Downregulation of NR3A-containing NMDARs is required for synapse maturation and memory consolidation

Abstract

NR3A is the only NMDA receptor (NMDAR) subunit that downregulates sharply prior to the onset of sensitive periods for plasticity, yet the functional importance of this transient expression remains unknown. To investigate whether removal/replacement of juvenile NR3A-containing NMDARs is involved in experience-driven synapse maturation, we used a reversible transgenic system that prolonged NR3A expression in the forebrain. We found that removal of NR3A is required to develop strong NMDAR currents, full expression of long-term synaptic plasticity, a mature synaptic organization characterized by more synapses and larger postsynaptic densities, and the ability to form long-term memories. Deficits associated with prolonged NR3A were reversible, as late-onset suppression of transgene expression rescued both synaptic and memory impairments. Our results suggest that NR3A behaves as a molecular brake to prevent the premature strengthening and stabilization of excitatory synapses and that NR3A removal might thereby initiate critical stages of synapse maturation during early postnatal neural development.

Figure 1. Expression of GFPNR3A in transgenic mice

(A) Constructs used for reversible production of GFPNR3A transgenic mice. The CaMKIIα promoter determines the temporal/spatial expression pattern of tTA. In the absence of doxycycline, the tTA protein drives expression of GFPNR3A linked to the tetO promoter (Gene ON). (B) Direct GFP visualization of GFPNR3A protein in the forebrain of double transgenic mice at postnatal day 25 (P25). Scale bar: 100 μm (25 μm in insets). Note high levels of expression in the striatum, specific cortical layers, and hippocampal CA1. (C) Dendritic targeting of GFPNR3A in pyramidal neurons of CA1 hippocampus (left, middle) and layer IV of the neocortex (right), shown by DAB-enhanced immunohistochemistry using anti-GFP antibody. Scale bars: 50 μm (left), 10 μm (middle, right). (D) Expression of transgenic GFPNR3A is regulated by doxycycline administration as shown by DAB-enhanced anti-GFP immunohistochemistry. Gene ON, P26 transgenic mice raised off doxycycline; Gene OFF, P26 transgenic mice shifted onto doxycycline at P21 (2 mg/ml in drinking water for 5 days) after allowing transgene expression at early stages. Abbreviations: St, striatum; SS1, primary somatosensory cortex; CA1, CA1 hippocampus; s rad, stratum radiatum; Cg, cingulate cortex; OT, olfactory tubercle; LS, lateral septum; Hi, hippocampus; Amy, amygdala. Scale bar: 500 μm.

Figure 2. GFPNR3A-containing NMDARs are functional and inserted into synapses

(A) Transgenic GFPNR3A assembles with endogenous NMDAR subunits. Total membrane extracts from hippocampi of P25 double transgenic mice were solubilized, immunoprecipitated with anti-GFP antibody and immunoblotted with the indicated antibodies. Additionally, 5% of the lysate (Input) used for immunoprecipitation was loaded. (B–C) Effect of prolonged NR3A expression on synaptic NMDAR currents. Data are from pharmacologically-isolated NMDAR-mediated EPSCs recorded from CA1 pyramidal neurons and evoked by Schaffer collateral stimulation. (B) Input-output (I–O) curve demonstrating that the amplitude of synaptic NMDAR currents recorded at +40 mV is reduced in neurons from P21–P25 GFPNR3A transgenic mice (dt GFPNR3A). RMANOVA revealed a significant main effect of genotype (F_{(1, 17)} = 10.56, p < 0.01). In this and all subsequent figures, error bars represent s.e.m. (C) Current-voltage (I–V) curve demonstrating reduced rectification of synaptic NMDAR currents measured in the presence of 4 mM Mg²⁺ in neurons from P21–P25 GFPNR3A transgenics. RMANOVA indicated a significant main effect of genotype (F_{(1, 25)} = 4.22, p < 0.05). (D) Synaptic NMDAR currents recorded in CA1 neurons from NR3A knockout mice after Schaffer-collateral stimulation exhibit stronger rectification at early postnatal stages (P6–P8) but normalize afterwards. (E) Comparison of sensitivity to Mg²⁺ block indexes (1/C₀) calculated for NMDARs in control, GFPNR3A transgenic and NR3A knockout mice at the ages indicated. Results are averages of 1/C₀ values calculated from I/V curves fitted as described in Appendix S1 (n= 6–14 mice per group).

