Depression biased non-Hebbian spike-timing-dependent synaptic plasticity in the rat subiculum

Abstract

The subiculum is a structure that forms a bridge between the hippocampus and the entorhinal cortex (EC), and plays a major role in the memory consolidation process. Here, we demonstrate spike-timing-dependent plasticity (STDP) at the proximal excitatory inputs on the subicular pyramidal neurons of juvenile rat. Causal (positive) pairing of a single EPSP with a single back-propagating action potential (bAP) after a time interval of 10 ms (+10 ms) failed to induce plasticity. However, increasing the number of bAPs in a burst to three, at two different frequencies of 50 Hz (bAP burst) and 150 Hz, induced long-term depression (LTD) after a time interval of +10 ms in both the regular-firing (RF), and the weak burst firing (WBF) neurons. The LTD amplitude decreased with increasing time interval between the EPSP and the bAP burst. Reversing the order of the pairing of the EPSP and the bAP burst induced LTP at a time interval of −10 ms. This finding is in contrast with reports at other synapses, wherein pre- before postsynaptic (causal) pairing induced LTP and vice versa. Our results reaffirm the earlier observations that the relative timing of the pre- and postsynaptic activities can lead to multiple types of plasticity profiles. The induction of timing-dependent LTD (t-LTD) was dependent on postsynaptic calcium change via NMDA receptors in the WBF neurons, while it was independent of postsynaptic calcium change, but required active L-type calcium channels in the RF neurons. Thus the mechanism of synaptic plasticity may vary within a hippocampal subfield depending on the postsynaptic neuron involved. This study also reports a novel mechanism of LTD induction, where L-type calcium channels are involved in a presynaptically induced synaptic plasticity. The findings may have strong implications in the memory consolidation process owing to the central role of the subiculum and LTD in this process.

Introduction

According to the Hebbian theory of plasticity, long-term potentiation is induced if the postsynaptic neuron fires within a narrow time window after the presynaptic firing (Hebb, 1949). This coincident activity modifies the synaptic strength that is manifested as a change in the amplitude or slope of the EPSPs recorded at the postsynaptic neuron. This change in synaptic strength depends upon the relative timing of pre- and postsynaptic activities (Bi & Poo, 1998) and is known as ‘spike-timing-dependent plasticity’ (Song 2000). Synaptic plasticity is widely believed to underpin memory formation (Martin et al. 2000). STDP is a form of synaptic plasticity that can explain both strengthening and weakening of synapses (Bi & Poo, 2001; Dan & Poo, 2006). Certain kinds of memories are formed in the hippocampus (Andersen, 2007) and stored in the cortex (Stern et al. 2001). Due to its strategic location, the subiculum funnels information from the hippocampus to various cortical sub-regions such as the EC layers IV and V, perirhinal cortex (Witter et al. 1989), amygdala and thalamus (Canteras & Swanson, 1992). Lesion studies have established the role of the subiculum in long-term spatial learning (Morris et al. 1990) and performance of working memory tasks (Galani et al. 1998) and, with CA1, it executes mnemonic functions of the brain (Deadwyler & Hampson, 2004). The majority of the neurons in the subiculum respond with burst(s) of multiple action potentials upon suprathreshold excitation (Staff et al. 2000). These neurons are different from typical regular firing neurons in certain aspects; they have a larger neuronal spike after-depolarization (ADP) and a higher sag ratio (van Welie et al. 2006). This difference in the intrinsic properties leads to diverse synaptic integration properties of the excitatory neurons in the subiculum (van Welie et al. 2006). Furthermore, the mechanism of synaptic plasticity varies with the intrinsic properties of the neurons as well (Wozny et al. 2008).

Although some properties of the subicular pyramidal neurons are well studied, other aspects like the expression profile and kinetics of various ion channels, including the magnesium unblock of NMDA receptors, are not well understood, which may influence the STDP plot of a synapse (Shouval et al. 2002; Johnston et al. 2003; Kampa et al. 2004). In the absence of this information, it is very difficult to predict and compare the STDP plot of the synapses on the subicular neurons with other synapses. Most of the synapses in different hippocampal subfields have been well studied for synaptic plasticity (Dudek & Bear, 1992; Bliss & Collingridge, 1993; Nicoll & Schmitz, 2005), and STDP (Debanne et al. 1998; Wittenberg & Wang, 2006; Astori et al. 2010). However, there are very few reports of synaptic plasticity in the proximal excitatory inputs on the subicular pyramidal neurons that predominantly originate from the CA1 (Kokaia, 2000; Fidzinski et al. 2008; Wozny et al. 2008) and STDP has not yet been reported for this synapse.

In this study, we asked if the proximal excitatory inputs on the subicular pyramidal neurons express STDP. Further, we explored the differences in the mechanism of such plasticity between different types of neurons, following earlier reports that the mechanism of synaptic plasticity in the subiculum depends on the type of postsynaptic neuron involved (Fidzinski et al. 2008; Wozny et al. 2008). Recently, it was demonstrated that the subiculum can generate gamma rhythms independently (Jackson et al. 2011); hence, we paired a single EPSP with a burst of three bAPs at 50 Hz (bAP burst). Interestingly, we found that causal pairing induced long-term depression (LTD), while anti-causal pairing induced long-term potentiation (LTP) in the subicular pyramidal neurons. In the WBF neurons, the negative side of the STDP curve showing LTP was slightly narrower than the positive side that showed LTD. The causal pairing-induced LTD was also observed in the RF neurons; however, there were prominent differences in the plasticity mechanism between these two neuronal sub-types.

Methods

Slice preparation

All the experiments were approved by the ‘Animal Ethics and Welfare Committee’ of the Indian Institute of Science, Bangalore, India, and all the guidelines of this committee were followed in all the experiments. Halothane-anaesthetized rats were decapitated and slices were cut in sucrose based ACSF (sACSF) (in mm): sucrose 250, KCl 2.5, MgCl₂ 3, CaCl₂ 1, NaHCO₃ 25, NaH₂PO₄ 1.25, glucose 10, ascorbic acid 0.4, bubbled with carbogen (95% O₂ + 5% CO₂), pH 7.4. In a subset of experiments, the slicing procedure was performed using NaCl-based ACSF (nACSF) where sucrose was completely replaced with 125 mm NaCl. The experiments were performed on 300 μm thick acute transverse hippocampal slices from Wistar rats (15–18 days old). The slices were cut using Leica VT 1200S vibratome, incubated at 34°C for 15 min, and thereafter maintained at room temperature (23 ± 1°C). The incubation solution contained (in mm): NaCl 125, KCl 2.5, MgCl₂ 2, CaCl₂ 2, NaHCO₃ 25, NaH₂PO₄ 1.25, glucose 25, ascorbic acid 1.0, pH 7.4, bubbled with carbogen throughout the incubation period.

Data acquisition and analysis

During the experimental recordings the slices were continuously superfused with carbogen saturated ACSF (in mm): NaCl 125, KCl 2.5, NaH₂PO₄ 1.25, NaHCO₃ 26, MgCl₂ 1, CaCl₂ 2, glucose 25, pH 7.4. Picrotoxin (10 μm) and CGP-55845 (1 μm) were added to block GABAergic inhibitory inputs. All the experiments were conducted at a bath temperature between 32 and 34°C. The patch pipettes (pipette resistance 3.5–5 MΩ) were filled with solution containing (in mm): potassium gluconate 125, KCl 20, Hepes 10, sodium phosphocreatine 10, Mg-ATP 5, Na-GTP 0.5, EGTA 0.2, pH was adjusted to 7.3 with KOH. In experiments with pharmacological blockers, 50 μm dl-AP5, 25 μm nifedipine, 50 μm verapamil hydrochloride, or 10 μm bicuculline were added to the bath solution. The physiological role of intracellular calcium was checked in some experiments by partially replacing potassium gluconate with 10 mm tetra-potassium salt of BAPTA in the internal solution.

Biocytin (0.5%) was included in the patch pipette to obtain information about location and morphology of the recorded neurons. After the experiment, each slice was incubated in 4% paraformaldehyde containing 0.1% glutaraldehyde in 0.1% PBS (pH 7.4) for at least 24 h. Thereafter, sections were washed three times in PBS for 10 min each. To eliminate endogenous peroxidase activity, slices were washed with 50% ethanol (v/v) for 10 min, 70% ethanol (v/v) for 15 min, and 50% ethanol (v/v) for 10 min followed by three PBS washes of 10 min duration each. Slices were permeabilized using 0.5% Triton X-100 in PBS for 1 h, which was then washed from the slices using PBS in three steps of 10 min each. Thereafter, slices were incubated in avidin–horseradish peroxidase (avidin-HRP) solution (Vectastain ABC-kit, Vector Labs, USA) for 3 h and then washed in three steps of 10 min each with PBS. Slices were then incubated in nickel intensification solution consisting of PBS with 0.05% of 3′-3-diaminobenzidine tetrahydrochloride (DAB), CoCl₂ (0.025%) and NiNH₄SO₄ (0.02%), (Vectastain ABC-kit, Vector Labs, USA) for 15 min, followed by addition of H₂O₂ (Vectastain ABC-kit, Vector Labs, USA) to a final concentration of 0.1%. This step was carefully monitored until the development of dark brown cells against a light brown background of the slice. Immediately after this the slices were washed in 0.1% PBS again, to stop the development of background colour. Slices were mounted on glass slides using chicken egg albumin and Vecta shield (Vector Labs).

