Two distinct and activity-dependent mechanisms contribute to autoreceptor-mediated inhibition of GABAergic afferents to hilar mossy cells

Abstract

We report that bath application of 3 μm carbachol (CCh), a muscarinic acetylcholine receptor agonist, reduces evoked IPSC amplitude recorded from hilar mossy cells in the rat dentate gyrus through a presynaptic mechanism. While CCh has been shown to inhibit evoked IPSCs in other systems, this effect is intriguing in that it does not require inhibitory action of either presynaptic muscarinic receptors or presynaptic cannabinoid receptors. Previous work from our lab has shown that identical application of CCh produces an action potential-dependent increase in ambient GABA in this system; however, inhibition of evoked IPSCs produced by both 3 and 10 μm CCh is insensitive to the GABA_B antagonist CGP52432. Therefore we hypothesized that CCh-mediated inhibition of evoked IPSCs might be produced by activity-dependent increases in ambient GABA and subsequent activation of presynaptic GABA_A receptors. Consistent with that hypothesis, we report that CCh-mediated inhibition of evoked IPSCs appears to be well correlated with CCh-mediated facilitation of spontaneous IPSCs and that CCh does not affect GABA_B-mediated IPSCs recorded in the presence of the GABA_A receptor antagonist picrotoxin. Intriguingly, however, we found that bath application of the GAT-1 transport blocker NO-711 (1 μm) produces inhibition of evoked IPSCs that is reversed by CGP52432, and that lower doses of CCh produce inhibition with greater CGP52432 sensitivity. These observations, combined with subsequent work on multiple pulse depression, reveal that feedback inhibition of GABAergic afferents to hilar mossy cells is governed by a complex relationship between two distinct and activity-dependent mechanisms.

Introduction

The effects of muscarinic agonists in the CNS are wide, complex and varied. At a systems level, antagonism of muscarinic acetylcholine receptors (mAChRs) can induce amnesia, while delivery of cholinesterase inhibitors can improve memory formation in animals with cholinergic lesions (Glick & Zimmerberg, 1972). Further, selective knockout of individual mAChR subtypes produces interesting and diverse behavioural effects (Bymaster et al. 2003; Wess et al. 2003). At a network level, it is apparent that selective muscarinic drive of specific interneurons and principle cells facilitates highly synchronous network oscillations in both cortical and subcortical structures (e.g. Traub et al. 1992; Lee et al. 1994; Fellous & Sejnowski, 2000). At a synaptic level, muscarinic receptors coupled to G_i/G_o (M2, M4) have been shown to inhibit calcium-dependent exocytosis from a variety of synapses. By contrast, across various cells and systems, mAChRs coupled to G_q/11 (e.g. M1, M3) promote somatic depolarization through activation of a mixed Na⁺/K⁺ current, alter intrinsic properties substantially by activation of a calcium-dependent but non-selective conductance (I_CAN), and promote phospholipase C (PLC)-dependent postsynaptic production of endogenous cannabinoids capable of acting as inhibitory retrograde messengers (Caulfield et al. 1993; Jones, 1993; Wess et al. 1996; Caulfield & Birdsall, 1998; Hajos et al. 1998; Kim et al. 2002; Fukudome et al. 2004; Edwards et al. 2006).

Indeed, in many areas of the CNS more than one effect of muscarinic agonists is apparent, often mediated by more than one specific site of action. For example, in the dentate gyrus, recent work from our lab has shown that bath application of 3 μm CCh (a non-selective muscarinic agonist) can enhance depolarization-induced suppression of inhibition by facilitating activity-dependent production of endocannabinoids in hilar mossy cells (Hofmann et al. 2006). Yet identical application of CCh also increases spontaneous GABAergic transmission and produces functionally relevant increases in ambient GABA (Nahir et al. 2007). In the current paper, we report that bath application of CCh also produces clear inhibition of evoked IPSCs recorded from hilar mossy cells. With that finding it became clear that the dentate gyrus is one of several areas in the CNS where bath application of muscarinic agonists both facilitates spontaneous GABAergic transmission and yet also inhibits evoked GABAergic transmission.

Other areas with such apparently disparate effects of muscarinic agonists include the lateral spiriform nucleus of the chick (Guo & Chiappinelli, 2001), area CA1 of the hippocampus (Pitler & Alger, 1992; Behrends & ten Bruggencate, 1993) and the entorhinal cortex (Xiao et al. 2009). In area CA1 of the hippocampus, these opposing effects have been shown to (at least in large part) be mechanistically distinct (Fukudome et al. 2004). However, in other cases the mechanisms through which muscarinic agonists inhibit evoked GABAergic transmission has remained unclear (see Discussion for more details). In this study we found that specific mechanisms reported to produce muscarinic inhibition of evoked GABAergic transmission in area CA1 of the hippocampus are not responsible for similar inhibition in the dentate. Therefore, we proceeded to examine the hypothesis that CCh-mediated facilitation of spontaneous IPSCs in the dentate gyrus may indirectly inhibit evoked IPSCs via functional modulation of inhibitory tone. Our results not only validate that hypothesis, but also reveal an unexpected activity-dependent switch in the specific mechanism of autoreceptor-dependent inhibition. As such, we believe the findings presented here are of significant interest on several levels. First, our results clearly emphasize the importance of extrasynaptic GABA in mediating the effects of CCh on inhibitory networks in the CNS. Second, our results provide the first clear mechanistic explanation for how action of CCh on a common pool of excitatory muscarinic receptors can simultaneously enhance spontaneous and inhibit evoked GABAergic transmission. Third, our results are to our knowledge the first to reveal that the specific mechanism of feedback inhibition at a central synapse is apparently subject to activity-dependent regulation. That last finding suggests the potential for development of therapeutics that selectively target, or effectively mimic, high activity conditions.

Methods

Hippocampal slice preparation

Male Sprague–Dawley rats between the ages of P18 to P25 received i.p. injections of ketamine (80–100 mg kg⁻¹) and were rapidly decapitated using a small animal guillotine. The brain was quickly removed and horizontal sections (300 μm thick) were made using a vibratome (Pelco, Redding, CA, USA). Slices were incubated for 30 min at 30–35°C and then allowed to reach room temperature (for 30 min). During cutting and incubation, slices were kept in dissecting solution that was saturated with 95% O₂–5% CO₂ and contained (in mm): 124 NaCl, 2.5 KCl, 1.23 NaH₂PO₄, 2.5 MgSO₄, 10 d-glucose, 1 CaCl₂ and 25.9 NaHCO₃. All animal procedures were approved by the Institutional Animal Care and Use Committee at the University of Florida, adhered to animal welfare guidelines issues by the National Institutes of Health, and are consistent with the policies and standards of The Journal of Physiology as outlined in Drummond (2009).

Whole-cell recording

After incubation, slices were transferred to a recording chamber and perfused at a constant rate of 2 ml min⁻¹ with ACSF containing (in mm): 126 NaCl, 3 KCl, 1.2 NaH₂PO₄, 1.5 MgSO₄, 11 d-glucose, 2.4 CaCl₂ and 25.9 NaHCO₃, that was heated to 30°C. Slices were visualized by infrared differential interference contrast (IR DIC) microscopy using an Olympus BX51WI microscope. Whole-cell voltage-clamp recordings were performed using borosilicate glass micropipette electrodes made from a Flaming-Brown electrode puller (Sutter P-97 Sutter Instruments, Novato, CA, USA). Recording pipettes had input resistances between 3 and 6 MΩ. For experiments recording GABA_A currents, recording pipettes were filled with an internal solution containing (in mm): 95 CsMeSO₃, 55 CsCl, 1 MgCl, 0.2 EGTA, 10 Hepes, 2 diNa-ATP, 0.3 NaGTP and 5 QX-314-Cl. For the experiments in Fig. 1D a ‘low Cl⁻’ version of this internal was used which contained (in mm): 140 CsMeSO₃, 3 NaCl, 0.2 EGTA, 10 Hepes, 2 diNa-ATP, 0.3 NaGTP and 5 QX-314-Cl. These solutions were pH adjusted to 7.3 using CsOH and volume adjusted to 300–315 mosmol l⁻¹. For experiments with GABA_B-evoked responses, recording pipettes were filled with an internal solution containing (in mm): 140 potassium gluconate, 8 KCl, 0.1 CaCl₂, 2 MgCl, 1 EGTA, 10 K-Hepes, 2 diNa-ATP and 0.3 NaGTP. This solution was adjusted to pH 7.3 using KOH, and to 295 mosmol l⁻¹. Sulforhodamine 101 was added each day to the internal solution to a concentration of 63 μm, and cells were examined using epifluorescence microscopy after whole-cell recording. The ionotropic glutamate antagonists 2-amino-5-phosphonovaleric acid (APV, 40 μm) and either NBQX (10 μm) or DNQX (20 μm) were added to the bath after obtaining a whole-cell recording but before attempting to isolate evoked IPSCs.

