Kv1.2 mediates heterosynaptic modulation of direct cortical synaptic inputs in CA3 pyramidal cells

Abstract

A short high frequency stimulation of mossy fibres (MFs) induces long-term potentiation (LTP) of direct cortical or perforant path (PP) synaptic inputs in hippocampal CA3 pyramidal cells (CA3-PCs). However, the cellular mechanism underlying this heterosynaptic modulation remains elusive. Previously, we reported that repetitive somatic firing at 10 Hz downregulates Kv1.2 in the CA3-PCs. Here, we show that MF inputs induce similar somatic firing and downregulation of Kv1.2 in the CA3-PCs. The effect of Kv1.2 downregulation was specific to PP synaptic inputs that arrive at distal apical dendrites. We found that the somatodendritic expression of Kv1.2 is polarized to distal apical dendrites. Compartmental simulations based on this finding suggested that passive normalization of synaptic inputs and polarized distributions of dendritic ionic channels may facilitate the activation of dendritic Na⁺ channels preferentially at distal apical dendrites. Indeed, partial block of dendritic Na⁺ channels using 10 nm tetrodotoxin brought back the enhanced PP-evoked excitatory postsynaptic potentials (PP-EPSPs) to the baseline level. These results indicate that activity-dependent downregulation of Kv1.2 in CA3-PCs mediates MF-induced heterosynaptic LTP of PP-EPSPs by facilitating activation of Na⁺ channels at distal apical dendrites.

Key points

• We investigated the cellular mechanisms underlying mossy fibre-induced heterosynaptic long-term potentiation of perforant path (PP) inputs to CA3 pyramidal cells.

• Here we show that this heterosynaptic potentiation is mediated by downregulation of Kv1.2 channels.

• The downregulation of Kv1.2 preferentially enhanced PP-evoked EPSPs which occur at distal apical dendrites.

• Such enhancement of PP-EPSPs required activation of dendritic Na⁺ channels, and its threshold was lowered by downregulation of Kv1.2.

• Our results may provide new insights into the long-standing question of how mossy fibre inputs constrain the CA3 network to sparsely represent direct cortical inputs.

Introduction

Direct and indirect synaptic inputs from the entorhinal cortex converge on apical dendrites of hippocampal CA3 pyramidal cells (CA3-PCs) via perforant path (PP) and mossy fibres (MFs), respectively. The MF, the axon fibre of dentate granule cells (GCs), sparsely innervates CA3-PCs, and forms a large non-Hebbian ‘conditional detonator’ synapse close to the soma of a CA3-PC (Bischofberger et al. 2006). It has been postulated that these properties of MFs together with the sparse firing of dentate GCs may help the CA3 network reduce the overlap between similar memory representations, and thus subserve pattern separation (O'Reilly & McClelland, 1994; Rolls, 2013). In vivo recordings of the hippocampal CA3 region revealed that spatial selectivity of a place cell firing and retrieval of stored memory are primarily governed by direct cortical inputs rather than by MF inputs, implying that PP-CA3 synapses are the primary sites for encoding and storage of cortical inputs in the CA3 network (McNaughton et al. 1989; Lee & Kesner, 2004). Previous computational models proposed that strong MF inputs facilitate the storage of new memories at Hebbian PP–CA3 synapses on the Hebb–Marr hetero-association network (McNaughton & Morris, 1987; Treves & Rolls, 1992). Furthermore, formal analysis predicted that the trade-off between pattern completion and pattern separation can be minimized when MF inputs participate in initial storage of direct cortical inputs but not in pattern completion-based retrieval (O'Reilly & McClelland, 1994). Therefore, understanding the cellular mechanism that underlies the heterosynaptic modulation of PP inputs by MF inputs may provide new insights into the long-standing question how MF inputs constrain the CA3 network to sparsely represent direct cortical inputs. It was shown that long-term potentiation (LTP) of field excitatory postsynaptic potentials at PP–CA3 synapses (PP-fEPSPs) is associatively facilitated by simultaneous high-frequency stimulation (HFS) of MFs (McMahon & Barrionuevo, 2002). A subsequent study showed, however, that a short tetanic stimulation of MF (100 Hz for 1 s) alone can induce heterosynaptic LTP of PP-fEPSPs (Tsukamoto et al. 2003), suggesting that heterosynaptic interactions between MF and PP inputs do not require the arrival of two inputs within a narrow time window. However, the cellular mechanisms underlying this type of heterosynaptic potentiation remain elusive.

Previously, we reported that the Kv1.2-mediated D-type K⁺ current activity (I_D) is significantly reduced by a short tetanic stimulation of the soma eliciting 20 action potentials (APs) at 10 Hz in CA3-PCs (Hyun et al. 2013). This activity-dependent downregulation of Kv1.2 lowered the input conductance (G_in) and the AP onset time, which we referred to as long-term potentiation of intrinsic excitability (LTP-IE). Here, we demonstrate that the activity-dependent downregulation of Kv1.2 underlies the MF-induced heterosynaptic LTP of direct cortical inputs (Tsukamoto et al. 2003). Furthermore, we show that LTP-IE results in preferential enhancement of PP-evoked EPSPs, whereas it little affects MF or associational/commissural (A/C) synaptic inputs. We present a plausible mechanism whereby the downregulation of Kv1.2 specifically enhances synaptic events that occur at the distal region of apical dendrites in a CA3-PC.

Methods

Animals and ethical approval

All animal studies and experimental protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at Seoul National University. The animals were maintained in standard environmental conditions (25 ± 2 °C; 12/12 h dark/light cycle) and were housed under veterinary supervision at the Institute for Experimental Animals, Seoul National University College of Medicine.

Slice preparation

Hippocampal slices were prepared from Sprague–Dawley rats (postnatal day (P)13 to P21) or C3HeB/FeJ Kcna2 mice (P13–P24) of either sex as described previously (Hyun et al. 2013). After rats or mice were anaesthetized by inhalation of 5% isoflurane, they were decapitated and the brain quickly removed and chilled in an ice-cold high-magnesium cutting solution containing the following (in mm): 116 NaCl, 26 NaHCO₃, 3.2 KCl, 0.5 CaCl₂, 7 MgCl₂, 1.25 NaH₂PO₄, 10 glucose, 2 sodium pyruvate, 3 ascorbate, with pH adjusted to 7.4 by saturating with carbogen (95% O₂, 5% CO₂), and with osmolarity of approximately 300 mosmol l⁻¹. The isolated brain was glued onto the stage of a vibrating blade microtome (Leica VT1200) and 300 μm-thick transverse hippocampal slices were cut. The slices were incubated at 34 °C for 30 min in the same solution, and thereafter maintained at room temperature. For experiments, we transferred the slice that recovered for at least an hour to a recording chamber superfused with artificial cerebrospinal fluid (ACSF) containing the following (in mm): 124 NaCl, 26 NaHCO₃, 3.2 KCl, 2.5 CaCl₂, 1.3 MgCl₂, 1.25 NaH₂PO₄, 10 glucose, bubbled with 95% O₂ and 5% CO₂.

Electrophysiological recordings

Whole-cell voltage- or current-clamp recordings from hippocampal CA3-PCs (one cell per slice) were carried out at 32 ± 2 °C while the recording chamber was perfused with ACSF at 1–1.5 ml min⁻¹. The recordings were made using a MultiClamp 700B amplifier controlled by Clampex 10.2 via Digidata 1440A data acquisition system (Molecular Devices, Sunnyvale, CA, USA). The pipette solution contained (in mm): 130 potassium gluconate, 7 KCl, 2 NaCl, 1 MgCl₂, 0.1 EGTA, 2 ATP-Mg, 0.3 Na-GTP, 10 Hepes (pH 7.30 with KOH, 295 mosmol l⁻¹ with sucrose). After forming a whole-cell patch on the soma of a CA3-PC, the membrane potential was maintained at −66.9 ± 1.3 mV under current-clamp conditions. Under this condition, we monitored input conductance (G_in) every 10 s before and after high frequency minimal stimulation of mossy fibres (MF conditioning) (Jonas et al. 1993). G_in was estimated from subthreshold voltage responses to −30 pA and +10 pA current steps (duration, 500 ms). Statistical values and significance of the conditioning effects on G_in was determined 40 min after MF conditioning. All recordings were performed in the presence a GABA_A receptor antagonist, picrotoxin (PTX, 100 μm) (except Fig.4B). In Figs 3I and 5A, synaptic blockers (PTX and CNQX) and inward current blockers were used to isolate the outward K⁺ current, the inward current blockers in a cocktail that consisted of 200 μm Ni²⁺, 125 μm Cd²⁺ and 0.5 μm tetrodotoxin (TTX) in order to block voltage-dependent Ca²⁺ and fast Na⁺ channels. The membrane potentials were not corrected for the liquid junction potential, which is predicted to be 12 mV. For local puff application of 100 μm 4-amionpyridine (4-AP), we used a pressure ejection system composed of a computer controlled pneumatic pump (Toohey Spritzer, Fairfield, NJ, USA) connected to the back of a puffing glass pipette (tip size: ∼3–4 μm). The puff area, which is defined as the area where the drug concentration is higher than 20% of the maximum, was determined from the fluorescence profile of Alexa Fluor 488 (Invitrogen), which was added to the puffing pipette (100 μm). The ejection pressures was adjusted (typically to ∼1–3 p.s.i.) such that the puff area became a circular region whose diameter was 140 μm from the centre (the puffing pipette tip) at 4 s after the start of ejection. Outward K⁺ currents were recorded between 3 and 5 s after the start of drug ejection. Puffing pipettes for distal and proximal dendritic regions were placed at 230 μm and ∼70–100 μm from the soma, respectively. Under these puff settings, a distal apical dendritic region of a CA3-PC spanning 160–300 μm from the soma was perfused by the distal puff, while the proximal dendritic region closer than 140 μm from the soma was perfused by the proximal puff.

Figure 4. Downregulation of Kv1.2 mediates MF input-induced non-Hebbian heterosynaptic LTP of PP–CA3 EPSPs.

Aa, left, experimental set-up. The incision was made along the hippocampal sulcus of the slice to prevent possible contamination of indirect cortical input. Stimulating electrodes were placed at the stratum lucidum and in close proximity to the hippocampal sulcus to activate mossy fibres (MF stim, red) and perforant path (PP stim, purple) inputs, respectively. Right, schematic diagram of synaptic inputs to apical dendrites of a CA3-PC. Inset: representative somatic recordings of EPSC evoked by stimulation of MF (grey) or PP (black trace) in the same CA3-PC. b, burst stimulation of MF induced LTP of PP–CA3 EPSPs. Stimulation of afferent MFs at 20 Hz for 2 s at t = 0 (blue arrow; MF conditioning) resulted in postsynaptic AP firing (blue trace in inset, denoted as ‘3’). Upper right two insets show representative PP-EPSP and PP-EPSC traces recorded at the time points indicated by the numbers (black, control; red, after MF conditioning; grey, DCG-IV). PP-EPSPs were attenuated by bath-applied DCG-IV to ca 30% of the baseline. Ac, summary of mean amplitudes of PP-EPSCs and PP-EPSPs before and 40 min after MF conditioning. B, MF conditioning-induced potentiation of PP-EPSPs occurs in PTX-free solution too, and was abolished by nimodipine (purple circles, 10 μm in the bath) and PP2 (green circles, 10 μm in the bath). C, the MF conditioning induced LTP of PP-EPSPs in CA3-PCs of WT mice, but not in those of Kcna2 HT mice. D, the MF conditioning potentiated PP-EPSPs in the presence of CPP (yellow circles, 20 μm in the bath) or MK-801 (light blue circles, 40 μm in the bath), but not in the presence of APV (light red circles, 40 μm in the bath). Bar graphs in B–D summarize the effects of the drugs or mouse genotypes on MF conditioning-induced relative changes in PP-EPSP amplitudes at 40 min.

Figure 3. High frequency minimal stimulation of MFs induces long-term reduction of G_in in CA3-PCs through Ca²⁺-dependent downregulation of Kv1.2.

