Selective muscarinic regulation of functional glutamatergic Schaffer collateral synapses in rat CA1 pyramidal neurons

Abstract

Analysis of the cholinergic regulation of glutamatergic neurotransmission is an essential step in understanding the hippocampus because it can influence forms of synaptic plasticity that are thought to underlie learning and memory. We studied in vitro the cholinergic regulation of excitatory postsynaptic currents (EPSCs) evoked in rat CA1 pyramidal neurons by Schaffer collateral (SC) stimulation. Using ‘minimal’ stimulation, which activates one or very few synapses, the cholinergic agonist carbamylcholine (CCh) increased the failure rate of functional more (36 %) than of silent synapses (7 %), without changes in the EPSC amplitude. These effects of CCh were insensitive to manipulations that increased the probability of release, such as paired pulse facilitation, increases in temperature and increases in the extracellular Ca²⁺:Mg²⁺ ratio. Using ‘conventional’ stimulation, which activates a large number of synapses, CCh inhibited more the pharmacologically isolated non-NMDA (86 %) than the NMDA (47 %) EPSC. The changes in failure rate, EPSC variance and the increased paired pulse facilitation that paralleled the inhibition imply that CCh decreased release probability. Muscarine had similar effects. The inhibition by both CCh and by muscarine was prevented by atropine. We conclude that CCh reduces the non-NMDA component of SC EPSCs by selectively inhibiting transmitter release at functional synapses via activation of muscarinic receptors. The results suggest that SCs have two types of terminals, one in functional synapses, selectively sensitive to regulation through activation of muscarinic receptors, and the other in silent synapses less sensitive to that regulation. The specific inhibition of functional synapses would favour activity-dependent plastic phenomena through NMDA receptors at silent synapses without the activation of non-NMDA receptors and functional synapses.

The hippocampus receives a cholinergic projection from the medial septal nucleus (Lewis & Shute, 1967; Frotscher & Leranth, 1985) and acetylcholine (ACh) is released in the hippocampus when the septum is stimulated (Dudar, 1977). Moreover, evidence has been provided of the existence of intrahippocampal cholinergic interneurons (Frotscher et al. 2000). Cholinergic interaction with glutamatergic transmission (Aigner, 1995) can influence certain forms of synaptic plasticity that are thought to underlie learning and memory (Madison et al. 1991; Goda & Stevens, 1996; Hasselmo, 1999) and lesions of the cholinergic septo-hippocampal projections produce memory and attention deficits (see e.g. Hasselmo, 1999). The cholinergic input from the medial septal nucleus is essential for triggering the ‘atropine-sensitive’ in vivo rhythmic network oscillations at 4–7 Hz called hippocampal theta (θ; Macadar et al. 1970; Dudar, 1977; Gaztelu & Buño, 1982), which have also been related to learning and memory processes (Huerta & Lisman, 1995). In addition, synaptically released ACh can induce transient Ca²⁺ elevations in CA1 astrocytes (Araque et al. 2002), and the astrocytic Ca²⁺ signal may induce the release of glutamate, which may influence synaptic transmission (Araque et al. 2001).

Extensively studied, the glutamatergic synapse between CA3 and CA1 pyramidal neurons via Schaffer collateral (SC) axons represents a classical model of synaptic plasticity. In this model, ACh, via activation of muscarinic cholinergic receptors (mAChR), has both pre- and postsynaptic actions. Postsynaptically, activation of mAChRs increases the excitability of CA1 pyramidal neurons by inhibiting several K⁺ currents (Benardo & Prince, 1982; Madison et al. 1987). Moreover, activation of mAChR increases postsynaptic Ca²⁺ and N-methyl-d-aspartate (NMDA) receptor (NMDAR)-mediated responses (Harvey et al. 1993; Marino et al. 1998). Presynaptically, activation of mAChRs inhibits excitatory afferents (Hounsgaard, 1978; Valentino & Dingledine, 1981), reducing the release of glutamate through the inhibition of voltage-gated Ca²⁺ channels at synaptic terminals via a G-protein-coupled signalling pathway (Qian & Saggau, 1997).

Several groups have described the existence of ‘silent synapses’ that contain NMDARs but no electrophysiologically detectable α-amino-3-hydroxy-5-methyl-4-isoxazole propionate (AMPA) receptors (AMPARs), and of ‘functional synapses’ (i.e. conducting synapses) that include both NMDARs and AMPARs (Isaac et al. 1995; Liao et al. 1995). Recent evidence has shown different modes of expression of AMPARs and NMDARs at SC synapses (Takumi et al. 1999), which may represent the structural counterpart of functional and silent synapses. Moreover, correlations between the type of glutamate receptor expressed and specific morphological features of dendritic spine (Toni et al. 1999), postsynaptic density (Takumi et al. 1999) and presynaptic arborization (Cantallops et al. 2000) have been reported. However, the possible differential cholinergic modulation of these morphologically and functionally dissimilar glutamatergic SC synapses that may underlie its cognitive influences is unknown.

We examined the action of the non-hydrolysable cholinergic agonist carbamylcholine-chloride (CCh) on the excitatory postsynaptic currents (EPSCs) evoked by SC ‘minimal’ stimulation, which activates a single or very few synapses, in functional and silent synapses. We also analysed the effect of CCh on the pharmacologically isolated NMDA and non-NMDA (i.e. AMPA/kainate) components of the EPSCs evoked by ‘conventional’ stimulation, which activates a large number of SCs, and compared them with the actions of adenosine, which also inhibits transmitter release (Wu & Saggau, 1994). We report that CCh specifically inhibits the non-NMDA component of SC EPSCs by selectively regulating transmitter release from functional synapses via activation of presynaptic muscarinic cholinergic receptors (mAChRs).

Methods

Procedures of animal care, surgery and slice preparation will be described briefly because they have been extensively detailed previously (Borde et al. 2000; Martín et al. 2001). All experiments in this study conformed to International Guidelines on the ethical use of animals and every effort was made to minimize the suffering and number of animals used.

Preparation and recordings

Fourteen- to 18-day-old Wistar rats were anaesthetized with ether, decapitated and the brain was removed and submerged in cold (≈4 °C) artificial cerebrospinal fluid (ACSF; for composition see below) maintained at pH 7.3 by bubbling with carbogen (95 % O₂-5 % CO₂). Transverse hippocampal slices (300-350 μm) were cut with a vibratome (model 101, Pelco, St Louis, MI, USA) and incubated in ACSF (>1 h at room temperature of 20–22 °C). Slices were transferred to an immersion recording chamber placed either on the stage of an inverted (Nikon Diaphot-TMD, Nikon, Tokyo, Japan) or an upright microscope (Olympus BX50WI, Olympus Optical, Tokyo, Japan) equipped with infrared and differential interference contrast imaging devices, and with a ×40 water immersion objective. Slices were superfused at a rate of 1–10 ml min⁻¹ with gassed (95 % O₂-5 % CO₂) ACSF and either maintained at room temperature or at 32–35 °C. At the highest superfusion rate total renewal of the solution in the chamber took ≈3 min.