Figure 3. Hippocampal LTP is attenuated in GFPNR3A transgenic mice

(A) LTP evoked by HFS of Schaffer collaterals and measured with field excitatory postsynaptic potentials (fEPSPs) in the CA1 region of the hippocampus is attenuated in P25 GFPNR3A transgenic mice (p < 0.005, Student’s t-test). Inset: Representative traces from the average of 30 traces before and after HFS in slices from control and GFPNR3A transgenic mice. (B) A stronger LTP protocol (3 epochs of HFS, arrows) was unable to rescue the deficit in GFPNR3A transgenic mice (p < 0.05, Student’s t-test). Inset: Representative traces from the average of 30 traces before and after repeated HFS in slices from control and GFPNR3A transgenic mice. (C) No apparent defects in LTD induced by low-frequency stimulation (LFS). Inset: Representative traces from the average of 30 traces before and after LFS in slices from control and GFPNR3A transgenic mice. (D) PPF (interstimulus interval: 25 ms to 250 ms) data indicate that release probability is similar in neurons from P25 control and transgenic mice. Inset: Representative traces from control and GFPNR3A transgenic experiments taken from the 25 ms interval. (E) Administration of doxycycline for 14 days (from P26–P40, see scheme) normalizes LTP deficits in GFPNR3A transgenics.

Figure 4. Decreased spine density in CA1 hippocampal neurons of GFPNR3A transgenic mice

(A) CA1 pyramidal neurons in acute hippocampal slices were filled with Alexa 568 (50 μM) via the patch pipette and transgenic neurons were identified by GFP-fluorescence. A maximum projection image of a filled control CA1 neuron is shown. Scale bar, 100 μm. (B) Only a fraction of CA1 pyramidal neurons express GFPNR3A. GFP-positive neurons (green) were counted on confocal hippocampal sections from double transgenic mice counterstained with TO-PRO-3 (blue) to label all nuclei. Scale bar, 50 μm. For quantitative analysis, brains from 10 double transgenic P25 mice were used and confocal CA1 images were acquired at various distances from the central line for neuronal counting (see schematic inset). Higher numbers of GFP-positive cells were found close to the subiculum. (C) Representative high magnification projections of Alexa 568-filled apical dendritic segments of P25 control and GFPNR3A-expressing CA1 neurons. Scale bar, 3 μm. (D) Mean spine densities in apical dendrites of P25 control and GFPNR3A-expressing CA1 neurons. Values from 3–10 secondary/tertiary dendrites were averaged to yield individual neuron parameters (* p < 0.05, ANOVA followed by Tukey’s t-test, n = 6–11 neurons per group). (E) Spines were grouped in different morphological categories and spine densities for the different categories were calculated as in D (see examples in C). (F) Representative apical dendritic segments of Golgi-impregnated CA1 pyramidal neurons from P25 control and GFPNR3A transgenic mice. Scale bar, 5 μm. (G) Mean spine densities in apical dendrites from CA1 pyramidal neurons. Values from 3–5 secondary/tertiary dendrites were averaged to calculate individual neuron parameters (* p < 0.005, ANOVA followed by Tukey’s t-test, n = 34–45 Golgi-stained neurons from 7–8 control and GFPNR3A transgenic mice).

Figure 5. GFPNR3A transgenic mice have fewer and smaller synapses

(A, B) Electron microscopy analysis shows reduced synapse density in the stratum radiatum of hippocampal CA1 from P25 GFPNR3A transgenic mice. s, spine. (C) Left, mean PSD length was significantly reduced in P25 GFPNR3A transgenics. Right, cumulative distribution plots reveal a shift towards synapses with shorter PSDs in P25 transgenic mice. (D, E) Synapse numbers (D) but not PSD length (E) returned to control levels by P35. Doxycycline administration for 10 days (from P26–35, see scheme) reversed the changes in PSD length. No differences in synapse number or size were observed between doxycycline-treated and untreated control mice (* p < 0.001, ANOVA followed by Tukey’s test, n = 6–8 mice per group, 10 fields of each animal were randomly acquired at the level of the central stratum radiatum and analyzed in a blind fashion).

Figure 6. Altered NR3A expression decreases functional synapse number

(A–C) mEPSCs from P25 GFPNR3A transgenic mice have decreased frequency but unchanged amplitude. (A) Representative traces of mEPSC recordings made at −80 mV in the presence of 0.2 mM TTX and 0.05 mM picrotoxin in neurons from control and GFPNR3A transgenic mice. The middle panel is the average of all mEPSCs from a representative experiment in control and transgenic mice. (B) Averaged data and cumulative probability histograms of mEPSC frequencies demonstrating a lower probability of spontaneous events in neurons from GFPNR3A transgenic mice (* p < 0.02, Student’s t-test). Grey trace indicates control data, black trace represents data from GFPNR3A transgenic experiments. (C) Averaged data and cumulative probability histograms of mEPSC amplitudes. (D) Ratio of NMDAR to AMPAR current does not differ between P25 control and GFPNR3A transgenics. Ratios were calculated from spontaneous synaptic events recorded at +40 mV before and after application of APV to estimate each component (D₁), or from synaptically evoked currents measuring the AMPAR component at −80 mV and the NMDAR component at +40 mV at a time (indicated by dotted line) when the AMPAR component had largely decayed (D₂).

