Authors' present addresses C. A. Chapman: Center for Studies in Behavioral Neurobiology, Department of Psychology, Concordia University, Montréal, Québec, Canada.
S. Bertrand: Laboratoire de physiologie et physiopathologie de la signalisation cellulaire, Université Bordeaux II, Bordeaux, France.
Repetitive stimulation of Schaffer collaterals induces activity-dependent changes in the strength of polysynaptic inhibitory postsynaptic potentials (IPSPs) in hippocampal CA1 pyramidal neurons that are dependent on stimulation parameters. In the present study, we investigated the effects of two stimulation patterns, theta-burst stimulation (TBS) and 100 Hz tetani, on pharmacologically isolated monosynaptic GABAergic responses in adult CA1 pyramidal cells. Tetanization with 100 Hz trains transiently depressed both early and late IPSPs, whereas TBS induced long-term potentiation (LTP) of early IPSPs that lasted at least 30 min. Mechanisms mediating this TBS-induced potentiation were examined using whole-cell recordings. The paired-pulse ratio of monosynaptic inhibitory postsynaptic currents (IPSCs) was not affected during LTP, suggesting that presynaptic changes in GABA release are not involved in the potentiation. Bath application of the GABAB receptor antagonist CGP55845 or the group I/II metabotropic glutamate receptor antagonist E4-CPG inhibited IPSC potentiation. Preventing postsynaptic G-protein activation or Ca2+ rise by postsynaptic injection of GDP-β-S or BAPTA, respectively, abolished LTP, indicating a G-protein- and Ca2+-dependent induction in this LTP. Finally during paired-recordings, activation of individual interneurons by intracellular TBS elicited solely short-term increases in average unitary IPSCs in pyramidal cells. These results indicate that a stimulation paradigm mimicking the endogenous theta rhythm activates cooperative postsynaptic mechanisms dependent on GABABR, mGluR, G-proteins and intracellular Ca2+, which lead to a sustained potentiation of GABAA synaptic transmission in pyramidal cells. GABAergic synapses may therefore contribute to functional synaptic plasticity in adult hippocampus.
Long-term potentiation (LTP) of excitatory synaptic transmission has been examined extensively in the hippocampus. In contrast, much less is known about plasticity of GABAergic inhibitory transmission. Since hippocampal GABAergic interneurons powerfully control the excitability of CA1 pyramidal neurons (Lacaille & Schwartzkroin, 1988; Freund & Buzsáki, 1996), plasticity of inhibitory synaptic transmission may exert major influences on hippocampal excitability and function. Growing evidence suggests that the efficacy of GABAergic synapses can be persistently enhanced in neonatal rat hippocampus (reviewed by Gaïarsa et al. 2002). However, contradictory results have been reported at mature GABAergic synapses. Inhibitory postsynaptic currents (IPSCs) have been shown to be depressed during the expression of excitatory LTP in CA1 pyramidal cells (Stelzer et al. 1994) and this long-term depression (LTD) of GABAergic inhibition may underlie the EPSP-to-spike coupling associated with LTP (Lu et al. 2000).
In contrast, the activation of local inhibitory circuits in area CA1 generates polysynaptic inhibitory responses in pyramidal cells that showed potentiation following a stimulation paradigm that induces LTP at excitatory synapses (Haas & Rose, 1982). The long-lasting plasticity in polysynaptic inhibition is highly stimulus-dependent, showing long-term potentiation after theta-burst stimulation (TBS), but not following 100 Hz tetanization (Perez et al. 1999; see also Chapman et al. 1998). This activity-dependent enhancement in polysynaptic inhibition may result from changes occurring at several synaptic sites in the hippocampal circuit. First, an increase in synaptic drive onto pyramidal cells that results in an increased excitatory drive on feedback interneurons (Maccaferri & McBain, 1995), and/or a direct strengthening of excitatory synapses onto interneurons (Ouardouz & Lacaille, 1995; Perez et al. 2001) could both increase the synaptic activation of interneurons leading to an increase in the amount of GABA released. Second, the reports of lasting potentiation of miniature IPSCs (Kang et al. 1998) and monosynaptic inhibitory responses (Shew et al. 2000) in developing hippocampus, as well as a long-term increase of GABAergic transmission following repetitive stimulation in the nucleus of the solitary tract (Grabauskas & Bradley, 1999), cerebellum (Kano et al. 1992) or developing visual cortex (Komatsu, 1996), all suggest that a strengthening of GABAergic synapses themselves may also contribute to enhance inhibition, and therefore modulate neuronal excitability in parallel with plasticity at excitatory synapses.
