Neural cell adhesion molecule-associated polysialic acid regulates synaptic plasticity and learning by restraining the signaling through GluN2B-containing NMDA receptors

Abstract

The neural cell adhesion molecule (NCAM) is the predominant carrier of alpha2,8 polysialic acid (PSA) in the mammalian brain. Abnormalities in PSA and NCAM expression are associated with schizophrenia in humans and cause deficits in hippocampal synaptic plasticity and contextual fear conditioning in mice. Here, we show that PSA inhibits opening of recombinant NMDA receptors composed of GluN1/2B (NR1/NR2B) or GluN1/2A/2B (NR1/NR2A/NR2B) but not of GluN1/2A (NR1/NR2A) subunits. Deficits in NCAM/PSA increase GluN2B-mediated transmission and Ca(2+) transients in the CA1 region of the hippocampus. In line with elevation of GluN2B-mediated transmission, defects in long-term potentiation in the CA1 region and contextual fear memory in NCAM/PSA-deficient mice are abrogated by application of a GluN2B-selective antagonist. Furthermore, treatment with the glutamate scavenger glutamic-pyruvic transaminase, ablation of Ras-GRF1 (a mediator of GluN2B signaling to p38 MAPK), or direct inhibition of hyperactive p38 MAPK can restore impaired synaptic plasticity in brain slices lacking PSA/NCAM. Thus, PSA carried by NCAM regulates plasticity and learning by inhibition of the GluN2B-Ras-GRF1-p38 MAPK signaling pathway. These findings implicate carbohydrates carried by adhesion molecules in modulating NMDA receptor signaling in the brain and demonstrate reversibility of cognitive deficits associated with ablation of a schizophrenia-related adhesion molecule.

Figure 1.

PSA inhibits GluN1/2B and GluN1/2A/2B receptors. A–C, Sample traces showing currents from GluN1/2B (A), GluN1/2A (B), and GluN1/2A/2B (C) receptors reconstituted in lipid bilayers and activated by 3 μm glutamate in the absence or presence of PSA and NMDA receptor antagonists. Channel activity is evident by upward transitions of the current, which represent the open state. Respective amplitude histograms, with peaks corresponding to the closed and the major open state are shown to the right of each sample trace. The traces illustrate the progressive reduction in the open probability (Po) of single GluN1/2B (A) and GluN1/2A/2B (C) channels in the presence of 1, 2, and 3 μg/ml PSA, respectively. Activities of GluN1/2B, GluN1/2A and GluN1/2A/2B channels were blocked by Ro 25-6981 (A), Zn²⁺ (B), and APV (C), respectively. D, Absolute values of mean + SEM of the open probability and chord conductance of GluN1/2B, GluN1/2A and GluN1/2A/2B channels in the absence of PSA. One-way ANOVA revealed a difference between three types of channels in terms of open probability (p < 0.001) and conductance (p < 0.001). The probability is higher for GluN1/2B channels compared with GluN1/2A and GluN1/2A/2B (p < 0.001), the conductances are distinct for all channels (p < 0.001; post hoc Bonferroni t test). E, Normalized values of mean + SEM of the open probability and chord conductance of GluN1/2B, GluN1/2A and GluN1/2A/2B channels as a function of PSA concentration. The values of parameters in the absence of PSA were set to 100%. Two-way repeated-measures ANOVA applied either to all three or to GluN1/2B and GluN1/2A/2B receptors revealed the effect on open probability (not on the chord conductance) of PSA (p < 0.01) and interaction between PSA and receptor composition (p < 0.05). ***p < 0.001, post hoc Bonferroni t test, significant effects of PSA on GluN1/2B and GluN1/2A/2B receptors compared with control.

Figure 2.

Increased responses mediated by GluN2B in hippocampal slices of NCAM^−/− mice. A, AMPA receptor-mediated fast component of fEPSP, the amplitude of which was taken as 100% [1], was blocked by perfusion with ACSF containing 0.25 Mg²⁺, the AMPA/kainate receptor antagonist NBQX and the GABA_A/glycine receptor antagonist picrotoxin (PiTX), leaving the GluN-mediated component [2]. Application of 0.25 μm NVP isolated fEPSPs mediated predominantly by GluN2B [3], which could be blocked to 1–1.5% level by the NMDA receptor antagonist APV (50 μm) [4]. To present changes in high-amplitude AMPA receptor-mediated fEPSPs and low amplitude NMDA receptor-mediated fEPSPs in one graph, a logarithmic Y-scale is used. The inset shows representative traces corresponding to time intervals color-coded in A. B, The NMDA receptor-mediated component of fEPSP [1] is more strongly reduced by 5 μm Ro 25-6981 [2] in NCAM^−/− mice than in NCAM^+/+ mice. The inset shows representative traces corresponding to time intervals color-coded in B. C, The NMDA receptor-mediated component of fEPSP [1] is more strongly reduced by NVP in NCAM^+/+ mice [2], whereas 10 μm PSA more strongly inhibits the residual component in NCAM^−/− mice [3]. The inset shows representative traces corresponding to time intervals color-coded in C. B, C, Note that data show only NMDA receptor-mediated component of fEPSP after 60 min in the presence of NBQX, PiTX and 0.25 mm Mg²⁺ and the amplitude of fEPSPs recorded 50–60 min in the presence of NBQX, PiTX and 0.25 mm Mg²⁺ was taken as 100%. The components left after APV application were subtracted for both genotypes. Bars represent means + SEMs of the amplitudes. Empty bars point to 10 min time intervals in which significant difference between genotypes was found, as indicated in the text. Numbers of recorded slices and mice, respectively, are indicated in parentheses.

