MicroRNA loss enhances learning and memory in mice

Abstract

Dicer-dependent noncoding RNAs, including microRNAs (miRNAs), play an important role in a modulation of translation of mRNA transcripts necessary for differentiation in many cell types. In vivo experiments using cell type-specific Dicer1 gene inactivation in neurons showed its essential role for neuronal development and survival. However, little is known about the consequences of a loss of miRNAs in adult, fully differentiated neurons. To address this question, we used an inducible variant of the Cre recombinase (tamoxifen-inducible CreERT2) under control of Camk2a gene regulatory elements. After induction of Dicer1 gene deletion in adult mouse forebrain, we observed a progressive loss of a whole set of brain-specific miRNAs. Animals were tested in a battery of both aversively and appetitively motivated cognitive tasks, such as Morris water maze, IntelliCage system, or trace fear conditioning. Compatible with rather long half-life of miRNAs in hippocampal neurons, we observed an enhancement of memory strength of mutant mice 12 weeks after the Dicer1 gene mutation, before the onset of neurodegenerative process. In acute brain slices, immediately after high-frequency stimulation of the Schaffer collaterals, the efficacy at CA3-to-CA1 synapses was higher in mutant than in control mice, whereas long-term potentiation was comparable between genotypes. This phenotype was reflected at the subcellular and molecular level by the elongated filopodia-like shaped dendritic spines and an increased translation of synaptic plasticity-related proteins, such as BDNF and MMP-9 in mutant animals. The presented work shows miRNAs as key players in the learning and memory process of mammals.

Figure 1.

Loss of Dicer1 gene and miRNAs in Dicer1^CaMKCreERT2 mutant mice. A, Levels of Dicer mRNA and protein in the hippocampus of mutant mice 9 weeks after induction of the Dicer1 gene mutation with tamoxifen (n = 4, *p < 0.05). B, Fluorescent in situ hybridization for Dicer mRNA in CA1 pyramidal cell layer (top) and dendritic layer of CA1–stratum radiatum (bottom) 9 weeks after tamoxifen treatment. C, Relative levels of selected mature miRNAs by qRT-PCR (TaqMan MicroRNA Array version 2.0) 9 weeks after induction of the Dicer1 gene mutation. D, In situ hybridization for Dicer mRNA and miR-124 with locked nucleic acid probes done 9 and 14 weeks after induction of mutation. A chart shows measured (4 and 9 weeks after tamoxifen) and postulated decrease in miRNA levels (as an example, miR-124 is shown) after induction on the Dicer1 gene mutation. Data are shown as mean ± SEM. GAPDH, Glyceraldehyde-3-phosphate dehydrogenase.

Figure 2.

Neuronal death in the hippocampus and cortex in Dicer1^CaMKCreERT2 mutant mice. A, Detection of neuronal cell death (TUNEL staining; bottom row) in the CA1 area of the hippocampus of control and mutant mice 2, 14, and 20 weeks after induction of Dicer1 gene mutation. Arrowheads indicate dying cells (in green). Morphology of nuclei is shown by DAPI staining (top row). CA1, Pyramidal cell layer; Or, stratum oriens; Rad, stratum radiatum. B, Quantification of cell number in CA1 region of the hippocampus and the cortex. Number of cell nuclei in defined area were counted and expressed as fold change to control (n = 3–8, *p < 0.05). Data are shown as mean ± SEM.

Figure 3.

Normal motor activity, exploration, and anxiety in new environment in Dicer1^CaMKCreERT2 mutant mice. A, Time to fall off rod rotating with increasing speed. The rotarod test revealed similar motor coordination and balance of mutant mice and their control littermates (F_(1,22) = 0.482; p = 0.495). B, Number of squares crossed in the open-field test. Five minutes of activity did not differ between genotypes because both groups moved comparable distance (data not shown) and crossed similar number of squares (F_(1,22) = 0.534; p = 0.473). C, Relative time (percentage) spent in zones of the elevated plus maze. Mutant and control mice explored the open arms of apparatus for most of the time (F_(1,22) = 0.307; p = 0.585). Time in center zone (F_(1,22) = 2.007; p = 0.171) and in closed arms (F_(1,22) = 0.022; p = 0.883) showed no difference in anxiety level. D, During adaptation to sweetened water in IntelliCage, mice of both genotypes spent similar time drinking water (F_(1,21) = 0.073; p = 0.789). E, Basal home cage circadian activity measured as an average number of visits from 7 d in each hour showed normal higher activity in active phase and lower activity in inactive phase in mice of both genotypes (F_(1,21) = 0.632; p = 0.436). Data are shown as mean ± SEM.