Patch-clamp recordings in the whole-cell configuration were performed using 3–6 MΩ fire-polished pipettes (pulled with a P87 puller, Sutter Instruments, Novato, CA, USA). Pipettes were connected to an Axoclamp 2B amplifier (Axon Instruments, Foster City, CA, USA) or a PC-ONE amplifier (Dagan Corp., Minneapolis, MN, USA), cells were recorded in the continuous single electrode voltage-clamp mode, and held at a membrane potential (V_m) of −60 or +60 mV (Liao et al. 1995) unless otherwise specified. The pipette was positioned in the stratum pyramidale under direct visualization of the CA1 region and pyramidal neurons were either viewed with the Olympus microscope or identified by the adaptation of regular spiking neurons with wider spikes in response to current injection in current-clamp recordings (Borde et al. 2000). Fast and slow capacitances were neutralized and series resistance was compensated (≈80 %). Patch recordings were rejected when the access resistance (7-15 MΩ) increased >20 % during the experiment.

Synaptic stimulation

The control ACSFs and the internal pipette solutions were different in experiments in which conventional and minimal stimulation were used. The solution used in conventional stimulation experiments was designed to reproduce a more physiological state without unwanted modifications of the resting potential and input resistance, except notably when CsCl was included in the pipette solution. This condition enabled recordings under voltage clamp and current clamp, the latter without the need of current injection to maintain a normal resting potential (see e.g. Barnes-Davies & Forsythe, 1995). The solution used in minimal stimulation experiments was the usual one used to reduce noise and isolate small EPSCs (see e.g. Isaac et al. 1995).

Minimal stimulation

Minimal bipolar stimulation of SCs was through a pipette pulled as a patch electrode using a theta capillary and placed in the stratum radiatum near (≈100 μm) the tip of the recording pipette. Single or paired (50 ms interval) pulse stimuli were delivered at 1.0, 0.33 or 0.1 s⁻¹ through two silver chloride electrodes placed in the pipette compartments. The control ACSF used in minimal stimulation experiments was (mm): 119.0 NaCl, 2.5 KCl, 1.0 KH₂PO₄, 1.3 MgSO₄, 26.2 NaHCO₃, 2.5 CaCl₂, 30.0 sucrose, 0.05 picrotoxin, 0.01 glycine and 11.0 glucose at pH 7.3 and the internal pipette solution contained (mm): 107.5 caesium gluconate, 8.0 NaCl, 0.2 EGTA, 20.0 Hepes, 10.0 TEA-Cl, 4.0 Mg-ATP and 0.3 GTP at pH 7.3 (adjusted with CsOH). In some experiments, to increase the release probability, the CaCl₂ concentration was raised to 5.2 mm without changing the concentration of MgSO₄, thus increasing the Ca²⁺:Mg²⁺ ratio.

Conventional stimulation

Stimulation of SCs was through bipolar nickel-chrome wire electrodes (80 μm diameter) placed at the stratum radiatum close to the pyramidal layer (≈100 μm) to evoke proximal excitatory input. Pulses separated by 50–100 ms (Grass S88 and SIU, Grass, Quincy, MA, USA), delivered at 0.33 s⁻¹, were used to evoke EPSCs. The internal patch pipette solution contained (mm): 100.0 potassium gluconate; 5.0 EGTA, 10.0 Hepes, 32.5 KCl, 1.0 MgCl₂ and 4.0 Na₂-ATP at pH 7.3 (adjusted with KOH). In some cases, 100.0 mm potassium gluconate was equimolarly substituted with caesium gluconate to block K⁺ conductances and no important differences were observed in the EPSC results. The control ACSF was as follows (mm): 124.0 NaCl, 2.69 KCl, 1.25 KH₂PO₄, 2.0 MgSO₄, 26.0 NaHCO₃, 2.0 CaCl₂, 0.05 picrotoxin, 0.01 glycine and 10.0 glucose at pH 7.3 (adjusted with NaOH). In the Mg²⁺-free solution, MgSO₄ was equimolarly replaced with CaCl₂. Muscarine (5 μm), atropine (10 μm), adenosine (50 μm) and CCh (5 μm) were added to the ACSF and superfused. Glutamate receptor antagonists, dl-2-amino-5-phosphonovaleric acid (APV; 50 μm) and 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX; 20 μm), were added to the ACSF and superfused.

Chemicals were obtained from Sigma-Aldrich Química (Spain), except for APV and CNQX, which were from Tocris Cookson (Bristol, UK).

Data acquisition and analysis

Data were high-pass filtered at 0.3 or 3.0 kHz and sampled at rates between 0.7 and 10.0 kHz through a Digidata 1200B interface (Axon Instruments) connected to a Pentium-based computer. The pCLAMP programs (Axon Instruments) were used to generate stimulus-timing signals and transmembrane current command pulses, and to record and analyse data. Results were expressed as means ± s.e.m. Data were compared using Student's paired or unpaired t test as appropriate (* P < 0.05, ** P < 0.01 and *** P < 0.001). No age-related differences were found in our sample.

The pre- or postsynaptic origin of the observed regulation of EPSC amplitudes by CCh was tested with conventional stimulation both by: (i) estimating changes in the paired pulse facilitation (PPF), which are considered to be of presynaptic origin (Creager et al. 1980; Clark et al. 1994; Kuhnt & Voronin, 1994) and which were quantified by calculating a PPF index ((R₂ - R₁)/R₁), where R₁ and R₂ are the peak amplitudes of the first and second EPSCs, respectively; and (ii) the modifications in the variance that parallel the EPSC amplitude changes, which reflect the changes in the probability of transmitter release (Bekkers & Stevens, 1990; Clements, 1990; Malinow & Tsien, 1990; Kuhnt & Voronin, 1994). To estimate the EPSC variance modifications, we first calculated the noise-free coefficient of variation (CV_NF) of the synaptic responses in CCh and control conditions with the formula:

where ${\delta }_{\text{EPSC}}^2$ and ${\delta }_{\text{noise}}^2$ are the variance of the peak EPSC and baseline, respectively, and m is the mean EPSC peak amplitude. The ratio of the CV in control and CCh conditions (CVR) was obtained then for each neuron as CV_CCh/CV_control (Clements, 1990). The CVR should be > 1 if the cause of the inhibitory effect is presynaptic. We finally constructed plots comparing variation in the normalized m (termed M) to the change in response variance of the EPSC amplitude (1/CV²), measured during CCh and normalized to the respective control value in each cell (Bekkers & Stevens, 1990; Malinow & Tsien, 1990). In these plots, values should follow the diagonal and the 1/CV² values remain under 1 if the inhibitory effect has a presynaptic origin. This method requires a binomial EPSC amplitude distribution, a condition that must be met for the synaptic variance to reflect the probability of transmitter release (i.e. the quantal variance). We could not directly test whether our data fitted a binomial distribution, but synaptic fluctuations were always evident and we assumed that synaptic release followed a binomial distribution.