Figure 3.

Acute removal of PSA upregulates GluN2B-mediated Ca²⁺ transients in CA1 pyramidal cells. A, Left: CA1 pyramidal neuron filled with 250 μm Fluo4. The sample image was obtained with high laser power to reveal the morphology of the cell. The dashed rectangle indicates the area of recording which is shown enlarged in the right image. The black line shows the line scan position. The circle indicates the position of the uncaged spot. B, A line-scan image obtained from the selected part of the dendrite. The white vertical line indicates the time of glutamate uncaging. Upper and bottom show the Ca²⁺ signal in control conditions (Control), after active endoNF application (endoNF), PSA application (PSA) and application of the GluN2B-specific antagonist Co101244. C, Representative traces from two neurons. Left: three sweeps showing Ca²⁺ transients in response to glutamate uncaging in control conditions, its increase by endoNF and reversal by PSA, corresponding to line-scan images in B. Right: Ca²⁺ transients recorded in another slice in the control condition and after inactive endoNF (endoNF−) application. D, Mean + SEM of measurements obtained in the experiments with application of active and inactive forms of endoNF (left and right panels, respectively). The active form of endoNF increased the integral of the Ca²⁺ signal more than twice. Two-way ANOVA with repeated measures revealed the effect of treatments on the integral of Ca²⁺ transients (p < 0.001). Comparisons with controls revealed the effects of endoNF and Co101244: *p < 0.05, **p < 0.01, ***p < 0.001; ⁺p < 0.05, significant difference between endoNF- and PSA-treated groups, post hoc Bonferroni t test; numbers of recorded cells are indicated in parentheses.

Figure 4.

Restoration of impaired LTP in NCAM^−/− mice through inhibition of GluN2B-containing receptors. Restoration of impaired LTP in NCAM^−/− mice (A, D) by application of the GluN2B-selective antagonist Ro 25-6981 (B), or the glutamate scavenger GPT (E). The mean slope of fEPSPs recorded 10 min before TBS was taken as 100% and arrows indicate delivery of TBS. Data represent mean + SEM, numbers of tested slices and mice, respectively, are indicated in parentheses. Insets show averages of 30 fEPSPs recorded 10 min before and 50–60 min after TBS. C, F, Mean + SEM of LTP levels recorded 50–60 min after TBS. Two-way ANOVA revealed the effect of NCAM (p < 0.001), and interaction between NCAM and Ro 25-6981 treatment (p < 0.05) in C and the effect of NCAM (p < 0.001), GPT treatment (p < 0.05) and interaction between NCAM and GPT (p < 0.05) in F. ***p < 0.001, significant difference between NCAM^+/+ and NCAM^−/− mice; ⁺p < 0.05, ⁺⁺p < 0.01, significant difference between untreated control and pharmacologically treated mice of the same genotype, post hoc Bonferroni t test.

Figure 5.

Restoration of impaired LTP in endoNF-treated hippocampal slices via inhibition of GluN2B-containing receptors. A, Normal levels of LTP in slices treated with the inactive form of endoNF (endoNF−) compared with sham controls. In contrast, LTP is impaired in slices treated with the active form of endoNF. LTP in endoNF-treated slices is restored to normal levels by application of the GluN2B-selective antagonist Ro 25-6981. B, Impaired LTP in endoNF-treated slices in the presence of GABA_A and GABA_B receptor antagonists bicuculline and CGP 55845. C, Mean + SEM of LTP levels recorded 50–60 min after TBS. One-way ANOVA revealed a difference (p < 0.001) between sham, inactive endoNF− and active endoNF (with and without Ro 25-6981)-treated groups. **p < 0.01, significant difference between slices treated with active and inactive endoNF; ⁺p < 0.05, significant effect of Ro 25-6981 on endoNF-treated slices, post hoc Bonferroni t test. Numbers of tested slices and mice, respectively, are indicated in parentheses in A, B.

Figure 6.

