Chronic GluN2B antagonism disrupts behavior in wild-type mice without protecting against synapse loss or memory impairment in Alzheimer's disease mouse models

Abstract

Extensive evidence implicates GluN2B-containing NMDA receptors (GluN2B-NMDARs) in excitotoxic-insult-induced neurodegeneration and amyloid β (Aβ)-induced synaptic dysfunction. Therefore, inhibiting GluN2B-NMDARs would appear to be a potential therapeutic strategy to provide neuroprotection and improve cognitive function in Alzheimer's disease (AD). However, there are no reports of long-term in vivo treatment of AD mouse models with GluN2B antagonists. We used piperidine18 (Pip18), a potent and selective GluN2B-NMDAR antagonist with favorable pharmacokinetic properties, for long-term dosing in AD mouse models. Reduced freezing behavior in Tg2576 mice during fear conditioning was partially reversed after subchronic (17 d) Pip18 treatment. However, analysis of freezing behavior in different contexts indicated that this increased freezing likely involves elevated anxiety or excessive memory generalization in both nontransgenic (NTG) and Tg2576 mice. In PS2APP mice chronically fed with medicated food containing Pip18 for 4 months, spatial learning and memory deficits were not rescued, plaque-associated spine loss was not affected, and synaptic function was not altered. At the same time, altered open field activity consistent with increased anxiety and degraded performance in an active avoidance task were observed in NTG after chronic treatment. These results indicate that long-term treatment with a GluN2B-NMDAR antagonist does not provide a disease-modifying benefit and could cause cognitive liabilities rather than symptomatic benefit in AD mouse models. Therefore, these results challenge the expectation of the therapeutic potential for GluN2B-NMDAR antagonists in AD.

Keywords: Alzheimer's disease; GluN2B; NMDAR; cognition; memory; spine.

Figure 1.

Pip18 is a potent and selective GluN2B antagonist with good brain availability. A, Potency and selectivity of Pip18 in cell-based assays. Top, Structure of Pip18. Center, Example experiment showing the dose–response curve for Pip18 using HEK cells stably expressing GluN1 and GluN2B NMDAR subunits. Bottom, Example experiment showing the lack of antagonism using HEK cells stably expressing GluN1 and GluN2A NMDAR subunits. Data are shown as mean ± SD. B, Top, Examples of isolated NMDAR fEPSPs recorded from CA1 stratum radiatum in response to Schaffer collateral stimulation for 2-week-old and 2-month-old mice before and after 1 μm Pip18 application. Bottom, Quantification of NMDAR EPSP inhibition by Pip18 in 2-week-old and 2-month-old mice (n = 4/group). As a control, inhibition by the nonselective NMDAR antagonist D-AP5 (50 μm) is shown. Data are shown as mean ± SEM. C, PK analysis of Pip18 brain levels after injections of 5 or 50 mg/kg Pip18 (n = 3 mice/time point/dose). Data are shown as mean ± SD. D, PK analysis of Pip18 brain levels during day 5 of treatment with Pip18 via medicated food using 100 or 300 mg/kg food (n = 3 mice/time point/dose). Data are shown as mean ± SD.

Figure 2.

Subchronic Pip18 partially reverses decreased freezing in Tg2576 mice during CFC. A, Freezing in response to the context, altered context, and cue. Planned comparisons showed significantly reduced freezing in vehicle-treated Tg2576 mice compared with NTG mice in response to the context, altered context, and cue (one-tailed t tests, black stars). ANOVA with the NTG group excluded revealed a significant effect of Pip18 treatment within Tg2576 mice for the altered context (F_(2,42) = 4.219, p = 0.021) and follow-up tests showed that the 15 mg/kg treatment was significantly different from vehicle (p < 0.05). ANOVA with the NTG group excluded also revealed a significant effect of Pip18 treatment within Tg2576 mice for the cue (F_(2,42) = 3.399, p = 0.043). Planned comparisons also showed increased freezing in Pip18-treated versus vehicle-treated Tg2576 mice (one-tailed t test, red stars). n = 14 NTG, n = 15 Tg2576/treatment group. *p < 0.05; **p < 0.01; ***p < 0.001. B, Freezing during training is shown. Comparison between vehicle-treated NTG and Tg2576 mice by ANOVA across time points revealed a significant interaction between genotype and time point (F_(3,112) = 4.458, p = 0.005) and follow-up tests showed a significant overall difference between NTG and Tg2576 mice and a significant effect of genotype within the last time point. *p < 0.05. ANOVA of the Tg2576 treatment groups across time points with the NTG group excluded revealed a significant effect of time point (F_(3,168) = 13.342, p < 0.001), but not treatment and no interaction between treatment and time point. C, Aβ levels measured by ELISA in the cortex and hippocampus of Tg2576 mice. There was no significant effect of treatment on Aβ40 or Aβ42 in either the cortex or hippocampus (n = 15 per group). All data are shown as mean ± SEM.