A, left, MF-EPSCs evoked at different stimulus intensities. Each EPSC trace represents the average of 10 responses of non-failure trials (except for 6 V) at the stimulus strength indicated. Middle, schematic diagram of MF synaptic inputs to the proximal region of apical dendrites in a CA3-PC (upper). Mean amplitudes of MF-EPSCs as a function of stimulus intensity (bottom). Note the abrupt increase in the peak EPSC amplitudes at 16 V. Right, monitoring of MF-EPSCs evoked by 18 V every 10 s. Traces in the upper panel show trial-to-trial changes of MF-EPSCs (superimposed grey traces) of non-failure trials (black, averaged trace). B, the number of spikes in postsynaptic CA3-PCs evoked by stimulation of afferent MFs (for 1 s or 2 s) as a function of stimulation frequency. Insets: somatic voltage recordings during MF stimulation (grey, stimulation for 1 s; black, 2 s). C, reduction in input conductance (G_in) after MF stimulation at different frequencies (20 Hz for 2 s or 50 Hz for 1 s, arrowhead). G_in values were normalized to the baseline value. D, MF conditioning-induced reduction of G_in was abolished by intracellular BAPTA (grey triangles, 10 mm) and bath-applied nimodipine (squares, 10 μm), but not by CPP (black circles, 20 μm in the bath). E, effects of dynasore (squares, 40 μm in the pipette) and PP2 (grey dots, 10 μm in the bath or pipette). F, MF conditioning reduced G_in in CA3-PCs of WT mice, but not in CA3-PCs of Kcna2 HT or homo KO mice. Insets in C to F, representative voltage responses to sub-threshold current injections (+10 and −30 pA) before (grey) and 20 min (C) or 40 min (D–F) after (black) MF conditioning. The control time courses of G_in (grey open circles) in D and E are reproduced from Fig.3C. G, summary for the relative changes in G_in (ΔG_in) under different conditions shown in C through F (bar graph). Mean values for the number of postsynaptic APs elicited by MF stimulation at 20 Hz for 2 s are superimposed (open circles, right axis). These values were not significantly different from each other (P > 0.1). H, left, plot of ΔG_in as a function of the number of APs in postsynaptic CA3-PCs elicited by afferent MF stimulation at different frequencies for 1 or 2 s. Data are collected from experiments using rats and WT mice under control or CPP conditions. A Gaussian function, y = ΔG_max[1 – exp(x/a)²], was fitted to the plot (grey line; ΔG_max = 32%; a = 10.4). Note the close correlation between ΔG_in and the number of APs regardless of MF stimulation frequencies. Right, the same plot of data collected from experiments of Kcna2 HT mice. I, effects of the MF conditioning on D-type K⁺ current (I_D) in the CA3-PCs. Outward K⁺ currents (I_K) elicited by a depolarizing step to −20 mV from −70 mV before and after the bath application of 30 μm 4-AP in the naïve and conditioned CA3-PCs. The subtraction of the latter (grey traces) from the former (black traces) I_K was regarded as I_D (black, control I_D; grey, I_D after MF conditioning). Bar graph, mean peak amplitudes of I_D induced by a step depolarization to −20, −30 and −40 mV before and after MF conditioning (n = 15).

Figure 5. Subcellular localization of Kv1.2 in CA3-PCs.

Aa, a schematic diagram and a composite fluorescence micrograph of the experimental set-up used for local puff application of 100 μm 4-AP to the proximal or distal apical dendritic region of CA3-PCs. The CA3-PC and the puffing pipettes were visualized by adding 100 μm Alexa 488 dye in the patch and puffing pipettes. Scale bars, 100 μm. A, a fluorescence image of Alexa dye pressure-ejected from the puffing pipette at 4 s after the start of the puff. A fluorescence line profile along the red dashed line is shown in the lower graph, where six other profiles are overlapped. The averaged profile is shown in red. c, outward K⁺ currents were elicited by a step depolarization to −20 mV before and after a local puff application of 100 μm 4-AP to a distal (left two traces) or proximal (right traces) dendritic region in the presence of an inhibitor cocktail. The 4-AP-sensitive difference currents shown below was regarded as D-type K⁺ current (I_D, red traces). The peak amplitude of I_D in a cell was measured from the averaged trace of 5 to 10 I_D traces. d, summary for the peaks of low 4-AP-sensitive current measured by local puff to proximal or distal dendrites in CA3-PCs. B, test for specificity of the anti-Kv1.2 antibody. a, hippocampal formations of WT and Kcna2 KO mice were immunolabelled with the anti-Kv1.2 antibody, visualized using Cy5-labelled anti-rabbit secondary antibody, and imaged under identical optical conditions. Scale bar, 200 μm. In the WT image, the hippocampal sulcus is outlined by a white dashed line. The dotted box area was magnified and shown as the right upper red colour image. Scale bar, 100 μm. b, anti-Kv1.2 antibody (red) specifically immunolabelled the HEK293 cell expressing Kv1.2-GFP (green). Note that untransfected cells are shown in the transmitted image. Ca, immunolocalization of surface Kv1.2 on a cultured putative hippocampal CA3-PC. Cultured neurons (DIV22) were immunolabelled for surface Kv1.2 (red) in a live cell state, and then for MAP2, a dendritic marker (green), after permeabilization. Scale bar, 100 μm. b, the distribution of Kv1.2-immunoreactive puncta (red dots) on the mask image of the cell shown in a (grey). c, the number of Kv1.2 puncta as a function of their curved distance from the soma (bin size, 5 μm). The dendritic area (dotted line) and the density of puncta (blue line) are superimposed on the graph (right axes). d, the summary graph for the puncta density profiles as a function of distance from the soma (grey lines, 16 cells). Each point in a density profile is an average of 10 points that were estimated every 5 μm. The density profile of the cell in a is denoted by a light blue line. Red symbols, averaged profile (mean ± SEM).

Isolation of monosynaptic MF-, PP- and A/C-CA3 responses

Afferent MFs were activated by monopolar stimulation delivered via an ACSF-filled glass pipette, which was positioned in the stratum (st.) lucidum (stimulus intensity, 6–20 V) using minimal stimulation techniques (Jonas et al. 1993) (Fig.3A). Monosynaptic direct cortical inputs in a CA3-PC were isolated using the methods pioneered by Berzhanskaya et al. (1998) and Tsukamoto et al. (2003). In both papers, direct cortical inputs were evoked by an electrode positioned at the st. lacunosum moleculare (SLM) of the CA1 area close to the hippocampal sulcus. To prevent contamination of polysynaptic inputs via entorhino-dentate connections, Berzhanskaya et al. (1998) made a cut through the hilus of the dentate gyrus (DG) (MF cut) (Berzhanskaya et al. 1998). Tsukamoto et al. (2003) made an additional cut along the hippocampal sulcus (sulcus cut), and named direct cortical inputs to the CA3-PCs as the temporoammonic (TA) path instead of PP. Following these studies, stimulation in CA1 SLM has been a standard method for stimulation of direct cortical inputs to the CA3 or CA2 area (Chevaleyre & Siegelbaum, 2010; Perez-Rosello et al. 2011; Sun et al. 2014). Because the present study aimed to elucidate the cellular mechanisms underlying the previous findings of Tsukamoto et al. (2003), we adopted the methods of Tsukamoto et al. (2003) in Figs 1, 2 and 8. After leaving the entorhinal cortex, PP fibres traverse the pyramidal cell layer of subiculum before entering the hippocampus (Tamamaki & Nojyo, 1993; Rowland et al. 2013). To validate that Tsukamoto's method activates direct cortical inputs, we compared the EPSCs evoked by this method and those by stimulation of subiculum in the slice without the sulcus cut. The rise times of EPSCs evoked by the two methods were not different (CAl SLM, 6.7 ± 0.4 ms, n = 16; subiculum, 6.1 ± 0.5 ms, n = 9, P = 0.35; Fig.2C), and were significantly slower than those evoked by stimulation of st. radiatum of area CA3 (slow A/C-EPSC, 2.9 ± 0.2 ms, n = 7, P < 0.005; Fig.2C), suggesting that synaptic inputs to distal apical dendrites are activated (Sjostrom & Hausser, 2006). Furthermore, the somatic conditioning enhanced EPSPs evoked by the two methods to a similar extent (Fig.2B), implying that the two methods activate the same kind of afferent fibres. Nevertheless, the anatomical identity of the direct cortical input activated by Tsukamoto's method is controversial. In contrast to the pervious conventional tracing study (Witter, 2007), it has been recently reported that the TA path does not reach the CA2 region, casting doubt on the existence of TA-CA3 synapses (Kohara et al. 2014). More recently, Sun et al. (2014) showed that stimulation in CA1 SLM evokes EPSPs not only in CA2-PCs but also in dentate granule cells, implying that PP fibres are activated. These studies imply that synaptic responses elicited by Tsukamoto's method most likely result from antidromic stimulation of cut PP fibres, and thus we refer to the synaptic inputs evoked by stimulation in CA1 SLM as PP inputs in the present study.

Figure 1. Repetitive somatic firing enhances PP-EPSPs but not MF-EPSPs.

Aa, tetanic stimulation of the soma at 10 Hz for 2 s (somatic conditioning; blue arrowhead at t = 0, inset) induced sustained enhancement of PP-EPSPs (red circles) but not MF-EPSPs (black circles) in CA3-PCs. b, schematic diagram of experimental setup and representative EPSPs and EPSCs evoked by stimulation of PP (left panel) or MFs (right panel). Averaged EPSP traces before (black) or 20 min after (red) conditioning are superimposed in each panel. The recording time points of the PP-EPSPs are denoted by ‘1’ and ‘2’ in panel a. Lower traces show PP- (left) and MF-EPSCs (right) before (black) and 25 min after (red) somatic conditioning. EPSCs were recorded at a holding potential of −60 mV. PP stimulation was confirmed by high sensitivity to DCG-IV of EPSPs (71% reduction on average) and slow rise time (>6.5 ms) of EPSCs. c, bar graphs show the mean for baseline values and somatic conditioning-induced changes of EPSPs and EPSCs. B, somatic conditioning-induced potentiation of PP-EPSPs is mediated by the same signalling pathway as long-term potentiation of intrinsic excitability (LTP-IE). Ba and b, a train of suprathreshold current pulses (1.2 nA for 2.2 ms) at 10 Hz was injected to the soma at t = 0 (somatic conditioning, arrowhead) in CA3 pyramidal cells. Time courses of PP-EPSPs are shown under different pharmacological conditions. The control time courses of PP-EPSPs are shown in grey symbols in each graph. c, somatic conditioning enhanced PP-EPSPs in the CA3-PC of WT mice but not in those of Kcna2 mutant mice. Insets in a–c, representative traces of PP-EPSPs before (grey) and after (black) somatic conditioning. d, plot of PP-EPSP amplitudes as a function of the amplitude of corresponding PP-EPSCs at the same synapses in CA3-PC of WT, HT and KO mice before (circles) and after somatic conditioning (triangles). Note that the EPSP-EPSC relationship of Kcna2 mutant mice overlaps the baseline relationship of WT mice and rats. Insets, mean values for the baseline amplitudes for PP–CA3 synaptic responses in different genotypes. e, summary graph for the effects of different pharmacological and genetic conditions on somatic conditioning-induced potentiation of PP-EPSPs. C, time courses of EPSP-to-spike coupling at the PP–CA3 synapse before and after somatic conditioning (arrows at t = 0). Somatic conditioning enhanced the number of spikes elicited by temporal summation of five EPSPs evoked by burst stimulation of PP at 20 Hz (a) or 50 Hz (b). Different symbols denote different cells. Insets: exemplar traces for summated PP-EPSPs before (black) and after (red) somatic conditioning.

Figure 2. Differential effects of somatic conditioning on PP-EPSPs and A/C fibre-evoked EPSPs (A/C-EPSPs).

In order to stimulate afferent fibres under a controlled intensity, synaptic responses were evoked by minimal stimulation techniques. A, EPSCs evoked by minimal stimulation of PP–CA3 synapses. Left, unitary EPSCs evoked by stimulation of PP–CA3 synapses at different stimulus intensities. Each EPSC trace represents the average of 20 responses of non-failure trials (except for 5.5 V) at the stimulus strength indicated. Middle, the graph for PP-EPSC amplitudes of non-failure trials vs. stimulus intensity shows an abrupt increase at a stimulus of 5.6 V. Right, monitoring of the peak amplitude of PP-EPSCs evoked by 5.6 V every 10 s. The failure rate was 52.75 ± 3.30% (n = 7), where EPSC < 1.5 pA is regarded as a failure. B, changes in PP-EPSP amplitudes of non-failure events evoked by minimal stimulation of PP before and after somatic conditioning (arrow and grey trace in the inset of each panel). Representative averaged traces for EPSCs or EPSPs are shown on the right (grey, control; black, after conditioning). Bar graphs show mean amplitudes of EPSCs and EPSPs evoked by stimulation of the subiculum (Sub.) or the CA1 SLM before and after somatic conditioning. Ca, no effects of somatic conditioning on A/C-EPSPs. A/C synaptic inputs were divided into two groups according to the rise time of EPSCs (b). Because the effects of somatic conditioning was not different between fast and slow A/C-EPSPs, data for relative changes in A/C-EPSPs of the two groups were pooled and shown as a function of time after somatic conditioning (left). Representative averaged traces for EPSCs or EPSPs are shown on the right (right; grey, control; black, after conditioning). b, mean values for the rise time of EPSCs evoked by stimulation of the CA3 stratum radiatum (SR), CA1 SLM and subiculum. Note that there is no significant difference in the rise time of EPSCs evoked by stimulation of CA1 SLM and subiculum, and that both rise times are slower than EPSCs by stimulation of the CA3 SR (left). Neither A/C-EPSPs nor A/C-EPSCs were altered after somatic conditioning (right).

Figure 8. Activation of dendritic Na⁺ channels is required for the somatic conditioning-induced potentiation of PP-EPSPs.