We also analysed the changes induced by CCh in the failure rate and the synaptic potency of EPSCs evoked by minimal stimulation. The synaptic potency is defined as the average of the peak EPSC amplitude when failures are excluded (Stevens & Wang, 1994). Failures were estimated by visual discrimination and only cells with unitary EPSCs that could be clearly discriminated from the background noise were considered (e.g. Fig. 1 and Fig. 2). Independent estimations of failures were made by two scientists who were not involved in this work and who did not know what to expect from the results, and there were no statistically significant differences between their estimations and the measurements made by one of us.

Figure 1. CCh increases the failure rate of functional synapses more than that of silent synapses.

A and B, superimposed EPSCs (10 sweeps) evoked by minimal stimulation at −60 and +60 mV in a representative functional synapse and a silent synapse, respectively. C and D, averaged EPSC (100 sweeps) at +60 mV in control conditions and in the presence of CCh (5 μm), in representative functional and silent synapses, respectively. E and F, time course of CCh effect on EPSC peak amplitudes at +60 mV in representative functional and silent synapses, respectively. Each square represents the time of occurrence and the peak amplitude of a single synaptic response (as in Fig. 2). G, summary data of average failure rate (%) in control conditions and in the presence of CCh in silent (n = 12) and functional synapses (n = 11).

Figure 2. Functional synapses: effects of CCh on NMDA and non-NMDA EPSC components.

A, superimposed EPSCs (10 sweeps) evoked by minimal stimulation at −60 and +60 mV in control conditions, in the presence of CCh (5 μm) and after a washout (Wash) in a representative functional synapse. B, averaged EPSCs (100 sweeps) at +60 and −60 mV in control conditions, in the presence of CCh and after a washout. C, time course of CCh effect on EPSC peak amplitudes at +60 (left) and at −60 mV (right) in a representative functional synapse. D, summary data of average failure rate (%) in control conditions and in the presence of CCh in functional synapses at −60 and +60 mV (n = 11).

Results

Selective inhibition of functional synapses by CCh

Recent evidence has shown the existence of silent synapses that do not respond at the resting membrane potential (V_r), but respond with a slowly decaying APV-sensitive EPSC when the neuron is depolarized (Isaac et al. 1995; Liao et al. 1995). This behaviour is interpreted to indicate that silent synapses contain only functional NMDARs that are blocked by extracellular Mg²⁺ at the V_r but respond when depolarization relieves the Mg²⁺ block (Isaac et al. 1995; however, see Discussion). Other synapses, termed functional (i.e. conducting) synapses, contain both NMDARs and AMPARs, and therefore respond at the V_r (Isaac et al. 1995).

In order to test whether silent and functional synapses showed different sensitivities to cholinergic agonists, we recorded the effects of CCh on EPSCs evoked in response to minimal stimulation. In these experiments, SC stimulation at 1.0, 0.3 and 0.1 s⁻¹ started immediately after entering the whole-cell configuration while the V_m was held at −60 mV. Control recordings of SC EPSCs evoked by the same stimulations were obtained 20 min later while clamping the cell at both −60 and +60 mV. This prolonged waiting period minimized the possible contribution of postsynaptically mediated plastic phenomena to the observed effects (Malinow & Tsien, 1990). Functional synapses showed EPSCs at both values of V_m in control ACSF (Fig. 1A), whereas silent synapses displayed EPSCs only when depolarized (Fig. 1B). In some experiments, there was a large variability in the amplitudes of EPSCs that we could not resolve from the noise, especially at +60 mV when noise was high. Therefore, for failures analysis we selected cells that showed EPSCs with an amplitude separation with background noise so that a visual discrimination between failures and successes was possible at both −60 and +60 mV. With stimulation at 1.0 and 0.3 s⁻¹ the analysis was performed with responses evoked by 100 successive stimuli in each experimental condition. A synapse was considered silent if it did not respond to 100 consecutive stimuli at −60 mV, and if EPSCs were evoked at +60 mV.

Stimulation of SCs at 1.0 s⁻¹ permitted a rapid analysis with a large number of responses, but may have reduced the probability of release (see e.g. Gasparini et al. 2000). Therefore, we also used a low stimulation rate of 0.1 s⁻¹ and restricted the analysis to 20 successive stimuli in each experimental condition to reduce the overall duration of the experiment in order to avoid artefacts. After a washout in control solution, the stimulation was changed to 1.0 s⁻¹ and 100 stimuli were delivered at −60 mV to confirm (as above) that the experiment was on a silent synapse.

In experiments performed at room temperature and with control ACSF, we found that 41.9 % of recorded synapses were silent (13 out of 31 synapses matched the criteria). There were abundant response failures at +60 mV and the proportion was similar in functional (48.6 ± 3.3 %; n = 11) and silent synapses (49.6 ± 3.5 %; n = 12; Fig. 1A, B and G). Moreover, in functional synapses the failure rate was similar at +60 and −60 mV (see Fig. 2D).

We analysed the changes in the failure rate of EPSCs and of the mean peak amplitude of successful EPSC responses, or synaptic potency (Stevens & Wang, 1994), evoked by minimal stimulation. Changes in the failure rate have typically been attributed to presynaptic mechanisms. Changes in synaptic potency may reflect both pre- and postsynaptic changes (see e.g. Atwood & Wojtowicz, 1999).

In functional synapses, superfusion with CCh (5 μm) reduced the mean EPSC peak amplitude measured at +60 mV by 67.8 ± 7.5 %, in averages calculated with responses to the 100 stimuli that included both EPSCs and failures. The reduction was from 12.3 ± 1.0 pA in the controls to 3.9 ± 1.1 pA in the presence of CCh (n = 11, P < 0.001, Student's paired t test; Fig. 1C). However, in silent synapses the mean EPSC peak amplitude reduction measured at +60 mV evoked by CCh (5 μm) was only 24.9 ± 9.1 % and not statistically significant (i.e. from 11.2 ± 1.7 pA in the controls to 8.4 ± 1.3 pA in the presence of CCh; n = 12, n.s., Student's paired t test; Fig. 1D). The effects of CCh lasted throughout the ≈15 min superfusion of CCh (Fig. 1E).