Figure 3.

Subchronic treatment increases fear memory generalization in both NTG and Tg2576 mice. A, Freezing during fear memory testing. ANOVA for freezing in the context revealed a significant effect of genotype (F_(1,83) = 37.69, p < 0.001), but not treatment (n = 14–15 mice per group). ANOVA for freezing in the altered context revealed a significant effect of treatment (F_(2,83) = 6.86, p = 0.002) and genotype (F_(1,83) = 23.51, p < 0.001), but no interaction. Follow-up tests showed that both 1 and 10 mg/kg were significantly different from vehicle (p < 0.01). ANOVA for freezing to the cue revealed a significant effect of treatment (F_(2,83) = 3.52, p = 0.034) and genotype (F_(1,83) = 22.85, p < 0.001), but no interaction. Follow-up tests showed that the 1 mg/kg group was significantly different from the vehicle group (p < 0.05). B, The discrimination index (DI) was calculated from the time spent freezing as follows: DI = (Context − Altered)/(Context + Altered). ANOVA revealed significant effects of treatment (F_(2,83) = 8.864, p < 0.001) and genotype (F_(1,83) = 5.95, p = 0.017), but no interaction. Follow-up tests showed that both 1 and 10 mg/kg were significantly different from vehicle (p < 0.05 and p < 0.001, respectively). The decreased DI values correspond to increased generalization. *p < 0.05; **p < 0.01; ***p < 0.001.

Figure 4.

Chronic treatment does not rescue spatial learning and memory deficits in PS2APP mice. A, Course of Pip18 treatment relative to the accumulation of plaque load (approximated as average plaque size × average plaque volume) in PS2APP mice. The timing of the behavioral testing before tissue harvest for subsequent spine analysis is also illustrated. B, Distance traveled to locate a hidden or visible platform in the water maze. Assessment of spatial learning during hidden training sessions (1–9) with a 2-way repeated-measures ANOVA revealed a significant effect of training session (F_(8,28) = 22.9; p < 0.0001) and genotype (F_(1,35) = 26.1; p < 0.0001), but no treatment effect or interaction (n = 9–10/group). There were no significant genotype or treatment differences during visible training sessions (v1–v3). C, Latency to locate a hidden or visible platform in the water maze. Two-way repeated-measures ANOVA revealed a significant main effect of training session (F_(8,28) = 37.5; p < 0.0001) and genotype (F_(1,35) = 21.1; p < 0.0001), but no treatment effect or interaction (n = 9–10/group). There were no significant genotype or treatment differences during visible training sessions (v1–v3). D, Spatial memory assessment in the water maze. Probe trials were given before training session 7 and 24 h after the final training session. Two-way ANOVA of the number of crossings over the target zone (200% larger than the platform) revealed a significant main effect of genotype (F_(1,35) = 4.3; p < 0.05), but no treatment effect or interaction (n = 9–10/group). All data are shown as mean ± SEM. E, Distance traveled to locate a hidden tunnel in the Barnes maze. Assessment of spatial learning by training trial with a two-way repeated-measures ANOVA revealed a significant main effect of training trial (F_(7,31) = 24.1; p < 0.001) and genotype (F_(1,37) = 9.4; p < 0.01), but no treatment effect or interaction (n = 10–11/group). F, Errors made during Barnes maze training. Assessment of spatial learning by training trial with a two-way repeated-measures ANOVA revealed a significant main effect of training trial (F_(7,31) = 19.7; p < 0.001) and genotype (F_(1,37) = 15.7; p < 0.001), but no treatment effect or interaction (n = 10–11/group). G, Spatial memory was tested 24 h after the final Barnes maze training trial by measuring the percentage of time searching over the former target/all hole locations. Two-way ANOVA revealed a significant main effect of genotype (p < 0.05), but no treatment effect or interaction (n = 10–11/group). All data are shown as mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001.