Figure 1. Carbachol produces a presynaptic inhibition of evoked IPSCs recorded from hilar mossy cells.

A, evoked IPSCs were recorded from hilar mossy cells voltage clamped at −70 mV in the presence of ionotropic glutamate receptor antagonists. IPSCs are observed as inward currents due to high (∼60 mm) Cl⁻ in the internal solution. Bath application of CCh (3 μm) produces a significant reduction in evoked IPSC amplitude (to 67.2 ± 2.6% of baseline, n= 34, P < 0.001). Insets are averages of ∼6 sweeps from a representative cell in baseline conditions and after application of CCh. B, CCh-mediated inhibition of evoked IPSCs was accompanied by a statistically significant increase in the coefficient of variation (CV, top), but did not have a significant effect on postsynaptic input resistance (bottom, probably because our Cs⁺-based internal solution blocks K⁺ conductances coupled to postsynaptic mAChRs). C, identical application of CCh has no effect on responses produced by local application of exogenous GABA (n= 9, P= 0.53). D, identical application of CCh still causes inhibition of evoked IPSCs nearly identical to that presented in panel A, even when recorded in mossy cells voltage clamped at 0 mV with a much lower Cl⁻ in the internal solution.

In most experiments, evoked IPSCs were generated at 0.33 Hz by a concentric bipolar stimulator (FHC, Bowdoin, ME, USA) that was connected to a constant current stimulus isolation unit (World Precision Instruments, Sarasota, FL, USA). Current intensity varied between 45 and 300 μA. In Figs 9 and 10 minimal stimulation techniques were used as previously described (Nahir et al. 2007; Hofmann et al. 2008) to generate responses carried by one or few afferents. In brief, standard recording pipettes were filled with ACSF and connected to a constant current stimulus isolator. Stimuli were delivered at 0.2 Hz for 0.1 ms. Intensity was increased slowly to identify responses that occurred with a sharp stimulus threshold (see Fig. 8A). Experiments were typically conducted with stimulation intensity of 5–10 μA above this threshold which was generally between 30 and 80 μA. For experiments involving multiple pulse depression, 10 such stimuli were delivered at 10 Hz, with 15 s between each train. For local application experiments, 100 μm GABA was ejected from a glass pipette (similar to that used for whole-cell recording) using a Picospritzer III (General Valve, Fairfield, NJ, USA). Typical pulse duration was 5 ms while average pressure was ∼138 kPa. The inter-spritz interval was 20 s.

Figure 9. Multiple pulse depression of minimally evoked IPSCs is partially occluded by CCh and partially blocked by CGP.

A, demonstration of isolation of unitary or near-unitary IPSCs. Insets show individual responses for a single cell at each stimulus intensity (grey) with an average overlaid in black. Scatter plot shows average amplitude and standard error of the peak for each stimulation intensity tested. A sharp amplitude threshold was noted between 55 and 60 μA, and failure rate effectively changed from ∼100% to ∼0%. Note that unitary IPSC amplitude (with high Cl⁻ internal) can be well over 100 pA. This process is representative of how unitary or near-unitary IPSCs were isolated for all experiments in this figure and in Fig. 10. Panels B and C are both summary plots of the normalized unitary IPSC amplitude in response to pulse 1 (filled circles) and pulse 10 (open circles) during 10 pulse train delivered at 10 Hz. B, multiple pulse depression is present during the baseline and is occluded by 10 μm CCh application which also reduces the P1 amplitude. In addition, the multiple pulse depression does not return to baseline with increased stimulus intensity (sufficient to restore the P1 amplitude to baseline level). The break in the timescale is due to the fact that CCh initially produces such a large increase in spontaneous IPSCs that picking out clean responses to the minimal stimulation is difficult in some cells. C, multiple pulse depression is present during the baseline and is partially blocked (rather than occluded) by application of 10 μm CGP. Insets in B and C: average P1 (black) and P10 (grey) traces from a representative cell during the last 3 min of baseline or drug treatment.

Figure 10. CCh-sensitive components of multiple pulse depression occur later in the pulse train than CGP-sensitive components.

A, bath application of CCh has minimal effect on the P2/P1 ratio but substantially reduces multiple pulse depression later in a pulse train. B, bath application of CGP eliminates paired pulse depression, but does not prevent the development of multiple pulse depression after pulse 2. C, bath application of CCh in the presence of CGP effectively eliminates multiple pulse depression produced by a 10-pulse train. In panels A–C raw data traces (top) are average traces from a representative cell during the last 3 min of baseline or drug treatment. Scatter plots (bottom) show the pulse ratio expressed as (Pn/P1) for all pulses in the pulse train in baseline conditions (filled circles) and after drug treatment (open circles). Baseline for panel C (filled circles) is recorded in the presence of CGP, while open circles in panel C indicate a lack of multiple pulse depression in CGP+CCh. It is worth noting that there was a high degree of variability in the total amount of multiple pulse depression observed in CGP (expressed as the P10/P1 ratio). This is apparent in the representative cells shown in panels B and C (bottom raw data trace in B vs. top raw data trace in C). It is also apparent in later pulses when comparing population data in these two panels (open circles in B vs. filled circles in C). However, despite variability in total multiple pulse depression observed in CGP, the key features relevant to evaluating our model, including a lack of paired pulse depression, and clear development of depression in pulses 3–10, were consistently maintained. Although not illustrated in this panel, multiple pulse depression was also effectively eliminated in the cells presented in C at the original stimulation intensity (P2/P1 ratio: 1.2 ± 0.1, P10/P1 ratio: 0.93 ± 0.07, n= 10 P > 0.2 in both cases). D, summary graphs comparing the pulse ratios of P2/P1 and P10/P1 in the baseline and CCh conditions. The statistically significant difference between the P2 and P10 pulse ratios in the baseline is absent in the CCh condition. In addition, the P10/P1 pulse ratio in CCh is significantly different from the P10/P1 ratio in the baseline (†) whereas the P2/P1 ratio in CCh is not significantly different than the P2/P1 ratio in the baseline. E, summary graphs comparing the pulse ratios of P2/P1 and P10/P1 in the baseline and CGP conditions. In contrast to D, there is a statistically significant difference between the P2 and P10 pulse ratios in both the baseline and CGP conditions. Furthermore, the P10/P1 and P2/P1 pulse ratios in CGP are significantly different from their counterpart ratios in the baseline (†). F, scatter plots indicating the CCh-sensitive (top) and the CGP-sensitive (bottom) components of multiple pulse depression. The CCh-induced change in pulse ratio was calculated by subtracting the open circles in C (CCh+CGP) from the filled circles in the same panel (CGP only). The CGP-induced change in pulse ratio was calculated by performing an identical analysis using the data presented in B. Overall, these panels highlight the fact that CGP-sensitive components of multiple pulse depression are maximal early in the pulse train, while CCh-sensitive components increase throughout the pulse train.

Figure 8. Baclofen (3 μm) inhibits evoked IPSC amplitude and this effect is occluded by prior application of 10 μm CCh.

A, summary plot of the effects of baclofen, a GABA_B agonist, on evoked IPSCs recorded from mossy cells. Baclofen (3 μm) reduced evoked IPSC amplitude to 57 ± 6.3% of baseline, n= 7, P < 0.001. B, the inhibitory effect of 3 μm baclofen is effectively occluded by prior application of 10 μm CCh. Insets: averaged evoked IPSC responses from a representative cell during last 3 min of baseline or drug treatment.

All voltage-clamp experiments were performed using an Axon Multiclamp 700A or 700B amplifier (Molecular Devices, Sunnyvale, CA, USA). Data were acquired at 20 kHz, filtered at 2 kHz and recorded to disk via a Digidata 1322A A/D converter using Clampex v. 9 (Molecular Devices). All chemicals used in these experiments were obtained from either Sigma-Aldrich (St Louis, MO, USA) or Tocris Cookson (Ellisville, MO, USA).

Identification of hilar mossy cells

Hilar mossy cells were identified using a number of anatomical and physiological characteristics, as previously described (Frazier et al. 2003; Hofmann et al. 2006; Nahir et al. 2007). Briefly, a typical mossy cell appeared significantly larger than neighbouring hilar cells when visualized under IR DIC, had a large whole-cell capacitance (∼200 pF), and displayed at least some large-amplitude (>100 pA), spontaneous excitatory postsynaptic currents (sEPSCs) when voltage clamped at −70 mV in the absence of glutamate receptor antagonists. Cells were considered mossy cells only if they met all of the above criteria and were both multipolar and spiny (with large proximal thorny excrescences) when viewed under fluorescence microscopy. A fluorescence image of these spines is available in Frazier et al. (2003).