Therefore, the aims of the present study were: (1) to examine whether long-term plasticity could occur directly at hippocampal GABAergic synapses by comparing monosynaptically evoked inhibitory responses in CA1 pyramidal neurons after TBS or 100 Hz tetani in the presence of ionotropic glutamate receptor antagonists; and if so, (2) to determine what mechanisms were responsible for such long-term changes. Our results show that TBS, but not 100 Hz trains, reliably induced LTP of monosynaptic GABAA responses. This stimulus-dependent potentiation was cooperative, requiring the activation of postsynaptic G-proteins, GABAB receptors, group I/II metabotropic glutamate receptors (mGluRs), and an increase in postsynaptic Ca2+. We also demonstrate that activation of these mechanisms is likely to require the simultaneous activation of several interneurons and/or excitatory afferents.
All experiments were carried out according to the guidelines laid down by our local Animal Care Committee at Université de Montréal. Adult male (4-6 weeks) Sprague-Dawley rats were anaesthetized with halothane (MTC Pharmaceuticals, Cambridge, Ontario, Canada) by inhalation (2 %). After decapitation, the brain was removed and placed in cold oxygenated (95 % O2-5 % CO2) artificial cerebrospinal fluid (ACSF) containing (mm): 124 NaCl, 5 KCl, 1.25 NaH2PO4, 2 MgSO4, 2 CaCl2, 26 NaHCO3 and 10 dextrose (pH ≈7.4, 295-310 mosmol l−1). Hippocampal slices were prepared with a vibrotome (Campden Instruments Ltd, Loughborough, UK) and allowed to recover in oxygenated ACSF at room temperature for at least 1 h before experiments.
Hippocampal slices (450 µm) were placed in a gas-fluid interface chamber and continuously perfused with ACSF (1.0 ml min−1) at room temperature while their upper surfaces were exposed to a humidified 95 % O2/5 % CO2 atmosphere. Intracellular recording electrodes were pulled from microfilament capillary glass on a Flaming-Brown micropipette puller (Sutter Instrument P-87, Novato, CA, USA) and filled with a solution containing 4 m potassium acetate and 0.01 m KCl. Recordings were performed at room temperature to improve the stability during long-term recordings. Membrane potential of CA1 pyramidal cells was held a few millivolts below spike threshold by constant depolarizing current injection to allow the recording of large-amplitude biphasic IPSPs. These IPSPs were amplified in bridge mode (Axoclamp 2A, Axon Instruments, Union City, CA, USA), displayed on a digital oscilloscope (model 1604; Gould, Ilford, UK), digitized using a data acquisition board (Axon Instruments TL-1-125) and stored on a microcomputer. Responses were also stored in digitized format using a video tape recorder (Neuro-Corder DR-886, Neuro Data Instruments Co., New York, USA) for later retrieval and analysis. Bridge balance was monitored regularly during recordings and adjusted as necessary.
Hippocampal slices (300 µm) were placed in a recording chamber continuously perfused with oxygenated ACSF at a rate of 2.0-3.0 ml min−1. An upright microscope (Zeiss Axioskop, Thornwood, NY, USA) equipped with a long-range water immersion objective (×40), Nomarski optics and an infrared camera (Cohu 6500, San Diego, CA, USA) was used for visual identification of cells. Patch pipettes (3-9 MΩ) were filled with a solution containing (mm): 140 potassium gluconate, 5 NaCl, 2 MgCl2, 10 Hepes, 0.5 EGTA, 10 phosphocreatine, 2 ATP-tris, 0.4 GTP-tris and 0.1 % biocytin (pH adjusted to 7.2-7.3 with KOH, 275-290 mosmol l−1). After obtaining tight seals (> 1 GΩ), whole-cell recordings were obtained from CA1 pyramidal cells in current-clamp or voltage-clamp configurations using an Axoclamp 2A or an Axopatch 1D amplifier, respectively (Axon Instruments). Experiments obtained in current-clamp were carried at 32 °C (Fig. 2) and voltage-clamp experiments were performed at 22-24 °C (Figs 3-6). Inhibitory responses were low-pass filtered at 10 KHz (−3 dB), analog filtered at 1 KHz (model 900, 8-pole bessel filter; Frequency Devices, Haverhill, MA, USA) and digitized at 20 KHz on a microcomputer using a data acquisition board (Axon Instruments TL-1-125) and pCLAMP6/7 software (Axon Instruments). In voltage-clamp experiments, recordings were discarded if series resistances increased > 25 % or decreased > 20 %.
Synaptic responses and data analysis
Synaptic responses were evoked by constant monophasic current pulses (50 µs, 25-300 µA) delivered through a stimulus isolation unit (WPI-A360, World Precision Instruments, Sarasota, FL, USA) to a stimulating electrode (fine monopolar, resin-coated tungsten electrode or concentric bipolar platinum-iridium electrode; Frederick Haer Co., Bowdoinham, ME, USA) placed in stratum radiatum of CA1 region. Monosynaptic inhibitory responses were pharmacologically isolated using 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) and (±)-2-amino-5-phosphopentanoic acid (AP5) to block non-NMDA and NMDA receptors, respectively. Stimulation intensity was adjusted so that the amplitude of the fast GABAA-mediated response was about 50 % of maximum.