The superfusion of CCh did not significantly modify the synaptic potency, either in functional (23.9 ± 1.8 pA in control conditions versus 18.8 ± 1.4 pA in the presence of CCh at +60 mV; n = 11, n.s., Student's paired t test) or in silent synapses (21.6 ± 3.1 pA in control versus 19.8 ± 3.1 pA under CCh at +60 mV, n = 12, n.s., Student's paired t test). However, superfusion with CCh increased the failure rate of functional (by 36.3 ± 4.5 %; n = 11, from 48.6 ± 3.3 % in the controls to 84.9 ± 3.4 % in the presence of CCh; n = 11, P < 0.001, Student's paired t test) more than of silent synapses (by 6.7 ± 3.2 %; from 49.6 ± 3.5 % in the controls to 56.3 ± 5.3 % in the presence of CCh; n = 12, n.s., Student's paired t test; Fig. 1E, F and G). Therefore, CCh inhibited EPSCs in functional synapses by increasing the failure rate without changing the synaptic potency.

Taken together, these results are consistent with the view that CCh acting presynaptically inhibited transmitter release at SC terminals of functional synapses more effectively than at SC terminals of silent synapses.

After return to the control ACSF, the mean EPSC peak amplitude (including failures) tended to recover rapidly in both synapse types (Fig. 1E and F). The failure rate during a washout was 66.2 ± 2.4 % in functional synapses (n = 11) and 53.3 ± 5.3 % in silent synapses (n = 12) and the synaptic potency was unchanged throughout the experiment.

Superfusion with CCh induced a similar mean EPSC peak amplitude reduction (including failures) in functional synapses held both at +60 and at −60 mV (Fig. 2A and B), and the effects lasted throughout the ≈15 min superfusion of CCh (Fig. 2C). In addition, although CCh markedly reduced the release probability of functional synapses, we never observed a switch from functional to silent synapses paralleling the drop in release probability.

The mean EPSC amplitude reduction induced by CCh was 67.8 ± 7.5 % at +60 mV (see values above) and 77.9 ± 7.2 % at −60 mV (i.e. from −8.7 ± 1.0 pA in the controls to −1.9 ± 0.5 pA in the presence of CCh; n = 11, P < 0.001, Student's paired t test), suggesting that CCh caused similar failure rate increments at both values of V_m in functional synapses. The failure rate of functional synapses in control conditions was 52.5 ± 2.9 and 48.6 ± 3.3 % at −60 and +60 mV, respectively (n = 11, n.s., Student's paired t test) and the increase in failure rate induced by CCh was 40.2 ± 3.3 and 36.3 ± 4.5 % at −60 and +60 mV, respectively (n = 11, P < 0.001, Student's paired t test; Fig. 2D).

The CCh challenge did not significantly modify the synaptic potency, either at +60 (see values above) or at −60 mV (peak EPSC values were −17.7 ± 1.8 pA in controls versus -15.6 ± 1.7 pA in the presence of CCh; n = 11, n.s., Student's paired t test). These results suggest that in functional synapses both AMPA and NMDA EPSC components were under the same presynaptic cholinergic regulatory mechanism.

Increasing the temperature to 31–34 °C did not modify the effects of CCh on the failure rate of functional synapses, which was increased by CCh from 35.8 ± 2.0 to 54.3 ± 1.7 % (n = 4, P < 0.05, Student's paired t test), whereas the failure rate of silent synapses was only increased from 38.5 ± 4.7 to 47.6 ± 6.4 % (n = 4, n.s., Student's paired t test; Fig. 3A and B). Moreover, at the higher temperature there was no significant modification of the synaptic potency by CCh in functional synapses (i.e. from 23.6 ± 8.4 pA in the controls to 13.5 ± 5.2 pA in the presence of CCh; n = 4, n.s., Student's paired t test) and silent synapses (i.e. from 18.3 ± 2.7 pA in the controls to 16.2 ± 2.9 pA in the presence of CCh; n = 4, n.s., Student's paired t test).

Figure 3. CCh effects are insensitive to manipulations that increase release probability.

A, top, control EPSC (left) and effects of 5 μm CCh (right) in a representative functional synapse at 31–34 °C recorded at +60 mV. A, bottom, summary data of average failure rate (%) in control conditions and in the presence of CCh in functional synapses at 31–34 °C, with a CaCl₂ concentration of 5.2 mm and with low-frequency stimulation (0.1 s⁻¹); all records at +60 mV (n = 4). B, same as A but silent synapses (n = 4). C, control EPSC (left) and effects of CCh (middle) on paired pulse facilitation (R₁ and R₂ responses to the first and second pulse, respectively) in a representative functional (top) and silent synapse(bottom). C, right, summary data of R₂ average failure rate (%) in control conditions and in the presence of CCh in functional (n = 5) and silent synapses (n = 7). In A, B and C, 10 traces are shown superimposed.

The effects of CCh were also insensitive to higher Ca²⁺ concentrations of 5.2 mm. The failure rate increased from 36.5 ± 8.4 to 54.9 ± 5.9 % (n = 4, P < 0.05, Student's paired t test) in functional synapses, whereas the failure rate of silent synapses only increased from 41.2 ± 3.2 to 51.2 ± 13.3 % (n = 4, n.s., Student's paired t test) with the CCh challenge (Fig. 3A and B), and there were no significant modifications of the synaptic potency in functional synapses (i.e. from 16.3 ± 2.1 pA in the controls to 14.2 ± 1.1 pA in the presence of CCh; n = 4, n.s., Student's paired t test) and silent synapses (i.e. from 14.3 ± 1.0 pA in the controls to 11.6 ± 2.6 pA in the presence of CCh; n = 4, n.s., Student's paired t test).

The effects of CCh were also similar when stimulation was at a lower rate of 0.1 s⁻¹ and failures in functional synapses increased from 32.5 ± 11.2 to 61.2 ± 12.9 % (n = 4, P < 0.01, Student's paired t test), whereas failures of silent synapses only increased from 44.3 ± 3.6 to 53.4 ± 10.4 % (n = 5, n.s., Student's paired t test; Fig. 3A and B). With low stimulation rate, application of CCh did not modify the synaptic potency in functional synapses (i.e. from 24.6 ± 3.3 pA in the controls to 16.0 ± 1.2 pA in the presence of CCh; n = 4, n.s., Student's paired t test) and silent synapses (i.e. from 14.6 ± 1.0 pA in the controls to 11.8 ± 1.9 pA in the presence of CCh; n = 5, n.s., Student's paired t test).