Electrophysiological recordings were performed using Multiclamp 700B amplifier (Molecular Devices, USA) and digitization was done with Digidata 1440A analogue–digital converter (Molecular Devices, USA), using Clampex 10 software (Molecular Devices, USA). Data were acquired at 20 kHz, and filtered at 6 kHz using low pass Bessel filter.

All the experiments were performed in the whole cell current clamp mode. Throughout the experiment, the cells were held at a membrane potential of −60 mV by constant current injection. The resting membrane potential (RMP) was monitored by removing the current injection briefly every third minute and the RMP values plotted in all the figures correspond to these measurements. The data were discarded if RMP depolarized to more than −55 mV or changed by more than 3–4 mV from the initial value. The series resistance (R_s) was compensated throughout the experiment and monitored every third minute. The experiments were terminated if the R_s changed by more than 20% of the initial value or crossed 15 MΩ. The input resistance (R_in) was monitored throughout the experiment, twice every minute, by injecting a 50 pA hyperpolarizing pulse of 500 ms duration.

The proximal excitatory inputs on the subicular pyramidal neurons were stimulated by placing an ACSF-filled theta electrode at a lateral distance of 50 μm from the apical dendrite within a radial distance of 100 μm from the soma of the recorded neuron (Fig.1A) in the layer stratum pyramidale of the subiculum. The stimulation strength was adjusted to evoke a monosynaptic response of 3.5–6 mV amplitude. In a subset of experiments performed to demonstrate input specificity of synaptic plasticity, a second stimulating electrode was placed distal to the soma and close to the apical dendrite. Both the inputs were stimulated alternately at 0.05 Hz during baseline and test stimulations. During the induction protocol the control input was not stimulated while the test input was paired with a burst of bAPs after a time interval of 10 ms (+10 ms). Cross-facilitation was performed to ascertain that both the inputs were independent. For cross-facilitation, both the pathways were stimulated at a time interval of 110 ms and the ratio of the EPSP slopes was compared with the ratio of the EPSP slopes when they were stimulated at a time interval of 20 s. The two pathway experiments were performed only when there was no difference between these two ratios. A burst of action potentials was evoked at the soma by three step depolarizing current pulses of 1.8 nA amplitude and 2 ms duration at 50 Hz, unless stated otherwise. This is referred as ‘bAP burst’ in the text. For the first 5 min the EPSPs were evoked at the rate of 0.05 Hz to record baseline, then one EPSP was paired with a bAP burst or a single action potential and this pair was repeated at the rate of 0.1 Hz for 10 min to induce STDP. The time interval between the EPSP and the bAP burst was varied in different experiments. After the pairing the synaptic strength was tested again at a stimulation frequency of 0.05 Hz for 30 min. The time interval between the EPSP and the bAP burst was defined by the time gap between the onset of the EPSP and the peak of the bAP nearest to it. The pairing was termed ‘causal’ or positive pairing when an EPSP was followed by the bAP burst, while it was termed ‘anti-causal’ or negative pairing when an EPSP followed the bAP burst. In some control experiments, pairing was replaced with only EPSPs or only bAP bursts repeated at 0.1 Hz for 10 min. The pairing protocol was replaced with high frequency stimulation (HFS) protocol in a few experiments. This protocol consisted of four tetanic stimuli, each of 1 s duration and 100 Hz frequency with an inter-tetanus interval of 10 s. Paired pulse ratio (PPR) experiments were performed by evoking two EPSPs at a time interval of 110 ms, as it was the minimum time interval that did not give rise to postsynaptic summation. The PPR test was performed twice, both before and after the induction period. A minimum time gap of 1 min was maintained between two consecutive PPR tests. The PPR was calculated as (EPSP₂/EPSP₁), where EPSP₂ and EPSP₁ are the second and first EPSP in the pair, respectively.

Figure 1. Experimental design and characterization of different subtypes of subicular pyramidal neurons.

A, schematic drawing illustrating electrode placement; the stimulating electrode was placed near apical dendrite and timing of the postsynaptic action potential(s) was controlled by current injection through somatic recording electrode. B, regular firing neuron: a, biocytin-stained neuron. b, voltage response to a step depolarization of 250 pA for 500 ms does not show any burst firing. c, a single AP in response to a 2 ms step pulse of 1.8 nA. Inset shows a single AP from the first spike in b on an expanded scale. C, weak burst firing neuron: a, biocytin-stained neuron; b, voltage response to a step depolarization of 250 pA for 500 ms shows burst of activity at the onset of depolarization; c, a single AP in response to a 2 ms step of 1.8 nA. Inset shows a burst of 2 APs from the first spike in b on an expanded scale. D, strong burst firing neuron: a, biocytin-stained neuron. b, voltage response to a step depolarization of 250 pA of 500 ms duration shows burst of activity at the onset of depolarization. c, burst of APs in response to 2 ms step of 1.8 nA. Inset shows a burst of three APs from the first spike in b on an expanded scale. Scale bar: 50 μm in biocytin-stained neurons; y-axis, 20 mV in all voltage traces; x-axis, 100 ms in 500 ms depolarization step, 25 ms in 2 ms depolarization pulse and 20 ms in inset.

Statistics

The EPSPs were quantified using the rising slope of initial 2 ms of the waveform. The slopes of all the EPSPs during baseline were averaged and the slopes of all other EPSPs following the induction were normalized with respect to this averaged value. This normalized change in the EPSP slope was multiplied by 100 to represent the percentage change in the EPSP slopes. The extent of plasticity was calculated by averaging the percentage change during the last 10 min of the experiment. All the data sets were tested for normal distribution using the D'Agostino–Pearson omnibus normality test. To assess the significance of plasticity induced by a particular protocol, Student's paired t test was used for the data sets found to be normally distributed; otherwise, Wilcoxon matched-pairs signed rank test (mentioned as Wilcoxon test in the text) was performed. These tests were performed between the average EPSP slopes during the baseline and the average EPSP slopes during the last 10 min of the experiments (test). The Wilcoxon signed rank test was performed between the PPR before and after plasticity induction since the data sets failed the normality test. Two populations were classified as statistically different if the P value was less than 0.05 for the comparisons where both the data sets were used only once for statistical comparison. For comparing the extent of plasticity induction between two different experiments, an unpaired t test or Mann–Whitney test was performed based on the outcome of the normality test. In the cases where one data set was used more than once for comparison, a Bonferroni–Dunn post hoc correction for multiple comparisons was applied and the threshold for significance was adjusted accordingly. Therefore, the P value threshold for the data used for multiple comparisons was set to be 0.01 (0.05/5) for the RF neurons (same control data set was used 5 times) and 0.012 (0.05/4) for the WBF neurons (same control data set was used 4 times). Since the ADP data were found to be normally distributed, an unpaired t test was performed between the ADP values of WBF and RF neurons. The R_in and RMP data did not pass the normality test; hence a Mann–Whitney test was performed for comparison between WBF vs. RF, WBF vs. SBF and RF vs. SBF neurons. For this comparison, the threshold significance was set to 0.025 (0.05/2) with the Bonferroni–Dunn post hoc correction.

Drugs

The following drugs were used: 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid tetrapotassium salt (BAPTA) 10 mm, bicuculline 10 μm, sodium ascorbate 0.4 and 1 mm, verapamil hydrochloride 50 μm (all from Sigma-Aldrich, USA); picrotoxin 10 μm, dl-2-amino-5-phosphonovalerate (dl-AP5) 50 μm, nifedipine 25 μm (all from Abcam, UK), (2S)-3-[[(1S)-1-(3,4-dichlorophenyl)ethyl]amino-2-hydroxypropyl] (phenylmethyl) phosphinic acid hydrochloride (CGP-55845) 1 μm (Tocris Biosciences, UK), biocytin staining kit (Vector Labs, UK).