Data analysis and statistics

Evoked IPSCs were analysed using ClampFit v. 9. Spontaneous IPSCs were analysed in OriginPro v. 8 (Originlab, Northampton, MA, USA) using parameter-based event detection software written in OriginC by C.J.F. Statistical tests were conducted using OriginPro v. 8. A two-tailed one-sample t test was routinely used to test the significance of the effects of bath-applied chemicals on evoked IPSCs (null hypothesis, mean = 1, for data normalized to the baseline mean). Paired Student's t tests were also used where appropriate, and P values are from two-tailed tests unless otherwise noted. Error bars in all figures represent the standard error of the mean (s.e.m.).

Results

Bath-applied carbachol produces a presynaptically mediated inhibition of evoked IPSCs recorded from hilar mossy cells

Hilar mossy cells were voltage clamped at −70 mV using a CsMeSO₃ internal solution that contained approximately 60 mm CsCl, and evoked IPSCs were generated using a concentric bipolar stimulator placed in the hilus (see Methods). Under these conditions, bath application of a low concentration of a muscarinic acetylcholine receptor (mAChR) agonist, carbachol (CCh, 3 μm), reduced the amplitude of evoked IPSCs to 67.2 ± 2.6% of baseline (Fig. 1A, n= 34, P < 0.001). Several lines of evidence strongly suggest that this effect of CCh on evoked IPSCs is mediated presynaptically. For example, CCh-induced inhibition of evoked IPSCs was accompanied by a significant increase in the coefficient of variation (P= 0.002, n= 34, Fig. 1B), and did not produce a statistically significant change in postsynaptic input resistance (from 130.9 ± 22.4 MΩ to 118.65 ± 27.1 MΩ, P= 0.79, n= 34, Fig. 1C). Further, identical application of CCh failed to reduce the amplitude of responses to exogenous GABA applied via a picospritzer (104 ± 6.05% of baseline, P= 0.53, n= 9 Fig. 1D). CCh-mediated inhibition of evoked IPSCs is also not dependent on high internal Cl⁻, or on negative holding potentials, as a similar inhibition of evoked IPSCs was observed in hilar mossy cells voltage clamped at 0 mV with a CsMeSO₃ internal solution that contained lower internal Cl⁻ (to 58 ± 5.4% of baseline, n= 4, see Methods, Fig. 1D).

CCh-induced inhibition of evoked IPSCs depends on activation of M1/M3 receptors, but does not require CB1 or GABA_B receptors

As a first step towards revealing the mechanism of CCh-mediated inhibition of evoked IPSCs in the dentate gyrus, we examined the sensitivity of the inhibition to a variety of muscarinic antagonists. We found the non-specific mAChR antagonist atropine (5 μm) provided strong reversal of CCh-induced inhibition of evoked IPSCs. Specifically, in a group of nine cells tested with 3 μm CCh and then 5 μm atropine, evoked IPSC amplitude in CCh and atropine was not significantly different than during the baseline (Fig. 2A, P= 0.31). In sharp contrast, a selective M2 receptor antagonist AFDX-116 failed to reverse the inhibitory effects of CCh on evoked IPSC amplitude (from 70 ± 5.1% of baseline in CCh to 67 ± 7.3% of baseline in CCh + AFDX-116, n= 8, P= 0.63, online Supplemental Fig. S1A). Further work revealed that CCh-mediated inhibition of evoked IPSCs is prevented by pre-application of the M1/M3 selective antagonist pirenzepine (6 μm, Fig. 2B) and is partially although not significantly reversed (from 74 ± 3.5% of baseline to 83 ± 5.7% of baseline, n= 6, P= 0.10) by the more M3 selective antagonist 4-DAMP at a concentration of 200 nm (data not shown). While there are multiple mechanisms through which muscarinic agonists have been shown to modulate synaptic transmission in other parts of the CNS, M1 and M3 receptors are typically coupled to G_αq/11, and not to G_i/o (Caulfield, 1993; Caulfield & Birdsall, 1998). As such, presynaptic inhibition of synaptic transmission that depends on M1/M3 receptors, while not conclusive, is suggestive of an indirect mechanism of action.

Figure 2. CCh-mediated inhibition of evoked IPSCs depends on activation of M1/M3 receptors but not CB1 or GABA_B receptors.

A, CCh-mediated inhibition of the evoked IPSC was strongly reversed by subsequent application of the non-selective mAChR antagonist atropine (Atr, 5 μm). Evoked IPSC amplitude in CCh and atropine was not significantly different to that in baseline conditions (n= 6, P= 0.16). By contrast, the M2 selective antagonist AFDX-116 was ineffective (see text). B, pretreatment with the M1/M3 selective antagonist pirenzepine (Pz) had no direct effect on evoked IPSC amplitude, but did prevent inhibition by subsequent application of CCh. C, summary plot of the effect of 5 μm AM-251, a CB1 receptor antagonist, on CCh-induced inhibition of evoked IPSCs. Blocking CB1 receptors did not significantly reverse the effects of CCh on evoked IPSC amplitude (71.8 ± 2.2% of baseline in CCh vs. 63.2 ± 5.2% of baseline after application of AM-251, n= 4, P= 0.28). Pretreatment with AM-251 was also ineffective (see text). D, blocking GABA_B receptors with 10 μm CGP also failed to reverse CCh-induced inhibition of evoked IPSCs (66.7 ± 4.7% of baseline in CCh vs. 65.0 ± 8.1% of baseline in CCh + CGP, n= 5, P= 0.73). CGP was equally ineffective when tested at 15–20 μm (n= 7 and 4, respectively, data not shown). E, rundown of evoked IPSC amplitude when stimulating at 0.33 Hz is variable in magnitude but consistently confined to the first 3 min. Grey bars indicate typical measurement periods used in this study. Labels as follows: B, baseline, 3–6 min, after rundown has stabilized; 1, test period 1, 15–18 min; 2, test period 2, 27–30 min. See text for specific values from this control dataset. F, strong reversal of CCh-mediated inhibition was also apparent following simple washout of CCh (recovery to 94 ± 5.7% of baseline, n= 7, P= 0.33). Insets in all panels are averages of sweeps collected over a 2–3 min period in the indicated condition (or time period).

Two specific indirect mechanisms of action are worthy of careful evaluation. First, activation of postsynaptic M1/M3 receptors on hilar mossy cells could conceivably facilitate release of endogenous cannabinoids and produce inhibition of GABAergic transmission via activation of presynaptic CB1 receptors. Indeed, M1/M3-mediated and PLC-dependent postsynaptic production of endocannabinoids has already been shown to produce CB1-dependent inhibition of GABAergic afferents to CA1 pyramidal cells (Fukudome et al. 2004). Further, prior work from our lab and others has indicated that a subset of GABAergic afferents to hilar mossy cells are CB1 positive (Acsady et al. 2000; Hofmann et al. 2006; Howard et al. 2007). Nevertheless, we found that CCh-mediated inhibition of evoked IPSCs as observed in hilar mossy cells is not reversed significantly by bath application of the CB1 receptor antagonist AM-251 (Fig. 2C). Although prior work from our lab has indicated similar bath application of AM-251 effectively blocks depolarization-induced suppression of inhibition in hilar mossy cells (Hofmann et al. 2006), we also examined the effect of CCh on evoked IPSC amplitude for slices pretreated with AM-251 for a minimum of 40 min. We found that bath application of 3 μM CCh reduced evoked IPSC amplitude by 27 ± 7.7% in AM-251-pretreated slices (n= 6), and that this level of inhibition is not significantly different to control (P= 0.5, see Supplemental Fig. S1B). Second, bath application of 3 μm CCh could plausibly raise ambient GABA levels enough to produce presynaptic inhibition through activation of high-affinity GABAergic autoreceptors. Consistent with this hypothesis, we have previously reported that identical application of 3 μm CCh increases spontaneous GABAergic synaptic activity in the hilus and raises ambient GABA enough to produce CGP52432 (CGP)-sensitive and GABA_B receptor-dependent inhibition of mossy fibre inputs to hilar mossy cells (Nahir et al. 2007). However, that mechanism is also not sufficient to explain our current observations, as 3 μm CCh produces inhibition of GABAergic afferents to hilar mossy cells that is insensitive to subsequent application of 10 μm CGP (Fig. 2D). Additional experiments indicated that CGP was equally ineffective when applied at 15 μm, 20 μm or 100 μm (n= 7, 4 and 5, respectively; see Supplemental Fig. S2). Further, we have also demonstrated that neither nicotinic acetylcholine receptors (nAChRs) nor glycine receptors are involved in CCh-mediated inhibition of evoked IPSCs. Specifically, we noted that CCh-mediated inhibition of evoked IPSCs is not reversed by simultaneous application of the nAChR antagonists mecamylamine (10 μm) and methyllycaconitine (70 nm, evoked IPSC amplitude 74 ± 1.5% of baseline in CCh vs. 71 ± 3.0% of baseline in nAChR antagonists, n= 4, P= 0.37; see Supplemental Fig. S3A). Similarly, bath application of the glycine receptor antagonist strychnine (250 nm; see Supplemental Fig. 3B) also fails to reverse CCh-mediated inhibition of evoked IPSCs. In fact, there was a mild reduction of IPSC amplitude when 250 nm strychnine was applied (from 54 ± 6.0% to 46 ± 5.9% of the pre-CCh baseline, P= 0.08). The IC₅₀ for strychnine inhibition of glycine receptors in this experiment is expected to be <50 nm (Jonas et al. 1998; Ghavanini et al. 2005). As expected, higher concentrations of strychnine (e.g. 3 μm) reduced evoked IPSC amplitude whether applied before or after CCh (data not shown) consistent with partial block of GABA_A receptors at that concentration.