Aa, the somatic conditioning (arrow at t = 0) enhanced EPSPs evoked by minimal stimulation of PPs. Bath application of 10 nm TTX brought back the enhanced PP-EPSPs to the baseline. A, summary for the effects of 10 nm TTX on the PP-EPSP amplitudes of non-failure events in naïve CA3-PCs and after the somatic conditioning. Bar graphs show mean amplitudes of EPSPs under indicated conditions. Upper traces, representative averaged traces for EPSPs (grey, baseline; red, after conditioning; black, 10 nm TTX). A, summary for the effects of 10 nm TTX on the PP-EPSC amplitudes of non-failure events in naïve CA3-PCs and after the somatic conditioning. Bar graphs show mean amplitudes of EPSCs under indicated conditions. Upper traces, representative averaged traces for EPSCs (black, baseline; red, after conditioning; grey, 10 nm TTX). In contrast to PP-EPSPs, neither 10 nm TTX nor the conditioning altered the PP-EPSCs. The statistical significance was tested in comparison with corresponding baseline amplitudes. B, a fluorescence image of a CA3-PC that was loaded with 100 μm Alexa Flour 488 via a somatic patch pipette. A glass pipette for field stimulation was filled with ACSF plus 100 μm Alexa dye, and placed close to a dendritic branch in SLM of the CA3-PC. Traces in the upper insets show exemplar somatic recordings elicited by a pulse stimulation (3.4 V / 0.1 ms) delivered to the glass electrode. These responses were completely blocked by 10 μm CNQX (grey trace). The bar graph in the lower inset shows the combined amplitude histogram of EPSCs collected from 8 cells evoked by various stimulation voltages (90 events; bin size, 0.25 pA). The superimposed red curve represents a multi-Gaussian function with uniform peak intervals (2.1 pA). Ca and b, an exemplar dependence of PP-EPSCs (a) and PP-EPSPs (b) on the stimulation voltage (V_stim) measured in a single CA3-PC (red symbols, control; blue symbols, 10 nm TTX). Upper traces, representative averaged traces for somatic PP-EPSCs and PP-EPSPs evoked by the V_stim denoted by the numbers (red, baseline; blue, 10 nm TTX). c, the amplitude of somatic PP-EPSPs as a function of the amplitude of somatic PP-EPSCs. Da and b, mean amplitudes of somatic PP-EPSCs and PP-EPSPs averaged from synaptic responses obtained in four cells are plotted as a function of the difference of stimulation voltage and threshold voltage (V_stim – V_thr; red symbols, control; blue symbols, 10 nm TTX). The threshold voltage (V_thr) is defined as a stimulation voltage at which an abrupt increase in the PP-EPSP occurred under control conditions. The statistical significance at each point was determined in comparison to the point at the 0.2 V lower V_stim. Upper traces, representative averaged traces for somatic PP-EPSPs and PP-EPSCs. c, the amplitudes of averaged somatic PP-EPSPs as a function of those of somatic PP-EPSCs. n.s., not significant; *P < 0.05. Ea, exemplar effects of the somatic conditioning on the dependence of somatic PP-EPSPs on V_stim (black symbols, control; red symbols, after somatic conditioning). Right traces, representative averaged traces for somatic PP-EPSPs at the V_stim denoted by the numbers (black, naïve; red, after conditioning). Eb, summary for the dependence of somatic PP-EPSPs on the V_stim under control conditions (black) and after the somatic conditioning (red). The relationships obtained from four cells are aligned to the reference voltage of the V_thr of each cell. The statistical significance was tested between two conditions at the same V_stim − V_thr. F, the failure rates of PP-EPSCs as a function of V_stim – V_thr (red, control; blue, 10 nm TTX).

PP synaptic responses were evoked every 8 or 10 s by extracellular stimulation using a concentric bipolar electrode (12.5 μm inner pole diameter, 125 μm outer pole diameter; FHC Inc. Bowdoin, USA) positioned at the CA1 area close to the hippocampal sulcus (see Fig.4A for a diagram of the experimental set-up). For the experiments of Fig.8B, PP synaptic inputs were activated by extracellular field stimulation using an ACSF-filled glass pipette positioned close to a single dendritic branch. To this end, we visualized dendrites of a CA3-PC in the SLM using Alexa Fluor 488, which was added to the somatic patch pipette and the stimulation electrode (100 μm). Stimulation pulses (monopolar pulses, 100 μs in duration) were generated by a digital stimulator (WPI DS8000; World Precision Instruments, Sarasota, FL, USA) and fed into the stimulation electrode via an isolation unit (WPI stimulus isolator DLS100). Consistent with previous reports (Kamiya et al. 1996; Tsukamoto et al. 2003), we found that bath-applied 2 μm DCG-IV (Tocris Bioscience, Bristol, UK), a group II mGluR agonist, reduced PP-EPSPs and MF-EPSPs by 71% and 88.4%, respectively (Shigemoto et al. 1997; Tsukamoto et al. 2003) (Figs 1 and 4), whereas it had no effect on A/C-EPSPs (106.0 ± 8.2%; Fig.2C). Since the MF was cut, the sensitivity of EPSPs evoked by the PP path for DCG-IV indicates little contamination of A/C inputs. Therefore, we concluded that PP–CA3 monosynaptic EPSPs were specifically isolated by the double incisions (MF cut plus sulcus cut; Figs 1 and 2). Afferent A/C fibres were activated by stimulation of the st. radiatum at intensities of 1–8 V. To confirm that the activated afferents were not significantly contaminated by MF and PP inputs, 2 μm DCG-IV was applied at the end of every experiment, and the data were accepted only if synaptic responses were not reduced.

Minimal stimulation techniques

In most of experiments, MF-, A/C-, and PP–CA3 synaptic inputs were evoked by minimal stimulation, as described previously (Jonas et al. 1993; Perez-Rosello et al. 2011). The stimulation intensity was gradually increased until EPSCs could be evoked in an all-or-none manner depending on the stimulation intensity. Just above the threshold, we found a range of stimulus intensities such that EPSCs/EPSPs were evoked with a failure rate of 40–60% and the average amplitude of EPSCs/EPSPs was relatively stationary (Figs 2A and 3A). When minimal stimulation techniques were not applied, stimulation intensity (4–12 V) was adjusted such that synaptic responses were lower than 30% of the maximum amplitude (Fig.1).

Kv1.2 mutant mice

Heterozygous Kv1.2 knock-out mice (Kcna2+/–) were purchased from The Jackson Laboratory (Bar Harbor, ME, USA; Donating Investigator: Dr Bruce Tempel, University of Washington School of Medicine) (Brew et al. 2007), and maintained at the approved specific-pathogen-free facilities. By inter-crossing Kcna2+/– mice, we bred homozygous knock-out mice (Kcna2–/–) and wild-type littermate (Kcna2+/+). For genotyping, DNA was isolated from the tail of each mouse in a litter aged 6–8 days as described by Brew et al. (2007). Detailed protocols are available online (http://depts.washington.edu/tempelab/Protocols/KCNA2.html). Although Kcna2–/– mice had a severely reduced life span (range P18–P23) (Brew et al. 2007), they appeared normal during the first 2 weeks of their life.

Cell culture

Hippocampal neurons were cultured on a coverslip suspended above an astrocyte feeder layer for low-density culture. Hippocampal neurons and glia were obtained from Sprague–Dawley rats according to the protocols approved by the Seoul National University Institutional Animal Care and Use Committee. Two weeks before neuronal culture, astrocytes were obtained by passing a cortical cell suspension from a postnatal day (P)1 rat through a cell strainer (40 μm mesh, BD Falcon, Franklin Lakes, NJ, USA), and then cultured in glial medium (minimum essential medium (MEM; Invitrogen) supplemented with 0.6% glucose, 1 mm pyruvate, 2 mm GlutaMAX-I (Invitrogen, Carlsbad, CA, USA), 10% horse serum (HS; Invitrogen) and 1% penicillin–streptomycin (PS; Invitrogen)). From embryonic day (E)18 fetal rats, the hippocampal CA1 region was surgically removed and discarded from the dentate gyrus (DG) and CA3 regions by dissecting along the hippocampal sulcus. The DG and CA3 tissues obtained were dissected in Hanks’ balanced salt solution (Invitrogen), digested with papain (Worthington, Freehold, NJ, USA), and then triturated with a polished half-bore Pasteur pipette. The neurons in the plating medium (the same composition with the glial medium except for 10% fetal bovine serum (FBS; Invitrogen) instead of HS) were plated on a 0.01% poly-d-lysine (Sigma, St Louis, MO, USA)-coated glass coverslip (Marienfeld, Lauda-Königshofen, Germany) in a 60 mm culture dish at a density of 1.0 × 10⁴ cells cm⁻². Paraffin dots (acting as ‘feet’ to suspend the coverslips above the glial feeder layer) were applied to the coverslips before neuron plating. The next day, coverslips were transferred above the glial culture pre-incubated in Neurobasal A medium (Invitrogen) supplemented with 0.5 mm GlutaMAX-I and 2% B-27 supplement (Invitrogen) for a day. To prevent the proliferation of glial cells, 5 μm of 1-β-d-cytosine-arabinofuranoside (Sigma) was added on the fourth day in vitro (DIV). HEK293 cells (ATCC) were plated at a density of 1 × 10⁵ cells per a 0.01% poly-l-lysine (Sigma)-coated glass coverslip (Marienfeld) in a 60 mm culture dish and maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% FBS and 1% PS.

Antibody

Rabbit polyclonal Kv1.2 antibodies were raised against a synthetic peptide corresponding to the extracellular region of mouse Kv1.2 (DENEDMHGGGVTFHTYSNS; amino acids 190–208). Peptide synthesis, immunization, sampling of rabbit anti-sera, and affinity purification were performed by a commercial facility (AbFrontier, Seoul, Korea). The specificity of anti-Kv1.2 antibody was verified by immunocytochemistry. The Kv1.2-GFP construct was created by subcloning mouse Kv1.2 cDNA (Addgene plasmid 30187) into the mammalian expression vector pEGFP-C1 (Clontech, Palo Alto, CA, USA) using XhoI and BamHI sites. The HEK293 cells expressing Kv1.2-GFP were specifically immunolabelled with anti-Kv1.2 antibody. Untransfected cells were visualized using differential interference contrast microscopy.

Immunocytochemistry

For detecting surface Kv1.2, live DIV22 hippocampal neurons were incubated with rabbit anti-Kv1.2 (1:200) in a serum-free culture medium for 15 min at 4°C, rinsed with phosphate-buffered saline (PBS, 10 mm, pH 7.4), fixed with 3.8% formaldehyde in PBS for 20 min at room temperature (RT), and washed with PBS for 15 min. For double-immunostaining with MAP2, cells were permeabilized with PBS plus 0.1% Triton X-100 (0.1% PBST) for 5 min at RT. The fixed cells were incubated in a blocking solution (5% donkey serum in 0.1% PBST) for 1 h at RT, and then with mouse anti-MAP2 (1:400, Millipore/Chemicon) diluted in the blocking solution for 1 h at RT. After three washes in 0.1% PBST, the cells were incubated with secondary antibodies diluted in the blocking solution for 1 h at RT in the dark. Secondary antibody dilutions were as follows: Cy5-conjugated anti-rabbit IgG (1:200, Abcam, Cambridge, MA, USA) or FITC-conjugated anti-mouse IgG (1:50, Jackson ImmunoResearch). Finally, the cells were washed three times in PBS and mounted with fluorescent mounting medium (Biomeda, Foster City, CA, USA). To immunostain total (cytosolic and surface) Kv1.2, cells were incubated in the blocking solution containing the rabbit anti-Kv1.2 (1:200) for 1 h at RT after the aforementioned fixation and permeabilization procedures. Subsequent steps were the same as those for surface Kv1.2 immunostaining. The immunostained cells were imaged with a FV300 (Olympus) confocal laser scanning microscope equipped with a 60× water-immersion objective.

Quantitative analysis of immunocytochemical data

The region of interest (ROI) for a single cultured neuron was defined as a binary mask covering the whole somatodendritic compartment of the neuron. It was made from the MAP2 fluorescence image of the neuron using the ‘ImageSeedFill’ routine supplied by Igor Pro (version 4.1; WaveMetrics, Lake Oswego, OR, USA) as previously described (Lee et al. 2012). To this end, we took a set of seed pixels from the pixels on the traced line drawn on the neurites of the same image using NeuronJ, and fed the seed pixels into the ‘ImageSeedFill’ routine. Once the ROI is determined, we counted the number of Kcna2-positive puncta on the ROI only if the pixel count of a puncta was higher than a threshold, which was set using an image histogram of the red fluorescence image (typically, between 1000 and 1500, 12 bit image). The curved distance from the soma (denoted by d) of each Kcna2-positive punctum was measured using the traced line on neurites. To determine the total area of neurites as a function of d, we measured the sum of widths of all neurites at the same d (denoted by Σwidth(d)) every 5 μm. The neurite area as a function of d (denoted by area(d)) was determined by 5 × Σwidth(d) (unit, μm²). The histogram of puncta numbers as a function of d was divided by area(d) to determine the density profile of Kcna2-positive puncta on the single neuron.