Results

Characterization of regular firing, weak burst firing, and strong burst firing neurons

As reported earlier (Staff et al. 2000), we found three subtypes of pyramidal neurons in the subiculum; the regular firing (RF) neurons did not fire a burst (Fig.1Bb), while the weak burst firing (WBF) neurons fired a single burst of action potentials (Fig.1Cb) and the strong burst firing (SBF) neurons fired multiple bursts (Fig.1Db) in response to a 500 ms long depolarizing current injection of 250 pA. In response to 1.8 nA depolarization for 2 ms, the RF and WBF neurons fired a single action potential (AP), while the SBF neurons fired a single burst consisting of two or three APs at a frequency higher than 100 Hz (Fig.1Bc, Cc and Dc, respectively). Thus, the SBF neurons could never elicit a single AP, while the WBF neurons could fire a single AP upon a short step depolarization, and consequently, could be evoked to fire at any required frequency, unlike the SBF neurons. Hence, all the subsequent experiments were performed on the WBF and RF neurons. The RMP of all the reported neurons was close to −63 mV with WBF neurons more depolarized compared to the RF neurons (RF neurons, −63.6 ± 0.3 mV, n = 51; WBF neurons, −62.8 ± 0.3 mV, n = 53; SBF neurons, –63.3±0.3, n = 46; RF vs. WBF, P < 0.025; RF vs. SBF, P = 0.17; WBF vs. SBF, P = 0.38; Mann–Whitney test with Bonferroni's correction, Table1). The R_in of the RF neurons was highest while that of the SBF neurons was lowest among the three neuronal subtypes (SBF neurons, 61.0 ± 4.1MΩ, n = 12; WBF neurons, 75.5 ± 3.0 MΩ, n = 52 and RF neurons, 91.3 ± 5.1 MΩ, n = 26; SBF vs. WBF, P < 0.025; WBF vs. RF, P < 0.025 and RF vs. SBF, P < 0.005; for SBF vs. WBF and WBF vs. RF, Mann–Whitney test and for SBF vs. RF, unpaired t test with Bonferroni's correction; Table1).The WBF neurons were found to have a higher ADP (17.6 ± 0.4 mV, n = 43) than the RF neurons (13.4 ± 0.4 mV, n = 36; P < 0.01, unpaired t test, Table1). However, the ADP could not be measured in the SBF neurons, since the ADP amplitude crossed the action potential threshold and fired a second spike immediately after the first, resulting in a burst of action potentials. No significant difference was found in the sag ratio of these three types of neurons (sag ratio for RF neurons, 0.75 ± 0.06, n = 20; WBF neurons, 0.75 ± 0.07, n = 19; SBF neurons, 0.80 ± 0.08, n = 20; SBF vs. WBF, P = 0.88; WFB vs. RF, P = 0.05 and RF vs. SBF, P = 0.07, unpaired t test with Bonferroni's correction; Table1), as reported earlier by Staff et al (Staff et al. 2000).

Table 1.

Different properties of the pyramidal neuron subtypes in the subiculum

	RF neurons	WBF neurons	SBF neurons
RMP (mV)	−63.6 ± 0.3 (n = 51)	–62.8 ± 0.3 (n = 53)	−63.3 ± 0.3 (n = 46)
Rin (MΩ)	91.3 ± 5.1 (n = 26)	75.5 ± 3.0 (n = 52)	61.0 ± 4.1 (n = 12)
Sag ratio	0.75 ± 0.06 (n = 20)	0.75 ± 0.07 (n = 19)	0.80 ± 0.08 (n = 20)
ADP (mV)	13.4 ± 0.4 (n = 36)	17.6 ± 0.4 (n = 43)**	—

Values are means ± SEM (n). For R_in: SBF vs. WBF, P < 0.025, Mann–Whitney test; WBF vs. RF, P < 0.025, Mann–Whitney test; RF vs. SBF, P < 0.005, unpaired t test with Bonferroni's correction. For RMP: RF vs. WBF, P < 0.025; RF vs. SBF, P = 0.17; WBF vs. SBF, P = 0.38; Mann–Whitney test with Bonferroni's correction. For sag ratio: SBF vs. WBF, P = 0.88; WFB vs. RF, P = 0.05 and RF vs. SBF, P = 0.07, unpaired t test with Bonferroni's correction. Following Bonferroni's correction, in all the above statistical tests the P value threshold for significant difference is 0.025 since each data set was used twice for comparison. For ADP; RF vs. WBF, P < 0.01(^**), unpaired t test.

Causal pairing of an EPSP with a bAP burst induced non-Hebbian t-LTD, while, high frequency stimulus (HFS) induced Hebbian LTP

The subicular pyramidal neurons show diverse firing patterns, as shown in Fig.1, and reported earlier (Staff et al. 2000). Some of them fire discrete single action potentials (RF), while others fire in burst mode (BF). We asked, if an EPSP paired with a single action potential in the RF and the WBF neurons can induce synaptic plasticity, as observed in the hippocampal synapses (Bi & Poo, 1998; Buchanan & Mellor, 2010); or would fail to induce plasticity as reported in the cortical layer V (Kampa et al. 2006), and Schaffer collateral–CA1 synapses in the hippocampus (Pike et al. 1999). We observed that causal pairing of an EPSP with a single bAP at a time interval of 10 ms and repeating the pair at 0.1 Hz for 10 min did not elicit any synaptic plasticity in the RF (–13.4 ± 9.2% of baseline, n = 6, P = 0.38 test vs. baseline, Wilcoxon test; Fig.2A) or the WBF (3.2 ± 7.8% of baseline, n = 8, P = 0.67 test vs. baseline, paired t test, Fig.2B) neurons. R_in and RMP were also stable throughout the experiment (Fig.2). Following an earlier report where a burst of action potentials in the postsynaptic neurons was found to be essential for pairing-induced synaptic plasticity in the CA3–CA1 synapses (Pike et al. 1999), we reasoned that the failure to observe plasticity in our case could be due to the single bAP used in the postsynaptic neurons. Hence, we paired a single EPSP with a burst of three bAPs at 50 Hz (bAP burst) at the same time interval of +10 ms and repeated this pair as described above. Contrary to the canonical STDP plots, where causal pairing induces LTP, such pairing-induced timing-dependent LTD (t-LTD) in both the RF (−26.2 ± 5.3% of baseline, n = 9, P < 0.01 test vs. baseline, paired t test with Bonferroni's correction, Fig.3A) and the WBF (−36.5 ± 5.0% of baseline, n = 9, P < 0.012 test vs. baseline, Wilcoxon test with Bonferroni's correction, Fig.3B) neurons. Since certain STDP induction protocols can also induce plasticity of the intrinsic properties of the neurons (Moore et al. 2009; Debanne & Poo, 2010), we monitored the R_in and RMP of the postsynaptic neuron throughout the experiments, and these were found to be stable in both the cell types (Fig.3). These synapses are known to express Hebbian LTP in response to high frequency synaptic stimulation (HFS; Wozny et al. 2008). However, Wozny et al. used 4 mm CaCl₂ and MgSO₄ each in the perfusate in some of their LTP experiments, while we used much lower concentrations of these salts (see Methods). Hence, we explored whether these synapses express HFS-induced LTP, and the results indicate a robust expression of LTP in the RF (45.5 ± 18.5%of baseline, n = 6, P < 0.05, Wilcoxon test, Fig.4A) as well as in WBF (67.8 ± 15.3% of baseline, n = 6, P < 0.05, Wilcoxon test, Fig.4B) neurons in response to HFS. Thus, in our experimental conditions, these synapses are capable of expressing Hebbian as well as non-Hebbian synaptic plasticity, depending upon the stimulation protocol.

Figure 2. Causal pairing of EPSP with a single bAP fails to induce LTD.

A, normalized EPSP slope time series for the experiments show that an EPSP paired with a single bAP failed to induce any plasticity (−13.4 ± 9.2%, n = 6, P = 0.38, Wilcoxon test for baseline vs. last 10 min of the experiment) in the RF subicular neurons. B, similar pairing experiments in the WBF subicular neurons failed to induce any plasticity (–3.2 ± 7.8%, n = 8, P = 0.67, paired t test for baseline vs. last 10 min of the experiment). In both the experiments, an EPSP was paired with a single bAP at a time interval of +10 ms and this pair was repeated at 0.1 Hz for 10 min. Scale bars for EPSP traces; x-axis, 50 ms; y-axis, 3 mV.

Figure 3. Pairing induced non-Hebbian plasticity.