An additional feature of the data worth noting is that in a number of panels in Figs 1 and 2, clear rundown of the response is apparent during the 6 min baseline period. This is potentially problematic because it could conceivably artificially exaggerate the apparent affect of CCh routinely measured at 15–18 min, or reduce the apparent effect of antagonists applied after CCh (such as atropine, AFDX-116, AM-251, etc.) routinely measured at 27–30 min. In order to carefully evaluate these possibilities, we examined rundown of evoked IPSC amplitude over a 30 min period absent of any drug application. Our results indicated that rundown of evoked IPSC amplitude when stimulated at 0.33 Hz is confined to the first 3 min of the baseline period. Specifically, in six cells tested, evoked IPSC amplitude measured between 3 and 6 min into the recording was just 83 ± 6.3% of the mean amplitude observed in the first five pulses (P= 0.03). By contrast, mean evoked IPSC amplitude at both 15–18 and 27–30 min was not altered from that observed at 3–6 min (98.9 ± 6.3% and 100.1 ± 7.9% of baseline, respectively, P≥ 0.8 in both cases, Fig. 2E). For that reason, effects of bath-applied compounds on evoked IPSC amplitude throughout this paper are quantified relative to the mean observed at the 3–6 min time period, and in our view no additional corrections for baseline rundown are required. Consistent with that conclusion, we noted that evoked IPSC amplitude recovered from 68 ± 5.5% of baseline during bath application of CCh to 94 ± 5.7% of baseline following simple washout (n= 7, P= 0.33 for recovery relative to baseline, Fig. 2F).

Overall we found the data presented in Figs 1 and 2 to be quite intriguing because collectively, they are indicative of a phenomenon that appears to be mechanistically distinct from most, if not all, well-described previous examples of CCh-mediated inhibition of synaptic transmission in the CNS. However, it is important to note that while insensitivity of CCh-mediated inhibition to AM-251 is a strong argument against an essential role for CB1 receptors, insensitivity to a selective GABA_B receptor antagonist, is not in and of itself sufficient to indicate an independence from CCh-induced changes in inhibitory tone.

Bath application of CCh produces opposing but related effects on evoked and spontaneous inhibitory synaptic transmission

We next sought to test more directly whether CCh-mediated inhibition of evoked IPSCs described here could, despite apparent insensitivity to CGP, be mechanistically related to the CCh-mediated facilitation of spontaneous IPSCs as described earlier (Hofmann et al. 2006; Nahir et al. 2007). Towards that end, we used a bipolar stimulator in the presence of DNQX and APV to evoke IPSCs at 5 s intervals, and then simultaneously monitored the effects of bath-applied CCh on both evoked and spontaneous IPSCs (Fig. 3). Analysis of the data from this experiment produced several interesting observations.

Figure 3. Bath application of CCh produces opposing but related effects on evoked and spontaneous inhibitory synaptic transmission.

A, large inset shows a typical sweep where the thick bar indicates the time period during which the evoked IPSC amplitude (and area) were measured, and the thin bar indicates the time period (3.5 s) in which sIPSCs were recorded. Similar sweeps were recorded every 5 s. Lower insets indicate that bath application of CCh simultaneously inhibits the evoked IPSC (left traces) and enhances the spontaneous IPSCs (right traces). B, summary plot indicating CCh-induced changes in sIPSC parameters, normalized to the baseline, in both control conditions (open bars), and in the presence of the M1/M3 muscarinic antagonist pirenzepine (Pz, 6 μm). Pirenzepine completely blocks the effects of CCh on both spontaneous IPSCs (presented here) and evoked IPSCs (Fig. 2). Insets to the right of the bar graph show the average of hundreds of individual spontaneous IPSCs recorded over a 3 min period both before (base, grey line) and after (CCh, black line) application of CCh in a representative cell (pirenzepine absent). The top inset indicates the degree to which CCh increased spontaneous IPSC amplitude and area. The bottom inset shows the same traces normalized to the peak amplitude. They are presented to emphasize the fact that spontaneous IPSC kinetics do not change appreciably, even as amplitude and area increase, with the application of CCh (see main text). C, summary plot showing CCh-induced changes in both spontaneous IPSCs (open circles) and evoked IPSCs (filled circles) over time. D, linear regression analysis of data from panel C. Average spontaneous IPSC area was quantified from 8 to 11 min (due to lack of stability in some cells), while evoked IPSC amplitude was quantified from 15 to 18 min. CCh-mediated inhibition of the evoked response was strongly correlated with average area of the sIPSCs observed in the presence of CCh. See text in Results for information on the relationship between other spontaneous IPSC parameters and the CCh-mediated inhibition of the evoked response.

First, we found that opposing effects of CCh on spontaneous and evoked IPSCs are readily apparent when measured simultaneously in a single experiment. Across seven cells, bath application of 3 μm CCh reduced evoked IPSC amplitude to 73 ± 5.1% of baseline (P > 0.001), while frequency, amplitude and area of sIPSCs were all significantly increased (frequency: from 6.6 ± 1.70 Hz to 18.0 ± 3.17 Hz, amplitude: from 27.2 ± 2.5 pA to 58.8 ± 9.64 pA, area: from 367 ± 59.7 ms pA to 718 ± 77.7 ms pA, P≤ 0.02 in all cases, Fig. 3A–B). The time-to-peak change was typically faster for all parameters of the spontaneous IPSCs than for inhibition of the evoked IPSC; however, the time of onset was not distinguishable with the 3.5 s bin times employed.

It is worth noting that all effects of CCh on spontaneous IPSCs described here (increases in frequency, amplitude and area) are likely to be mediated presynaptically. A postsynaptic mechanism responsible for increases in spontaneous IPSC amplitude and area would require either an appropriate direct effect of CCh on the intrinsic properties of hilar mossy cells, or a clear change in the gating properties of the postsynaptic GABA_A receptors. The first possibility was ruled out under our experimental conditions by data presented in Fig. 1. The second hypothesis can largely be eliminated based on the observation that spontaneous IPSC kinetics do not change appreciably as amplitude and area increase. For example, for the seven cells examined in Fig. 3C and D, spontaneous IPSC decay time measured as the time in milliseconds from the peak to the point where current decayed to 37% of peak amplitude was 15.6 ± 1.0 ms in control conditions and 15.4 ± 1.1 ms after application of CCh (n= 7, P= 0.77). The general stability of spontaneous IPSC kinetics can be seen for a representative cell in the inset traces in Fig. 3B. Importantly, we can also report that miniature IPSC kinetics (observed in the presence of 1 μm TTX) are not altered by bath application of CCh (monoexponential decay time constant: 8.68 ± 0.66 ms in baseline vs. 7.45 ± 0.79 ms in CCh, n= 17, P= 0.29, data not shown). Overall these findings suggest that increased spontaneous IPSC amplitude and area are probably due to increased synchrony of presynaptic release; however, further investigation of that specific mechanism is outside the scope of the current study.