Electron microscopy

Three SD rats (9 weeks old, 290–300 g) and two Kcna2–/– mice (P17) of either sex were anaesthetized with sodium pentobarbital (80 mg kg⁻¹, intraperitoneal injection) and perfused transcardially with 100 ml of heparinized normal saline followed by 500 ml of a freshly prepared mixture of 4% paraformaldehyde and 0.01% glutaraldehyde in a 0.1 m phosphate buffer (PB, pH 7.4). The hippocampus was removed and postfixed in the same fixative for 2 h at 4°C. Sections were cut sagittally on a vibrating blade microtome at 60 μm and cryoprotected in 30% sucrose in PB overnight at 4°C. Sections were frozen on dry ice for 20 min and thawed in 10 mm phosphate buffered saline (PBS, pH 7.2) to enhance penetration of antibody. Hippocampal sections were blocked with 10% normal donkey serum (NDS; Jackson Immunoresearch, West Grove, PA, USA) for 30 min to mask secondary antibody binding sites. For immunostaining of Kv1.2, hippocampal sections were incubated overnight in a rabbit anti-Kv1.2 antibody (32 μg ml⁻¹ in PBS). After rinsing with PBS, sections were incubated with 1 nm gold-conjugated donkey anti-rabbit (1:50; EMS, Hatfield, PA, USA) antibodies for 2 h. The sections were post-fixed with 2% glutaraldehyde in PB for 10 min, rinsed with PBS several times, incubated for 4 min in silver intensification solution (HQ silver™ Enhancement Kit, Nanoprobes, Yaphank, NY, USA), and rinsed with 0.1 m sodium acetate and PB. Sections were further rinsed with PB, osmicated (0.5% osmium tetroxide in PB) for 1 h, dehydrated in graded alcohols, flat embedded in Durcupan ACM (Fluka, Buchs, Switzerland) between strips of Aclar plastic film (EMS), and cured for 48 h at 60°C. Thin sections were cut with a diamond knife, mounted on formvar-coated single slot nickel grids, and stained with uranyl acetate and lead citrate. Grids were examined on a Hitachi H-7500 electron microscope (Hitachi, Tokyo, Japan) at 80 kV accelerating voltage. Images were captured with DigitalMicrograph software driving a cooled CCD camera (SC1000; Gatan, Pleasanton, CA, USA) attached to the microscope, and saved as TIFF files. We assessed the specificity of the primary antibodies in three different ways. Firstly, we processed hippocampal sections from three rats according to the above-described protocols, except that primary or secondary antibodies were omitted. Omission of primary or secondary antibodies completely abolished specific staining. Secondly, the specificity of the immunoreaction was confirmed by the consistency of immunostaining in adjacent serial thin sections of the same pyramidal soma. Finally, we assessed the specificity of immunostaining using the hippocampal sections from two Kcna2−/− mice, and confirmed that no immunogold particles were detected in those that were processed by the same protocol.

Quantitative analysis of the immunogold particle density

Samples representing soma and proximal and distal apical dendrites of CA3-PCs were taken from the st. pyramidale, st. lucidum and st. lacunosum moleculare, respectively. Although we took samples for distal apical dendrites from the regions including the SLM close to the hippocampal sulcus and the suprapyramidal blade of the dentate gyrus (>350 μm from the soma), it was not easy to differentiate the SLM from the most distal part of the st. radiatum, and thus the latter may be included in the SLM samples. For proximal and distal basal dendrites, samples were taken from the st. oriens. We compared the immunogold particle densities between ultrathin sections at a depth of between 1 and 2 μm from the tissue surface. In each electron micrograph, we identified pyramidal somata, dendritic shafts and spines, and counted gold particles on or close to the plasma membrane. Gold particles located close to the cell membrane (<25 nm) were considered to be associated with the membrane. The immunogold particle density on the membrane was calculated as the number of particles divided by the length of the plasma membrane (μm), which were both measured using ImageJ software (NIH, Bethesda, MD, USA). Statistical data are presented as means ± SEM.

Compartmental simulation of CA3 pyramidal cells

Simulations of cellular ionic currents and postsynaptic potentials were performed using NEURON software (version 7.1 running on the MS Windows 7). Model files are available in the public database (https://senselab.med.yale.edu/modeldb/). The morphology of a reconstructed CA3-pyramidal cell was downloaded from http://senselab.med.yale.edu (CA3c(L22).hoc). Voltage-dependent fast Na⁺ current (I_Na), transient outward current (I_A), D-type K⁺ current (I_D), and delayed rectifying K⁺ current (I_DR) were incorporated in the NEURON model of a CA3-PC. The gating kinetics of voltage-dependent K⁺ channels were modelled after Lazarewicz et al. (2002) and Hemond et al. (2008), and that of Na⁺ channels after Hu et al. (2009). We adjusted the gating parameters for K⁺ currents such that the simulated somatic outward K⁺ current best reproduces the experimental somatic recordings of the outward K⁺ current under the same depolarization steps. The parameters for the activation and inactivation curves of K⁺ and Na⁺ channels were adopted from reported values as shown in Table 1 (Saviane et al. 2003; Kim et al. 2012). The equations describing the time and voltage dependence of K⁺ and Na⁺ conductances are listed in Table 2. The distributions of Na⁺ and K⁺ conductances along the somatodendritic axis were set essentially according to Kim et al. (2012), but slight changes of the current densities were made such that the simulation results are compatible with our previous somatic recordings of the fast-, slow- and non-inactivating outward K⁺ currents (Fig.7B). The channel densities for I_Na, I_A, and I_DR in the soma were set as 20, 23 and 3.6 mS cm⁻², respectively. As Kim et al. (2012) reported earlier, we assumed that I_DR activity is uniformly distributed, and that the channel density for I_A increases along the apical dendrite with a slope of 0.055 mS cm⁻² μm⁻¹. In addition, we assumed that Na⁺ channel density in the proximal half of apical dendrites (<150 μm) is 20% of that in the soma but increases with a slope of 0.096 mS cm⁻² μm⁻¹ in the distal half, resulting in the same dendritic Na⁺ channel density as the soma at 316 μm (Kim et al. 2012). The channel density for I_D has not been previously studied_. Because we discovered through the present study that Kv1.2 density is 10 times higher at distal apical dendrites (Fig.6D), we assumed that the density of K_D channels in the soma and proximal dendritic regions (< 150 μm from the soma) is 10 times lower than that in distal apical dendrites, where the K_D channel density was set to 3 mS cm⁻². The axonal conductances for I_Na, I_A, I_DR and I_D were assumed to be 100, 20, 10 and 3 mS cm⁻² (Kim et al. 2012).

Table 1.

Parameters used in the compartmental simulation

Model parameter	Values
Holding potential for voltage clamp	−80 mV
RMP for current clamp	−70 mV
Passive properties
Membrane capacitivity (Cm)	0.8 μF cm−2
Membrane resistivity (Rm)	220 kΩ cm−2
Dendritic axial resistivity (Ra)	200 Ω cm
Spine membrane correction factor	2
Ionic current parameters	V1/2 / slopea
IDb
Activation	−48 / 5.22 mV
Inactivation	−90 / −8.7 mV
IAc
Activation	−15 / 8.7 mV
Inactivation	−70 / −5.22 mV
IDRd
Activation	−3.8 / 8.7 mV
INad
Activation	−30.4 / 6.2 mV
Inactivation	−73 / −8.0 mV
Density of ionic currentsd	Soma/Prox./Dist. dendritee
Gmax of ID (GKd,max)	0.3 / 0.3/ 3 mS cm−2
Gmax of IA (GKa,max)	23 / 23 / 33.8 mS cm−2
Gmax of IDR (GKdr,max)	3.6 / 3.6/ 3.6 mS cm−2
Gmax of INa (GNa,max)	20 / 4 / 18.4 mS cm−2
Synaptic parameters
NMDA/AMPA ratiof	0.2
Rise and decay τ of AMAPRc	1.5 ms
Rise τ of NMDARf	4 ms
Decay τ of NMDARf	16 ms
[Mg2+]	1 mM
Recovery τ of SV poolg	100 (20) ms
Facilitation magnitudeg	1 (2)
Decay τ of faciltationg	50 (50) ms
Release probabilityg	0.2 (0.05)
Synaptic weight at MF/AC/PP synapses	0.5/0.1/0.19 nS

Abbreviations: RMP, resting membrane potential; I_D, D-type K⁺ current; I_A, A-type K⁺ current; I_DR, delayed rectifier K⁺ current; I_Na, fast Na⁺ current; τ, time constant; SV, synaptic vesicle.

^a1/ [1 + exp(–(V – V_1/2) / Slope)].

^bAdapted from Saviane et al. 2003 after correction for liquid junction potential.

^cFit to the data.

^dMinor adjustments of parameters reported by Hu et al. 2009 (parameters for Na⁺ channel kinetics) and Kim et al. 2012 (the rest).

^eProx., proximal dendrite at 50 μm; Dist., distal dendrite at 300 μm from the soma.

^fBaker et al. 2011.

^gNumbers in parentheses are the parameters for MF synapses.

Table 2.

The time and voltage dependence of Na⁺ and K⁺ conductances

IA = GKa(Vm – EK)
GKa = GKa,max n h
αn = exp[0.0383 z (Vm +15)]
βn = exp[0.021 z (Vm +15)]
where z = −3 – 1 / [1 + exp((Vm + 40)/5)]
n∞ = 1 / (1 + αn)
τn = max{0.1, βn / [0.05 (1+ αn)]}
αh = exp[0.1915 (Vm + 70)]
h∞ = 1 / (1 + αh)
τh = max[13.4, 0.52 (Vm + 50)]
ID = GKd(Vm – EK)
GKd = GKd,max x y
αx = exp[(Vm + 48) / 10.44]
βx = exp[–(Vm + 48) / 10.44]
x∞ = αx / (αx + βx)
τx = 1 / (αx + βx) + 1
αy = 10−3 exp[–(Vm + 90) / 17.4]
βy = 10−3 exp[(Vm + 90) / 17.4]
y∞ = αy / (αy + βy)
τy = 1 / (αy + βy) + 100
IDR = GKdr(Vm – EK)
GKdr = GKdr,max n
αn = exp[–(Vm + 3.8) / 8.7]
βn = exp[–(Vm + 3.8) / 17.1]
n∞ = 1 / (1 + αn)
τn = 50 βn / (1 + αn)
INa = GNa(Vm – ENa)
GNa = GNa,max m3 h
αm = 0.182(Vm + 38) / [1 – exp(–(Vm + 38) / 8)]
βm = −0.124 (Vm + 38) / [1 – exp((Vm + 38) / 8)]
τm = 1 / (αm + βm)
m∞ = αm / (αm + βm)
αh = 0.024(Vm + 50) / [1 – exp(–(Vm + 50) / 5)]
βh = −0.0091(Vm + 75) / [1 – exp((Vm + 75) / 5)]
τh = 1 / (αh + βh)
h∞ = 1/ [1 + exp((Vm + 73) / 8)]

The values for current density (G_max) are shown in Table 1. The units for rate constants and voltage are 1/ms and mV, respectively.

Figure 7. NEURON simulation for the differential effects of I_D reduction on MF-, A/C- and PP-EPSPs in a CA3-PC model.