Inhibitory responses were evoked every 30 s and cells were kept for further analysis only if recordings were stable over a 10 min baseline period. Slices then received either three trains of high-frequency stimulation (HFS) or three episodes of theta-burst stimulation (TBS) delivered at 30 s intervals. HFS consisted of 50 µs duration pulses repeated at 100 Hz for 1 s and an individual episode of TBS was composed of five bursts separated by 200 ms intervals (5 Hz), each burst consisting of four 100 µs pulses at 100 Hz. Following HFS or TBS, inhibitory responses were recorded for 30 min, and untetanized control cells were recorded for a similar 40 min period. Tests in various conditions were interleaved with control experiments.
The peak amplitude of early and late GABAergic responses was measured relative to the pre-stimulus baseline using Clampfit 6/7 (Axon Instruments). Potentiation of early and late responses in intracellular recordings with sharp electrodes was assessed by comparing the average responses recorded during the last 5 min of baseline period (n= 10 traces) to those acquired after 30 s, 5-10 min, 15-20 min and 25-30 min post-tetanus (n= 10 traces) using repeated measures ANOVAs followed by Newman-Keuls tests (Fig. 1). In whole-cell recordings, early IPSP/C amplitudes were compared between the last 5 min of baseline period and 25-30 min post-TBS using Student's t test. Data are expressed as means ±s.e.m. and differences were considered significant when P < 0.05.
Paired-recordings of GABAA unitary IPSCs
Unitary inhibitory postsynaptic currents (uIPSCs) were evoked during paired whole-cell recordings from a synaptically connected interneuron, located in stratum lacunosum-moleculare near its border with stratum radiatum, and a CA1 pyramidal cell, as previously described (Bertrand & Lacaille, 2001). Briefly, interneurons were recorded in whole-cell current-clamp mode using an Axoclamp 2A amplifier and pyramidal cells were voltage clamped at -40 mV with an Axopatch 200B amplifier (Axon Instruments). uIPSCs were elicited by a single presynaptic action potential triggered by somatic depolarization of the interneuron every 5-7 s. To investigate the plasticity of the interneuron-pyramidal cell synapse, the TBS was delivered to the interneuron intracellularly using current injection. During TBS, the postsynaptic pyramidal cell was held in current-clamp mode. uIPSCs were analysed as previously described (Bertrand & Lacaille, 2001). In summary plots, data from 13 to 18 consecutive IPSCs were binned and averaged across experiments.
Concentrated stock solutions of CNQX, AP5 (Tocris Cookson, Ellisville, MO, USA), CGP55845 (Novartis, Basel, Switzerland), acetazolamide and (RS)-α-ethyl-4-carboxyphenylglycine (E4-CPG; Tocris Cookson) were prepared in advance, frozen at -20 °C, and diluted to their final concentrations on the day of experiment. These agents were bath-applied throughout the recording period, except acetazolamide which was perfused starting 20 min before the beginning of experiment. For experiments with blocking activation of postsynaptic G-proteins, GTP was replaced by 100 µm guanosine 5′-O-(2-thiodiphosphate) (GDP-β-S) in the intracellular solution. In experiments where the Ca2+ chelator BAPTA was added to the patch solution, potassium gluconate was reduced to 110 mm and EGTA was removed. All chemicals were purchased from Sigma (St Louis, MO, USA) unless otherwise indicated.
In most whole-cell recordings, biocytin was added to the patch solution to confirm morphologically that cells were pyramidal neurons. After recordings, slices containing biocytin-filled cells were fixed at 4 °C overnight with 4 % paraformaldehyde in 0.1 m phosphate buffer (PB) and were then washed and stored in 0.1 m PB. To reveal biocytin, the slices were processed using the Vectastain ABC kit (Vector Laboratories, Burlingame, CA, USA) as previously described (Chapman & Lacaille, 1999a), air-dried for 24 h, mounted with DPX (Electron Microscopy Sciences, Fort Washington, PA, USA) and observed under a light microscope.