The second important finding apparent in Fig. 3 is that when examined on a cell-by-cell basis, CCh-mediated increases in spontaneous IPSCs are strongly correlated with CCh-mediated inhibition of the evoked IPSC (Fig. 3C and D). When evaluating these results it is important to ask which specific parameters of spontaneous IPSCs might reasonably be expected to have the strongest correlation with CCh-induced changes in evoked IPSC amplitude. Since our overall hypothesis is that CCh-mediated inhibition of evoked IPSCs depends on increases in ambient GABA, then it stands to reason that spontaneous IPSC parameters that are most closely related to the amount of GABA released (either per event or per unit time) ought to have the strongest correlation with changes in evoked response amplitude. Indeed, we find that the average area of the spontaneous IPSCs as observed several minutes after application of CCh, and the integrated area under all spontaneous IPSCs over the same time period, are both very strongly correlated with CCh-mediated decreases in evoked IPSC amplitude (adjusted R²= 0.74 and 0.60, P < 0.01 and P= 0.02, respectively, n= 7, Fig. 3D). For this analysis, area of each event is calculated as the area under the curve measured between the event start time and the point after the peak where event amplitude has decayed by 63% (to 37% of peak amplitude). Integrated area under all spontaneous IPSCs for a given time period is calculated as the average area of all events multiplied by the total number of events. We believe these parameters are the best correlates because they are likely to be the parameters that most directly reflect activity-dependent increases in ambient GABA outside the synaptic cleft. Nevertheless, spontaneous IPSC frequency and amplitude also showed similar trends that were nearly statistically significant (adjusted R² values were 0.39 and 0.32, respectively, n= 7, P= 0.08 and 0.10). While correlation does not prove causation, overall we believe these results are clearly highly consistent with the hypothesis that CCh-mediated inhibition of evoked IPSCs is mechanistically related to CCh-induced facilitation of spontaneous IPSCs (and resulting increases in ambient GABA).

Third, we found that pirenzepine blocked not only CCh-mediated inhibition of the evoked IPSC (as presented in Fig. 2), but also completely eliminated CCh-mediated facilitation of spontaneous IPSCs (Fig. 4B, black bars). This is an important observation because it suggests that action of muscarinic agonists on a common pool of M1/M3 receptors is responsible for both facilitation of spontaneous IPSCs and inhibition of evoked IPSCs.

Figure 4. CCh fails to inhibit evoked IPSCs reported by postsynaptic GABA_B receptors in the presence of the GABA_A receptor antagonist picrotoxin.

A, A pulse train (6 pulses at 200 Hz, arrowhead) was used to illicit GABA_B evoked responses in the presence of 50 μm picrotoxin. Flyout (between the dotted lines) shows the response to the pulse train in control conditions (baseline), after bath application of 3 μm CCh (CCh), and then after subsequent application of 10 μm CGP (CGP, this trace is indicated by the horizontal arrow). Lower traces in panel A indicate the CCh-sensitive and CGP-sensitive components of the currents evoked by the pulse train. These traces were obtained by subtracting the response in CCh from the baseline response, and by subtracting the response in CGP+CCh from that in CCh only, respectively. B, summary bar graph of the average amplitude of the CCh- and CGP-sensitive currents. GABA_B-mediated evoked responses recorded from mossy cells in the presence of 50 μm picrotoxin were not inhibited by 10 μm CCh, but were blocked by 10 μm CGP (n= 8, CCh-sensitive current: −6.02 ± 2.65 pA, P= 0.06; CGP-sensitive current: 19.5 ± 4.75 pA, P= 0.004). Similar results were obtained in identical experiments that used BMI instead of PTX (see text).

CCh fails to inhibit evoked IPSCs reported by postsynaptic GABA_B receptors in the presence of the GABA_A receptor antagonist picrotoxin

If CCh-mediated inhibition of the evoked IPSC does in fact depend on increases in ambient GABA, and yet is insensitive to CGP, it seems natural to hypothesize that high-affinity presynaptic ionotropic GABA_A receptors might be involved. However, this is a difficult hypothesis to test experimentally because there are no well-established and reliable methods for blocking putative presynaptic GABA_A receptors involved in autoreceptor inhibition without also blocking the postsynaptic GABA_A receptors necessary to observe the evoked response. In order to work around this problem, we decided to test the effect of CCh on GABA_B-mediated IPSCs recorded from hilar mossy cells in the presence of the GABA_A receptor antagonist picrotoxin (PTX). We reasoned that if GABA_A receptor activation is required for CCh-mediated inhibition of evoked responses, then CCh should have no effect on GABA_B-mediated evoked responses observed in the presence of the GABA_A receptor antagonist PTX. Indeed, we found this to be the case. GABA_B responses were evoked in hilar mossy cells using a train of six pulses delivered at 200 Hz. Importantly, this train was evoked using a bipolar stimulator placed in the hilus exactly as in previous experiments, and thus evoked GABA release is likely to originate from the same afferents that contributed to the evoked IPSC in earlier experiments (see also control below). GABA_B-dependent evoked responses were recorded as outward currents from hilar mossy cells voltage clamped at −60 mV using a potassium gluconate-based internal solution (see Methods). These outward currents were not significantly reduced by bath application of 3 μm CCh in the presence of 50 μm PTX but were reduced by subsequent application of CGP (Fig. 4A and B). This result strongly suggests that GABA_A receptors are indeed required to observe CCh-mediated inhibition of evoked IPSCs. While we attempted to use a KFl-based internal solution containing disodium 4,4′-diisothiocyanatostilbene-2,2′-disulfonate (DIDS) to selectively block GABA_A receptors in the postsynaptic cell, measurement of GABA_B-evoked responses was not feasible with that internal solution because postsynaptic GABA_B receptors were also blocked (data not shown). Similarly, we were unable to get robust isolation of GABA_B IPSCs using 1–2 mm DIDS in a potassium gluconate-based internal solution (either small residual GABA_A currents remained, or both GABA_A and GABA_B responses were absent, data not shown). We also attempted to isolate GABA_B responses in the absence of PTX, either after the faster GABA_A-mediated responses or in cells voltage clamped at the chloride equilibrium potential. However, neither of these strategies was sufficient to reliably determine if GABA_B-mediated currents are in fact inhibited by CCh in the absence of PTX. However, we were able to demonstrate that the same pulse train that produces GABA_B-mediated postsynaptic currents in the presence of PTX, also produces CCh-sensitive and GABA_A-dependent postsynaptic currents in the absence of PTX. Specifically, evoked IPSC amplitude as measured following a six pulse train at 200 Hz in the absence of PTX was reduced to 69 ± 9.7% of baseline by bath application of 3 μm CCh (n= 6), and this effect is not significantly different than that reported on evoked IPSCs produced by single pulses in Fig. 1 (P= 0.89). That result indicates, at least, that the same population of afferents that produce GABA_B-sensitive currents in the presence of PTX in Fig. 4, are sensitive to CCh-mediated inhibition in the absence of PTX, and that the pulse train used to evoke GABA_B responses does not, in and of itself, reduce sensitivity to CCh. Similarly, in order to rule out the remote possibility of a direct antagonistic action of PTX on mAChRs, we repeated the experiments shown in Fig. 4 using 20 μm bicuculline (BMI) instead of PTX. We found that CCh also failed to have any significant effect on GABA_B-mediated IPSCs in the presence of BMI (mean CCh-sensitive current amplitude was –6.8 ± 3.1 pA, P= 0.09 vs. CGP-sensitive current of 19.1 ± 1.2 pA, P < 0.001, n= 5). Cumulatively, we believe these experiments provide very compelling if not definitive evidence that CCh-mediated inhibition of evoked IPSCs is likely to depend on activation of presynaptic GABA_A receptors.

NO-711 and CCh-mediated increases in ambient GABA display differential sensitivity to CGP

Next we reasoned that if CCh-mediated inhibition of evoked IPSCs does indeed depend on activation of presynaptic GABA_A receptors, then other means of increasing ambient GABA around the terminals ought to inhibit evoked IPSCs in a manner that is functionally occluded by CCh. Consistent with that hypothesis we found bath application of the GAT-1 transport blocker NO-711 (1 μm) causes an inhibition of evoked IPSCs (to 71.9 ± 12.5% of baseline, n= 5, P < 0.01, Fig. 5A) that is very similar to that produced by 3 μm CCh. Further, we found that the inhibitory effects of NO-711 on the evoked IPSC amplitude were partially occluded by 3 μm CCh and fully occluded by 10 μm CCh. Specifically, the evoked IPSC amplitude in the presence of both NO-711 and 10 μm CCh was 99.5 ± 6.03% of the amplitude in 10 μm CCh alone (Fig. 5B, n= 8, P= 0.94). Overall, these results are consistent with the idea that CCh and NO-711 both inhibit evoked IPSCs by producing increases in the concentration of ambient GABA. Surprisingly, however, we found that NO-711-mediated inhibition of evoked IPSCs is completely reversed by bath application of CGP (to 96.4 ± 3.2% of baseline, Fig. 5C, n= 5, P= 0.33), and that 10 μm CCh, despite fully occluding the effects of NO-711, is completely CGP insensitive (Fig. 5D, n= 5).