Figure 4.

miRNAs loss in the forebrain of adult mice enhances learning and memory. A, Spatial learning task in the Morris water maze 5 and 12 weeks after the Dicer1 gene disruption. Relative time spent in the target quadrant (TQ) containing a hidden platform and an average of other quadrants is shown. Memory test, 24 h after the last training day, showed an enhanced memory for platform location in mutant compared with control mice (F_(1,22) = 8.355; **p < 0.01) only in tests performed 12 weeks after induction of mutation. Randomness level (25%) is indicated by dotted line. B, Place learning with appetitive reinforcement in IntelliCage. Ratio (percentage) of visits in correct corner to visits in all other corners in the IntelliCage. Before the task, the least preferred corner was chosen for every individual mouse (24 h before, F_(1,20) = 1.003; p = 0.328). In the subsequent days, mutant mice developed much higher preference to correct corner (days 1–4, F_(1,20) = 4.925; *p < 0.05; **p < 0.01). C, Trace fear conditioning memory test. At 48 h after training, mutant mice displayed a higher freezing response compared with control animals (F_(1,21) = 4.775; *p < 0.05; ***p < 0.001) in new context; trace, 1 min interval including 10 s CS (tone) and freezing was measured for the next 50 s corresponding to a trace and US (shock) in training; ITI, 3 min intertrial interval between CS presentations. D, Context memory test. The level of freezing (percentage) was assessed in re-exposition to the training context 5 d after training. Mutant mice showed no decline in freezing behavior across 6 min of exposure to the same context (F_(1,21) = 7.674; *p < 0.05). Data are shown as mean ± SEM.

Figure 5.

Synaptic transmission and plasticity in hippocampal slices. A, Left, Representative fEPSPs recorded in area CA1 of both genotypes when stimulating Schaffer collaterals with increasing stimulation intensities. Right, Input/output relationships 12–15 weeks after induction of the Dicer mutation (each n = 3). B, At 5–6 weeks after induction of mutation, tetanic stimulation [i.e., high-frequency stimulation (HFS)] induced PTP (control, 1.88 ± 0.08, n = 8 slices from 4 mice; mutant, 1.94 ± 0.11, n = 8 slices from 4 mice; p = 0.59) and LTP (control, 1.29 ± 0.05, n = 8 slices from 4 mice; mutant, 1.34 ± 0.08, n = 8 slices from 4 mice; p = 0.62). Bottom, Expanded time course. C, At 12–15 weeks after induction of mutation, PTP was higher in the mutant than in control (PTP: control, 1.76 ± 0.10, n = 8 slices from 3 mice; mutant, 2.39 ± 0.14, n = 8 slices from 3 mice; p < 0.01; LTP: control, 1.34 ± 0.05, n = 8 slices from 3 mice; mutant, 1.39 ± 0.08, n = 8 slices from 3 mice; p = 0.6). D, At 12–15 weeks after induction of mutation, repeated HFS induced higher PTP each time after HFS, but LTP remained comparable in both genotypes (control, 1.84 ± 0.1, n = 9 slices from 3 mice; mutant, 1.98 ± 0.13, n = slices from 3 mice; p = 0.41). In both genotypes, four times HFS induced higher LTP than single HFS (p < 0.05). The traces show the mean of 45 and 30 consecutive fEPSPs before and 40–50 min after single HFS in both genotypes, respectively. TAM, Tamoxifen.

Figure 6.

Altered morphology of dendritic spines and the expression level of synaptic proteins in mutant animals. A, Three-dimensional reconstruction of dendrites from control and mutant animals 12 weeks after induction of the Dicer1 gene mutation. A visualization of synaptic connections by the presynaptic marker Bassoon (in blue). B, An increase in the length of dendritic spines in CA1 hippocampal neurons of Dicer1^CaMKCreERT2 mutant mice compared with controls (control, n = 5; mutant, n = 6; average total length of analyzed dendrites per animal was 140 μm). C, Representative pictures showing morphology of control and mutant CA1 area neuron (filled with biocytin) 15 weeks after induction of the Dicer1 gene mutation. D, Sholl analysis of dendritic tree of CA1 neurons performed 15 weeks after induction. For analysis, all dendrites of individual neurons were traced, and statistical analysis was performed by repeated-measures ANOVA. There is no significant difference in dendritic tree arborization between control and mutant mice (p = 0.4; F_(1,172) = 0.78). An interaction between the genotype and the distance from the soma (dependent variable categorical predictor, p < 0.01; F_(1,172) = 1.53) indicates a shift in the neuronal arborization toward the more distal dendrites in mutants.

Figure 7.

Expression level of synaptic proteins in mutant animals. A, C, Relative protein levels of synaptic plasticity-related genes: proBDNF, BDNF, PSD95, GluR1, and GluR2 (control, n = 8; mutant, n = 5; *p < 0.05, ***p < 0.001). All values are presented in relation to the neuronal marker NeuN. B, Gel zymography for MMP-9 showing a level of the enzyme activity; Western blot showing representative samples for β-DG (MMP-9 substrate) cleavage. Quantification of a product of cleavage β-DG₃₀ (control, n = 8; mutant, n = 5; *** p < 0.001). D, A table shows potential interacting pairs of miRNA–mRNA in synaptic plasticity-related genes. Colors represent number of predicted miRNA binding sites in target mRNA. Data are shown as mean ± SEM.