TBS induces LTP at GABA synapses
Stable intracellular recordings were obtained with sharp electrodes from 30 CA1 pyramidal cells that had a mean resting membrane potential of -56.9 ± 0.9 mV, a mean input resistance of 63 ± 3 MΩ and a mean action potential amplitude of 85.6 ± 1.4 mV. When cells were held at subthreshold membrane potentials (−52.2 ± 0.8 mV), Schaffer collateral stimulation in the presence of CNQX (20 µm) and AP5 (50 µm) evoked biphasic IPSPs composed of an early GABAA-mediated component (peak latency: 27.5 ± 1.1 ms) and a late component dependent on GABAB receptors (peak latency: 269.0 ± 5.6 ms). In untetanized cells, the amplitude of early IPSPs was stable for the 40 min recording period (99 ± 3 % of baseline), but the amplitude of late IPSPs slowly decreased to 86 ± 5 % of baseline after 40 min (Fig. 1A and C; n= 10). A similar small ‘run-down’ of late IPSPs was also observed previously (Perez et al. 1999). When TBS was delivered, the amplitude of early IPSPs increased gradually to 114 ± 3 % of baseline at 5-10 min post-TBS, remained elevated at 15-20 min post-TBS (110 ± 2 %; P < 0.05, repeated measures ANOVA) and at 25-30 min post-TBS (106 ± 3 % of baseline; P < 0.05, one-tailed; Fig. 1A and C), whereas the amplitude of late IPSPs declined to 87 ± 7 % of baseline, a level similar to that observed in untetanized cells. In contrast, both early and late IPSPs were depressed to 68 ± 8 and 52 ± 17 % of baseline, respectively, immediately after high-frequency stimulation (HFS) (Fig. 1B; P < 0.05). IPSPs then returned to control levels within 5 min post-HFS and remained constant through 30 min (early: 97 ± 5 % of baseline, late: 79 ± 4 % of baseline; Fig. 1C). Thus, while a selective LTP of early GABAA-mediated IPSPs was induced by TBS, HFS resulted in a transient depression of both early and late IPSPs and did not exert long-term changes.
Postsynaptic G-protein activation is required for LTP
We next investigated the mechanisms of TBS-induced potentiation of early GABAA responses using whole-cell recordings. To determine whether LTP could be elicited at a temperature closer to physiological conditions, we carried out current-clamp experiments at 32 °C. Similar to that observed with intracellular recordings using sharp microelectrodes at room temperature, the amplitude of early IPSPs increased progressively and was potentiated to 122 ± 9 % of baseline levels 30 min after TBS (n= 11; Fig. 2), suggesting that LTP of GABAA synaptic transmission can be expressed at a near-physiological temperature. Since G-protein-coupled receptors are involved in tetanization-induced LTP at inhibitory synapses in visual cortex (Komatsu, 1996), we next examined whether blocking postsynaptic G-proteins affected the potentiation. This was done by replacing GTP in the patch solution with GDP-β-S (0.1 mm). This substitution did not alter the baseline IPSPs (Fig. 2A) but abolished TBS-induced LTP (101 ± 6 % of baseline 30 min after TBS; n= 5), indicating that activation of postsynaptic G-proteins in CA1 pyramidal cells is required for the induction of LTP of GABAA synaptic responses.
LTP of synaptic currents does not involve changes in GABA release
Whole-cell voltage-clamp recordings were next conducted to verify that GABAA inhibitory postsynaptic currents (IPSCs) could also be potentiated (holding potential (Vh) = -55 to -60 mV). Cells (n= 59) had electrophysiological properties (input resistance, spike accommodation) characteristic of pyramidal cells and for 44 of them, biocytin-labelling confirmed their pyramidal nature. Monosynaptic early IPSCs were completely blocked by bath application of bicuculline (25 µm), confirming that they were mediated by GABAA receptors (data not shown). Consistent with current-clamp results, we found that early IPSCs gradually increased in the first 2-5 min following TBS (Fig. 3B) and were significantly potentiated to 141 ± 13 % of baseline levels 30 min post-TBS (n= 12). This potentiation was not attributed to changes in intrinsic properties of pyramidal cells since input resistance (158 ± 11 MΩ during baseline vs. 163 ± 18 MΩ post-TBS) and holding current (−31 ± 11 pA during baseline vs. -40 ± 14 pA post-TBS) were not affected by LTP (P > 0.3). The reversal potential of IPSCs (EIPSC) was shifted from -64.9 ± 0.5 mV during baseline to -69.1 ± 0.6 mV during LTP (P < 0.05; n= 5). However, it is unlikely that displacement in EIPSC contributed to LTP because EIPSC was also significantly shifted towards hyperpolarized potentials at the end of recording in untetanized cells (baseline: -64.6 ± 1.2 mV; 30 min later: -66.8 ± 1.3 mV; P < 0.05; n= 7), in which IPSCs were not potentiated (102 ± 5 % of baseline; n= 11). Although the shift in EIPSC appears more pronounced after TBS than in untetanized cells (−4.2 mV for TBS cells vs. -2.2 mV for untetanized cells), this difference was not significant (t test; P > 0.1). Therefore, these results argue against a modification of Cl− gradient in LTP, and suggest that the potentiation is mediated by an increase in GABAA receptor activity.
We wondered whether the potentiation of IPSCs was accompanied by presynaptic changes in GABA release. To examine this question, we tested whether changes in the paired-pulse ratio (PPR), which is representative of changes in the probability of transmitter release (Zucker, 1989), could be observed during LTP. Paired-pulse stimulation given at a 75 ms interval generated two successive IPSCs that showed paired-pulse depression. In untetanized cells, IPSC amplitude and PPR were both stable throughout the recording period (Fig. 3Cc1 and D; PPR during baseline: 0.77 ± 0.10; 30 min later: 0.86 ± 0.09; P > 0.1; n= 5). During LTP, the increase in IPSC amplitude was not accompanied by a change in PPR (Fig. 3Cc2 and D; baseline ratio: 0.73 ± 0.03; 30 min post-TBS: 0.74 ± 0.04; n= 5), suggesting that presynaptic mechanisms governing the release of GABA were not affected during LTP.