A, the normalized EPSP slope time series for the experiment in the top panel shows the induction of t-LTD in the RF neurons (−26.2 ± 5.3%, n = 9, P < 0.01, paired t test with Bonferroni's correction for baseline vs. last 10 min of the experiment) when an EPSP was paired with a burst of three bAPs at a time interval of +10 ms. The next two panels below show the stability of R_in and RMP throughout the experiment in panels A–D. The dashed line at −60 mV in the RMP plots indicates the membrane potential at which the neurons were held by current injection, while the data points show the actual RMP values without current injection at different time points in this and the RMP plots of other figures. B, a similar protocol as in A induced t-LTD in the WBF neurons (−36.5 ± 5.0%, n = 9, P < 0.012, Wilcoxon test with Bonferroni's correction for baseline vs. last 10 min of the experiment) at +10 ms timing, when an EPSP was paired with a burst of three bAPs at a time interval of +10 ms. C, in the RF neurons, pairing an EPSP with a burst of 3 bAPs at a time interval of −10 ms induced small LTP (27.4 ± 12.8%, n = 7, P = 0.11, Wilcoxon test for baseline vs. last 10 min of the experiment) that was statistically not significant. D, in the WBF neurons, pairing an EPSP with the bAP burst at −10 ms timing interval showed prominent LTP (37.0 ± 10.7%, n = 7, P < 0.05, Wilcoxon test for baseline vs. last 10 min of the experiment). In all the experiments, the EPSP–bAP burst pair was repeated at 0.1 Hz for 10 min, indicated by horizontal double-headed arrows in the top panels of A–D which also apply to the bottom panels showing normalized R_in and RMP in a given figure. The action potentials shown in the representative traces for the pairing protocol are truncated for the sake of clarity. The traces shown during the induction period represent only one EPSP–bAP burst pair; 60 such pairs were used in each induction protocol in this and other figures. Scale bars for these representative traces: x-axis, 50 ms; y-axis, 20 mV. The average EPSP traces shown at the top correspond to the parts of the time series denoted by numbers (1, baseline; 2, test for this and other figures on which the statistical tests were done). The dotted traces in the overlapped representative EPSP traces (1+2) in this and other figures are from the last 5 min of the experiment. Scale bars for EPSP traces: x-axis, 50 ms; y-axis, 3 mV.

Figure 4. High frequency synaptic (HFS) stimulation without pairing induced Hebbian LTP.

A, normalized EPSP slope time series shows induction of robust LTP in the RF neurons (45.5 ± 18.5% of baseline, n = 6, P < 0.05, Wilcoxon test for baseline vs. last 10 min) following high frequency presynaptic stimulation (4 tetanic stimuli, each of 1 s duration and 100 Hz frequency with inter-tetanus interval of 10 s) was applied on the presynaptic inputs. B, robust LTP was observed in the WBF neurons following HFS stimulation (67.8 ± 15.3% of baseline, n = 6, P < 0.05, Wilcoxon test for baseline vs. last 10 min). The next two panels below show the stability of R_in and RMP throughout the experiments. Scale bars for the EPSP traces: x-axis, 50 ms; y-axis, 3 mV. The vertical thick arrows indicate the application of HFS protocol.

Further, we examined whether sucrose-containing ACSF as the dissection medium influenced the STDP plot in favour of LTD in response to a causal pairing protocol, as observed in the rat CA3–CA1 synapses (Edelmann & Lessmann, 2011). Sucrose was replaced with NaCl in the dissection solution and the pairing of an EPSP with a bAP burst at +10 ms time interval was performed. Such pairing induced robust t-LTD in the slices prepared in NaCl-containing ACSF as well (−31.7 ± 7.9% of baseline, n = 6; P < 0.05 test vs. baseline, Wilcoxon test, data not shown). We used a lower concentration of picrotoxin (10 μm) to block inhibitory inputs in most of the experiments. Hence, we further confirmed our key finding of causal pairing-induced t-LTD with additional block of inhibitory inputs by pairing an EPSP with a bAP burst at a time interval of +10 ms in the presence of 10 μm bicuculline, in addition to 10 μm picrotoxin. In conditions of additional block of inhibitory transmission, induction of tLTD was observed in both the neuronal subtypes (RF neurons, −33.7 ± 9.1% of baseline, n = 6; WBF neurons, −23.2 ± 5.8% of baseline, n = 6, data not shown). This plasticity was similar to that induced in the presence of picrotoxin and CGP-55485 only (in the RF neurons, P = 0.53 and in the WBF neurons, P = 0.11, Mann–Whitney test with Bonferroni's correction). Subsequently, we tested the timing dependence of this t-LTD in the WBF neurons by increasing the time interval between the EPSP and the burst of bAPs. We observed that causal pairing at a time interval of +30 ms induced small t-LTD (−13.5 ± 9.5% of baseline, n = 7, P = 0.29 test vs. baseline, Wilcoxon test, Fig.8A) that was not significant. Further, increasing the time interval to +50 ms abolished the t-LTD (−1.6±5.2% of baseline, n = 6, P = 0.91, test vs. baseline, Wilcoxon test, Fig.8A).

Figure 8. Pairing an EPSP with a burst of action potentials leads to depression biased reverse STDP rule in the weak burst firing neurons of the subiculum.

A, overlap of the time series of normalized EPSP slopes for different positive time intervals of the STDP protocol shows that in the WBF neurons an EPSP paired with a bAP burst induced LTD that decreased with increasing time interval; the extent of t-LTD was −36.5 ± 5.0% (n = 9, P < 0.012, Wilcoxon test with Bonferroni's correction) for +10 ms (), −13.5 ± 9.5% (n = 7, P = 0.29) for +30 ms () and −1.6 ± 5.2% (n = 6, P = 0.91, Wilcoxon test) for +50 ms (). The statistical comparison was performed between EPSP slopes during baseline vs. EPSP slopes during last 10 min of the recording in all the three experiments. B, at negative pairings, longer than −10 ms no plasticity was induced (−3.9 ± 6.5%, n = 7, P = 0.69 for −20 ms (); −12.6 ± 14.8%,n = 6, P = 0.44 for −50 ms, (); and 2.5 ± 7.8%,n = 5, P = 0.82 for −85 ms, (); Wilcoxon test for baseline vs. last 10 min of the recording in all the three experiments). A robust LTP (37.0 ± 10.7%, n = 7, P < 0.05, Wilcoxon test) was induced, when a burst of bAPs was followed by an EPSP at a time interval of 10 ms (data not shown here for the sake of clarity, refer to Fig.2C). R_in and RMP were stable throughout the experiments. The symbols used for normalized R_in and RMP correspond to that used in the normalized EPSP slope time series. During the plasticity induction protocol, an EPSP was paired with a bAP burst at varying time intervals, as indicated in the figures, and the pairs were repeated at 0.1 Hz for 10 min. Scale bars for EPSP traces: x-axis, 50 ms; y-axis, 3 mV. C, amplitude of plasticity is plotted for various timing intervals (Δt) on the x-axis for the data shown in A and B and Fig.2C and D. The timing is given a positive value if the EPSP preceded the bAP burst, while a negative value was given if the EPSP followed the bAP burst. Δt was calculated as described in the Methods section. The plot shows a non-Hebbian STDP profile expressing t-LTD on the positive side of the curve and t-LTP on the negative side. While t-LTP is limited to a near coincident time interval of −10 ms on the negative side of the STDP curve, a small amount of LTD can be observed till a time interval of +30 ms on the positive side. Open circles, individual neurons; filled circles, mean ± SEM values for a particular time interval.

Anti-causal pairing of an EPSP with a bAP burst induced t-LTP at a narrow time interval in the WBF neurons

Next we asked if these synapses also express t-LTD with anti-causal pairing as reported earlier for excitatory synapses (Caporale & Dan, 2008; Buchanan & Mellor, 2010). This was addressed by reversing the order of the EPSP and the bAP burst in the pair. Contrary to the canonical Hebbian STDP plot in excitatory synapses, we observed LTP at anti-causal pairing at a close time interval of −10 ms in the WBF neurons (37.0 ± 10.7% of baseline, n = 7, P < 0.05, test vs. baseline, Wilcoxon test, Fig.3D). Although such pairing induced small LTP in the RF neurons as well, it was not significant (27.4±12.8%, of baseline, n = 7, P = 0.11 test vs. baseline, Wilcoxon test, Fig.3C). However, in the WBF neurons similar pairings failed to induce any plasticity at longer time intervals (−3.9 ± 6.5% of baseline, n = 7, P = 0.69, test vs. baseline, for −20 ms; −12.6 ± 14.8% of baseline, n = 6, P = 0.44 test vs. baseline, for −50 ms; 2.5 ± 7.8% of baseline, n = 5, P = 0.82, test vs. baseline, for −85 ms, Wilcoxon test, Fig.8B).

The STDP plot of the paired inputs on the subicular WBF neurons showed that this synapse behaves in a non-Hebbian manner (Fig.8C). The amplitude of t-LTD decreased with increasing time interval between the EPSP and the burst of action potentials (Fig.8A and C), while LTP was induced only at −10 ms anti-causal pairing (Figs 3D and 8B and C). Therefore, we observed a reverse STDP rule contrary to most of the hippocampal and cortical excitatory synapses (Caporale & Dan, 2008). The asymmetric STDP plot was different from earlier reports where the negative side of the STDP plot is wider than the positive side in most of the excitatory synapses (Bi & Wang, 2002). Moreover, in the STDP curve reported here, the plasticity was limited to the time interval of −10 ms on the negative side, whereas on the positive side plasticity could be induced up to +30 ms. The negative side of this STDP curve is narrower than the positive side, hence the STDP curve is biased towards depression.