We also analysed the changes in functional and silent synapses induced by CCh associated with modifications of the PPF. Only measurements performed at +60 mV will be mentioned to simplify the description since the effects of CCh were identical at both potentials in functional synapses. The PPF is a form of short-term presynaptic plasticity characterized by an increased peak amplitude of the second EPSC (R₂) when it is elicited shortly after (<100 ms) a preceding EPSC (R₁; Creager et al. 1980; Hess et al. 1987; Clark et al. 1994; Kuhnt & Voronin, 1994). In control conditions, functional synapses showed higher R₂ mean amplitudes than R₁ mean amplitudes in averages including failures (R₂ = 19.2 ± 1.1 pA and R₁ = 14.5 ± 1.1 pA; P < 0.05, n = 5) resulting in a PPF index of 0.36 ± 0.15. This increased R₂ relative to R₁ is thought to be caused by the increased release probability that results from Ca²⁺ remaining in the terminal after the first AP (Kamiya & Zucker, 1994). However, silent synapses did not show higher R₂ than R₁ amplitudes (R₂ = 9.8 ± 1.1 pA and R₁ = 9.3 ± 1.1 pA; n.s., n = 7), giving a PPF index of 0.05 ± 0.05. This result suggests that synaptic terminals of functional and silent synapses could be functionally different.

In most functional and silent synapses, the synaptic potency of R₂ was higher than that of R₁ (3 out of 5 or 60 % functional and 3 out of 7 or 43 % silent synapses), suggesting an increase by the second relative to the first AP in the number of synapses activated, a rise in the quantal size of individual vesicles or an enlargement in the number of vesicles released, as has been shown by other groups (e.g. Isaac et al. 1998; Atwood & Wojtowicz, 1999; Oertner et al. 2002). In functional synapses, the potency of R₁ was 28.3 ± 2.4 pA and that of R₂ was 33.3 ± 2.8 pA (P < 0.05, n = 5) and in silent synapses the synaptic potency of R₁ was 15.9 ± 1.4 pA and that of R₂ was 21.1 ± 2.9 pA (P < 0.05, n = 7). The synaptic potency of R₁ and R₂ was the same in the rest of the synapses, as expected if only one synapse was stimulated (Fig. 3C, control). The effects of the CCh ‘pulse’ were similar in the two groups, in the sense that functional synapses were markedly inhibited and silent synapses were not, and therefore both groups will be treated together.

The PPF index (i.e. peak values of averages including failures) was significantly increased from 0.36 ± 0.15 to 0.70 ± 0.18 by CCh in functional synapses (P < 0.05, n = 5). These modifications were paralleled by a greater increase in the proportion of failures of R₁ compared with R₂ (from 49.2 ± 5.6 to 77.2 ± 5.2 % for R₁; P < 0.05; and from 34.4 ± 4.2 to 63.4 ± 12.4 % for R₂; n.s.; same cells; Fig. 3C), suggesting that presynaptic mechanisms underlie the inhibitory effects of CCh (Creager et al. 1980; Clark et al. 1994; Kuhnt & Voronin, 1994). The CCh challenge reduced the synaptic potency of R₂ from 33.0 ± 2.8 to 23.4 ± 2.4 pA (P < 0.05, n = 5), suggesting that other presynaptic mechanisms were also at work.

Bath-applied CCh did not significantly modify the PPF index in silent synapses (from 0.05 ± 0.05 to 0.31 ± 0.11; n.s., n = 7) and there were no significant changes in failure rates of R₁ and R₂ (from 47.1 ± 4.1 to 54.8 ± 9.2 % for R₁ and from 53.3 ± 4.6 to 52.7 ± 6.1 % for R₂; n.s., n = 7), or of the synaptic potency (from 21.1 ± 2.9 to 18.7 ± 2.5 pA for R₂; n.s., n = 7; Fig. 3C).

Taken together, these results suggest that the differential presynaptic effect of CCh on functional and silent synapses is not modified by any of the manipulations that increase the probability of transmitter release.

Presynaptic muscarinic inhibition

The presynaptic effects of CCh on the EPSC may be mediated through muscarinic or nicotinic receptors (Hounsgaard, 1978; Valentino & Dingledine, 1981; Hefft et al. 1999). Therefore, under conventional stimulation we analysed the effects of CCh and muscarine on the PPF.

The mean peak amplitude of R₁ was −162.2 ± 27.2 pA and that of R₂ was −232.1 ± 38.4 pA in control ACSF (n = 10), and superfusion with CCh (5 μm) reduced R₁ to −24.6 ± 6.4 pA and R₂ to −67.7 ± 13.6 pA. Therefore, CCh inhibited R₁ more than R₂ (84.8 ± 3.2 versus 70.8 ± 5.7 %; n = 10, P < 0.01, Student's paired t test), resulting in an average increase of the PPF index from 0.40 ± 0.07 to 1.6 ± 0.17 (n = 10, P < 0.001, Student's paired t test; Fig. 4A, left and middle panels).

Figure 4. Conventional stimulation: effects of CCh and muscarine.

A, left, superimposed, normalized (to R₁), averaged (10 sweeps) EPSCs evoked by paired pulse stimulation (90 ms delay) in control conditions (black trace) and in the presence of CCh (5 μm; grey trace); middle, summary data showing average PPF index ((R₂ - R₁)/R₁) in control and CCh solutions (n = 10); and right, plot of the variance (1/CV²) as a function of the mean peak EPSC amplitude (M) in the presence of CCh, normalized to control conditions. B and C, same as A, but in the presence of muscarine (5 μm, n = 10) and CCh in Mg²⁺-free solutions (n = 6) and paired pulses at 70 and 240 ms delay, respectively.

Similar results were obtained when muscarine was superfused, which reduced R₁ by 82.8 ± 4.2 % and R₂ by 72.7 ± 6.4 % (n = 10, P < 0.01, Student's paired t test), increasing the PPF index from 0.77 ± 0.09 to 2.47 ± 0.13 (n = 10, P < 0.001, Student's paired t test; Fig. 4B, left and middle panels). Moreover, the effects of CCh and muscarine were blocked by previous superfusion with the unspecific muscarinic antagonist atropine (10 μm, n = 10, data not shown).

