Acetylcholine (ACh) has a crucial role in mediating learning and memory1 2 3 4. A number of ACh receptors (AChRs) have been identified and are classified into two large families, muscarinic AChR (mAChRs) and nicotinic AChRs (nAChRs). Although the mAChRs are G-protein-coupled receptors, the nAChRs form ligand-gated ion channels1. The cholinergic modulation of synaptic plasticity is well described, including in long-term potentiation (LTP), a cellular model of learning and memory1 5 6 7 8. However, it is still unclear whether ACh mediates the learning-induced synaptic changes. Plasticity at excitatory and inhibitory synapses is involved in learning and memory. Experience, such as learning, strengthens excitatory glutamatergic synaptic transmission by driving AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid)-type glutamate receptors (AMPARs) into synapses9 10 11 12 13 14 15 16 17 18 19. The inhibitory avoidance (IA) task, a contextual fear-learning task, increases AMPARs in CA3-CA1 hippocampal pyramidal synapses, which is required for memory formation16. Spatial learning in a water maze increases the frequency but not the amplitude of miniature inhibitory post-synaptic currents (mIPSCs) at hippocampal synapses20. Here we find that mAChR activation mediates the IA learning-induced synaptic delivery of AMPARs to hippocampal CA3-CA1 synapses. IA learning also strengthens inhibitory hippocampal synapses through the activation of nAChRs but not mAChRs. Further, we find significant correlation between the IA-induced increase in miniature excitatory post-synaptic current (mEPSC) and mIPSC amplitudes at individual pyramidal neurons. Thus, ACh balances the excitatory and inhibitory synaptic inputs onto CA1 pyramidal neurons in IA learning through the activation of distinct sets of AChRs. Results Extracellular ACh level in CA1 increases during learning To investigate learning-induced synaptic modification in the hippocampus, we used the IA task (Fig. 1). In this paradigm, rats are allowed to cross from an illuminated box to a dark box where an electric foot shock is delivered. Thus, rats learn to avoid the dark box and stay in the lighted one, which they would normally not prefer16 17. The tendency to avoid the dark box therefore indicates the acquisition of contextual memories. The rats avoided entering the dark box when it was associated with a mild electric shock (IA trained), but not those given foot shock without any contextual experience (unpaired) or those allowed to simply explore the experimental cage (walk through) (Fig. 1). Untrained control rats were kept in their home cages and were not exposed to the IA apparatus. As ACh induces LTP in hippocampal slices21, we hypothesized that ACh release into the hippocampus triggers the delivery of AMPARs in vivo. We therefore examined the extracellular ACh levels in the animals under different learning conditions by in vivo microdialysis of the dorsal CA1 (Fig. 2a). Although significant but transient increases in the extracellular ACh levels were observed in the unpaired (F 9, 72=3.830, P=0.0005, n=9, one-way analysis of variance (ANOVA)) and walk-through animals (F 9, 63=2.137, P=0.039, n=8, one-way ANOVA), the IA-trained rats displayed a prolonged rise in ACh levels (Fig. 2b, F 9, 63=6.694, P ~10 pA was recorded. When recording simultaneously from two cells, the stimulus intensity was increased until both cells showed a response >~10 pA. Synaptic AMPA receptor-mediated responses at –60 and +40 mV were averaged over 50–100 trials, and their ratio (averaged response at −60 mV/+40 mV) was used as an index of rectification. For paired recordings, infected and nearby uninfected cells (~100 μm) were accessed as whole cells, and the synaptic response to a stimulus was recorded from both cells simultaneously. The AMPA/NMDA ratio was calculated as the ratio of the peak current at −60 mV to the current at +40 mV 150 ms after stimulus onset (40–60 traces averaged for each holding potential). For the miniature recordings, the mEPSC (−60 mV holding potential) and mIPSC (0 mV holding potential) were recorded for 5 min in the same CA1 neuron. Bath application of an AMPA receptor blocker (CNQX, 10 μM) or γ-aminobutyric acid-A receptor blocker (bicuculline methiodide, 10 μM) completely blocked the mEPSC (at −60 mV) or mIPSC (at 0 mV) events, respectively. To evaluate the paired-pulse ratio from the EPSC or IPSC average, 30–60 sweeps were recorded with paired stimuli at 100-ms intervals. The EPSC or IPSC amplitudes were measured from the peak of the post-synaptic current to the basal current level immediately before the electrical stimulation. In vivo microdialysis Under sodium pentobarbital anaesthesia (30–50 mg kg−1, i.p.), a stainless steel guide cannula (outer diameter, 0.51 mm) was implanted stereotaxically into the right side of the dorsal hippocampus. After cannula implantation, a stylet was inserted into the guide until the microdialysis was performed. Although the rats were reared and housed socially, after surgery each rat was individually housed in a cylindrical plastic cage (diameter=35 cm, height=45 cm). The experiment was performed in an electromagnetic- and sound-shielded room (length 1.2 m, width 