A, a reconstructed CA3 pyramidal neuron used in this simulation. MF, A/C and PP synaptic sites on apical dendrites are indicated by arrows of different colours (blue, green and red, respectively). The distances of synaptic sites from the soma were 47, 146 and 330 μm for the MF, A/C and PP synapses, respectively. Ba, densities of ionic conductance incorporated in this simulation. Densities of Na⁺ current (G_Na, red), A-type K⁺ current (G_Ka, blue) and D-type K⁺ current (G_Kd, black) are shown as a function of distance from the soma. G_Na at distal apical dendrites 344 μm from the soma was assumed to be the same as that in the soma (red asterisk). Conductance for delayed rectifier K⁺ current was assumed to be evenly distributed (3.6 mS cm⁻²). b, experimental recordings of outward K⁺ currents at the soma elicited by a step depolarization from holding potential of −70 mV under the control conditions (upper) and after the somatic conditioning (lower; noisy grey traces). Each grey trace represents an averaged trace of K⁺ currents recorded from four CA3-PCs induced by the same pulse protocol. The depolarization steps ranged from −50 to −10 mV with increments of 10 mV. The superimposed continuous lines represent outward K⁺ currents that were simulated using the same depolarizing step under control conditions (upper) and after reducing G_KD to 10% of the control (lower). The liquid junction potential was assumed to be 10 mV. c, simulated voltage responses to somatic current injection of +10 and −30 pA under the control conductances (grey) and after the reduction of G_Kd to 10% (black), which reduced the G_in estimate from 3.53 to 2.98 nS. Ca, simulated EPSPs evoked by activation of MF, A/C or PP synapses. Their locations on apical dendrites are denoted in A. Four EPSP traces superimposed in each panel represent somatic EPSPs evoked by activation of the same synapses under four different settings of ionic conductances (light blue, control; red, reduction of G_Kd to 10% of the control value; dashed blue line, reduction of G_Na to 50%; dashed red line, reduction of G_Kd and G_Na to 10% and 50%, respectively). b, the simulated PP-EPSPs shown in panel a are superimposed on the experimental recordings of PP-EPSPs under the control conditions (black noisy trace), after somatic conditioning (grey noisy trace) and after adding 10 nm TTX to the conditioned CA3-PC (grey noisy trace of lower amplitude; thin scale bar, 0.2 mV; reproduced from the right trace in Fig.8A,b). Simulated traces for the concomitant dendritic local EPSPs are superimposed (faster decaying traces denoted by arrow; thick scale bar, 5 mV). c, comparison of local dendritic PP- and A/C-EPSPs at the synaptic sites. Da, somatic EPSPs evoked by a burst activation (five pulses at 50 Hz) of MF (lower), A/C (middle) or PP (upper) synapse. The superimposed EPSP traces represent EPSPs simulated under different ionic conductances (the same line symbols as C). b, concomitant dendritic EPSPs evoked by a burst activation of PP or A/C synapse. E, the amplitude of simulated somatic PP-EPSPs as a function of synaptic conductance under different settings of ionic conductances (blue, control; red, 10% G_Kd; green, 50% G_Na). The right part shows somatic (upper row) and local dendritic PP-EPSPs (lower row) under different synaptic conductances (left, 0.15 nS; middle, 0.2 nS; right, 0.25 nS). On each panel, EPSPs simulated under different ionic conductances are superimposed (blue, control; red, 10% G_Kd; green, 50% G_Na).

Figure 6. Electron-microscopic localization of immunogold particles for dendritic Kv1.2 in different strata of the hippocampal CA3 region.

A, PP-EPSPs (left), but not A/C-EPSPs (right), were enhanced by somatic conditioning in the CA3-PCs of 6 week-old SD rat (grey, control; black, conditioned). Upper traces are exemplar EPSPs (grey, control; black, conditioned). B, electron micrograph of the stratum (st.) lacunosum moleculare of hippocampal CA3 region of Kv1.2 KO mice. Note that no immunogold particle was detected. The dendritic structures are outlined by white dotted lines. d, dendrite; scale bar, 500 nm. C, electron micrographs of the st. pyramidale (a), distal st. oriens (b), st. lucidum (c) and st. lacunosum moleculare (d) show cytosolic (arrow) or membrane-associated (arrowhead) gold particles on the putative soma, distal basal dendrites, proximal apical dendrites and distal apical dendrites of CA3-PCs, respectively (outlined with white dashed lines). Scale bars, 500 nm. D, summary for immunogold particle densities on the plasma membrane of the soma or dendrites in different strata of the hippocampal CA3 region. Data were obtained from hippocampal preparations from three rats. The numbers in parentheses are the number of somata or dendrites that were observed.

The membrane capacitivity (C_m) and the dendritic axial resistivity (R_a) were set as 0.8 μF cm⁻² and 200 Ω cm, respectively (Major et al. 1994). To correct the spine membrane area on dendrites, the C_m of dendrites was multiplied by 2. Because I_D contributes to the resting membrane conductance in our simulation, the passive membrane resistivity (R_m) was set slightly higher than the previous estimate (Major et al. 1994; ∼120–200 KΩ cm⁻²). When R_m was set to 220 KΩ cm⁻², the simulated G_in (estimated from simulated voltage responses to −30 pA and +10 pA current steps) was similar to the mean value of our experimental estimates (3.5 nS; Fig.7Bc; Hyun et al. 2013).

Glutamate synapses were modelled after Baker et al. (2011) except the rise time of AMPA receptor (AMPAR) activation. The kinetics of I_AMPA and I_NMDA were modelled using alpha and biexponential functions, respectively. The rise time of the simulated PP-EPSP was best fit to the experimental data by setting the rise time of AMPAR to 1.5 ms. The rise and decay time constants of N-methyl-d-aspartate receptors (NMDARs) were set to 4 and 16 ms, respectively (Baker et al. 2011). The AMPA current (I_AMPA) to NMDA currents (I_NMDA) ratio was set to 0.2. Voltage dependence of the Mg²⁺ block of NMDAR was modelled to 1 / [1 + exp(−0.062 V_m) × ([Mg²⁺] / 3.57)] (Jahr & Stevens, 1990). Short-term depression and facilitation were formulated as follows:

where D_n and D′_n are the readily releasable vesicle pool size just before and after arrival of n-th AP, respectively, and

where P_b and A_f are basal release probability (P_r) and an increase in P_r immediately after an AP, respectively. The parameters for synaptic weight, P_b, A_f, τ_D and τ_F at MF, A/C and PP synapses are listed in Table 1. Synaptic weights were determined such that the amplitudes of simulated somatic EPSPs are similar to those of EPSPs evoked by minimal stimulation in the experiments (Fig.2). We assumed that minimal stimulation of PP and A/C fibres activates two synapses at different branches in st. lacunosum moleculare and st. radiatum, respectively, while MF stimulation activates a single synapse at st. lucidum (Fig.7C). For simulation of burst PP- and A/C-EPSPs, six synapses in different branches were activated (Fig.7D). Under these parameters, the baseline EPSP amplitudes of MF-, A/C- and PP-evoked synaptic responses at the soma were 1.3, 0.4 and 0.4 mV, respectively. For simulation of EPSPs, the resting membrane potential was set at −70 mV.

Data analysis

Electrophysiological data were analysed using IgorPro (version 6.10A; WaveMetrics, Lake Oswego, OR, USA). Statistical data were presented as means ± standard error of mean (SEM, denoted as error bars), and n indicates the number of cells or animals studied. The differences between statistical data under two experimental conditions were evaluated using Student's t test, the Mann-Whitney U test, or Wilcoxon's signed rank test after testing normality using the Kolmogorov–Smirnov test. Differences were considered as significant when P < 0.05. n.s., no statistical significance; *P < 0.05; **P < 0.01; ***P < 0.005.

Results

Repetitive somatic firing enhances preferentially perforant path synaptic inputs

Previously, we reported that repetitive somatic firing (20 APs at 10 Hz) induces downregulation of D-type K⁺ channel subunits, Kv1.2 (gene name, Kcna2), in the CA3-PCs (so-called LTP-IE) (Hyun et al. 2013). Because LTP-IE depended on dendritic Ca²⁺ signalling and was abolished by surgical truncation of distal apical dendrites (Hyun et al. 2013), we hypothesized that LTP-IE may be accompanied by enhancement of dendritic excitability and synaptic potentials. To test this hypothesis, we made whole-cell patch recordings in the soma of a CA3-PC, and examined the effects of short tetanic stimulation of the soma (2 s at 10 Hz; referred to as ‘somatic conditioning’) on EPSPs evoked by stimulation of direct cortical afferent fibres, which innervate distal apical dendrites. We isolated monosynaptic direct cortical synaptic inputs to a CA3-PC by stimulation of CA1 SLM as described in Methods. We monitored the amplitude of PP-EPSPs every 8 s. Somatic conditioning enhanced PP-EPSPs by 1.8 ± 0.14-fold (P < 0.005, n = 22) but not PP-evoked excitatory postsynaptic currents (PP-EPSCs; Fig.1Aa; baseline, 27.8 ± 3.7 pA; 109 ± 14%, P = 0.3, n = 10) in the same synapses. Somatic conditioning enhanced not only the amplitude of PP-EPSPs but also the number of spikes elicited by temporal summation of PP-EPSPs (Fig.1C). Since somatic recordings of EPSPs that occur at synapses on distal dendrites may be heavily influenced by active dendritic conductances (Urban et al. 1998; Kim et al. 2012), the enhancement of PP-EPSPs with little changes in PP-EPSCs may be interpreted as an increase of dendritic intrinsic excitability rather than synaptic strength itself. LTP-IE is mediated by Ca²⁺ influx through L-type Ca²⁺ channels (but not through N-methyl-d-aspartate receptors (NMDARs)), activation of protein tyrosine kinase (PTK), and endocytosis of Kv1.2 (Hyun et al. 2013). We tested whether the same cellular signalling pathway underlies somatic conditioning-induced enhancement of PP-EPSPs. The enhancement of PP-EPSPs was not affected by a PKA inhibitor, KT-5720 (1.2 μm), but was inhibited by nimodipine (10 μm in the bath), PP2 (10 μm in the bath or pipette) and dynasore (40 μm in the pipette), which block intracellular [Ca²⁺] increase, L-type Ca²⁺ channels, PTK and endocytosis, respectively (Fig.1B). Furthermore, in contrast to CA3-PCs of wild-type (Kcna2+/+ or WT) mice, such PP-EPSP changes were not observed in CA3-PCs of Kv1.2 knock-out mice (Kcna2–/– or KO) or heterozygous littermate (Kcna2+/– or HT) (Fig.1B), indicating that the enhancement of PP-EPSPs is mediated by the same mechanisms as LTP-IE (Hyun et al. 2013). As expected, CA3-PCs of Kcna2 mutant mice displayed lower baseline G_in, faster AP onset time and reduced D-type K⁺ current (I_D) (Table 3), raising the possibility that the effect of somatic conditioning on PP-EPSPs may be occluded in the CA3-PCs of Kcna2 mutant mice. To test this possibility, the baseline and conditioned PP-EPSP amplitudes were plotted as a function of corresponding EPSC amplitudes (Fig.1Bd). For WT mice and rats, the baseline EPSPs and EPSCs displayed a linear relationship, and most conditioned EPSPs were above the baseline relationship. The EPSP amplitudes of Kcna2 mutant mice were on the baseline relationship of WT animals rather than conditioned ones, suggesting that the lack of PP-EPSP potentiation in CA3-PCs of Kcna2 mutant mice is not caused by occlusion of the somatic conditioning effect on PP-EPSPs. The mean values for baseline EPSPs and EPSCs were not different between animal groups (bar graphs in Fig.1Bd).

Table 3.

Baseline intrinsic properties of CA3-PCs of wild type (WT) and Kcna2 mutant mice

Genotype	Input conductance (nS)b	First spike latency (ms)b	Peak ID (pA)c	AP threshold (mV)
Kcna2+/+a	3.63 ± 0.26	465.8 ± 26.8	313.8 ± 28.3	−36.1 ± 0.7
(WT)	(n = 17)	(n = 7)	(n = 7)	(n = 12)
Kcna2+/–	2.92 ± 0.16	306.5 ± 16.2*d	180.4 ± 15.8**	−35.9 ± 0.9
(HT)	(n = 11)	(n = 7)	(n = 10)	(n = 8)
Kcna2–/–	2.49 ± 0.41*	278.1 ± 30.1*	148.9 ± 14.7**	−37.7 ± 0.9
(KO)	(n = 7)	(n = 8)	(n = 7)	(n = 8)

^aData for WT and homozygous KO mice are reproduced from Hyun et al. (2013).

^bInput conductance and first spike latency were measured using the same protocol as described in Hyun et al. (2013).

^cI_D, 30 μm 4-AP-sensitive D-type K⁺ current at the test potential of −20 mV.

^dStatistical significance was tested in comparison with WT mice; means ± SEM; *P < 0.05, **P < 0.01.

Intriguingly the same conditioning enhanced neither MF-EPSPs (baseline, 3.93 ± 0.48 mV; 0.84 ± 0.13-fold, P = 0.94, n = 6) nor MF-EPSCs (141.8 ± 14.1 pA to 138.0 ± 14.9 pA, P = 0.67, n = 5; Fig.1A), suggesting that somatic conditioning enhances preferentially synaptic inputs to distal apical dendrites. To further test this idea, we compared the effect of somatic conditioning on PP-EPSPs with that on A/C fibre-evoked EPSPs (A/C-EPSPs). A/C synaptic inputs could be differentiated from PP inputs by their insensitivity to DCG-IV (106.0 ± 8.2%, n = 9). In order to stimulate afferent fibres under a controlled intensity, we restricted the analysis to synaptic inputs evoked by minimal stimulation techniques (Fig.2A) (Jonas et al. 1993). A/C-EPSPs were divided into two groups according to the rise time of their corresponding EPSCs (fast, 1.5 ± 0.2 ms, n = 5; slow, 3.0 ± 0.2 ms, n = 4; Fig.2C) (Perez-Rosello et al. 2011). There was no difference in the baseline amplitudes between PP-, fast A/C- and slow A/C-EPSPs evoked by minimal stimulation (PP, 0.49 ± 0.08 mV; slow A/C, 0.66 ± 0.09 mV; fast A/C, 0.54 ± 0.06 mV, P = 0.401). Under the minimal stimulation conditions, somatic conditioning enhanced PP-EPSPs evoked by stimulation of either CA1 SLM or subiculum (CA1 SLM, 187.1 ± 18%, n = 5; subiculum, 186.7 ± 12.7%, n = 7, Fig.2B). In contrast, A/C-EPSPs were not altered by the same somatic conditioning regardless of corresponding EPSC rise time (fast, 94.1 ± 4.1%, P = 0.19, n = 5; slow, 108.1 ± 6.6%, P = 0.41, n = 4; Fig.2C). These results indicate that the excitability change caused by somatic conditioning increases preferentially PP-EPSPs with little effect on MF- and A/C-EPSPs. Furthermore, the somatic conditioning altered neither the amplitude of PP-EPSCs nor their failure rate, arguing against possible involvements of retrograde signalling-induced presynaptic mechanisms (48.6 ± 1.39% to 47.7 ± 5.34%, n = 6, P = 0.89).

