Cyclin-dependent kinase 5 governs learning and synaptic plasticity via control of NMDAR degradation

Abstract

Learning is accompanied by modulation of postsynaptic signal transduction pathways in neurons. Although the neuronal protein kinase cyclin-dependent kinase 5 (Cdk5) has been implicated in cognitive disorders, its role in learning has been obscured by the perinatal lethality of constitutive knockout mice. Here we report that conditional knockout of Cdk5 in the adult mouse brain improved performance in spatial learning tasks and enhanced hippocampal long-term potentiation and NMDA receptor (NMDAR)-mediated excitatory postsynaptic currents. Enhanced synaptic plasticity in Cdk5 knockout mice was attributed to reduced NR2B degradation, which caused elevations in total, surface and synaptic NR2B subunit levels and current through NR2B-containing NMDARs. Cdk5 facilitated the degradation of NR2B by directly interacting with both it and its protease, calpain. These findings reveal a previously unknown mechanism by which Cdk5 facilitates calpain-mediated proteolysis of NR2B and may control synaptic plasticity and learning.

Figure 1

Conditional loss of Cdk5 in adult mouse hippocampus. (a) Cdk5 gene targeting strategy. Inset, PCR genotyping of wild-type (+) and floxed (fl) alleles. (b) Radiolabeled in situ with quantification for hippocampus (Hip), striatum (Str), cortex (Ctx), cerebellum (Cer) and hippocampal layers (n = 6–8) of wild-type (WT) and knockout (KO) mice. (c) Quantitative immunoblots of Cdk5, Cre and α-tubulin (Tub) in hippocampal homogenates (n = 20). (d) Cdk5 kinase activity immunoprecipitated from hippocampus. Radiolabeled (³²P-H1) and Coomassie-stained (CBB) H1 histone from WT and KO mice were compared with roscovitine (Ros) and IgG controls (n = 6). (e) Radiolabeled in situ for hippocampal Cdk5 mRNA with pseudocolor quantification (+ = high, − = low) (f) Cdk5 and Cre mRNA fluorescent in situ hybridizations with Nissl counterstains in WT and KO CA1 pyramidal neurons. Arrows show neurons with no Cre mRNA or no Cdk5 loss. Data represent mean ± s.e.m.; *P < 0.05, **P < 0.01, versus WT; Student’s t-test.

Figure 2

Superior performance of conditional Cdk5 KO mice in contextual fear conditioning, extinction and water-maze reversal learning and memory tasks. (a,b) Context- and cue-elicited memory 24 h after training and extinction. Freezing times before conditioning (Pre) and after context (n = 20) or cue (n = 11–12) re-exposure are shown with extinction learning to contextual environment (n = 10). (c,d) Learning and memory in water maze. Escape latencies of training and reversal (four trials per day) are shown with visual acuity assessment (V), and individual trials 1–8 on reversal days 1–2 (left). Spatial preferences and representative paths after initial and reversal training are shown as the percentage of time spent in target quadrant (T, right). L, O and R indicate left, opposite and right quadrants, respectively; n = 10–11. Data represent mean ± s.e.m.; *P < 0.05 post hoc, ***P < 0.01 versus WT, **P < 0.05 versus other quadrants, ^†P < 0.05 versus WT in T; Student’s t-test.

Figure 3

Enhanced synaptic plasticity and NMDAR-mediated currents in the hippocampal SC-CA1 pathway of conditional Cdk5 KO mice. (a) Basal input/output curve and PPF (50–800-ms interstimulus interval); n = 7–9 slices, 5 animals per genotype. (b) LTP after 2× TBS and 3× TBS plotted as percent of baseline (−10 to 0 min). Representative traces at baseline (1,2) and 60 min (3,4) are shown. (c) LTP 57–60 min after 2× TBS, 50 Hz, 3× TBS and 100 Hz represented as percent change from baseline (n = 7–8 slices, 4–5 animals per group). (d) NMDAR-mediated fEPSP (fEPSP NMDA) input/output relationship and NMDA/AMPA ratio. AMPA- and NMDA-mediated fEPSPs were measured in 2 and 0 mM Mg²⁺ and 0 and 20 μM DNQX, respectively (n = 6 slices, 3 animals per genotype). (e) Whole-cell NMDA/AMPA EPSC charge ratio with representative traces. AMPA- and NMDA-mediated EPSCs were measured at −70 mV and +40 mV/20 μM DNQX, respectively (n = 7–10 cells, 5–7 animals per genotype). The τ is shown for AMPA-mediated EPSCs. Data represent mean ± s.e.m.; *P < 0.05 post hoc, **P < 0.05 versus WT; Student’s t-test.

Figure 4

Increased ifenprodil-sensitive NMDAR-mediated current and NR2B levels as a result of reduced calpain activity account for enhanced synaptic plasticity in Cdk5 KO mice. (a) Effect of ifenprodil (ifen) on NMDAR-mediated EPSCs. Representative traces of NMDAR-mediated EPSCs before (I_NMDA) and after (I_{ifen-resistant}) treatment with 10 μM ifenprodil are shown with quantification for I_{ifen-sensitive} and τ for I_NMDA; n = 6–8 cells, 6–7 animals per genotype. (b,c) Hippocampal glutamate-receptor–subunit immunoblots (b, n = 9–12) and NR2B surface/total ratio (c, n = 6 slices, 3 animals per genotype) are shown. (d) Effects of ifenprodil (20 μM) and AP5 (50 μM) on LTP enhancement, with representative baseline (1,2) and 60-min (3,4) traces (n = 6). (e) Calpain-mediated NR2B cleavage in hippocampal slices after NMDA treatment. Cleaved NR2B (cNR2B) was detected with Tub loading control after 30- or 60-min treatment with buffer, 50 μM NMDA or 50 μM NMDA/20 μM calpeptin (Cal; n = 4–6). (f) Calpain-mediated degradation of NR2B in vitro in the presence or absence of Cdk5/p25 (n = 3). (g) Coimmunoprecipitation of NR2B, Cdk5 and calpain-1 from WT hippocampus. Lysates or immunoprecipitates derived using the antibodies indicated (IP) were immunoblotted for the proteins denoted (left). Representative of eight experiments. (h) Cytoplasmic NR2B, Cdk5 and calpain binding assays. Immobilized NR2B C terminus (top) or Cdk5/p25 (bottom) were incubated with Cdk5/p25, NR2B and/or unactivated calpain. Immunoblots show selective binding of Cdk5, NR2B and calpain. Representative of 12 experiments. Data represent mean ± s.e.m.; *P < 0.05, **P < 0.01, ***P < 0.05 post hoc versus WT; Student’s t-test.