Figure 5. Block of GAT-1 GABA transporters inhibits evoked IPSCs in a manner that is occluded by bath application of CCh and yet is also reversed by CGP.

A, summary plot of the effect of a GAT-1 blocker on evoked IPSCs recorded from mossy cells. NO-711 (1 μm) produced an inhibition similar to that observed in CCh (71.9 ± 5.56% of baseline, n= 5, P= 0.007). B, summary plot of the effects of 1 μm NO-711 on inhibition mediated by 10 μm CCh. The inhibition previously seen with NO-711 application was occluded by the CCh-mediated inhibition (evoked IPSC amplitude in CCh + NO-711 was 99.5 ± 17.1% of CCh alone, n= 8, P= 0.94). C, summary plot of the effects of 10 μm CGP on inhibition mediated by 1 μm NO-711. CGP treatment reverses the NO-711-mediated inhibition (144.1 ± 12.10% of NO-711, n= 3, P= 0.07). D, summary plot of the effects of 10 μm CGP on inhibition mediated by 10 μm CCh. CGP had no effect on the CCh-mediated inhibition (111.08 ± 7.62% of CCh, n= 5, P= 0.22). Insets: averaged evoked IPSC responses from a representative cell during last 3 min of baseline or drug treatment.

A novel model for activity-dependent autoreceptor-mediated feedback inhibition

The experiments presented above raise a challenging yet crucial question: if NO-711 and CCh both inhibit evoked IPSCs by raising ambient GABA, how can NO-711-mediated inhibition be completely reversed by CGP while CCh-mediated inhibition is completely unaffected? In an effort to reconcile these apparently disparate results we have developed a hypothetical model for how GABA_A and GABA_B autoreceptors might interact to provide complex regulation of inhibitory transmission to hilar mossy cells (Fig. 6). According to the model, GABAergic afferents express presynaptic or axonal GABA_A receptors, presynaptic GABA_B receptors and presynaptic NO-711-sensitive GABA transporters. A key anatomical feature of the model is that the GABA_B receptors and the NO-711-sensitive transporters are located both near to one another, and closer to the release site, than the more distal (possibly axonal) GABA_A receptors. A key functional variable relevant to understanding the experimental results presented above is the background level of synaptic activity in the inhibitory network. Thus, we hypothesize that when synaptic activity is low, bath application of the GAT-1 blocker NO-711 is able to cause only a small increase in ambient GABA around the terminal, sufficient to activate GABA_B receptors located close to the release site, but insufficient to activate the more distal GABA_A receptors (Fig. 6A). This would explain why the NO-711-mediated inhibition presented in Fig. 5 is CGP sensitive. By contrast, when CCh is applied, spontaneous IPSC amplitude and frequency are increased, so much so that a larger field of ambient GABA develops around the terminals, despite the fact that GAT-1 transporters are not blocked (Fig. 6B). We hypothesize that this CCh-mediated increase in ambient GABA is sufficient to activate the GABA_A receptors located farther from the release site. We also hypothesize that this CCh-mediated inhibition is CGP insensitive because axonal or preterminal shunting of the action potential by GABA_A receptors effectively reduces the impact of GABA_B-mediated shunting in the terminal (as it relates to exocytosis). In fact, overall this model envisions GABA_B receptors as providing rather conventional feedback inhibition under light to moderate load, while GABA_A receptors function more as an analog gate, capable of scaling the action potential before it enters the terminal under conditions when the network is extremely active.

Figure 6. A hypothetical model for activity-dependent regulation of GABA_A and GABA_B autoreceptor- mediated inhibition.

A, when spontaneous activity is low, pharmacological blockade of GAT-1 transporters creates a small increase in ambient GABA around the terminals sufficient to activate proximal GABA_B autoreceptors. B, in the presence of CCh spontaneous activity is high. This increased activity produces a larger increase in ambient GABA, despite the fact that GAT-1 transporters are not blocked. The larger increase in ambient GABA is sufficient to activate more distal GABA_A receptors. C, depiction of an interneuron with minimal stimulation of 10 pulses at 10 Hz (see Figs 10 and 11). The concentration of the released GABA increases and spreads with each additional pulse in the train describing how the GABA_B autoreceptors are responsible for the multiple pulse depression in the beginning of the train and the GABA_A autoreceptors are responsible for the multiple pulse depression at the end of the train.

In the remaining section of this study we will present eight specific predictions about the physiology of this system that are based on the model as described above. All eight predictions are readily accessible for experimental evaluation, and indeed, experimental work in all eight instances has produced results that are consistent with expectations.

Predictions of the model vs. experimental data

An initial yet essential prediction of the model as described above is that lower doses of CCh should produce inhibition of evoked IPSCs with greater CGP sensitivity. The reasoning behind this hypothesis is straightforward. If the key difference between the effects of 1 μm NO-711 and 3–10 μm CCh on evoked IPSC amplitude is due to underlying differences in the extent to which they raise ambient GABA around the terminal, then lower concentrations of CCh should produce smaller increases in ambient GABA and correspondingly more NO-711-like (e.g. CGP-sensitive) inhibition. Indeed, we found that bath application of 500 nm CCh produced a much smaller increase in spontaneous IPSC frequency and integrated area than 3 μm CCh (Fig. 7A and B), and also resulted in mild (but clear) inhibition of evoked IPSCs (to 81.8 ± 3.2% of baseline, n= 9, P < 0.001, Fig. 7C). Importantly, the inhibition of evoked IPSCs produced by 500 nm CCh, unlike that produced by 3–10 μm CCh, was strongly reversed by subsequent application of CGP (to 95.4 ± 4.1% of baseline, n= 9, P < 0.001, Fig. 7C). Further, among cells tested with 500 nm CCh, there was a clear correlation between the degree of inhibition of the evoked IPSC and the extent of CGP-mediated recovery, such that cells that showed less inhibition in response to CCh were more likely to return to or exceed the baseline amplitude during application of CGP (Fig. 7E, open circles, dashed regression line, adjusted R²: 0.65, P= 0.006). This relationship was notably absent in cells tested with 3–10 μm CCh (Fig. 7E, filled circles, continuous regression line, adjusted R²: −0.08, P= 0.57). In order to confirm that these results truly indicate CGP-mediated reversal of the inhibition produced by 500 nm CCh, we further demonstrated that there was no significant increase in evoked IPSC amplitude when CGP is applied in the absence of CCh, and conversely, that 500 nm CCh has no effect on evoked IPSC amplitude when applied to slices pretreated with 10 μm CGP (CGP: −0.49 ± 3.8% change from baseline, P= 0.9; CCh: −2.92 ± 3.6% change from CGP, P= 0.5, n= 6, data not shown). Cumulatively, these data indicate that as per the model above, a lower concentration of bath-applied CCh does indeed produce smaller facilitation of spontaneous IPSCs and also more CGP-sensitive inhibition of evoked synaptic transmission.

Figure 7. Although inhibition of evoked IPSCs produced by 3 μm CCh is insensitive to CGP, inhibition of evoked IPSCs produced by 500 nm CCh is reversed by CGP.

A, large inset shows a typical sweep where the bar indicates the time period (3.5 s) in which spontaneous IPSCs were recorded. Similar sweeps were recorded every 5 s. Lower insets indicate that the effect of 500 nm CCh on spontaneous IPSCs in a representative cell. B, effects of CCh (500 nm, open circles; or 3 μm, filled circles) on spontaneous IPSCs parameters. CCh (500 nm) produced much milder increases in spontaneous IPSC frequency (left panel) and integrated area (right panel) than 3 μm CCh, although effects on spontaneous IPSC area (middle panel) were comparable. C, summary plot comparing the effect of 10 μm CGP on the inhibition produced by 3 μm CCh (filled circles – data from Fig. 6D, n= 5) and 500 nm CCh (open circles, n= 9). CGP reverses CCh-mediated inhibition only in response to 500 nm CCh. D, cell by cell summary plots of the data presented in panel C. Open circles represent individual cells and filled circles indicate the average value. E, linear regression analysis of the evoked IPSC amplitude in CCh as a % of baseline vs. the % recovery of the evoked IPSC amplitude in CGP (500 nm CCh: open circles and dotted regression line; 3–10 μm CCh: filled circles and continuous regression line). There is a strong correlation between these variables in cells treated with 500 nm CCh but not in cells treated with 3 μm CCh.