High frequency minimal stimulation of MFs induces long-term reduction of input conductance in CA3-PCs

Given that the MF–CA3 synapse is a conditional detonator synapse (Henze et al. 2002; Bischofberger et al. 2006), we posed a question of whether HFS of MFs can result in excitability changes similar to those induced by the repetitive somatic firing in a postsynaptic CA3-PC. To address this question, we made whole-cell patch recordings in the soma of a CA3-PC, and examined postsynaptic AP discharge evoked by 1 or 2 s stimulation of afferent MFs at different frequencies (10, 20, 30, 50 and 100 Hz) using minimal stimulation techniques (Fig.3A). Minimal stimulation of MFs elicited repetitive firing of the postsynaptic CA3-PC at 10–20 Hz when the stimulation frequency was 20 Hz or higher (Fig.3B), raising the possibility that high frequency MF inputs may induce LTP-IE in a postsynaptic CA3-PC.

To test this possibility, we examined relative changes of G_in in CA3-PCs before and after conditioning with high frequency minimal stimulation of afferent MFs. G_in was determined every 10 s from the voltage deflections elicited by positive and negative current injections (+10 and −30 pA, respectively) at the resting membrane potential (−66.9 ± 1.3 mV; exemplar voltage traces, Fig.3C). After assessment of the baseline G_in, we stimulated afferent MFs at different frequencies, and then resumed monitoring of G_in. MF stimulations at 20 Hz for 2 s, at 30 Hz for 2 s, or at 50 Hz for 1 s all reduced G_in to similar extents 40 min after the stimulations (about 70% of the control value; Fig.3C and G). Either no MF stimulation or stimulation at 20 Hz for a shorter period (1 s) did not significantly alter G_in (Fig.3G). Since dentate GCs fire in vivo at a peak frequency of up to 20 Hz in their place fields (Leutgeb et al. 2007; Neunuebel & Knierim, 2012), we adopted the stimulation of afferent MFs at 20 Hz for 2 s as the ‘MF conditioning’ protocol in subsequent studies.

We tested whether the same cellular signalling pathway responsible for the LTP-IE underlies the MF conditioning-induced reduction in G_in (Hyun et al. 2013). We first tested a possible involvement of NMDARs. The addition of 3-(2-carboxypiperazin-4-yl)propyl-1-phosphonic acid (CPP, 20 μm) to the bath, an NMDAR antagonist, did not prevent the MF conditioning effect in CA3-PCs (73.4 ± 2.8%, P < 0.005, n = 9; Fig.3D). Next, we tested the effects of BAPTA (10 mm in the pipette), nimodipine (10 μm in the bath), PP2 (10 μm in the bath or pipette) and dynasore (40 μm in the pipette), which block intracellular [Ca²⁺] increase, L-type Ca²⁺ channels, PTK and endocytosis, respectively. All of these drugs abolished the MF conditioning-induced reduction of G_in. Finally, to test the involvement of Kv1.2, we examined the effect of MF conditioning in CA3-PCs of wild-type (WT), Kcna2 knock-out mice (KO, Kcna2–/–) and heterozygous littermates (HT, Kcna2+/–). MF conditioning reduced G_in in the WT CA3-PCs (70.0 ± 1.7%, P < 0.005, n = 5; Fig.3F). However, no significant changes in G_in were made by the same MF conditioning in CA3-PCs from KO and HT mice (KO, 101.6 ± 2.9%, n = 8; HT, 109.9 ± 5.2%, n = 7; Fig.3F), suggesting that downregulation of Kv1.2 mediates MF conditioning-induced reduction of G_in. Figure 3G summarizes the effects of the drugs and Kcna2 deletion on relative changes of G_in (ΔG_in) measured 40 min after MF conditioning, showing that the cellular mechanism underlying the G_in reduction caused by MF conditioning is the same as that caused by somatic conditioning. Given that somatic firing mediates the MF conditioning-induced reduction of G_in, it is expected that the extent of ΔG_in depends on the AP number elicited during MF conditioning. To test this view, we pooled all the data obtained from experiments of rats and WT mice under the control or CPP conditions, and plotted ΔG_in as a function of the AP number in the same CA3-PCs elicited by MF stimulation at different frequencies (Fig.3H, left). There was a close correlation between these two parameters, regardless of the stimulation frequencies, indicating that the number of somatic APs is the key determinant of ΔG_in. In contrast, CA3-PCs of Kcna2 HT mice exhibited little correlation between ΔG_in and the AP number (Fig.3H, right). The number of APs induced by MF conditioning under the experimental conditions that suppressed LTP-IE was not different from the control conditions, arguing against the possibility that lack of LTP-IE may result from a decrease in AP firing during MF conditioning (Fig.3G).

To confirm the involvement of D-type K⁺ current (I_D), we compared the outward K⁺ current (I_K) after the MF conditioning with that under the control conditions. I_D was isolated by digital subtraction of the outward K⁺ currents in the presence of 30 μm 4-aminopyridine (4-AP) from the total outward K⁺ currents under control conditions (Storm, 1988). Figure 3I shows representative I_K traces under control conditions and in the presence of 4-AP at the test potential of −20 mV. Because I_K can be isolated in the presence of the Na⁺ and Ca²⁺ channel blockers that inhibit the conditioning, it was not possible to examine effects of the conditioning on the whole-cell K⁺ currents in the same cell. The statistical mean of I_D was significantly lower in the CA3-PCs that underwent the MF conditioning, compared to that in the control CA3-PCs (Fig.3I).

Downregulation of Kv1.2 is responsible for the non-Hebbian heterosynaptic LTP of PP-EPSPs induced by burst stimulation of MFs

Tsukamoto et al. (2003) showed that a short tetanic stimulation of MFs (100 Hz for 1 s) induces non-Hebbian LTP of PP-fEPSPs, but the cellular mechanism remains elusive. We posed a question of whether downregulation of Kv1.2 is involved in this type of heterosynaptic interaction between MF and PP synapses. Before addressing this issue, we tested whether the same MF conditioning protocol as in Fig.3 minimal stimulation of afferent MFs at 20 Hz for 2 s, induces LTP of PP-EPSPs (Fig.4A). We isolated monosynaptic PP synaptic inputs to a CA3-PC as described in Methods, and monitored the amplitude of PP-EPSPs every 8 s. The baseline PP-EPSPs were in the range of 0.6 to 1.7 mV (1.1 ± 0.1 mV, n = 18). MF conditioning resulted in a significant enhancement of PP-EPSPs (220 ± 16% of control, n = 18, P < 0.005; Fig.4Ab and Ac). Similar results were observed in the absence of PTX (235 ± 24%, n = 6; Fig.4B). In contrast to EPSPs, the same conditioning did not alter PP-evoked excitatory postsynaptic currents (PP-EPSCs; baseline, 17.4 ± 2.4 pA; 117.6 ± 16.9%, n = 14, P = 0.19, Fig.4A). Because somatic recordings of PP-EPSPs that occur at distal dendrites are heavily influenced by active conductance of dendrites (Urban et al. 1998; Kim et al. 2012), the enhancement of PP-EPSPs with little change in concomitant PP-EPSCs could be interpreted as an increase in dendritic intrinsic excitability rather than in the synaptic strength itself. Therefore, the MF conditioning-induced LTP of PP-EPSPs may be closely related to the LTP-IE. We further investigated the signalling mechanisms underlying the heterosynaptic LTP of PP-EPSPs. The MF conditioning-induced LTP of PP-EPSPs was abolished by bath application of nimodipine or PP2, like LTP-IE (Fig.4B). Furthermore, MF conditioning did not significantly alter PP-EPSPs in the CA3-PCs of Kcna2 HT mice (77.2 ± 6.7%, P = 0.11, n = 8), but induced LTP in those of WT mice (232.4 ± 9.6%, P < 0.005, n = 7; Fig.4C). These results indicate that the MF conditioning-induced heterosynaptic LTP of PP-EPSPs is mediated by the same cellular mechanisms as LTP-IE, the downregulation of Kv1.2.

Finally, we studied the possible involvement of NMDARs by testing the effects of three different NMDAR antagonists in the rat CA3-PCs. Whereas neither 20 μm CPP (230 ± 12%, n = 9, P < 0.005) nor 40 μm MK-801 (200 ± 27%, n = 5, P < 0.01) affected the MF conditioning-induced LTP of PP-EPSPs, 40 μm d,l-2-amino-5-phosphonopentanoic acid (APV) significantly suppressed it (Fig.4D). Such discrepancy between the effects of NMDAR blockers could be interpreted in light of the effects of APV on the somatic conditioning-induced changes in G_in (Hyun et al. 2013). Neither 30 μm dl-APV (100.26 ± 4.12%, n = 9, P = 1.00) nor 30 μm d-APV (93.10 ± 2.89%, n = 11, P = 0.64) induced a significant change in G_in after somatic conditioning in the presence of CNQX and PTX. Both dl- and d-APV reduced the baseline G_in (dl-APV, 70.3 ± 7.6%, n = 16, P = 0.001; d-APV, 75.9 ± 5.0%, n = 6, P = 0.008), suggesting that APV occludes the effects of somatic conditioning by reducing the baseline G_in. These results imply that the effect of APV on heterosynaptic LTP of PP-EPSPs is not directly related to the blockade of NMDARs.

The distribution of Kv1.2 is polarized to the distal apical dendrites of CA3-PCs

Previously, we reported that LTP-IE requires backpropagating AP-induced dendritic Ca²⁺ signalling and is abolished by surgical truncation of distal apical dendrites (Hyun et al. 2013). Given that the effects of somatic conditioning are mediated by downregulation of Kv1.2, the preferential effect of somatic conditioning on PP-EPSPs raises the possibility of a polarized distribution of Kv1.2 in the apical dendrites of CA3-PCs. We further investigated possible polarization of I_D activity. We tested whether I_D is differentially affected by a local puff application of 100 μm 4-AP to either the proximal or the distal half of the apical dendrites (Fig.5Aa). We estimated the spatial profile of 4-AP concentration around the puffing pipette tip based on the fluorescence profile of Alexa Fluor 488 under the same conditions. The line profile of fluorescence measured at 4 s after ejection indicates that the puffing area, which is defined as the area where the drug concentration is higher than 20% of the maximum, spans 70 μm from the pipette tip (Fig.5Ab; see Methods). Outward K⁺ current was elicited by a somatic step depolarization to −20 mV in the presence of a cocktail of synaptic blockers plus inward current blockers. We compared the effectiveness of the suppression of outward K⁺ current caused by a local puff of 4-AP to a distal dendritic region (ca 230 μm from the soma) and to a proximal dendritic region (∼70–100 μm from the soma) in the same CA3-PC (Fig.5Ad). Despite the voltage clamp being less effective in the distal region than the proximal one (Baker et al. 2011), the distal puff caused a stronger suppression of K⁺ current than the proximal puff, indicating that the 4-AP-sensitive current was significantly larger in the distal puff (Fig.5Ac and d; distal puff, 127 ± 16.2 pA, n = 6; proximal puff, 66 ± 11 pA, P = 0.014; paired t test, n = 6). The 4-AP-sensitive current was not detected when the puffing pipette was placed at the point 300 μm away from the soma, probably because of the spatial clamp problems (data not shown). These results support the hypothesis that the majority of D-type K⁺ (K_D) channels are expressed in the distal half of apical dendrites in CA3-PCs.

Although it has been reported that dendrites of CA3-PCs are especially heavily immunostained in the hippocampus (Sheng et al. 1994), surface expression of Kv1.2 on the plasma membrane, which should be more correlated to the I_D activity, has not been studied. To test whether the surface expression of dendritic Kv1.2 is polarized in CA3-PCs, we raised an antibody targeting the extracellular epitope of Kv1.2, and tested its specificity by comparing immunoreactivity between hippocampi from WT and KO mice (Fig.5B). To detect Kv1.2 subunits on the plasma membrane, the immunolabelling should be done on live cells. The live cell immunolabelling was not possible on a brain slice, and thus we examined the surface expression of Kv1.2 in cultured putative CA3-PCs (DIV22, see Methods). We labelled the surface Kv1.2 molecules on live cultured rat CA3-PCs, and then labelled MAP2, a dendritic marker, after permeabilization of the same cells. An exemplar immunograph of a putative CA3-PC is shown in Fig.5Ca. The distribution of Kcna2 puncta was determined on the mask image of the same cell (Fig.5Cb). From this figure, we estimated the density of Kcna2 puncta as a function of their distance from the soma (see Methods, Fig.5Cc). The density profiles of Kcna2 puncta estimated from 16 cells are summarized in Fig.5Cd. The mean density profile indicates that the density of Kcna2 puncta is higher in the distal dendritic regions than the proximal regions.