Properties of the non-Hebbian t-LTD

In the cortical layer V pyramidal neurons, Kampa et al. (2006) demonstrated that EPSPs paired with bAP bursts of higher frequencies are more potent in inducing spike-timing-dependent LTP compared to an EPSP paired with a single bAP or a burst(s) of bAPs at lower frequencies. We therefore tested whether changing the frequency of action potentials in the burst affects the non-Hebbian t-LTD. This was done by pairing an EPSP with a burst of three bAPs at a higher frequency (150 Hz) instead of 50 Hz, at a time interval of +10 ms, where the t-LTD was most prominent. We found that both the RF (−24.6 ± 8.7% of baseline, n = 6; Fig.5A) and the WBF (−36.8% ±7.7 of baseline, n = 6, Fig.5B) neurons showed t-LTD. Further, this t-LTD was similar to that obtained with an EPSP paired with a 50 Hz bAP burst (pairing with 50 Hz bAP burst vs. pairing with 150 Hz bAP burst; P = 0.70 for RF neurons and P = 0.98 for WBF neurons, Mann–Whitney test with Bonferroni's correction). Our results, along with previous findings (Pike et al. 1999; Kampa et al. 2006), suggest that an EPSP paired with a single bAP may fail to induce timing-dependent plasticity in a system where pairing with a burst of bAPs can induce plasticity. However, we failed to observe any t-LTP induced by causal pairing with a burst of bAPs at a higher frequency in the subicular pyramidal neurons.

Figure 5. Causal pairing of EPSP with a 150 Hz bAP burst also induces t-LTD.

A, normalized EPSP slope time series for the experiments show that an EPSP paired with a high frequency (150 Hz) bAP burst induced t-LTD (−24.6 ± 8.7%, n = 6) in the RF subicular neurons. B, normalized EPSP slope time series from similar experiments in the WBF subicular neurons shows t-LTD induction when an EPSP was paired with a higher frequency bAP (150 Hz) burst (−36.8 ± 7.7%, n = 6). (Pairing with 50 Hz bAP burst vs. pairing with 150 Hz bAP burst, P = 0.70 for RF neurons, P = 0.98 for WBF neurons, Mann–Whitney test with Bonferroni's correction.) In both the experiments, pairing was performed at a time interval of +10 ms and this pair was repeated at 0.1 Hz for 10 min indicated by double-sided arrow. Scale bars for EPSP traces: x-axis, 50 ms; y-axis, 3 mV.

The non-Hebbian t-LTD observed here requires both pre- and postsynaptic activities, as reported earlier in the rat cortical layer II/III and layer V synapses (Birtoli & Ulrich, 2004). In the absence of either of these activities, this phenomenon was not observed in both the neuronal subtypes. Repetition of bAP bursts at 0.1 Hz alone, or repetition of only EPSPs at 0.1 Hz for 10 min did not induce any plasticity in the RF neurons (4.7 ± 14.0% of baseline, n = 7, P = 0.89 test vs. baseline, paired t test for repeated bursts of bAPs and −1.0 ± 8.8% of baseline, n = 8, P = 0.69 test vs. baseline for EPSPs, Wilcoxon test; Fig.6A) and the WBF neurons (−5.0 ± 5.8% of baseline, n = 6, P = 0.56 test vs. baseline, Wilcoxon test, for repeated bursts of bAPs; 0.6 ± 9.6% of baseline, n = 7, P = 0.94 test vs. baseline for EPSPs, Wilcoxon test; Fig.6B).

Figure 6. The non-Hebbian LTD requires both pre- and postsynaptic activity.

A, normalized EPSP slope time series shows that the RF subicular neurons do not show plasticity when only bAP bursts were repeated at 0.1 Hz for 10 min (; 4.7 ± 14.0%, n = 7, P = 0.89, paired t test for baseline vs. last 10 min of the experiment), or only EPSPs were repeated at 0.1 Hz for 10 min (; −1.0 ± 8.8%, n = 8, P = 0.69, Wilcoxon test for baseline vs. last 10 min of the experiment). B, similar experiments in the WBF neurons showed failure of plasticity induction when only bAP bursts were repeated at 0.1 Hz for 10 min (; −5.0 ± 5.8% n = 6, P = 0.56, Wilcoxon test for baseline vs. last 10 min of the experiment), or only EPSPs were repeated at 0.1 Hz for 10 min (; 0.6 ± 9.6% n = 7, P = 0.94, Wilcoxon test for baseline vs. last 10 min of the experiment). In both the figures the double-headed horizontal arrows show the time when EPSPs or bAP bursts were repeated. The two bottom panels show the stability of R_in and RMP throughout the experiment in both the neuronal subtypes. Scale bars for EPSP traces: x-axis, 50 ms; y-axis, 3 mV.

Next, we tested whether the t-LTD was input specific. This was experimentally verified by stimulating two independent inputs on the same postsynaptic neuron. While, only one of the inputs was paired with the bAP burst (test pathway); the other was silent during the pairing and acted as the control pathway. In the RF neurons, only the test pathway showed t-LTD (−26.0 ± 6.9% of baseline, n = 9, P < 0.05 test vs. baseline, paired t test), while the control pathway was unaffected (−1.0 ± 9.4% of baseline, n = 9, P = 0.52 test vs. baseline, paired t test; Fig.7A). Similarly, the t-LTD in the WBF neurons was also specific to the paired pathway (−28.2 ± 3.5% of baseline, n = 7, P < 0.05 test vs. baseline, Wilcoxon test), while the control pathway did not express any plasticity (−2.8 ± 3.0% of baseline, n = 7, P = 0.30 test vs. baseline, Wilcoxon test; Fig.7B). This confirmed that the t-LTD is specific to the inputs paired with the bAP burst in both the neuronal subtypes.

Figure 7. Pairing-induced non-Hebbian LTD is specific to the paired synapse.

A, pairing-induced LTD in the RF neurons is specific to the input paired with the burst of bAP, as only the input paired with the bAP burst (paired pathway, ) showed t-LTD (−26.0 ± 6.9%, n = 9, P < 0.01, paired t test for baseline vs. last 10 min of the experiment), while the control pathway (), which was silent during induction, did not show any plasticity (−1.0 ± 9.4%, n = 9, P = 0.52, paired t test for baseline vs. last 10 min of the experiment). B, in the WBF neurons as well, only the paired pathway () showed t-LTD (−28.2 ± 3.5%, n = 7, P < 0.05, Wilcoxon test for baseline vs. last 10 min of the experiment), while the control pathway () did not show any plasticity (−2.8 ± 3.0%, n = 7, P = 0.30, Wilcoxon test for baseline vs. last 10 min of the experiment). The lower two panels below show the stability of R_in and RMP throughout the experiments. Scale bars for the EPSP traces: x-axis, 50 ms; y-axis, 3 mV.

LTD is postsynaptic in the WBF neurons, while presynaptic in the RF neurons

An earlier report suggests that in the subiculum the site of induction of synaptic plasticity can be either pre- or postsynaptic depending on the involvement of the postsynaptic neuronal subtype (Wozny et al. 2008). The involvement of the postsynaptic neuron in synaptic plasticity is associated with changes in intracellular calcium, and its chelation leads to loss of plasticity. We included 10 mm BAPTA in the patch pipette to chelate calcium in the postsynaptic neuron, and paired an EPSP with a bAP burst at the time interval of +10 ms. The t-LTD was blocked in the WBF neurons (−3.9 ± 8.9% of baseline, n = 7, P = 0.94 test vs. baseline, Wilcoxon test; Fig.9B), whereas the extent of plasticity did not change in the RF neurons (control −26.2 ± 5.3% of baseline, n = 9; with BAPTA −40.0 ± 9.7% of baseline, n = 7, P = 0.21 Mann–Whitney test for control vs. BAPTA with Bonferroni's correction; Fig.9A). Additionally, we performed paired pulse ratio (PPR) analysis in both the neuronal subtypes, as it indicates the change in release probability. A change in PPR after the induction protocol suggests that the plasticity involves presynaptic rather than postsynaptic mechanisms (Schulz et al. 1994). Since we did not observe any difference in the plasticity phenomenon between experiments performed in ACSF containing picrotoxin and ACSF containing picrotoxin + bicuculline, we pooled the PPR data from both these experiments. We observed an increase in PPR due to induction of t-LTD in the RF neurons (1.4 ± 0.1, before pairing and 1.8 ± 0.1 after pairing, n = 14; P < 0.01, Wilcoxon test; Fig.9Ca), while it remained unchanged in the WBF neurons (1.6 ± 0.1 before pairing and 1.7 ± 0.1 after pairing, n = 17, P = 0.46, paired t test; Fig.9Cb). The intracellular BAPTA experiments suggest that the induction of spike-timing-dependent LTD does not require postsynaptic calcium in the RF neurons, while postsynaptic calcium is required for LTD induction in the WBF neurons. The PPR experiments indicate that t-LTD is expressed via some presynaptic mechanism in the RF neurons while postsynaptic mechanisms are involved in the WBF neurons. These results are contrary to the HFS-induced LTP, where induction and expression of LTP is presynaptic in bursting, while postsynaptic in regular firing subicular neurons (Wozny et al. 2008).