Figure 5.

Chronic treatment alters anxiety-like behavior and impairs active avoidance performance in NTG mice. A, Percent activity in the center of the open field. ANOVA revealed a significant difference between the control and Pip18-treated NTG mice groups (F_(3,37) = 5.7; p < 0.01). *p < 0.05 by Tukey's post hoc test. B, Number of rearings in the open field. ANOVA revealed a significant difference between the control and Pip18-treated NTG mice groups (F_(3,37) = 10.8; p < 0.0001). ***p < 0.001 by Tukey's post hoc test. C, Total distance traveled in the open field. There were no significant differences between the groups. D, Percent avoidance responses during active avoidance training. Two-way repeated-measures ANOVA revealed a significant effect of genotype (F_(1,31) = 9.5; p < 0.01), but no treatment effect or interaction. However, analysis of performance on day 3 of training (blocks 11–15) found a significant effect of treatment within the NTG mice (F_(1,15) = 5.1; *p < 0.05; n = 8–9/group). All data are shown as mean ± SEM. E, Percentage of escape responses during active avoidance training. Two-way repeated-measures ANOVA revealed a significant main effect of genotype (F_(1,31) = 8.6; p < 0.01), but no treatment effect or interaction. However, analysis of performance on day 3 of training (blocks 11–15) found a significant main effect of treatment in the NTG mice (F_(1,15) = 5.2; *p < 0.05; n = 8–9/group). All data are shown as mean ± SEM.

Figure 6.

Chronic Pip18 treatment does not rescue spine loss in PS2APP mice. A, Example images of GFP-labeled spines that were imaged at the end of the study in NTG mice or PS2APP mice either away from plaques or near plaques (methoxy-X04-labeled in blue). Scale bar is 20 μm and applies to all panels. B, Pip18 treatment did not alter the density of plaques or the average volume of plaques (n = 7–8 mice per group). C, Although there was a significant effect of genotype on spine density, with reduced density in PS2APP mice, there was no effect of treatment (n = 7–8 mice per group). Spine density was significantly reduced in PS2APP mice both near plaques and away from plaques in both the vehicle and Pip18 treatment groups compared with vehicle-treated NTG mice. *p < 0.05; **p < 0.01; ***p < 0.001. D, There was no effect of genotype or treatment on spine volume. All data are shown as mean ± SEM.

Figure 7.

Chronic treatment does not alter synaptic function in PS2APP mice. A, Input–output relationships for EPSPs measured in response to logarithmically increasing stimulus intensities (n = 26–32 slices from six to seven animals per group). The corresponding stimulus intensities were as follows: 0.004, 0.006, 0.008, 0.012, 0.017, 0.024, 0.034, 0.040, 0.057, 0.082, 0.117, 0.168, 0.240, 0.344, 0.493, 0.706, and 1.000 mA. B, The data from the maximal stimulus intensities in A replotted for clarity. There we no effects of genotype or treatment on the maximal EPSP response. C, EPSPs were measured in CA1 in response to Schaffer collateral stimulation. The PPR is shown as a function of interstimulus interval (n = 26–32 slices from six to seven animals per group). Assessment by 3-way ANOVA found a significant interaction between interval and genotype (DF = 6, F_(6,749) = 4.76, p < 0.001), but no effect of treatment. All data are shown as mean ± SEM. ***p < 0.001.

Figure 8.

Early treatment with Pip18 fails to prevent spine loss in PS2APP mice. A, Course of Pip18 treatment relative to the accumulation of plaque load (approximated as average plaque size × average plaque volume) in PS2APP mice and the timing of tissue harvest for spine analysis. B, Pip18 treatment did not alter the density of plaques or the average volume of plaques (n = 4–5 mice per group). C, Although there was a significant effect of genotype on spine density with reduced density in PS2APP mice, there was no effect of treatment (n = 4–5 mice per group). Spine density was reduced in PS2APP mice both near plaques and away from plaques. *p < 0.05; **p < 0.01; ***p < 0.001 relative to vehicle-treated NTG mice. D, There was no effect of genotype or treatment on spine volume. All data are shown as mean ± SEM.