GABAB receptor activation in LTP
G-protein-coupled receptors play important roles in the modulation of GABA synaptic transmission (Cartmell & Schoepp, 2000). Notably, GABAB receptors have been reported to regulate inhibitory synaptic transmission following repeated stimulation and/or postsynaptic depolarization (Komatsu, 1996; Kawaguchi & Hirano, 2000; Kotak et al. 2001). We therefore wanted to determine whether G-protein-dependent mechanisms in LTP involved GABAB receptors. The application of the GABAB receptor antagonist CGP55845 (1-5 µm) affected GABA transmission in two ways. First, the amplitude of IPSCs was transiently depressed to 85 ± 8 % of baseline immediately after TBS in the presence of CGP55845 (n= 10), as compared to a small increase to 109 ± 5 % of baseline in its absence (Fig. 4B; n= 12; t test, P < 0.05). Second, CGP55845 produced variable effects on the potentiation of IPSCs, blocking LTP in half of cells tested (5/10) but not in the remaining ones. Overall, however, TBS did not produce significant LTP in the presence of CGP55845 (118 ± 12 % of baseline; Fig. 4B and C; P > 0.1), suggesting that LTP of GABA synapses depends on GABAB receptor activation.
The transient depression of IPSCs following TBS in the presence of the GABAB receptor antagonist may result from an alteration of synaptic transmission during the episodes of TBS because a slow inward current reaching -23.0 ± 5.9 pA (n= 4) was generated in presence of CGP55845, but not in its absence (−3.3 ± 1.1 pA; n= 5; Fig. 4D). A similar inward current has been described in CA1 pyramidal cells following intense stimulation and prolonged activation of GABAA receptors (Kaila et al. 1997). It has been attributed to bicarbonate (HCO3−) efflux through GABAA receptors (Grover et al. 1993) or HCO3−-dependent increase in extracellular K+ concentration (Smirnov et al. 1999). Analogous mechanisms appear responsible for the generation of the inward current during TBS because this current almost completely disappeared in the presence of acetazolamide (2 µm), an inhibitor of carbonic anhydrase, the enzyme synthesizing HCO3− (Fig. 4D; -6.7 ± 2.4 pA; n= 8). It is therefore possible that CGP55845, by blocking GABAB receptor-mediated presynaptic inhibition, may increase GABA release and prolong GABAA receptor activation. This may in turn result in dendritic Cl− accumulation and collapse of Cl− gradient (Staley et al. 1995; Staley & Proctor, 1999), which may promote HCO3−-dependent currents.
To determine whether this inward current is responsible for the transient depression of IPSCs and inhibition of LTP, we examined whether simultaneous application of CGP55845 and acetazolamide would eliminate the depression and restore LTP. While the inward current disappeared, the transient depression (83 ± 7 % of baseline) and inhibition of LTP (122 ± 13 % of baseline; n= 8) persisted in the presence of CGP55845 and acetazolamide (Fig. 4B and C). The results indicate that inhibition of presynaptic GABAB receptors generate a HCO3−-dependent inward current during TBS, but this is not responsible for the inhibition of LTP of IPSCs. Thus, GABAB receptors located postsynaptically, rather than presynaptically, may be those critical for LTP at GABA synapses.
Group I/II mGluR antagonist attenuates LTP
As described above, GABAB receptor blockade did not always inhibit LTP, suggesting that other mechanisms might also participate in this potentiation. Because fast excitatory transmission by ionotropic glutamate receptors was blocked, but metabotropic glutamate receptors (mGluRs) could still be activated, we tested whether mGluR-dependent mechanisms could be involved in LTP. The delivery of TBS during the application of 500 µm E4CPG, a group I/II mGluR antagonist, resulted in an initial gradual increase in IPSCs that was similar to that observed in control (Fig. 5B). However, this potentiation reached a plateau about 10-15 min after TBS such that IPSCs were only enhanced to 115 ± 7 % of baseline 30 min post-TBS (n= 7), a level that was not significantly different from baseline (Fig. 5B and C). This inhibition of LTP therefore indicates that group I/II mGluRs may also participate in LTP of inhibitory synaptic transmission.