A second related prediction of our model is that NO-711-mediated inhibition of evoked IPSCs should demonstrate reduced sensitivity to CGP if it could be manipulated to cause a larger increase in ambient GABA. To test this hypothesis we evaluated the ability of 10 μm CGP to reverse inhibition produced by increasing concentrations of NO-711 from 1 to 50 μm. When 50 μm NO-711 was applied in the presence of a modified ACSF that increases neuronal excitability, an inhibition of evoked IPSCs was produced that was, consistent with expectations, largely insensitive to CGP. These experiments, with additional details, are presented in the online supplemental data (Supplemental Fig. S4). At this point it seems apparent that whether inhibition of evoked IPSCs is produced by NO-711 or CCh, CGP-mediated reversal is apparent under conditions that are expected to produce a relatively modest increase in ambient GABA, and is reduced or absent when a larger increase in ambient GABA is expected. Finally, a third prediction in this area is that CGP-mediated reversal of inhibition should be apparent even when a large increase in ambient GABA is produced, if the effect of presynaptic GABA_A receptors could be effectively minimized without completely eliminating the postsynaptic IPSC. To evaluate this hypothesis we tested 10 μm CGP against inhibition of evoked IPSCs produced by 3 μm CCh in the presence of 5 μm strychnine. This concentration of strychnine is expected to fully block glycine receptors, and also to partially antagonize GABA_A receptors (Jonas et al. 1998; Ghavanini et al. 2005). Specifically, we noted that bath application of 5 μm strychnine reduced the amplitude of the GABA_A-mediated postsynaptic IPSC to 37.2 ± 4.9% of baseline levels (n= 6, P < 0.001). Importantly, we also found that inhibition of residual evoked IPSCs produced by 3 μm CCh in the presence of 5 μm strychnine is in fact significantly reversed by bath application of 10 μm CGP (from 74.8 ± 1.7 to 93 ± 3.6% of baseline, n= 5, P < 0.01). Overall, this is the only condition in the entire study in which CGP effectively reverses inhibition of evoked IPSCs produced by a low micromolar concentration of CCh. These experiments are also illustrated in Supplemental Fig. S5.

A fourth key prediction of the model is that bath application of a GABA_B receptor agonist should also produce inhibition of evoked IPSCs recorded from hilar mossy cells, while a fifth related prediction is that such GABA_B-mediated inhibition should be functionally occluded by bath application of low micromolar doses of CCh. Again, these predictions stand up well to experimental examination. Specifically, we found bath application of 3 μm baclofen caused a clear inhibition of evoked IPSC amplitude (to 57 ± 6.3% of baseline, Fig. 8A, n= 7, P < 0.001) without significantly altering postsynaptic holding current or input resistance (96.8 ± 2.47 and 101 ± 8.90% of baseline, respectively, P= 0.9 in both cases). Further, we noted that this effect is dramatically reduced (and no longer statistically significant) in cells pretreated with 10 μm CCh (Fig. 8B, 84.4 ± 8.5% of baseline, n= 6, P= 0.12). It is interesting to speculate on the specific mechanism of this occlusion. One straightforward possibility is that this occlusion occurs because GABA_B receptors are desensitized in the presence of CCh by higher local concentrations of GABA near the release sites. We believe we can effectively eliminate this hypothesis because we found that GABA_B-mediated evoked IPSCs are in fact insensitive to CCh in the presence of PTX or BMI (Fig. 4). Similarly, we noted that a high concentration of baclofen (10 μm) will still produce inhibition of evoked IPSCs in 10 μm CCh (Supplemental Fig. S6). Thus, it appears that CCh-mediated increases in ambient GABA increase competition for the GABA_B receptor and simply shift the dose–response curve for baclofen to the right.

A sixth important prediction of the model is that high-frequency pulse trains, delivered in the absence of CCh or CGP, should produce multiple pulse depression that is partially occluded by CCh and partially blocked by CGP. In order to test this hypothesis we used minimal stimulation techniques to produce evoked IPSCs to hilar mossy cells that depend on activation of one or few afferents. The techniques for isolating these unitary or near-unitary responses have been described in detail in some of our prior publications, are reviewed in the Methods section of the current manuscript, and are illustrated in Fig. 9A (see Methods and Fig. 9 legend for more details). Once a minimally evoked IPSC was isolated, a 10-pulse train was delivered at 10 Hz. As expected, the amplitude of the minimally evoked IPSC showed depression during this train which we refer to as multiple pulse depression. For the group of cells illustrated in Fig. 9A and B, in control conditions, the P10/P1 ratio was 0.53 ± 0.03 (n= 15, P < 0.001). In order to test the current hypotheses we examined, in separate experiments, the effects of bath application of CCh and CGP on the P10/P1 ratio. We found that bath application of CCh both decreased the P1 amplitude and increased the P10/P1 ratio from 0.51 ± 0.02 to 0.65 ± 0.06 (n= 5, P= 0.04, 1-tail paired t test). This change in the P10/P1 ratio was not mediated solely by a reduction in the amount of synaptically released GABA available for spillover because multiple pulse depression was not restored when stimulus intensity was increased as needed to restore the P1 amplitude to baseline levels (Fig. 9B, P10/P1 with high stimulation: 0.80 ± 0.12, n= 5, P= 0.03, 1-tail paired t test). If, as earlier data suggest, the CCh-mediated inhibition of P1 is produced by increases in ambient GABA, then this result demonstrates that increases in ambient GABA functionally occlude the development of multiple pulse depression by activating high-affinity GABAergic autoreceptors prior to delivery of the pulse train. By contrast, bath application of CGP on otherwise naïve slices had no significant effect on the P1 amplitude (Fig. 9C, n= 7, P= 0.28), but increased the P10 amplitude, which again leads to a significant increase in the P10/P1 ratio from 0.54 ± 0.05 to 0.75 ± 0.05 (n= 10, P < 0.001). This result is also consistent with our model in indicating that a significant portion of multiple pulse depression is blocked by prior antagonism of GABA_B receptors.

A seventh essential prediction of our model is that careful examination of the development of multiple pulse depression during the pulse train should reveal an early dependence on CGP-sensitive mechanisms and a later dependence on CCh-sensitive mechanisms. Indeed, this too seems to be the case. In Fig. 10A we show the development of multiple pulse depression across all 10 pulses in control conditions, and again following bath application of CCh. Similar data are presented in Fig. 10B for a separate group of cells that were exposed (after baseline recording) to CGP instead of CCh. The first important point to make is that although bath application of CCh produced a small increase in the P2/P1 ratio (from 0.78 ± 0.5 to 0.86 ± 0.04), the effect was not statistically significant (n= 5, P= 0.13, Fig. 10D, open vs. filled bars for P2/P1). In sharp contrast, bath application of CGP effectively eliminated paired pulse depression measured at this time point (0.81 ± 0.03 in control conditions vs. 1.03 ± 0.07 in CGP, n= 10, P= 0.007, Fig. 10E, open vs. filled bars for P2/P1). This simple analysis is sufficient to establish that CGP affects multiple pulse depression at an earlier time point than CCh. Importantly, however, after P2, further development of multiple pulse depression had much greater sensitivity to CCh than to CGP. This later conclusion can be validated by comparing changes in the P2/P1 ratio with changes in the P10/P1 ratio in Fig. 10A and B. Specifically we found that in control conditions significant multiple pulse depression develops after P2 (e.g. from a P2/P10 ratio of 0.78 ± 0.05 to a P10/P1 ratio of 0.51 ± 0.02, n= 5, P= 0.02, in panel 10A). Further, we noted that CCh, despite having no significant effect on the P2/P1 ratio, effectively prevented development of further multiple pulse depression after P2. This is indicated by the fact that the P10/P1 ratio in CCh is not significantly different to the P2/P1 ratio observed in the same condition (0.78 ± 0.05 vs. 0.80 ± 0.12, n= 5, P= 0.45, Fig. 10D, open bars). In sharp contrast, CGP effectively eliminated development of multiple pulse depression at pulse 2, and yet did not interfere with the development of multiple pulse depression in pulses 3–10. The later point is reinforced by the observation that, in the presence of CGP, and in contrast to the CCh dataset, the P10/P1 ratio is still significantly smaller than the P2/P1 ratio (0.75 ± 0.05 vs. 1.03 ± 0.07, respectively, n= 10, P= 0.002, Fig. 10E, open bars). These important differences are highlighted in the open bars in Fig. 10D and E, and in our view, provide a compelling argument in support of this seventh prediction.