Figure 9. LTD is presynaptic in the RF neurons while postsynaptic in the WBF neurons.

A, the time series of normalized EPSP slopes for an EPSP paired with a bAP burst at a time interval of +10 ms shows that, in the RF neurons, dialysis of the postsynaptic neuron with 10 mm BAPTA could not block t-LTD (−40.0 ± 9.7%, n = 7 with BAPTA, (); −26.2 ± 5.3%, n = 9, without BAPTA (); P = 0.21 control vs. BAPTA, Mann–Whitney test with Bonferroni's correction). B, the t-LTD was blocked in the WBF neurons with 10 mm intracellular BAPTA in the postsynaptic neuron (−3.9 ± 8.9% of baseline, n = 7, P = 0.94, with BAPTA (), Wilcoxon test for baseline vs. last 10 min of the experiment; −36.5 ± 5.0%, n = 9 without BAPTA ()). In both the figures, the bottom panels show the stability of R_in and RMP throughout the experiments. Plasticity was induced by an EPSP paired with bAP burst at a time interval of +10 ms. The control data appearing in panels A and B have been adapted from panels A and B, respectively, of Fig.2. Scale bars for the EPSP traces: x-axis, 50 ms; y-axis, 3 mV. Ca, the RF neurons show increase in PPR after t-LTD induction (1.4 ± 0.1, before pairing and 1.8 ± 0.1 after pairing, n = 14; P < 0.01(**), Wilcoxon test for PPR before induction vs. PPR at the end of the experiment). Cb, PPR did not change in the WBF neurons after t-LTD induction (1.6 ± 0.1 before pairing and 1.7 ± 0.1 after pairing, n = 17, P = 0.46 (ns, not significant), paired t test for PPR before induction vs. PPR at the end of the experiment). PPR measurements from individual neurons before and after induction are connected with a line. Filled circles with bars represent mean ± SEM for all the experiments. Scale bars for EPSP traces: x-axis, 50 ms; y-axis, 3 mV.

t-LTD is NMDAR dependent in the WBF neurons, while L-type calcium channel dependent in the RF neurons

Neurons mobilize calcium from different sources during the induction of synaptic plasticity. The most common source of calcium is through NMDA receptors due to their dependence on glutamate binding, modulation by voltage, and calcium permeability (Bliss & Collingridge, 1993). The importance of Ca²⁺ mobilization through NMDA receptors in the RF and WBF neurons was tested by inducing t-LTD with causal pairing at a +10 ms time interval in the presence of the NMDA receptor blocker dl-AP5 (50 μm) in the bath solution. In the WBF neurons, blocking NMDA receptors abolished t-LTD (−7.7 ± 8.6% of baseline, n = 8, P = 0.42 test vs. baseline, paired t test; Fig.10), while in the RF neurons dl-AP5 did not affect the t-LTD (control −26.2 ± 5.3%, n = 9; with dl-AP5 −27.5 ± 5.2%, n = 7, P = 0.99, Mann–Whitney test for control vs. dl-AP5 with Bonferroni's correction; Fig.10A).

Figure 10. t-LTD is NMDA receptor mediated in the WBF neurons.

A, the time series of normalized EPSP slopes shows that blocking NMDA receptors with 50 μm dl-AP5 did not influence t-LTD in the RF neurons (−27.5 ± 5.2%, n = 7 with dl-AP5 (); −26.2 ± 5.3%, n = 9 without dl-AP5 (); P = 0.99 control vs. dl-AP5, Mann–Whitney test with Bonferroni's correction for control vs. dl-AP5). B, blocking NMDA receptors using 50 μm dl-AP5 blocked t-LTD in the WBF neurons (−7.7 ± 8.6%, P = 0.42, n = 8 with dl-AP5 (), paired t test for baseline vs. last 10 min of the experiment; −36.5 ± 5.0%, n = 9 without dl-AP5 ()).The control data appearing in panels A and B have been adapted from panels A and B, respectively, of Fig.2. In both the figures, bottom panels show the stability of R_in and RMP throughout the experiments. Scale bar for EPSP traces: x-axis, 50 ms; y-axis, 3 mV.

Another source of calcium mobilization is through L-type calcium channels, due to their voltage dependency and calcium permeability. Their involvement in synaptic plasticity has been documented in the hippocampus and amygdala (Huang & Malenka, 1993; Kapur et al. 1998; Weisskopf et al. 1999). The involvement of the L-type calcium channels in the t-LTD of RF and WBF neurons was tested with the L-type calcium channel blocker nifedipine (25 μm) in the bath solution. Nifedipine in the bath solution did not affect t-LTD in the WBF neurons (control −36.5 ± 5.0%, n = 9; with nifedipine −34.6 ± 8.0%, n = 6, P = 0.99, Mann–Whitney test for control vs. nifedipine with Bonferroni's correction; Fig.1B), while it blocked the t-LTD in the RF neurons (2.5 ± 3.9% of baseline, n = 9, P = 0.47 test vs. baseline, paired t test; Fig.1A). This observation was further confirmed by using another L-type calcium channel blocker verapamil hydrochloride (50 μm) in the bath solution (Bauer et al. 2002; Fourcaudot et al. 2009), which was effective in blocking the t-LTD in the RF neurons (−4.2 ± 4.1% of baseline, n = 6, P = 0.44 test vs. baseline, Wilcoxon test; Fig.1A).

Discussion

In the present study, we have demonstrated that near coincident activity of proximal excitatory inputs on the subicular pyramidal neurons and action potentials in the post synaptic neuron induces synaptic plasticity. The amplitude and direction of this plasticity is determined by the order and time interval between the two activities. This was studied in detail in the WBF neurons and the results were used to generate the STDP plot. The STDP curve was non-Hebbian, as it showed LTD at causal and LTP at anti-causal pairings; and was biased towards depression, as it showed t-LTD for a wider time interval between the EPSP and the bAP burst (Fig.8). Similar inputs on the RF neurons also induced non-Hebbian t-LTD with causal pairing of an EPSP and a bAP burst. In both the neuronal subtypes, the non-Hebbian t-LTD was specific to the synaptic input that was paired with the bAP burst (Fig.7). Interestingly, the HFS to similar inputs induced LTP in both the neuronal subtypes (Fig.4), confirming the previous observations of Wozny et al (Wozny et al. 2008). Although causal pairing induced t-LTD in both the neuronal subtypes, the mechanism associated with the induction of plasticity was different in these neurons. In the case of WBF neurons, t-LTD was induced postsynaptically and required calcium influx through the NMDA receptors (Figs 9 and 10), while t-LTD was induced and maintained presynaptically in the RF neurons and required the activation of L-type calcium channels (Figs 9 and 1). Such mechanistic differences in synaptic plasticity between two different cell types in the rat subiculum have been reported previously for HFS-induced LTP (Wozny et al. 2008). Though the involvement of L-type calcium channels in presynaptic plasticity has been reported previously in the amygdala (Fourcaudot et al. 2009), to the best of our knowledge the present report is the first of its kind in the hippocampus.

Comparison of the STDP plot in the subiculum with other synapses

Contrary to the canonical STDP plots in the excitatory synapses, wherein causal pairing induces t-LTP and anti-causal pairing t-LTD (Caporale & Dan, 2008), causal pairing induced t-LTD in the proximal excitatory inputs on the WBF and RF neurons in the subiculum. At many synapses, pairing an EPSP with a single bAP induces STDP, pairing with a single bAP failed to induce plasticity at the proximal synapses on the subicular WBF and RF neurons (Fig.2); however, when paired with a burst of bAPs, it induced plasticity (Figs 3 and 5). Similar observations have been reported previously in the hippocampus (Pike et al. 1999) and the cortex (Kampa et al. 2006). The burst of bAPs may be more efficient than a single bAP in overcoming dendritic filtering to invade the dendritic arborization and depolarize the dendrites (Lisman, 1997). Indeed, in rat cortical pyramidal neurons, while pairing a single bAP with an EPSP failed to open synaptic NMDARs, pairing a burst of bAPs with an EPSP caused significant activation of synaptic NMDARs (Kampa et al. 2006). The frequency of APs in the bAP burst also appears to be critical in inducing plasticity at certain synapses. For example, t-LTD induced by anti-causal pairing is limited to a lower frequency of action potentials in the bAP burst at the synapses between layer V cortical pyramidal neurons in the rat visual cortex (Sjöström et al. 2001). However, both the WBF and the RF neurons showed t-LTD upon pairing an EPSP with both low (50 Hz), and high (150 Hz) frequency action potential bursts (Figs 3 and 5). Inhibitory activity in the network can influence the plasticity induction rules (Meredith et al. 2003). The induction of robust t-LTD at a +10 ms time interval in both the subtypes of neurons in the presence of bicuculline, picrotoxin and CGP-55845, which block inhibitory transmission, suggests that the non-Hebbian plasticity observed in the present study was not due to the influence of inhibitory activity. In an earlier study, the presence of sucrose has been demonstrated to restrain pairing-induced LTP induction in CA1 pyramidal neurons (Edelmann & Lessmann, 2011). However, in our study, removing sucrose from the dissection solution did not change the direction of plasticity.