A postsynaptic Ca2+ rise is needed at GABA synapses to elicit LTP
Given that mGluRs can trigger postsynaptic Ca2+ elevations in hippocampal neurons (Murphy & Miller, 1988), we next wanted to test whether an increase in postsynaptic Ca2+ is required to elicit LTP of IPSCs. As shown in Fig. 5B, LTP was induced when the basal postsynaptic Ca2+ level was only moderately buffered by a low concentration (0.5 mm) of the Ca2+ chelator EGTA present in the patch pipette (see Methods; same as in Fig. 3B). In contrast, when a high concentration of the Ca2+ chelator BAPTA (10 mm) was substituted for EGTA to prevent Ca2+ elevation, LTP was abolished (Fig. 4B and C; 95 ± 6 % of baseline; n= 9). These results therefore indicate that, as found for plasticity at excitatory synapses (Malenka, 1991), a rise in postsynaptic Ca2+ is also required for the induction of LTP at GABA synapses onto CA1 pyramidal cells.
Short-term plasticity elicited by TBS in interneuron-pyramidal cell pairs
The results described so far were obtained using extracellular stimulation in stratum radiatum which activates several inhibitory/excitatory fibres. To determine whether LTP could also be induced by activation of a single interneuron, we obtained paired whole-cell recordings from synaptically connected interneurons in stratum lacunosum-moleculare (LM) and pyramidal cells (PYR). Five of 93 LM-PYR pairs obtained were synaptically connected, which corresponds to 5.4 % of all cell pairs tested (Bertrand & Lacaille, 2001). The intracellular delivery of TBS to a single LM interneuron resulted in short-term potentiation (STP) of average unitary IPSC (uIPSC) amplitude in three of the five pairs tested (Fig. 6), and no changes in the remaining two pairs. In the three pairs with changes, the average amplitude of uIPSCs (including failures) was significantly increased until 180 s post-TBS and returned to control values 360 s post-TBS (Fig. 6Aa1 and Bb1). This STP of uIPSCs was associated with a significant decrease in failure rate, as the percentage of failures decreased from 75.3 ± 3 to 62.2 ± 4 % 90 s post-TBS and further to 60.7 ± 13 % 180 s post-TBS, before returning to control level 360-450 s after TBS (Fig. 6Bb2). In contrast, the amplitude of uIPSC was not affected when failures of synaptic transmission were excluded (Fig. 6Aa2 and Bb3). These results demonstrate that at synapses between individual LM interneuron and pyramidal cells, TBS activation elicits solely short-term changes, which are likely to involve presynaptic mechanisms. Thus, cooperativity among inhibitory and excitatory inputs appears to be required for LTP.
The major findings of the present study are that tetanization at 100 Hz produced short-term depression of both GABAA and GABAB responses of CA1 pyramidal cells, whereas theta-patterned stimulation induced selective long-term potentiation of monosynaptic GABAA responses. This is markedly different from plasticity at excitatory synapses of CA1 pyramidal cells, where both tetanic and theta-patterned stimulation reliably induce LTP (Larson et al. 1986; Davies et al. 1991). The examination of the signalling pathways mediating TBS-induced LTP of inhibition revealed that postsynaptic G-protein-dependent mechanisms are necessary for LTP induction, probably through activation of postsynaptic GABAB receptors and group I/II mGluRs, and that a rise in postsynaptic Ca2+ is also required. Moreover, we found that concomitant activation of several interneurons/ presynaptic fibres appears necessary to induce LTP since the intracellular delivery of TBS to a single interneuron resulted only in a transient increase in GABAA responses at individual interneuron-pyramidal cell synapses.
Synaptic inhibition regulated at several sites in the hippocampal circuit
Perez et al. (1999) previously reported that both early GABAA and late GABAB components of polysynaptic IPSPs, evoked in pyramidal cells by stimulation of local inhibitory circuits, displayed long-lasting enhancement following theta-patterned stimulation. A long-lasting enhancement was present following 100 Hz tetanization only when postsynaptic Ca2+ elevation was prevented. This activity-dependent plasticity of polysynaptic responses was also prevented by NMDA receptor blockade. In contrast, the present study of pharmacologically isolated monosynaptic GABAergic responses has determined that only GABAA-mediated monosynaptic IPSPs are enhanced after theta-patterned stimulation by NMDA receptor independent processes. The different mechanisms involved in the long-term plasticity of polysynaptic versus monosynaptic responses indicate that these phenomena may involve changes at distinct sites in the circuit. For monosynaptic responses, direct modulation of GABA synapses may be induced by TBS. For polysynaptic responses, since feedforward and feedback excitatory inputs to interneurons are functional, TBS-induced LTP at excitatory synapses of pyramidal cells may passively propagate to interneurons (Maccaferri & McBain, 1995) and/or LTP may occur directly at excitatory synapses of interneurons (Ouardouz & Lacaille, 1995; Perez et al. 2001). These changes may lead to an increase in synaptic activation of interneurons, resulting in an enhancement of GABA release and potentiation of inhibition in pyramidal cells. Thus, theta-patterned activation may influence inhibition of pyramidal neurons by regulating synaptic transmission at many levels in the hippocampal circuit. Moreover, the activation of multiple processes by theta-patterned stimulation could explain why a potentiation of late IPSPs was observed only for polysynaptic and not for monosynaptic responses, since changes leading to enhanced activation of interneurons and GABA release are observed solely in polysynaptic recording conditions.