Expression of Kv1.2 subunits in both of somatodendritic and axonal compartments complicates the interpretation of immunohistochemistry data (Sheng et al. 1994; Grosse et al. 2000). To differentiate whether Kcna2 puncta are associated with pre- or postsynaptic compartments, we tried to identify the Kv1.2 molecules associated with the dendritic plasma membrane using immunogold electron microscopy. We used adult rats for this experiment, because the preferential enhancement of PP-EPSPs after the somatic conditioning was found in the CA3-PCs of adult rats like in those of juvenile rats (Fig.6A; PP-EPSPs, 198.5 ± 19.4%, P < 0.005, n = 6; A/C-EPSPs, 91.2 ± 13.7%, P = 0.74, n = 4). The specificity of immunogold staining was confirmed using Kv1.2 KO mice (Fig.6B). The immuno-EM study on WT rats revealed that many immunogold particles are closely associated with the plasma membrane of dendrites in the SLM of the CA3 region (Fig.6Cd). For quantitative analysis, we counted the number of immunogold particles attached to the plasma membrane (<25 nm) of somata or dendrites in different strata of the hippocampal CA3 region. The immunogold particle density, expressed as the number of particles per unit length of somatic or dendritic plasma membrane, was highest in the SLM among different strata of the CA3 region (Fig.6D), suggesting that the surface expression of Kv1.2 on the somatodendritic compartment is polarized to distal apical dendrites in CA3-PCs.

Compartmental simulation of the specific enhancement of PP inputs by downregulation of Kv1.2

Figures 1 and 2 show that a short train of somatic firing and consequent downregulation of Kv1.2 result in the preferential enhancement of synaptic inputs to distal apical dendrites. We tested whether this result can be reproduced in the NEURON simulation. Recently, Kim et al. (2012) elucidated the densities of ionic currents (fast Na⁺ current (I_Na), A-type K⁺ current (I_A) and delayed rectifier K⁺ current (I_DR)) along the apical dendrites of CA3-PCs. The density of I_D, however, has not been previously studied, and thus we assumed that the density of I_D conductance (G_Kd) is 10 times higher in the distal half of apical dendrites than other subcellular regions of a CA3-PC according to Fig.6D. We incorporated conductance densities of these ionic currents in the reconstructed morphology of a CA3-PC (Fig.7A). The conductance values reported by Kim et al. (2012) and G_Kd were adjusted such that the simulated somatic outward K⁺ current best reproduces the experimental somatic recordings of the outward K⁺ current under the same depolarization steps. Figure 7Ba shows the conductance densities in apical dendrites used in the simulation plotted as a function of the distance from the soma. In Fig.7Bb (upper panel), the simulated K⁺ currents were superimposed on the averaged traces of outward K⁺ currents that were experimentally recorded under control conditions (from a holding potential of −70 mV to depolarization steps ranging from −50 mV to −10 mV in 10 mV increments). The outward K⁺ currents after somatic conditioning were simulated by spatially uniform reduction of G_Kd to 10% of the control (lower panel of Fig.7Bb). The membrane resistivity (R_m) was set to 220 KΩ cm² to reproduce the typical value for experimental estimates of G_in (3.5 nS). Figure 7Bc shows the simulated voltage responses to the current injection of −30 and +10 pA at the soma under control conditions and after reducing G_Kd to 10% of the control, which reduced G_in from 3.53 to 2.98 nS. The latter is close to the mean value for G_in measured in the presence of 50 nm DTX-I, a selective blocker of Kv1.1/2/6-containing channels (Hyun et al. 2013). The density of I_Na (G_Na) was set according to Kim et al. (2012) (Fig.7Ba).

The AMPAR and NMDAR currents were simulated by alpha and bi-exponential functions, respectively (Baker et al. 2011). We simulated synaptic responses evoked by synaptic current injection at different locations along the apical dendrites of the CA3-PC model (denoted in Fig.7A; the proximal, intermediate and distal dendritic sites for simulation of MF, A/C and PP synapses). We assumed that minimal stimulation of PP or A/C fibres activates two synapses at different dendritic branches. The baseline synaptic conductances for MF, A/C and PP synapses were set such that the somatic EPSP amplitudes are 1.3, 0.4 and 0.4 mV, respectively (see Methods and Table 1 for synaptic parameters in detail). The adequacy of synaptic parameters was tested by superimposing the simulated PP-EPSP traces on the experimental recordings of PP-EPSPs (Fig.7Cb; reproduced from the right traces in Fig.8Ab). Under the control distribution of G_Na along the apical dendrites (Fig.7Ba), the uniform reduction of G_Kd to 10% of the control value enhanced the PP-EPSP from 0.4 mV to 0.6 mV, but affected neither MF- nor A/C-EPSPs (Fig.7Ca). This effect of the G_Kd reduction on the PP-EPSP was abolished by 50% reduction of G_Na (Fig.7Ca). In addition, when the G_Na was kept low (50% of the control), the same G_Kd reduction (to 10%) caused only a small increase in the PP-EPSP amplitude (0.07 mV; red dashed line in Fig.7Cb) compared to the mean increase in PP-EPSPs induced by the conditioning (0.3 mV; Fig.8Ab). Therefore, even though the reduction of G_Kd may reinforce the passive synaptic normalization (Jaffe & Carnevale, 1999), the enhancement of PP-EPSPs caused by somatic conditioning cannot be fully explained by an alteration of passive properties alone. Nevertheless, the baseline passive normalization may play an essential role in the activation of dendritic Na⁺ channels at distal dendrites upon the G_Kd reduction, because high input impedance of distal dendrites enables a unitary synaptic input to induce a strong local dendritic depolarization close to the threshold for Na⁺ channel activation (Jaffe & Carnevale, 1999; Baker et al. 2011). The dendritic unitary EPSPs at the A/C and PP synaptic sites are compared under the control conditions in Fig.7Cc, showing that the dendritic PP-EPSP is higher than the dendritic A/C-EPSP at their synaptic sites (3 vs. 16 mV). Even when the A/C-EPSPs were temporally summated by synaptic activations five times at 50 Hz, the summation did not elicit a dendritic Na⁺-spike under the low G_Kd condition, whereas the summation of PP-EPSPs did (Fig.7Db). Accordingly, like unitary EPSPs, the temporal summation of somatic PP-EPSPs was preferentially enhanced in a G_Na-dependent manner by the reduction of G_Kd (Fig.7Da). These simulation results suggest that activation of dendritic Na⁺ channels is required for the preferential enhancement of PP-EPSPs induced by downregulation of G_Kd.

To predict the change of a somatic PP-EPSP caused by generation of a dendritic Na⁺ spike, we simulated somatic PP-EPSPs in response to a gradual increase in the synaptic conductance by a step of 0.05 nS at the PP synaptic site on the branch numbered ‘5’ under three different conditions of ionic conductances (control, 10% G_Kd or 50% G_Na) with other simulation parameters kept constant (Fig.7E). Under control conditions, a dendritic Na⁺-spike and consequent step-like enhancement of somatic PP-EPSP occurred when the synaptic input was increased from 0.2 nS to 0.25 nS. Once a dendritic Na⁺-spike was generated at a single branch, there was no further increase in the dendritic and somatic PP-EPSPs by an increase in the synaptic conductance in the same branch, resulting in the all-or-none behaviour of the somatic PP-EPSPs. Such all-or-none behaviour of PP-EPSPs was abolished by the reduction of G_Na to 50%. After the reduction of G_Kd to 10%, the dendritic Na⁺-spike was elicited at lower synaptic input (0.2 nS) than the control conditions.

Somatic conditioning-induced LTP of PP-EPSPs depends on local activation of dendritic Na⁺ channels

The generation of dendritic Na⁺-spikes is more vulnerable to partial block of Na⁺ channels using 10–20 nm tetrodotoxin (TTX) than axonal conduction (Mackenzie & Murphy, 1998). Previously, we confirmed this result in the CA3-PCs by showing that 10 nm TTX significantly attenuates backpropagating AP-induced Ca²⁺ transients at the distal apical dendritic region but has little effect on the concomitant somatic Ca²⁺ transients (Hyun et al. 2013). The simulation results in Fig.7 predict that the potentiation of PP-EPSPs caused by downregulation of Kv1.2 depends on the activation of dendritic Na⁺ channels at the local PP synaptic site. To test this prediction, we examined the effect of 10 nm TTX on PP-EPSPs under control conditions and after somatic conditioning. We employed the minimal stimulation technique to minimize spatial summation of local PP-EPSPs, because a dendritic Na⁺ spike is predicted to be generated by a local EPSP at individual synaptic sites rather than by spatial summation of local EPSPs at a conduit dendrite. As expected from Fig.2B, somatic conditioning enhanced EPSPs evoked by minimal stimulation of PPs (baseline, 0.44 ± 0.04 mV; 163 ± 12%, n = 11, P < 0.005; Fig.8Aa). This enhancement of PP-EPSPs was abolished by 10 nm TTX, which brought the EPSP amplitude back to the baseline level (after conditioning, 0.72 ± 0.06 mV; TTX, 0.42 ± 0.01 mV; P < 0.005, Fig.8Aa and b). In contrast, neither somatic conditioning nor 10 nm TTX had significant effects on PP-EPSCs (baseline, 8.1 ± 0.3 pA; after conditioning, 102.8 ± 5.7%; TTX, 100.7 ± 4.5%, n = 11, P = 0.8; Fig.8Ac). These results are consistent with the simulation results, and support the hypothesis that the potentiation of PP-EPSPs caused by somatic conditioning depends on activation of dendritic Na⁺ channels.

The simulation results in Fig.7E predict that somatic PP-EPSPs will display an all-or-none behaviour depending on the generation of a dendritic Na⁺ spike as the PP-synaptic input to a single dendritic branch increases. We tested whether this simulation result can be experimentally reproduced. To this end, we had to avoid activating synapses on multiple dendritic branches, because the synaptic input-dependent all-or-none behaviour of EPSPs cannot be observed when EPSPs from different branches are spatially summated (Polsky et al. 2004). To stimulate a single dendritic branch, we visualized dendrites of a CA3-PC in SLM using a fluorescent dye (Alexa Fluor 488) that was introduced via a somatic patch pipette, and then positioned a monopolar glass electrode for extracellular stimulation close to a single dendritic branch (typically 20 μm; Fig.8B). This focal stimulation elicited synaptic responses that were completely blocked by CNQX, ruling out the direct stimulation of a dendritic branch (inset in Fig.8B). A multi-Gaussian function was best fitted to the combined histogram of EPSC amplitudes collected from eight cells by assuming that a unitary EPSC amplitude is 2.1 pA (lower panel of Fig.8B), suggesting that the focal stimulation activates multiple units of synapses. The probability that the focal stimulation fails to elicit synaptic responses depended on the stimulation voltage (V_stim), but the dependence was not altered by 10 nm TTX arguing against its effect on presynaptic compartments (Fig.8F; repeated measures ANOVA, F_(1,6) = 3.26, P = 0.12). For the following analyses of synaptic responses, only non-failure events were taken into account. Figure 8C shows the representative EPSC and EPSP responses to the increases in V_stim by a step of 0.2 V. The scatter plot in each panel shows individual synaptic responses (Fig.8Ca and b). The mean EPSC amplitude was gradually increased by the increase in V_stim, and was not altered by 10 nm TTX at each V_stim (Fig.8Ca). In contrast, somatic PP-EPSPs showed a step-like increase in the mean amplitude between V_stim of 2.2 and 2.4 V, and this all-or-none response was abolished by 10 nm TTX. The mean EPSPs as a function of EPSC amplitude at the same V_stim values are plotted in Fig.8Cc, showing that there is a step-like increase in the EPSP between the EPSCs of 4.6 and 5.7 pA under control conditions but not in the presence of 10 nm TTX. Because there was a cell-to-cell variation in the threshold stimulation voltage (V_thr) at which the step-like increase in the EPSP amplitude occurs (typically ∼4–5 V), the synaptic responses from different cells were plotted as a function of V_stim – V_thr (Fig.8D). Each point in these plots was compared to the synaptic response evoked by one step (0.2 V) lower V_stim. Both EPSC and EPSP were significantly increased at the step to the V_thr (EPSC, P = 0.045; EPSP, P = 0.015; n = 4). Such a significant increase in the EPSP was abolished by 10 nm TTX (P = 0.76), whereas that in the EPSCs was not affected (P = 0.015). The effect of TTX is re-illustrated in the EPSP vs. EPSC plot in Fig.8Dc. These results suggest that activation of dendritic Na⁺ channels is required for the step-like increase in the EPSP amplitude in response to the gradual increase in the synaptic input. Finally, we examined whether the V_thr is lowered by somatic conditioning. Consistent with the simulation results, the EPSP vs. V_stim relationship was left-shifted by the somatic conditioning (Fig.8Ea). The effects of somatic conditioning are summarized in Fig.8Eb by plotting the mean amplitudes of EPSPs as a function of V_stim – V_thr, where V_thr is a threshold stimulation voltage of each cell under control conditions. This plot shows the somatic conditioning lowers the V_thr, suggesting that the magnitude of synaptic inputs required for the generation of a Na⁺-spike at distal apical dendrites is lower in the conditioned CA3-PCs than in the naïve CA3-PCs.