In Mg²⁺-free solution and with CNQX (20 μm), a condition in which the NMDA conductance is isolated (Hestrin et al. 1990), superfusion with CCh evoked a significantly smaller EPSC reduction (R₁, 47.0 ± 8.4 % and R₂, 47.9 ± 5.8; n = 6, n.s., Student's paired t test) and no modification of the PPF index (0.16 ± 0.03 in Mg²⁺-free conditions and 0.18 ± 0.04 in the presence of CCh; n = 6, n.s., Student's paired t test; Fig. 4C, left and middle panels).

We also constructed plots of the 1/CV² ratio as a function of the mean peak EPSC amplitude, M (see Methods) for the effects of CCh in normal ACSF (Fig. 4A, right panel), muscarine (Fig. 4B, right panel) and CCh in Mg²⁺-free solution with CNQX (20 μm; Fig. 4C, right panel). These plots revealed that values grouped following the diagonal and that 1/CV² values were less than 1.0, implying that the CCh reduced the probability of transmitter release and thus that the effects were presynaptic. This conclusion was also apparent from the coefficient of variation ratio (CVR, see Methods). The CVR was consistently >1 with CCh, in normal ACSF (2.4 ± 0.2; n = 10), in Mg²⁺-free solution with CNQX (20 μm; 1.4 ± 0.3, n = 6) and when muscarine was superfused (2.6 ± 0.4; n = 10). These results indicate that a change in the release parameter is the most likely explanation for the reduction of the EPSCs amplitude.

Taken together, the above results imply that CCh, via activation of presynaptic mAChRs, did not much affect silent synapses, but inhibited synaptic transmission at functional synapses by reducing the reliability of release without changing the synaptic potency.

Effects of CCh on NMDA and non-NMDA EPSC components

If CCh acting presynaptically more effectively inhibits transmitter release at SC terminals of functional synapses than of silent synapses, we would expect further inhibition of the non-NMDA than NMDA EPSC component evoked by conventional stimulation. To test this hypothesis, the NMDA and non-NMDA EPSC components were isolated with CNQX (20 μm) and APV (50 μm), respectively, in Mg²⁺-free external solution (Hestrin et al. 1990) and the effects of CCh (5 μm) were tested in both conditions of block of EPSC components (Fig. 5). The isolated non-NMDA component showed a fast decay time constant (τ = 18.3 ± 3.2 ms; n = 6), while the decay time constant of the NMDA component was slower (τ = 70.5 ± 4.5 ms; n = 6; Fig. 5A and B). The isolated non-NMDA component was reduced by 86.6 ± 3.9 % (Fig. 5A and C), whereas the isolated NMDA component was only reduced by CCh by 47.0 ± 8.4 % (n = 6, P < 0.01, Student's unpaired t test; Fig. 5B and C). These results imply that CCh more effectively inhibits the non-NMDA than the NMDA component of the SC EPSCs.

Figure 5. Effects of CCh on isolated NMDA and non-NMDA EPSCs.

A and B, averaged (10 sweeps) non-NMDA EPSCs (50 μm APV, Mg²⁺ free) and averaged NMDA EPSCs (20 μm CNQX, Mg²⁺ free), respectively, before (control), in the presence of CCh (5 μm) and after a 40 min washout. Paired pulse stimulation at 150 ms delay in A and 240 ms delay in B. C, summary data showing mean peak amplitude reduction (%) of R₁ induced by CCh, in APV (n = 6) and CNQX (n = 6) in Mg²⁺-free solutions.

The CCh challenge has postsynaptic effects that may modify the membrane conductance and change the voltage control of the dendritic sites where EPSCs are generated, thus introducing possible recording artefacts. In cells recorded with potassium gluconate-based solution, holding current was increased by superfusion with CCh (5 μm; from 30.9 ± 10.3 pA in control conditions to 150.5 ± 11.2 pA in the presence of CCh; n = 10, P < 0.001) and there was a parallel increase in the input resistance (from 180.2 ± 11.7 MΩ in control conditions to 234.2 ± 13.9 MΩ in CCh; n = 10, P < 0.01). These postsynaptic effects were abolished when the caesium gluconate-based pipette solution was used and 5 μm CCh neither modified the holding current (from 230.9 ± 20.3 pA in control conditions to 240.5 ± 11.2 pA in CCh; n = 10, n.s.) nor the input resistance (from 243.2 ± 15.7 MΩ in control conditions to 254.2 ± 11.8 MΩ in CCh; n = 10, n.s.). However, there were no differences in the effects of CCh on EPSCs in control conditions or when K⁺ conductances were blocked with Cs⁺ in the pipette solution. In addition, we found no changes in the rise and decay time constants of EPSCs (fits to single exponential functions) that may reflect membrane conductance modifications at, or near, the sites of EPSC generation (Mainen et al. 1996). In control conditions, the rise and decay time constants of the non-NMDA EPSCs were 5.9 ± 1.0 and 20.2 ± 2.4 ms (n = 6) and those of the NMDA EPSCs were 14.3 ± 2.6 and 71.6 ± 1.2 ms (n = 6), respectively. Under the CCh challenge those of the non-NMDA EPSCs were 4.2 ± 1.3 and 18.2 ± 1.8 ms (n = 6) and those the NMDA EPSCs were 16.2 ± 3.3 and 70.1 ± 8.9 ms (n = 6), respectively (n.s.).

Therefore, these results eliminate the possibility of an artefact due to changes in the dendritic voltage control and provide further evidence indicating that the NMDA component of SC EPSCs is less sensitive to the presynaptic inhibition by CCh than the non-NMDA component.

The selective inhibition is specific to cholinergic modulation

To test whether the selective inhibition of the non-NMDA EPSC component was specific to cholinergic agonists, we analysed the effects of adenosine, another known presynaptic inhibitor of SC synapses (Wu & Saggau, 1994). Experiments were performed under block of NMDA and non-NMDA EPSC components with APV (50 μm) and CNQX (20 μm), respectively, in Mg²⁺-free external solution before and during superfusion with adenosine (50 μm). Adenosine reduced both EPSC components equally (Fig. 6). The non-NMDA component was reduced by 88.0 ± 5.3 % (Fig. 6A and C) and the NMDA component was decreased by 80.8 ± 1.6 % (n = 5, n.s., Student's unpaired t test; Fig. 6B and C).

Figure 6. Effects of adenosine on isolated NMDA and non-NMDA EPSCs.

A and B, averaged (10 sweeps) non-NMDA EPSCs (50 μm APV, Mg²⁺ free) and averaged NMDA EPSCs (20 μm CNQX, Mg²⁺ free), respectively, before (control), in the presence of adenosine (50 μm) and after a 40 min washout. C, summary data showing mean peak amplitude reduction (%) of R₁ induced by CCh, in APV- and CNQX-containing Mg²⁺-free solutions (n = 5).

