice are not due to reduced neural activity in the CA1 region of the hippocampus.
Knockout of GluR2 in CA1 does not affect immediate early gene expression or seizure susceptibility.
NMDAR-independent learning is not impaired in GluR2-cKO mice
Our results demonstrate that deletion of GluR2 in the CA1 region of the hippocampus produces AMPAR-mediated plasticity that impairs learning and memory. However, AMPAR-mediated plasticity may not be detrimental to all types of learning. In fact, a number of different experiences (e.g. learning, drug exposure, sensory deprivation) increase the expression of GluR2-lacking Ca2+
-permeable AMPARs 
. It is possible, therefore, that the experience-dependent expression of these receptors plays a functional role in subsequent plasticity and learning (i.e. experience-dependent learning). The following experiments examined this idea.
Experience-dependent learning can be studied in the hippocampus using the ‘upstairs/downstairs’ procedure. In this task, an initial learning event on day 1 (context A; ‘upstairs’) is NMDAR-dependent, while a subsequent learning event the next day (context B; ‘downstairs’) is NMDAR-independent 
. We first established the ‘upstairs/downstairs’ effect in wild-type mice using the NMDAR antagonist CPP. 129S6 mice were conditioned sequentially in two different environments (design illustrated in ). One group of mice received saline injections before training in context A on day 1 and CPP before training in context B on day 2. A second group of mice received CPP prior to training in context A on day 1, followed by an injection of saline before training in context B on day 2. Both groups of mice were then tested on days 3 and 4 (no injections) to assess memory for each context. Similar to previous reports, we found that injections of CPP given prior to training were more effective at blocking learning in context A than context B (significant context x drug interaction F (1, 14)
11.824, p<.05) () 
. This demonstrates that NMDAR activation is required for initial learning in context A, but not subsequent learning in context B.
NMDAR and GluR2-lacking AMPAR make distinct contributions to learning.
We next studied the effect of the GluR2-cKO on fear learning in the ‘upstairs/downstairs’ paradigm. We found that deletion of GluR2 was more effective at blocking initial learning in context A than subsequent learning in context B (significant context x genotype interaction (F (1, 47)
5.802, p<0.05)) (). This suggests that GluR2-cKO mice are more impaired on hippocampus-dependent learning tasks that require the NMDAR (context A; day 1) than tasks that are independent of the NMDAR (context B; day 2).
In the next experiment we verified that GluR2-cKO mice were indeed using an NMDAR-independent learning mechanism in context B (). To do this we used the same behavioral design as in our first experiment (illustrated in ). In control animals, we once again found that CPP was more effective at blocking learning in context A than context B (significant context x drug interaction F (1, 35)
12.728, p<.05). Similarly, GluR2-cKO mice that received saline were once again impaired in context A, but not context B, relative to control mice that received saline (significant context x genotype interaction, F (1, 21)
10.347, p<.05). Lastly, the increased freezing GluR2-cKO mice exhibited in context B was not reduced by the administration of CPP (effect of context, F (1, 18)
46.036, P<.05; no context x drug interaction, F<1). These results demonstrate that learning in context B occurs via an NMDAR-independent plasticity mechanism in both control and GluR2-cKO mice. In the next experiments we examined cellular mechanisms that could potentially mediate learning in the absence of NMDAR activation.
Our electrophysiological data indicate that Ca2+-permeable AMPARs are present in the CA1 region of our GluR2-cKO mice and are able to mediate LTP in the absence of NMDAR activation. It is possible, therefore, that these receptors contribute to NMDAR-independent learning in context B. To examine this idea we trained GluR2-cKO animals using the ‘upstairs/downstairs’ design. We trained the mice in context A and then administered saline, CPP, or CPP + IEM-1460 (a Ca2+-permeable AMPAR antagonist) immediately prior to learning in context B (). Consistent with our previous results, planned comparisons (Fisher's PLSD) showed that blocking NMDARs with CPP did not impair learning in context B relative to saline controls (p>.05). In contrast, the addition of a Ca2+-permeable AMPAR antagonist produced a significant memory deficit relative to controls that received saline (p<.05). This data suggests that GluR2-lacking receptors contribute to NMDAR-independent learning in our knockout animals.
In our last experiment we determined if Ca2+
-permeable AMPARs also contribute to NMDAR-independent learning in wild-type mice (). Planned comparisons (Fisher's PLSD) showed that after training in context A, blocking NMDARs did not affect learning in context B relative to mice that received saline (p>.05). In addition, just as in the GluR2-cKO mice, the addition of IEM-1460 significantly impaired learning in context B (p<.05). This suggests that GluR2-lacking AMPARs play a role in NMDAR-independent learning in wild-type mice. This result is consistent with studies showing that activation of NMDARs can induce the synaptic expression of Ca2+
-permeable AMPARs