Elevated signaling via p38 MAPK in NCAM^−/− mice and endoNF-treated slices. A, Western blots showing levels of phosphorylated (phospho) and total p38 and p42/p44 MAPK proteins in the hippocampi of NCAM^+/+ versus NCAM^−/− mice, and in active (+) versus inactive (−) endoNF-treated slices. Blots were stripped and reblotted with GAPDH antibody as a loading control. The intensity of GAPDH signal was used for normalization of corresponding MAPK signals. B, Mean + SEM of the levels of phosphorylated p38, p42 and p44 MAPKs in NCAM^+/+ versus NCAM^−/− hippocampi. Band intensities for NCAM^−/− and NCAM^+/+ probes from the same blot were normalized using an average intensity of NCAM^+/+ probes. Two-way ANOVA with repeated measures revealed the effects of NCAM (p < 0.01), MAPK identity (p < 0.001), and interaction between NCAM and MAPK (p < 0.001). ***p < 0.001, **p < 0.01, significant difference between genotypes, post hoc Bonferroni t test. C, Mean + SEM of the levels of phosphorylated p38, p42 and p44 in inactive (endoNF−) and active endoNF-treated hippocampal slices. Each band intensity corresponding to active endoNF-treated slices derived from one mouse was normalized using the band intensity corresponding to inactive endoNF-treated slices from the same mouse (set to 100%), with both bands being from the same blot. Two-way ANOVA with repeated measures revealed the effect of MAPK identity (p < 0.01) and interaction between MAPK and endoNF treatment (p < 0.01). *p < 0.05, significant difference between slices treated with active and inactive endoNF, post hoc Bonferroni t test.

Figure 7.

Restoration of impaired LTP in NCAM^−/− and/or EndoNF-treated hippocampal slices via inhibition of p38 or ablation of Ras-GRF1. Application of the p38 inhibitor, SB 203580 (B), restores impaired LTP in slices from NCAM^−/− mice (A) to the levels seen in NCAM^+/+ mice. E, SB 203580 also restores impaired LTP in endoNF-treated slices (D) to the levels seen in sham controls. F, EndoNF does not affect LTP in Ras-GRF1^−/− mice. The mean slope of fEPSPs recorded 10 min before TBS was taken as 100% and arrows indicate delivery of TBS. Data represent mean + SEM, numbers of tested slices and mice, respectively, are indicated in parentheses. Insets show averages of 30 fEPSPs recorded 10 min before and 50–60 min after TBS. C, G, Mean + SEM of LTP levels recorded 50–60 min after TBS. Two-way ANOVA revealed the effect of NCAM (p < 0.01) and interaction between NCAM and SB 203580 treatment (p < 0.05) in C and the effect of endoNF (p < 0.001) and interactions between endoNF and SB 203580 treatment (p < 0.05) and endoNF and Ras-GRF1 (p < 0.01) in G. ***p < 0.001, **p < 0.01, significant differences between slices from NCAM^+/+ versus NCAM^−/− mice or between sham versus endoNF-treated slices, ⁺⁺p < 0.01, ⁺p < 0.05, significant effects of SB 203580 or Ras-GRF1 ablation, post hoc Bonferroni t test.

Figure 8.

Intrahippocampal injection of Ro 25-6981 (Ro25) restores contextual memory in NCAM^−/− mice. A, Scheme of experiments with intrahippocampal injection of Ro25. H & H, Habituation and handling of mice before the training session; TR, training; T, test session. Injection of either drug or vehicle (Veh) was done 15 min before the beginning of the training session. Time is given in days (−d7 and −d2, 7 and 2 d before training; d1 and d7, days 1 and 7 after training); CC+ and CC−, conditioned and neutral contexts; CS+ and CS−, conditioned and neutral stimuli. B, Mean + SEM of freezing and discrimination levels before and after fear conditioning combined with intrahippocampal injection of Ro25. Contextual memory test (on the left): Levels of freezing before training (B, baseline) and on the two posttraining days, d1 and d7, in the conditioned context (filled bars) and in the control neutral context (empty bars). Tone memory test (on the right): Levels of freezing during presentation of the conditioned tone (filled bars) and the control tone (empty bars) at the two posttraining days, d1 and d7. Discrimination of CC+ versus CC− and CS+ versus CS− was computed as the difference between CC+ and CC− or CS+ and CS− freezing responses, respectively. Four-way ANOVA of freezing time at different contexts revealed the effects of genotype (p < 0.001), Ro25 (p < 0.01), day (p = 0.001), context (p < 0.001) and interactions between context and Ro25 (p < 0.01); context, genotype and Ro25 (p < 0.05); and day and context (p < 0.001). ANOVA of context discrimination revealed the effects of Ro25 (p < 0.01), day (p < 0.001) and interaction between genotype and Ro25 (p < 0.05). ANOVA of freezing time to different tones revealed the effects of genotype (p < 0.001), day (p < 0.001), tone (p < 0.001) and interactions between tone and genotype (p = 0.001); tone, genotype and Ro25 (p < 0.01); and day and tone (p < 0.01). ANOVA of tone discrimination revealed the effects of genotype (p = 0.001), day (p < 0.01) and interaction between genotype and Ro25 (p < 0.01). *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.001, significant differences between NCAM^+/+ and NCAM^−/− untreated mice; ⁺p < 0.05, ⁺⁺p < 0.01, ⁺⁺⁺p < 0.005, ⁺⁺⁺⁺p < 0.001, significant differences between vehicle and Ro25-treated NCAM^−/− mice, Bonferroni post hoc t test; ^#p < 0.05, significant differences between freezing on CC+ and CC− or CS+ and CS−, Wilcoxon test. Numbers of tested mice are indicated in parentheses.