2.2 m and height 2.3 m)43. The stylet was replaced with a microdialysis probe the day before the experiment (outer diameter=0.31 mm, AI-4-0.5, Eicom Co., Kyoto, Japan). A two-channel fluid swivel device (SSU-20; Eicom Co.) was connected to the inlet and outlet of the probe. During the experiment, an artificial cerebrospinal fluid solution (147 mM NaCl, 4 mM KCl, 1.2 mM CaCl2 and 0.9 mM MgCl2) was infused through the dialysis probe with a 0.5-mm-long semipermeable membrane at a rate of 1.2 μl min−1 using a microdialysis pump (CMA/102, Carnegie Medicin, Stockholm, Sweden). The rats were housed individually in a cage, and the dialysis was performed under unanesthetized, freely moving conditions. After an overnight stabilization period, the dialysates were automatically collected into an autoinjector (24 μl; EAS-20, Eicom Co.) every 15 min, and the same volume of ethylhomocholine solution (100 nM) was mixed in as the internal standard. This mixture was injected directly onto an HPLC column every 15 min44. IA learning On the training day, the rats were moved into an electromagnetic- and sound-shielded room (length 1.2 m, width 2.2 m and height 2.3 m) with an IA training apparatus (length 25 cm, width 62 cm and height 45 cm). The apparatus is a two-chambered Perspex box consisting of a lighted safe side and a dark shock side, separated by a trap door. During training, the IA-trained rats were placed in the safe side of the box facing a corner opposite the door. After the trap door was opened, rats could enter the dark box at will. The latency before entering the novel dark box was measured as a behavioural parameter (latency before IA learning). Four seconds after the animals entered the dark side, we closed the door and applied a scrambled electrical foot shock (2 s, 1.6 mA) via electrified steel rods in the floor of the box. The rats were kept in the dark compartment for 10 s before being returned to their home cage. The training procedure for the control groups was as follows: the unpaired control rats (foot shock only) were placed in the box in the dark side and subjected to a scrambled electrical foot shock (2 s, 1.6 mA) without any contextual experience. The walk-through control rats were placed in the IA training apparatus and allowed to explore for 1 min, without shock. The untrained control rats were not removed from their home cages. Thirty minutes after the procedure described above, the rats were placed in the lighted side. The latency before entering the dark box was measured as an indicator of learning performance (latency after IA learning). The rats were then killed with an overdose of pentobarbital. For the untrained control, rats were injected with the same dose of anaesthesia in their home cage. Biochemical analysis of ACh ACh was quantified by combining HPLC, an enzyme reaction, and electrochemical detection (HTEC-500, Eicom Co.). A solution of 0.1 mM Na2HPO4 (pH 8.5) and 200 mg l−1 sodium 1-decanesulfonate (Aldrich Chemical Company, Inc., Milwaukee, WI) was delivered as the HPLC mobile phase at 150 μl min−1. After sample separation in a styrene polymer column (AC-GEL, Eicom Co.), the ACh was converted to hydrogen peroxide by a post-column enzyme reactor (AC-ENZYMPAK, Eicom Co.) containing immobilized acetylcholinesterase and choline oxidase. The hydrogen peroxide was detected with an electrochemical detector, with a minimum detectable amount of 5–10 fmol per sample. To calculate the recovery rate of each dialysis probe, standard samples were also infused through the probe in vitro. The amount of ACh collected every 15 min was divided by the in vitro recovery rate to estimate the extracellular ACh level. Statistics Extracellular ACh levels were analysed by one-way ANOVA with repeated measures followed by post hoc analysis with the Fisher-protected least significant difference test, where the variable was time. To evaluate the difference between groups, we calculated the baseline levels of ACh before the behavioural test, and the response was calculated as the AUC for each rat. The AUC, AMPA/NMDA ratio, IA latency, mEPSC and mIPSC data were analysed by one-way factorial ANOVA where the variable was the treatment group. The ANOVA was followed by post hoc analysis with the Fisher’s protected least significant difference test. The rectification index was analysed by a paired t-test. Cumulative distribution was analysed by the Kolmogorov–Smirnov test. The paired recording data were analysed by the Wilcoxon non-parametric test. The ratio of PPF or PPD (R2/R1) was analysed by unpaired two-tailed t-test. To evaluate the correlation between mEPSC and mIPSC parameters, Spearman’s rank correlation coefficient was calculated. P<0.05 was considered statistically significant. Author contributions D.M. and A.S. performed the experiment. T.T. and D.M. designed the experiment and wrote the manuscript. T.T and D.M. co-managed the project. Additional information How to cite this article: Mitsushima, D. et al. A cholinergic trigger drives learning-induced plasticity at hippocampal synapses. Nat. Commun. 4:2760 doi: 10.1038/ncomms3760 (2013). Supplementary Material Supplementary Information Supplementary Figures S1-S2 and Supplementary Tables S1-S2