There is ample diversity in STDP plots in the hippocampus (Buchanan & Mellor, 2010); wherein causal pairing induces LTP (Magee & Johnston, 1997; Debanne et al. 1998; Bi & Poo, 2001; Astori et al. 2010), no plasticity (Pike et al. 1999) or even LTD (Christie et al. 1996). There are also reports from the somatosensory cortex, where causal pairing induces LTD (Egger et al. 1999; Birtoli & Ulrich, 2004; Letzkus et al. 2006; Shindou et al. 2011). Such causal pairing-induced non-Hebbian t-LTD, observed here in the CA1–subiculum synapses, has also been reported in the distal synapses of layer II/III pyramidal neurons on the layer V pyramidal neurons in the rat somatosensory cortex (Letzkus et al. 2006), spiny stellate neurons in the layer IV of barrel cortex of young rats (Egger et al. 1999), and cerebellum like structure of a weakly electric fish (Bell et al. 1997). Whereas, Bell et al. (1997) observed LTP by anti-causal pairing, Egger et al. (1999) observed LTD by both causal and anti-causal near coincident pairings. In the study by Letzkus et al. (2006), causal pairing caused LTD at the distal synapses and LTP in the proximal synapses, while anti-causal pairing caused LTD at the proximal synapses and LTP at the distal inputs (Letzkus et al. 2006). Various factors such as dendritic location of synaptic inputs, calcium dynamics, and neuromodulators have been attributed to such non-Hebbian STDP in different synapses (Kampa et al. 2007; Froemke et al. 2010).

By repetitive pairing of an EPSP with a bAP burst, Birtoli and Ulrich have reported an ‘LTD only’ STDP plot in the layer V of rat somatosensory cortex (Birtoli & Ulrich, 2004). However, our STDP plot primarily showed LTP with anti-causal pairing. They also reported LTP induction by repeated pairing of an EPSP with a single bAP, which failed to induce any plasticity in our system (Fig.2). In a later study by the same group, cellular mechanisms of causal pairing-induced LTD were described in neocortical pyramidal neurons (Czarnecki et al. 2007), where t-LTD was found to be mGluR (metabotropic glutamate receptors)-mediated and abolished by buffering postsynaptic calcium. However, in our study t-LTD was mediated by postsynaptic calcium and required NMDAR activation in the subicular WBF neurons (Figs 9 and 10) while in the RF subicular neurons, a postsynaptic calcium-independent role for L-type calcium channels was found to be important in the induction of t-LTD (Figs 9 and 11). Unlike the typical asymmetric STDP curves where the positive side of the curve (causal pairing) is narrower than the negative side (anti-causal pairing), in the subicular synapses the positive side of the STDP curve is broader than the negative side (Fig.8). Along with several other reports (Sjöström et al. 2001; Rumsey & Abbott, 2004; Wittenberg & Wang, 2006), our finding also endorses the view that timing rules for the STDP are malleable, and may vary from synapse to synapse.

Figure 11. t-LTD is dependent upon L-type calcium channels in the RF neurons.

A, the time series of normalized EPSP slopes shows that induction of t-LTD was blocked in the presence of L-type calcium channel blocker nifedipine (25 μm) in the RF neurons (2.5 ± 3.9%, P = 0.47, n = 9 with nifedipine (), paired t test for baseline vs. last 10 min of the experiment; −26.2 ± 5.3%, n = 9 without nifedipine,()). In the presence of another L-type calcium channel blocker verapamil hydrochloride (50 μm) EPSPs paired with postsynaptic bAP bursts also failed to induce t-LTD in the RF neurons (−4.2 ± 4.1% of baseline, n = 6, P = 0.44, (), Wilcoxon test for baseline vs. last 10 min of the experiment). B, blocking L-type calcium channels with 25 μm nifedipine did not affect t-LTD induction by pairing EPSPs with postsynaptic bAP bursts in the WBF neurons (−34.6 ± 8.0%, n = 6 with nifedipine (); −36.5 ± 5.0%, n = 9 without nifedipine (); P = 0.99, Mann–Whitney test with Bonferroni's correction for control vs. nifedipine).The control data appearing in panels A and B have been adapted from panels A and B, respectively, of Fig.2. In both the figures the bottom panels show the stability of R_in and RMP throughout the experiments. Scale bar for EPSP traces: x-axis, 50 ms; y-axis, 3 mV.

The atypical STDP plot of the proximal excitatory synapses on the WBF and RF subicular pyramidal neurons is rather intriguing, and may have to do with intrinsic properties such as the ADP (Karmarkar et al. 2002) and sag ratio (van Welie et al. 2006), which affect coincidence detection and synaptic integration, respectively. Another factor that can influence STDP and has not been studied in the subicular pyramidal neurons is the distribution and kinetics of the potassium channels (Zhou et al. 2005). The STDP plot reported here resembles the STDP plot reported on the distal inputs of the cortical neurons (Letzkus et al. 2006; Kampa et al. 2007). It appears that the proximal dendrites of the subicular pyramidal neurons behave like distal dendrites of the cortical neurons. In summary, the non-Hebbian t-LTD biased STDP plot observed in the present study can either be attributed to certain peculiar intrinsic properties of the subicular pyramidal neurons or the induction protocol itself.

Mechanistic differences in the t-LTD between the two different types of subicular neurons

The results of experiments with intracellular BAPTA (Fig.9) indicate that induction of t-LTD is postsynaptic in the WBF neurons, but presynaptic in the RF neurons. Our observations related to the site of origin of plasticity are different from the HFS-induced LTP reported earlier (Wozny et al. 2008). While HFS-induced LTP was presynaptic in the burst firing neurons and postsynaptic in the RF neurons, our results suggest that the pairing-induced t-LTD is presynaptic in the RF neurons and postsynaptic in the WBF neurons.

One of the puzzling findings was the resistance of t-LTD to intracellular loading of BAPTA in the RF neurons. Since the synaptic plasticity is expressed by change in presynaptic release probability and requires near coincident activation of pre- and postsynaptic neurons, we believe some kind of retrograde messenger is involved during the repeated coincident activations that is different from the mechanisms so far identified. A possible role of K⁺ as a retrograde messenger in presynaptic plasticity has been suggested previously (Sastry et al. 1986; Matyushkin et al. 1995). An earlier report by Poolos et al. (1987) showed that repetitive stimulation of Schaffer collaterals cause a considerable rise in extracellular K⁺ in the rat CA1 stratum radiatum. Recently, K⁺ released from postsynaptic neurons due to synaptic activation has been shown to influence presynaptic activity by acting as a retrograde messenger (Shih et al. 2013). Bursts of action potentials can cause considerable change in the extracellular K⁺ due to the limited volume of the interstitial space and low baseline levels of extracellular K⁺ (Kume-Kick et al. 2002). Similar rise in extracellular K⁺ can be expected around the presynaptic terminal while evoking the EPSCs and the bAP burst; together these currents would result in an increase in extracellular K⁺ in the synaptic cleft resulting in strong depolarization of the presynaptic terminals. Further depolarization can occur due to the accumulation of intracellular calcium during presynaptic activity that occurs 10 ms before the bAP burst. An increase in extracellular K⁺ due to postsynaptic activity was previously shown to modify presynaptic activity (Malenka et al. 1981). Depression of glutamatergic synaptic release due to strong depolarization by high extracellular K⁺ has also been demonstrated in dissociated hippocampal neuronal cultures (Moulder et al. 2006). Such changes in the presynaptic depolarization during the plasticity induction period in our experiments may be involved in the expression of t-LTD through presynaptic changes in protein phosphorylation/synthesis (Collingridge et al. 2010). However, we cannot exclude the possible involvement of some other retrograde messenger(s) that does not require postsynaptic calcium for its release (Ramikie et al. 2014).