Therefore, both EPSC components were blocked to a similar degree by adenosine, suggesting that the differential presynaptic regulation of non-NMDA and NMDA components of SC EPSCs is specific to cholinergic agonists.

Discussion

We describe a new form of differential control by CCh of the ionotropic components at the excitatory glutamatergic synapses of CA3 with CA1 hippocampal pyramidal neurons via SCs. This cholinergic control is characterized by a selective regulation of the non-NMDA component of SC EPSCs. We provide evidence implying that CCh exerts its effects via activation of mAChRs by reducing the probability of glutamate release from the presynaptic terminals of functional but not of silent synapses.

Differential cholinergic regulation of functional and silent synapses

Using minimal stimulation, we demonstrate that functional synapses show a much higher sensitivity to inhibition by CCh than silent synapses. The cholinergic inhibition of functional synapses is caused by an increase in the proportion of failures without changes of the amplitude of the EPSCs. The failure rate of silent synapses is much less affected by CCh. Modifications in the synaptic failure rate have been classically considered to indicate variations in the release probability (p) at synaptic terminals (see e.g. Redman, 1990). Therefore, the increased failure rate of functional synapses by CCh implies a reduced p (Redman, 1990; Kuhnt & Voronin, 1994; Schulz et al. 1994; Stevens & Wang, 1994; Atwood & Wojtowicz, 1999).

Application of CCh did not modify the synaptic potency at the functional synapses considered (Stevens & Wang, 1994), implying no modification in the quantal content (m) or in quantal amplitude (q) of EPSCs. Therefore, in these synapses CCh did not induce changes in the number of presynaptic release sites (n) or decreases in postsynaptic response to a transmitter quantum, q (Atwood & Wojtowicz, 1999). In addition, in functional synapses the increased failure rate was identical when the neuron was held at −60 and +60 mV, implying a similar decrease in p independent of postsynaptic membrane potential, and suggesting identical underlying presynaptic cellular mechanisms. Moreover, manipulations that increase the release probability, such as decreasing the stimulation rate, increasing the bath temperature and increasing the Ca²⁺:Mg²⁺ ratio, did not significantly modify the effects of CCh.

When paired minimal stimulation was used, an increased release probability of the second relative to the first stimulation was observed in control conditions, in accordance with the evidence indicating that the PPF is produced by an increased p as a result of the Ca²⁺ remaining in the terminal when the second action potential arrives quickly after the first AP (Kamiya & Zucker, 1994). These effects were observed in all functional synapses tested under paired stimulation. However, silent synapses did not show a clear PPF, suggesting that presynaptic terminals of silent and conducting synapses may be functionally different.

In most of the functional and silent synapses that we recorded, the synaptic potency of the second EPSC was larger than that of the first response. This R₂ > R₁ difference in synaptic potency contributed to the PPF, and has previously been demonstrated using recording methods that increase the signal-to-noise ratio and enable a separation of small EPSCs from failures (e.g. Isaac et al. 1998; Oertner et al. 2002). This behaviour has been interpreted as indicating an increase in the number of synapses activated by the second AP, an enlargement in the quantal size of individual vesicles or a rise in the number of vesicles released (Isaac et al. 1998; Atwood & Wojtowicz, 1999; Oertner et al. 2002).

The CCh challenge increased the PPF index in functional synapses by reducing the failure rate of R₁ more than that of R₂. However, these effects of CCh were much smaller, or altogether absent, in silent synapses. Therefore, CCh acts via activation of presynaptic mAChRs by selectively decreasing the release probability at functional synapses. The CCh ‘pulse’ also reduced the synaptic potency of the second EPSC (R₂), but the amplitude contribution of this effect to the changes in PPF induced by CCh did not counteract the larger R₁ to R₂ failure rate reduction. The change in synaptic potency is interesting, but is out of the scope of the present work and will be analysed in the future.

A postsynaptic potentiation of NMDA responses by ACh has been demonstrated (Harvey et al. 1993; Marino et al. 1998), which could complicate our presynaptic interpretation of the differential muscarinic effects on NMDA and non-NMDA EPSC components. However, the postsynaptic potentiation of the NMDA conductance peaks in about 2 min and decays to control values in <10 min (Marino et al. 1998), whereas the differential inhibitory effects described here peaked after total renewal of the chamber (≈5 min) and reached a steady state that remained unchanged up to 15 min after. In addition, the intracellular dialysis occurring during the prolonged waiting period preceding the CCh challenge (>20 min) minimized the possible contribution of postsynaptic mechanisms (Malinow & Tsien, 1990).

It has been shown that the presynaptic inhibition by CCh is caused by a selective blockade of voltage-gated ω-conotoxin-GVIA-sensitive Ca²⁺ channels via a G-protein-coupled muscarinic mechanism (Qian & Saggau, 1997). Therefore, dissimilarity in the proportion of (i) functional mAChRs, (ii) ω-conotoxin-GVIA-sensitive calcium channels, or (iii) the intracellular cascade that links both elements could underlie the differences in CCh sensitivity of functional and silent synapses.

Both presynaptic and postsynaptic mechanisms have been proposed as cellular mechanisms underlying silent synapses. The evidence in favour of a postsynaptic origin implies that functional AMPARs are absent in silent synapses (Isaac et al. 1995; Liao et al. 1995; Durand et al. 1996; Petralia et al. 1999). The ideas in favour of a presynaptic origin of silent synapses are based on the higher affinity of NMDARs compared with AMPARs for the natural agonist, glutamate (Patneau & Mayer, 1990). Functional AMPARs and NMDARs would be present in all synapses and the glutamate concentration in functional contacts would be sufficient to activate both receptor types, whereas a lower glutamate concentration would activate only NMDARs in silent synapses (Kullmann & Asztely, 1998). Recently, the concentration of glutamate in the cleft of silent synapses has been estimated to be low (< 170 μm; Choi et al. 2000) and it has been shown that manipulations which increase the release probability can convert silent synapses into conducting ones (Gasparini et al. 2000).

We employed manipulations that increased release probability, but with the methodology we used, which was solely designed to test the effects of CCh, we could not observe whether silent synapses were switched to conducting ones when the release probability was incremented, as reported by Gasparini et al. (2000). However, our results suggest that the selective inhibition of functional synapses is not linked to differences in the release probability because manipulations that increased the release probability did not significantly modify the effects of CCh. Moreover, we should like to emphasize that, although CCh markedly reduced the release probability of functional synapses, we never observed a switch from functional to silent synapses in the presence of CCh, as would be expected if the mechanism underlying silent synapses was exclusively a low release probability.