Frequency-dependent plasticity of monosynaptic GABAA responses
The modulation of GABAergic synapses in the adult hippocampal CA1 region appears to be highly dependent on tetanization parameters and may switch from a sustained enhancement of GABAA synaptic function to a transient depression of both GABAA and GABAB responses with increased frequency of repetitive stimulation. The depression following 100 Hz tetani could involve presynaptic mechanisms because both early and late components of IPSPs were affected. A transient depression of early IPSCs was also observed when TBS was given in presence of a GABAB receptor antagonist. Although we did not investigate the underlying mechanisms, we believe that a shift in the Cl− gradient leading to dendritic inward current and suppression of GABA responses is unlikely to contribute since the depression persisted even though the inward current was prevented with the carbonic anhydrase inhibitor acetazolamide. Alternatively, a depletion of releasable vesicles caused by increased GABA release during sustained tetanization (HFS), as well as during TBS when GABAB receptor-mediated presynaptic inhibition was removed, could account for the short-term depression (Jensen et al. 1999b). In turn, this depletion of releasable vesicles may have precluded the activation of postsynaptic mechanisms, which could explain the absence of LTP of monosynaptic GABAA responses following HFS.
The effectiveness of theta-patterned stimulation in inducing LTP at excitatory synapses has been linked to a transient reduction in GABAergic transmission via presynaptic GABAB autoreceptors (Davies et al. 1991). For the potentiation of GABAA transmission, however, TBS may result in LTP because of an effective activation of postsynaptic GABAB receptors, mGluRs and Ca2+-dependent processes, while not activating other mechanisms that contribute to depression (Stelzer et al. 1994; Stelzer & Shi, 1994).
LTP involves postsynaptic mechanisms
The suggestion that postsynaptic changes are involved in LTP is supported by the observations that: (1) TBS potentiates selectively GABAA, but not GABAB responses, (2) PPR is unaffected by LTP, (3) the reduction of LTP by a GABAB receptor antagonist persisted after the elimination of the slow inward current resulting from presynaptic GABAB receptor blockade, (4) blocking activation of postsynaptic G-proteins prevents LTP and (5) the LTP is also prevented by postsynaptic injection of BAPTA. GABAA receptors contains phosphorylation sites for protein kinases and their function may be regulated depending on their phosphorylation/dephosphorylation state (Smart, 1997). Since GABAB receptors are linked to cyclic AMP (cAMP) production (Cunningham & Enna, 1996), signalling via cAMP/protein kinase A (PKA) pathways could participate in LTP. Although reports concerning the potentiation of hippocampal GABAA receptor function by PKA-dependent phosphorylation remain controversial (Kapur & Macdonald, 1996; Poisbeau et al. 1999), such effects on GABAA synaptic transmission need to be explored in detail in CA1 region. Alternatively, modifications in protein phosphatases activity might also be involved (Wang et al. 2003).
A displacement of EIPSC toward hyperpolarized potentials was observed during whole-cell recordings, which should increase the Cl− driving force. Activity-dependent shifts in Cl− reversal potential have been reported, but these changes were toward positive membrane potentials which decreased inhibitory transmission (Sun et al. 2000; Gusev & Alkon, 2001). During development, the action of GABA switches from excitatory to inhibitory because of an increased expression of the K+-Cl− cotransporter 2 (KCC2), which shifts the Cl− reversal potential toward more hyperpolarized potentials (Clayton et al. 1998). However, it is unlikely that the shifts in EIPSC were critically involved in the potentiation of IPSCs in the present study since similar shifts were also present in untetanized cells that did not show potentiation of IPSCs.
While long-term GABAergic plasticity has been reported to result from postsynaptic changes in layer V neurons of the visual cortex (Komatsu, 1996) and in layer II/III neocortical pyramidal cells (Holmgren & Zilberter, 2001), LTP of GABAergic transmission in the hippocampus has mainly been linked to presynaptic changes. Shew et al. (2000) demonstrated that 100 Hz tetanization induces a GABAB receptor-dependent LTP of early IPSPs in rat CA1 pyramidal neurons during development. The GABAB receptors that mediate this effect are thought to be located presynaptically because irreversible activation of postsynaptic G-proteins by GTP-γ-S did not prevent the potentiation. In addition, Kang et al. (1998) described a potentiation of GABAA synaptic transmission induced by repetitive activation of interneurons that triggers GABAB receptor-dependent intracellular Ca2+ elevation in astrocytes. This potentiation was also induced by direct stimulation of astrocytes and blocked by preventing astrocytic Ca2+ rises. However, since ionotropic glutamate receptor antagonists blocked this astrocyte-mediated potentiation, it is unlikely that such mechanisms contribute to the TBS-induced potentiation of monosynaptic IPSP/Cs observed here. Thus, the present results are, to our knowledge, the first indication that long-term potentiation at adult hippocampal GABA synapses involves postsynaptic GABAB receptor activation. These results are consistent with a growing body of evidence reflecting the importance of GABAB receptors for the long-term regulation of inhibitory synaptic transmission in different brain areas (Komatsu, 1996; Kawaguchi & Hirano, 2000, 2002; Kotak et al. 2001).