Figure 7. Genetic deletion of NR3A accelerates the developmental onset of LTP and synaptic maturation

(A) LTP evoked by HFS and measured in the CA1 region of the hippocampus is enhanced in P6-8 NR3A knockout mice when compared to control wild-type mice (p < 0.05, Student’s t-test), while the magnitude of LTP at P14–P18 and in adults is similar between genotypes. (B) Immunoblot analyses of NMDAR subunit abundance show that NR1 levels are significantly increased in forebrain synaptic fractions from P6-8 NR3A knockout mice (* p < 0.01 vs wild-type, Student’s t-test, values within bars represent sample sizes). Protein expression data are averages of densitometric values (OD) relative to protein loads. Representative immunoblots are shown. PNS, postnuclear supernatant, SPM, synaptic plasma membrane, PSDI, postsynaptic fraction. (C) Input-output (I–O) curve demonstrating that the amplitude of synaptic NMDAR currents recorded at +40 mV is enhanced in neurons from P6-8 NR3A knockout mice. RMANOVA revealed a significant main effect of genotype (F_{(1, 15)} = 6.65, p < 0.05).

Figure 8. Disrupted long-term storage of spatial memory in GFPNR3A transgenic mice

(A) Diagram showing training protocols used for assessment of spatial memory in the Morris water maze; remote memory tests are indicated by vertical arrows. (B, C) Adult GFPNR3A transgenic mice display normal acquisition of spatial reference memory, as shown by declines in swim distances and times to locate the hidden platform, but their performance is poorer than controls after platform reversal. Remote memory testing was conducted 21 days after the initial 6 day training period (day 27), and then mice were retrained for 3 days to restore performance. Transgenic mice showed longer swim distances during retraining, but by day 30 both control and transgenic mice returned to day 6 levels. After changing the platform location from the NE to the SW quadrant, swim distances and times were longer for both genotypes on day 31 and 32 relative to day 6 or 30. By day 33, controls returned to levels observed on day 30, but GFPNR3A transgenics needed more days of training to achieve these levels. (D) Control and transgenic mice showed no differences in learning to swim to a visible platform. (E) Probe trials for acquisition given at the end of test days 3, 6, and 27. Dashed lines indicate chance levels (25%). Note that both control and transgenic mice learned the hidden platform location, but remote memory was significantly impaired in GFPNR3A transgenics. (F) Probe trials for reversal testing given at the end of test days 33 and 36; remote memory was assessed 21 days later (day 57). Both control and transgenic mice learned the new platform location, as shown by a preference for the SW compared to NE quadrants on day 36 (* p < 0.05 control vs transgenics, Bonferroni corrected pairwise comparisons, n = 10 mice/genotype).

Figure 9. Deficits in long-term memory are reversible

(A, B) Adult GFPNR3A transgenic mice exhibit anomalies in long-term memory in social transmission of food preference and object recognition tests. (A) Transgenic mice demonstrate reduced preferences for the familiar diet at 20 min and 24 hrs, and no preference for either the familiar or novel diet at 10 days relative to controls. Bonferroni post-hoc comparisons found significant differences between control and transgenic mice at 20 min (p < 0.036), 24 hrs (p < 0.006), and 10 days (p < 0.001). Control preference scores did not differ between test times, but transgenics showed a reduction in preference at 10 days compared to 20 min (p < 0.002) and 24 hrs (p < 0.001). (B) Transgenic mice display intact object recognition at 20 min but are impaired at 24 hrs and 10 days. Control preference scores did not differ between test times, whereas mutants showed a graded loss in preference over time. Hence, preferences for control and transgenic mice were similar at 20 min, but differed at 24 hrs (p < 0.016) and 10 days (p < 0.001). (C) Defects in long-term memory are reversible. Transgenic mice were raised off-doxycycline to allow GFPNR3A expression, and treated with doxycycline for 14 days before testing. Treatment was continued during the consolidation phase (see scheme, vertical arrows indicate test times). Note that adult transgenic mice display normal object recognition memory at all times following doxycycline treatment. Moreover, when compared to untreated transgenic mice, doxycycline-treated transgenics showed the same performance at 20 min, but enhanced memory for the novel object at 24 hrs and 10 days (n = 9–10 mice/genotype/treatment; *p < 0.05 from controls, +p < 0.05 from 20-min test, ^p < 0.05 from 24-hr test, #p < 0.05 from water-treated transgenics).