Therefore, the present results are consistent with the existence of different presynaptic terminals with specific sensitivity to cholinergic modulation. In this scenario, we propose that terminals of functional synapses have the molecular machinery necessary for the presynaptic muscarinic inhibition by CCh, whereas this machinery would be altogether absent or less expressed in terminals of silent synapses.

It has recently been shown that newly created synapses have only NMDARs and that the proportion of synapses expressing functional AMPARs increases with synapse maturation (Petralia et al. 1999). In addition, activity-dependent increases in synaptic efficacy also increase the AMPAR content of silent synapses, thus converting them into conducting synapses (Liao et al. 1999). Moreover, silent and functional SC synapses also show morphological pre- and postsynaptic differences (Takumi et al. 1999; Cantallops et al. 2000). Therefore, there is ample evidence indicating a solid match between the development of synaptic structure and function.

Our hypothesis of different presynaptic terminals for synapses expressing or not expressing functional AMPARs would require that when AMPARs are recruited by development or activity into the spine of silent synapses, the presynaptic terminal is modified in parallel to express the molecular machinery that enables the cholinergic modulation. Consequently, postsynaptic modifications that convert silent into functional synapses (Isaac et al. 1995) should be tightly coordinated with presynaptic changes. Interestingly, such a synchronized development of synapses regulated by an activity-dependent gene has recently been reported in the tectum of Xenopus, where coordinated pre- and postsynaptic morphological modifications match the maturation into functional synapses (Cantallops et al. 2000). Moreover, different classes of presynaptic calcium channels are expressed at different times during synapse development and maturation in the hippocampus (Scholz & Miller, 1995), again indicating a close match between structural and functional development. In addition, a rapid increase in the expression of presynaptic proteins paralleling the onset of long-term potentiation has been shown in hippocampal neurons in culture (Antonova et al. 2001), implying coordinated presynaptic and postsynaptic changes during plasticity. Therefore, our results suggest that the targeting of the molecular machinery necessary for the cholinergic presynaptic inhibition may be related to the development, maturation and activity-dependent plasticity of synaptic connections in the CA1 region.

Differential regulation of AMPA and NMDA EPSC components

The glutamatergic SC EPSCs evoked in CA1 pyramidal neurons by conventional stimulation exhibit a non-NMDA component and an NMDA component (Collingridge et al. 1983; Hestrin et al. 1990). We show that CCh more effectively inhibits the non-NMDA than the NMDA component of the SC EPSCs. This selective effect is clearly apparent when EPSCs are evoked in Mg²⁺-free external solution, a condition in which the NMDA response loses its voltage dependence. The selective inhibition is also present when K⁺ conductances are blocked by intracellular Cs⁺, a condition in which the voltage control of dendritic regions is favoured. Therefore, the selective inhibition of non-NMDA EPSCS by CCh is insensitive to both Mg²⁺-free external and intracellular Cs⁺-containing solutions, manipulations that minimize the possible effects of dendritic depolarization. There were no changes in the rise and decay time constants of EPSCs, suggesting no membrane conductance modifications at, or near, the sites of EPSC generation (Mainen et al. 1996). Moreover, the selective regulation of the non-NMDA EPSC component by CCh agrees with the specific presynaptic inhibition of functional synapses observed when minimal stimulation was used.

Therefore, the present results suggest that higher sensitivity of functional synapses to activation of mAChRs is the cellular mechanism responsible for the more potent effects of CCh on the non-NMDA than the NMDA component of SC EPSCs observed when conventional stimulation is used.

We also show that the differential control by CCh is specific because adenosine, another known presynaptic inhibitor at the SC synapses (Wu & Saggau, 1994), similarly reduced both NMDA and non-NMDA EPSC components.

Functional implications

Activation of mAChRs depolarizes, reduces spike frequency adaptation and increases the excitability of CA1 pyramidal neurons by blocking several K⁺ conductances (Benardo & Prince, 1982; Madison et al. 1987; Borde et al. 2000) and selectively inhibits the non-NMDA component of SC EPSCs by reducing the probability of transmitter release at functional synapses without much affecting the terminals of silent synapses (present results). Therefore, in physiological conditions of concurrent activation of cholinergic inputs and CA3 pyramidal neurons, there would be both (i) a preferred activation of NMDARs relative to non-NMDARs via the selective inhibition of glutamate release at the terminals of functional synapses and (ii) a depolarization of the CA1 pyramidal neuron induced by the ACh released by cholinergic fibres. The ACh-mediated depolarization would relieve the Mg²⁺ block of NMDA channels in the absence of an AMPA-mediated depolarization in CA1 pyramidal cells. Silent synapses are converted into functional synapses by activity-dependent plastic phenomena related to synapse maturation, learning and memory (Isaac et al. 1995, 1998; Liao et al. 1995; Kullmann & Asztely, 1998; Gasparini et al. 2000). In this scenario, cholinergic input would disconnect the mature and potentiated functional synapses while favouring transmission through immature and non-potentiated silent synapses. The resulting specific activation of NMDARs would promote Ca²⁺ inflow and synaptic responses with slow decays that would assist the summation of EPSPs, thus favouring activity-dependent plastic phenomena mediated via NMDAR activation at silent synapses. The consequences of this funnelling of activity through silent synapses would induce their maturation and favour their potentiation by converting them into functional synapses. This may relate to the laminar distribution of cholinergic effects on CA1 neurons, where the presynaptic inhibition of the input from CA3 is stronger than the entorrhynal input (Hasselmo & Schell, 1994), a selective suppression that has been linked by extensive modelling to associative memory functions of the hippocampus (Hasselmo, 1999).

Iontophoretic application of glutamate and stimulation of SC in vitro induces rhythmic theta-like oscillations in some CA1 pyramidal neurons by activation of NMDARs (Bonansco & Buño, 2002). Interestingly, blocking AMPARs with CNQX extends the oscillatory behaviour to all pyramidal neurons, indicating that the activation of AMPARs masks the NMDA-induced oscillations. This effect of AMPAR activation is caused by a reduction of the negative slope conductance of the net current, which is a prerequisite for an oscillatory behaviour and which is otherwise provided by the NMDA conductance when activated in isolation (Bonansco et al. 2002). The above-described effects of differential glutamate receptor activation suggest that block of functional synapses by cholinergic inputs would favour NMDA oscillations and the theta rhythm by inhibiting functional synapses, and thus excitation through activation of AMPARs, and by favouring the isolated activation of silent synapses and excitation of NMDARs and allows participation in the induction of plastic phenomena (Huerta & Lisman, 1995).