Cooperativity requirement for LTP
The intracellular delivery of TBS to a single interneuron resulted in a short-term increase in uIPSCs in pyramidal cells. A similar transient IPSC increase has been observed following tetanic stimulation in paired recordings in culture (Jensen et al. 1999a). The absence of long-term increase during paired recordings suggests that the activation of a single interneuron is insufficient for LTP induction at GABA synapses. This may result from the fact that simultaneous stimulation of several interneurons is necessary to activate postsynaptic GABAB receptors (Scanziani, 2000; Bertrand et al. 2001) or that mechanisms other than those activated by GABA are involved. Interestingly, Komatsu (1996) reported that tetanization-induced LTP of inhibitory transmission in the visual cortex required GABAB receptor activation as well as postsynaptic Ca2+ rise linked to α1-adrenoceptor and/or 5-HT2 receptor activation. The implication of GABA-independent mechanisms in TBS-induced LTP is supported by the observation that the potentiation of IPSCs is prevented by a group I/II mGluR antagonist. mGluRs play important roles in excitatory LTP in area CA1 (Bashir et al. 1993) as well as in LTP at excitatory synapses of oriens-alveus interneurons (Perez et al. 2001). However, while it is well known that mGluRs can decrease the release of GABA (Desai et al. 1994; Gereau & Conn, 1995), information concerning their role in long-term regulation of inhibitory transmission is scarce (Liu et al. 1993; Chevaleyre & Castillo, 2003).
How can mGluRs affect the transmission at GABA synapses? TBS also triggers the release of glutamate from excitatory fibres, which may spillover and activate perisynaptic mGluRs located at proximity of GABAergic synapses on pyramidal cells (Baude et al. 1993; Lujan et al. 1997). Given that group I mGluRs are linked to inositol 1,4,5-trisphosphate (IP3) formation and release of Ca2+ from internal stores (Murphy & Miller, 1988), they could trigger Ca2+ signals leading to LTP. Because in some cells, a residual LTP persisted in presence of the GABAB receptor antagonist, it is possible that the strength of stimulation of Schaffer collateral fibres which impinged on those cells may have been sufficient to provide appropriate mGluR-mediated signalling to induce potentiation. Further, it remains to be examined if, as reported for astrocytes (Kang et al. 1998), GABAB receptors can generate Ca2+ rises in CA1 pyramidal cells which may be sufficient for LTP induction. Thus, we suggest that the co-activation of GABAergic and glutamatergic mechanisms, either in parallel or in synergy (Hirono et al. 2001), may cooperatively lead to LTP of GABAA synaptic transmission in adult pyramidal cells.
The hippocampal theta rhythm may contribute to learning and memory by promoting plasticity at excitatory synapses. Our work indicates that theta-patterned activity also induces plasticity at GABAergic synapses. What is the physiological relevance of this strengthening of GABA synapses ? LTP can be induced at Schaffer collateral-pyramidal cell synapses by either TBS or 100 Hz tetani. However, LTP of field EPSPs (fEPSPs) was of larger magnitude following TBS than 100 Hz tetani when GABAA inhibition was blocked by bicuculline (Chapman et al. 1998). This unmasking of LTP of fEPSPs by bicuculline may reflect a sustained increase in inhibition induced by TBS, opposing the LTP at excitatory synapses (Chapman et al. 1998). Thus, the physiological relevance of LTP at GABA synapses may be an activity-dependent regulation of excitability of pyramidal cells. In addition, since inhibitory interneurons contribute to pacing of theta oscillations in the hippocampus (Chapman & Lacaille, 1999a,b), theta-dependent strengthening of GABA synapses may serve to reinforce their role in hippocampal theta activity. Finally, rhythmic activation of excitatory and inhibitory inputs to pyramidal cells appears necessary to cooperatively promote plasticity of GABA synapses. Such cooperative induction of plasticity at inhibitory synapses may provide a novel mechanism through which hippocampal excitability can be modulated adaptively in an activity-dependent manner, and thus participate in the processes mediating learning and memory.
This research was funded by grants to J.-C.L. from the Canadian Institutes of Health Research (CIHR; MT-10848), Fonds de la Recherche en Santé du Québec, and Fonds pour la Formation de Chercheurs et l'Aide à la Recherche. J.-C.L. is the recipient of a Canada Research Chair in Cellular and Molecular Neurophysiology, and a member of the CIHR Group on Synaptic Transmission and Plasticity.