Discussion

Our previous and present studies demonstrated that repetitive somatic firing at 10–20 Hz in CA3-PCs, regardless of whether it is elicited by injection of suprathreshold current pulses (somatic conditioning) or by burst stimulation of afferent MFs (MF conditioning), results in the same consequences: sustained reduction of G_in (Hyun et al. 2013) (Fig.3) and non-Hebbian LTP of PP-EPSPs but not PP-EPSCs (Figs 1 and 4). Both of these effects depended on the activation of L-type Ca²⁺ channels and PTK and normal expression of Kv1.2 but not on NMDARs (Hyun et al. 2013) (Figs 1, 3 and 4). Furthermore, the specific enhancement of EPSPs with little alteration of EPSCs implies that the enhancement of dendritic excitability rather than the synaptic strength itself underlies the somatic and MF conditioning-induced LTP of PP-EPSPs. Therefore, we conclude that repetitive somatic firing and consequent downregulation of Kv1.2 mediate the burst MF input-induced non-Hebbian heterosynaptic LTP of PP-EPSPs. Nevertheless, we cannot rule out involvements of other ion channels in the LTP-IE and the heterosynaptic interactions, which may explain why the simulated ΔG_in (15.6%, Fig.7Bc) is smaller than the actual one (30%). Despite a line of evidence that LTP-IE underlies the heterosynaptic LTP of PP-EPSP, it is puzzling why the potentiation of PP-EPSPs developed more rapidly than the reduction of G_in (Fig.4). It is unlikely that metabotropic role of neurotransmitters or aberrant synaptic stimulation are responsible for the discrepancy in the developing time courses, because not only MF conditioning but also somatic conditioning resulted in such enhancement of PP-EPSPs in the early phase before the full development of LTP-IE (Fig.1). Interestingly, the enhancement of PP-EPSPs in the early phase after conditioning was more prominent when the baseline PP-EPSP was higher (2 mV and 1 mV for Fig.1 and 4, respectively) than when PP-EPSPs were evoked using minimal stimulation techniques (0.5 mV in Fig.2B). Given that stronger stimulation recruits higher number of synapses and results in spatial summations of synaptic inputs arriving at multiple branches of distal apical dendrites, the early phase enhancement might involve excitability changes related to spatial summation of multiple synaptic inputs that are not reflected in the G_in monitoring.

Possibility for conditioning of a CA3-PC by synaptic inputs other than MFs

Given that the number of somatic APs elicited within 1 or 2 s was regarded as the key factor in determining the excitability change in response to the MF stimulation (Fig.3H), it is plausible that any kind of synaptic input, as long as it induces such repetitive somatic firing, can induce LTP-IE. The anatomical notion that CA3-PCs receive extensive inputs from A/C synapses raises the possibility that a CA3-PC may be primed by somatic firing caused by spatiotemporal summation of massive A/C inputs. Nevertheless, Tsukamoto et al. (2003) failed to induce heterosynaptic LTP of PP-EPSPs by HFS of A/C fibres. Consistent with Tsukamoto et al. (2003), we found that the changes in ΔG_in induced by HFS of A/C fibres was not significant even at very high stimulation intensity (ΔG_in = 8.6 ± 2.5%, n = 5, P > 0.05; data not shown). The mechanisms underlying the different efficiency between MF and A/C fibre stimulations remain to be elucidated. We imagine that the difference in the firing patterns elicited by HFS of MFs and A/C fibres may be one cause of the discrepancy. Although HFS of A/C fibres was not so efficient as MF stimulation in the induction of LTP-IE, this result does not guarantee that A/C synaptic inputs do not induce LTP-IE in vivo. It is still possible that LTP-IE could be induced by A/C synaptic inputs as long as they elicit the AP firing pattern similar to the somatic conditioning (10 Hz for 2 s). Therefore, this possibility may be more precisely evaluated by A/C fibre stimulation mimicking A/C synaptic input patterns in vivo.

Cellular mechanisms underlying the specific enhancement of PP inputs by downregulation of Kv1.2

Recent studies using direct dendritic patch recording and the glutamate uncaging on apical dendrites of CA3-PCs have showed that the proximal apical dendrites of CA3-PCs exhibit lower Na⁺ channel density than the axosomatic compartment (Kim et al. 2012; Makara & Magee, 2013). Surprisingly, however, Kim et al. (2012) showed that the Na⁺ channel density at distal dendrites is as high as in the soma. As a result, the threshold current required for eliciting dendritic Na⁺-spikes was lower at distal dendrites than proximal ones (Kim et al. 2012). Previous studies in CA3, CA1 and neocortical PCs revealed that dendritic spikes involve voltage-dependent Na⁺ and Ca²⁺ channels and/or NMDARs expressed on dendrites (Golding & Spruston, 1998; Larkum et al. 2009; Makara & Magee, 2013). Low-threshold dendritic K⁺ conductance, which opposes these depolarization-activated channels, potentially regulates the threshold of dendritic spikes (Golding et al. 1999). I_D is uniquely positioned for the regulation of somatic or dendritic spike threshold, because it is activated by low-voltage depolarization and already slightly active in the resting membrane potential range (Storm, 1988; Golding et al. 1999; Hyun et al. 2013).

Our simulation results show that PP-EPSPs can be preferentially enhanced even if G_Kd is reduced over all the dendritic arbor by the same proportion, and suggest that concerted actions of passive normalization and polarized distributions of ionic channels are responsible for this preferential enhancement. The structural feature of CA3-PCs of severe tapering of dendrites from proximal to distal terminations provides high input impedance at distal dendrites, and thereby a unitary PP synaptic input is expected to induce a higher local depolarization than proximal synaptic inputs (passive normalization; Fig.7C) (Jaffe & Carnevale, 1999; Baker et al. 2011). Furthermore, the Na⁺-to-K⁺ current ratio is higher at the distal region of apical dendrites than other dendritic regions (Kim et al. 2012). These two factors may allow the baseline PP-EPSPs at the synaptic sites to be closer to the threshold for the dendritic Na⁺ channel activation than at proximal synaptic sites. Finally, the threshold for a local dendritic Na⁺ spike may be more lowered at distal dendrites by a uniform reduction of G_Kd because of higher expression of Kv1.2 (Figs 5 and 6). From these reasons, the reduction in G_Kd may enhance the possibility for Na⁺ channels to be activated preferentially at distal apical dendrites.

Involvement of NMDARs in the MF conditioning

The present study elucidated the cellular mechanisms underlying the findings of Tsukamoto et al. (2003). Nevertheless, our results are not consistent with Tsukamoto et al. (2003) regarding the issue of NMDAR involvement. Tsukamoto et al. (2003) concluded that activation of NMDARs mediates the heterosynaptic interaction, based on the observation that APV, an NMDAR blocker, abolished the heterosynaptic interaction. In contrast, the present study showed that other NMDAR blockers, CPP and MK-801, had little effect on MF conditioning-induced LTP of PP-EPSPs (Fig.4D). This discrepancy may result from an adverse effect of APV. We have previously examined the effect of APV on G_in of CA3-PCs. To our surprise, APV alone gradually reduced the baseline G_in by ca 30%, and occluded further reduction of G_in after somatic conditioning (Fig. S2 in Hyun et al. 2013). MK-801 and CPP, however, did not alter the baseline G_in (data not shown) nor the effects of somatic and MF conditionings on G_in and PP-EPSPs (Figs 1B, 3D and 4D). These results indicate that the occlusion effects of APV are not caused by blocking of NMDARs but by an off-target effect of APV. Recently, however, it has been reported that potentiation of NMDA-EPSC at MF synapses enhances E-S coupling at MF synapses (Hunt et al. 2013). The enhancement of E-S coupling of MF inputs can be important for heterosynaptic effects given that repetitive somatic firing mediates burst MF input-induced heterosynaptic LTP of PP-EPSPs. Therefore, NMDARs may be indirectly involved in the MF-induced heterosynaptic LTP of PP-EPSPs.

Physiological implications

Many types of learning accompany an increase in neuronal excitability, and it is evolutionarily well conserved, suggesting that it is an essential part of memory formation together with changes in synaptic strength (Zhang & Linden, 2003; Disterhoft & Oh, 2006; Matthews et al. 2008). However, an unopposed increase in excitability will not only saturate the capacity for information storage but also eventually result in epileptiform activity. In vivo examination of learning-associated excitability changes revealed that such learning-associated excitability changes disappear over days (Moyer et al. 1996), implying that there may be mechanisms that reverse the increased excitability of CA3-PCs. It remains to be elucidated whether downregulation of Kv1.2 in CA3-PCs is also restored and, if so, how it is restored.

It is noteworthy that burst MF inputs do not directly potentiate the PP synaptic strength but enhance the dendritic responses at the PP synaptic sites. PP–CA3 synapses display the Hebbian property (McMahon & Barrionuevo, 2002). Dendritic spikes or APs backpropagated from the soma facilitate the induction of LTP at Hebbian synapses. Given that downregulation of Kv1.2 enhances E-S coupling and facilitates dendritic spikes (Fig.1C), it is expected that it may also facilitate LTP of synaptic strength upon arrival of high frequency PP synaptic inputs afterwards, and thus lower the threshold of synaptic inputs required for induction of LTP or metaplasticity. Assuming the metaplastic role of MF conditioning and consequent downregulation of Kv1.2, we imagine that once a subset of CA3-PCs experiences a short period of high frequency MF inputs, subsequent PP synaptic inputs will be preferentially stored in the subset of CA3-PCs. In line with this prediction, it has been found that behavioural expression outlasts learning-related intrinsic plasticity, suggesting that the mnemonic role of intrinsic plasticity consists in metaplastic changes of synapses rather than in coding and maintaining the memory itself (Sehgal et al. 2013).

Furthermore, it should be noted that the priming of a CA3-PC does not modulate A/C-EPSPs. A/C fibres constitute an autoassociative network that may subserve pattern completion of a partial cue. The lack of long-term modulation of A/C synapses may prevent the CA3 autoassociative network from being vulnerable to MF inputs. Therefore, MF inputs may guide CA3-PCs to store new memories at PP synapses with no deterioration of the pre-existing autoassociative network. It remains to be elucidated whether mutation of Kv1.2 and consequent lack of MF-induced heterosynaptic modulation of PP synaptic inputs cause any impairments in behavioural pattern separation task and/or pattern completion-based recall.

Glossary

A/C

associational/commissural

ACSF

artificial cerebrospinal fluid

AP

action potential

4-AP

4-aminopyridine

APV

2-amino-5-phosphonopentanoic acid

CA3-PC

CA3 pyramidal cell

CPP

3-(2-carboxypiperazin-4-yl)propyl-1-phosphonic acid

DIV

days in vitro

EM

electron microscopy

EPSC

excitatory postsynaptic current

EPSP

excitatory postsynaptic potential

E-S coupling

EPSP-to-spike coupling

fEPSP

field EPSP

GC

granule cell

G_in

input conductance

ΔG_in

relative changes of G_in

HFS

high-frequency stimulation

HT

heterozygous knock-out

I_D

D-type K⁺ current

K_D channel

D-type K⁺ channel

KO

knock-out

LTP

long-term potentiation

LTP-IE

LTP of intrinsic excitability

MF

mossy fibre

NMDA

N-methyl-d-aspartate

NMDAR

NMDA receptor

PKA

protein kinase A

PP

perforant path

PP-EPSP

perforant path-evoked excitatory postsynaptic potentials

PTK

protein-tyrosine kinase

PTX

picrotoxin

SLM

stratum lacunosum moleculare

TA

temporoammonic

TTX

tetrodotoxin

WT

wild-type

Additional information

Competing interests

The authors declared no competing financial interests.

Author contributions

J.H.H., K.E. and S.H.L. conceived and designed the experiments. J.H.H., K.E., K.H.L., J.Y.B. and S.H.L. performed the experiments. J.H.H., K.E., K.H.L., Y.C.B., S.Y.K., M.H.K., W.K.H. and S.H.L. analysed the data. J.H.H., K.E. and S.H.L. wrote the paper.

Funding

This study was supported by a grant from the National Research Foundation of Korea (grant number 20120009135 and NRF-2014R1A2A2A01003167) and Seoul National University Hospital (2015). J.H.H. and K.E. are graduate students supported by BK21plus.