The source of calcium for t-LTD induction is also different in these two types of neurons. In the WBF neurons, NMDA receptors provide the calcium required by the postsynaptic neurons (Fig.10) whereas for the RF neurons, the source of calcium is L-type calcium channels (Fig.11). This observation suggests that there are mechanistic differences between the LTD induced by low frequency stimulation (LFS) and pairing of an EPSP with a burst of bAPs. LFS-induced LTD has been demonstrated to be NMDA receptor dependent in both the neuronal subtypes (Fidzinski et al. 2008). However, Li et al. (2005) observed subicular LTD to be dependent upon mACh receptors, when induced by pairing EPSPs with postsynaptic depolarization. HFS-induced LTP is also NMDA receptor dependent in both the neuronal subtypes (Wozny et al. 2008), although Kokaia (2000) observed HFS-induced LTP to be NMDAR independent and presynaptically induced in the mice subiculum. These reports suggest substantial diversity in the mechanisms of synaptic plasticity in the subiculum, depending on the induction protocols and neuronal subtypes. While L-type calcium channels have been associated with postsynaptic plasticity (Huang & Malenka, 1993; Kapur et al. 1998), we show blockade of postsynaptic calcium-independent t-LTD with L-type calcium channel blockers (Fig.11), which has not been reported earlier in the STDP literature. Experimental evidence of the role of L-type calcium channels in plasticity of the CA1–subiculum synapse is scant. Shor et al. (2009) have reported bi-directional modulation of synaptic plasticity by L-type calcium channels that is dependent on the type of postsynaptic neuron involved. In their study, blocking L-type calcium channels with nifedipine resulted in the conversion of LTD into LTP in burst firing neurons, and LTP into LTD in RF neurons. However, the location of the L-type calcium channels in the pre-or postsynaptic sites contributing to the plasticity changes were not discussed in their work. Few reports suggest the involvement of L-type calcium channels in presynaptic plasticity in other brain areas like the amygdala (Fourcaudot et al. 2009). In the mossy fibre synapses, L-type calcium channels have been found to play a role in synaptic facilitation and presynaptic LTP under certain conditions (Lauri et al. 2003). In Schaffer collateral–CA1 synapses in the rat hippocampus, LTP induced by short depolarization steps in the presence of a high concentration of extracellular calcium was found to be induced presynaptically and sensitive to L-type calcium channel block (Hendricson et al. 2003), although their study relied on analysis of mEPSCs only. In a recent study, the role of L-type calcium channels in the presynaptic mossy fibre LTP was proposed (Nistico et al. 2011). L-type calcium channels have been found to be expressed at the distal axons of the CA1 pyramidal neurons (Tippens et al. 2008) that synapse on the subicular pyramidal neurons. However, these studies do not specify the subicular neuronal subtypes on which the CA1 axon terminals synapse and it is difficult to generalize this observation to our findings that L-type calcium channels were specifically important in the induction of plasticity only in the RF neurons of the subiculum, and not in the WBF neurons.

Functional implications of non-Hebbian STDP in the subicular excitatory synapses

The kind of STDP curve observed here may have an important role in modulating hippocampal output since the subiculum projects to many cortical and subcortical regions, thus acting as a bridge between the hippocampus proper – CA1 in particular – and the EC (O'Mara, 2005). The subiculum has been implicated in memory formation, particularly in the memory for temporal order and novelty detection (Potvin et al. 2010), and contextual exploration (Chang & Huerta, 2012). Most of the theories related to memory formation and hippocampus argue for a hippocampal and cortical interface that transforms the short-term temporary memory traces into long-term permanent memories (Rolls, 1996; Nadel & Moscovitch, 1997), and anatomically the subiculum is perfectly placed for such a role. Hippocampal LTD has been found to be associated with novelty acquisition (Manahan-Vaughan & Braunewell, 1999), and necessary for spatial memory consolidation (Ge et al. 2010), indicating that the non-Hebbian STDP phenomenon in the subiculum may play a pivotal role in the process of learning and memory formation. The subiculum shows higher theta–gamma coherence with the area CA1 during contextual exploration (Chang & Huerta, 2012), leading to strengthening of these synapses due to heightened Hebbian LTP. The reversed STDP curve observed in our study may decelerate the synapse saturation and adjust the synaptic strength such that it does not reach the extremes of the BCM (Bienenstock, Cooper, and Munro curve for sliding threshold model of synaptic plasticity) curve and change the plasticity thresholds drastically (Stanton, 1996). For high frequency (50 Hz) synaptic inputs, the subicular BF neurons have different temporal integration properties to the RF neurons, owing to different levels of I_h expression (van Welie et al. 2006). STDP has also been demonstrated to influence the dendritic integration properties of a neuron locally (Campanac & Debanne, 2008). According to cable theory, the proximal synaptic inputs have an advantage over the distal synapses in terms of their efficacy at eliciting an action potential at the soma. Therefore, decreasing their synaptic strength via a reverse STDP rule can establish dendritic democracy (Rumsey & Abbott, 2006). The reverse STDP rule can also reduce the probability of an action potential firing by the postsynaptic cell in response to presynaptic neuronal activity (Rumsey & Abbott, 2006). This was demonstrated previously in a pair of spiny stellate neurons, wherein presynaptic action potentials failed to generate a postsynaptic action potential after the induction of reverse STDP (Egger et al. 1999). Using computer simulations, Rumsey & Abbott (2004) have predicted that a larger area under the curve for LTD in the non-Hebbian STDP plot would buffer the postsynaptic neuron from excessive firing. Since the majority of the excitatory neurons in the subiculum are intrinsically bursting in nature (Staff et al. 2000), it is important to have a mechanism to buffer excessive firing in these neurons and reduce the spread of hyperexcitability. The major route for the spread of epileptiform activity from the hippocampus to the cortical areas like the amygdala is through the subiculum (Stoop & Pralong, 2000). This observation becomes more relevant in the hyperactive networks seen in epileptic conditions, where such pairings and an HFS-like scenario may exist together. Thus, reverse STDP may decrease the probability of postsynaptic neuronal spiking and thus filter the hyperactivity of the hippocampus to prevent it from invading the entorhinal cortex and other areas. However, it is not clear if such a reverse STDP phenomenon exists under hyperactive network conditions.

In conclusion, we demonstrate an STDP plot in the proximal excitatory synaptic inputs on subicular excitatory neurons that shows LTD with causal pairing and LTP with anti-causal pairing. This t-LTD uses different calcium sources in different subtypes of postsynaptic neurons. In the RF neurons, the t-LTD is mediated through L-type calcium channels and is resistant to calcium chelation in the postsynaptic neuron, whereas in the WBF neurons calcium increase in the postsynaptic neuron is required along with the activation of NMDA receptors.

Glossary

ADP

after depolarization

AP

action potential

bAP

back propagating action potential

BF

burst firing

EC

entorhinal cortex

HFS

high frequency stimulus

I_h

hyperpolarization-activated non-selective cation channels

LFS

low frequency stimulus

LTD

long-term depression

LTP

long-term potentiation

NMDAR

NMDA receptor

PPR

paired pulse ratio

RF

regular firing

R_in

input resistance

RMP

resting membrane potential

R_s

series resistance

sACSF

sucrose based ACSF

STDP

spike-timing-dependent plasticity

t-LTD

timing-dependent LTD

WBF

weak burst firing

mGluR

metabotropic glutamate receptor

BCM

Bienenstock, Cooper, and Munro

Key points

• Spike-timing-dependent plasticity (STDP) is the induction of synaptic plasticity by coincident activity of pre- and postsynaptic neurons.

• In most of the excitatory synapses, an EPSP immediately followed by a back-propagating action potential (bAP) enhances the synaptic efficacy, whereas the reverse weakens it.

• Contrary to the above observation, we demonstrate that, at the proximal excitatory synapses on the subicular pyramidal neurons, an EPSP immediately followed by a burst of bAPs weakens the synaptic strength, whereas the reverse strengthens the synapse in both bursting and regular firing neurons.

• This reverse STDP rule may have strong implications in synaptic integration and information outflow from the hippocampus.

• Interestingly, the mechanisms associated with synaptic depression using the same induction protocol were different in the two neuronal subtypes, being postsynaptic in the bursting neurons requiring NMDA receptor activity, but presynaptic in the regular firing neurons involving L-type calcium channels.

Additional information

Competing interests

The authors declare that they have no conflict of competing interest.

Author contribution

A.P. performed the experiments and analysed the data. A.P. and S.K.S. designed the experiments, interpreted the data and wrote the manuscript.

Funding

A.P. was supported by research fellowships from CSIR, India. The research was partly supported by funds from the DBT-IISc Partnership program for Advanced Research in Biological Sciences and Bioengineering (DBT/BF/PRIns/2011–12/IISc).