An eighth and final key prediction of our model is that multiple pulse depression should be effectively eliminated by the combined effects of CCh and CGP. To test this hypothesis, slices were exposed to CGP prior to isolation of a minimally evoked IPSC. The key features of multiple pulse depression as observed in CGP, including lack of depression at P2, and clear development of multiple pulse depression in pulses 3–10 are apparent in the raw data trace and the summary plot presented in Fig. 10C (see Fig. 10 legend for additional comments on variability of total multiple pulse depression in CGP). Importantly, addition of CCh to these CGP-treated slices effectively eliminated multiple pulse depression in all remaining pulses (Fig. 10C). This observation simultaneously confirms the eighth prediction of our model, and reinforces the earlier observation that CCh-mediated increases in ambient GABA predominantly effect multiple pulse depression after pulse 2. Next, we used the data presented in Fig. 10B and C to quantify the CCh-sensitive and the CGP-sensitive components of multiple pulse depression at all pulses throughout the train. Consistent with expectations, we found CCh-sensitive components of multiple pulse depression increase steadily throughout the pulse train (Fig. 10F top panel) while CGP-sensitive components of multiple pulse depression are maximal after pulse 2, and do not change significantly in pulses 3–10 (Fig. 10F, bottom panel, see legend for further details).

Finally, in a last series of experiments, we asked whether key features of this model would still be apparent when examining the effect of CCh and CGP on multiple pulse depression generated using conventional bulk stimulation techniques. Identical pulse trains were delivered via a concentric bipolar stimulator (as opposed to a minimal stimulator, see Methods) and evoked IPSCs were recorded from hilar mossy cells as in previous experiments. Although additional complexities, such as increased heterosynaptic activation of presynaptic receptors produced by near-simultaneous release from multiple afferents, are likely to be involved here, several key features of our model were indeed still apparent. Most notably we found that CGP still reduced the P2/P1 ratio significantly (from 0.80 ± 0.07 in control conditions to 0.93 ± 0.03, n= 6, P= 0.02, Fig. 11A and B) and that subsequent application of CCh to the same cells still caused a reduction in multiple pulse depression that was clearly specific to later pulses in the train. For example, the P10/P1 ratio increased from 0.94 ± 0.04 in the presence of CGP to 1.05 ± 0.05 in the presence of both CGP and CCh (n= 6, P= 0.05, Fig. 11B). Further, an analysis of the development of multiple pulse depression throughout the train continued to indicate a rapid rise in the CGP-sensitive component that was near-maximal after two to three pulses, and a slow increase in the CCh-sensitive component that continued throughout the pulse train (Fig. 11C, see legend for further details).

Figure 11. Macroscopic stimulation also produces multiple pulse depression of evoked IPSCs with distinct CCh-sensitive and CGP-sensitive components.

A, bath application of 10 μm CGP dramatically reduces multiple pulse depression of evoked IPSCs produced with a concentric bipolar stimulator. Scatter plot shows average P10 amplitude (open circles) and average P1 amplitude (filled circles, n= 6). Subsequent application of 10 μm CCh reduces the P1 amplitude and eliminates the remaining multiple pulse depression. B, data from A are plotted to show the pulse ratio (expressed as Pn/P1) for all pulses in the pulse train during baseline (filled circles), after application of CGP (open downward triangles) and after application of both CGP and CCh (filled upward triangles). Note that in these experiments CCh and then CGP were applied consecutively to each cell. As with the experiments on unitary IPSCs, CGP has maximal effect early in the pulse train, while CCh-sensitive components of multiple pulse depression are not apparent until later in the train. C, summary plot of CCh-sensitive (filled upward triangles) and CGP-sensitive (open downward triangles) components of multiple pulse depression. Again, CGP-sensitive components are apparent early, while CCh-sensitive components of multiple pulse depression increase throughout the pulse train. Calculations for this panel were done as in Fig. 10F except that the CGP-sensitive component was better isolated by subtraction of the CCh-sensitive component (CGP sensitive = baseline CGP – CCh sensitive). This was not possible in Fig. 10 because CGP and CCh were tested in separate groups of cells.

Discussion

In this study we investigated the mechanism through which bath application of CCh inhibits evoked GABAergic transmission in the hilar region of the dentate gyrus. Our results suggested that the mechanism involved CCh-mediated increases in spontaneous GABA release and subsequent activation of presynaptic GABAergic autoreceptors. Intriguingly, further work revealed that a low dose of CCh (500 nm) produced increases in ambient GABA that primarily targeted presynaptic GABA_B receptors, while higher doses of CCh (3–10 μm) produced activity-dependent inhibition of evoked responses that is independent of GABA_B receptors and probably depends on activation of presynaptic or axonal GABA_A receptors. The potential significance of that finding is highlighted by the fact that we also observed a similar switch in the mechanism of feedback inhibition in unitary or near-unitary evoked responses during the development of multiple pulse depression. Overall, we are able to make several major conclusions based on the data presented here, each of which is discussed in greater detail below.

One basic but important conclusion of this work is that consideration of extrasynaptic GABA is essential to a complete understanding of the effects of muscarinic agonists on synaptic transmission. Clearly, inhibitory effects of CCh on evoked synaptic transmission have been frequently reported in the CNS. However, in many cases these effects are produced by activation of presynaptic muscarinic receptors probably coupled to inhibitory G-proteins such as G_i/o. In other cases, muscarinic agonists have been shown to produce indirect inhibition of synaptic transmission by facilitating cannabinoid-dependent retrograde inhibition (Kim et al. 2002; Fukudome et al. 2004). However, we found that neither of these mechanisms contributed substantially to CCh-mediated inhibition of evoked IPSCs recorded from hilar mossy cells. Instead, we found that inhibition of evoked IPSCs produced by 500 nm CCh depended on activation of GABA_B receptors. That finding alone, in a broad sense, reinforces the conclusions of a small group of prior studies that have linked bath-applied CCh to measurable changes in inhibitory tone (Vogt & Regehr, 2001; Kerr & Capogna, 2007; Nahir et al. 2007). However, intriguingly, we also noted that inhibition of evoked IPSCs produced by 3–10 μm CCh was both activity dependent and yet GABA_B receptor independent.

That observation is closely related to a second major conclusion of this study, which is that activation presynaptic or axonal GABA_A autoreceptors (as well as presynaptic GABA_B receptors) by extrasynaptic GABA is likely to be important to the regulation of inhibitory synaptic transmission in the dentate gyrus. Prior examples of presynaptic ionotropic GABA_A receptors mostly describe heteroreceptors on glutamatergic or glycinergic terminals or axons (Jang et al. 2002; Turecek & Trussell, 2002; Ruiz et al. 2003; Jang et al. 2005, 2006; Koga et al. 2005; Alle & Geiger, 2007). However, there have been a few prior studies that suggested the existence of presynaptic ionotropic GABA_A autoreceptors, most notably in hippocampus (Axmacher & Draguhn, 2004; Axmacher et al. 2004), hippocampal culture (Vautrin et al. 1994) and cerebellum (Pouzat & Marty, 1999; Mejia-Gervacio & Marty, 2006). Such studies are generally difficult due to the lack of pharmacological tools that reliably and selectively modulate presynaptic as opposed to postsynaptic GABA_A receptors. This problem is exacerbated by the potential for multiple specific subunit combinations to create functional high-affinity receptors. For reviews on these topics see Kullmann et al. (2005; Glykys & Mody (2007 and Draguhn et al. (2008). In the present study, the possible involvement of presynaptic or axonal GABA_A autoreceptors was suggested by our observation that low micromolar concentrations of CCh produce inhibition of evoked IPSCs that is clearly dependent on robust synaptic activity in GABAergic axons (e.g. Figs 3 and 7), and yet is also clearly independent of GABA_B receptors (Figs 2 and 5). The hypothesis was further reinforced by experiments in Fig. 4, which indicated that a concentration of CCh that reliably inhibits GABA_A-mediated evoked IPSCs is unable to inhibit GABA_B-mediated evoked IPSCs in the presence of PTX or bicuculline. It is also reinforced by evidence showing that there are no apparent postsynaptic effects of CCh on intrinsic properties of hilar mossy cells under our experimental conditions, and by additional analysis indicating that the kinetics of miniature IPSCs are not significantly altered by CCh. Together these results suggest that there is no direct effect on gating of postsynaptic GABA_A receptors. Finally, this hypothesis is additionally reinforced by our development of a novel model for activity-dependent interaction of presynaptic GABA_A and GABA_B autoreceptors (Fig. 6). This conceptual model was useful for generating numerous predictions about the physiology of the system, all of which held up well to experimental examination. One particularly novel aspect of our work with respect to presynaptic ionotropic GABA_A autoreceptors is the implication that they are likely to be activated only when there is robust activity in the inhibitory network. Nevertheless, we readily acknowledge that some of the key lines of evidence suggesting the existence of GABA_A autoreceptors required indirect measurements and/or relied on a process of elimination. Therefore, we firmly believe that it will be important for future studies to use a combination of direct physiological measurements from presynaptic axons and/or terminals, as well as subcellular immunohistochemical and anatomical techniques, to provide more definitive evidence that presynaptic or axonal ionotropic GABA_A receptors are present. However, in our view, at present, all conceivable alternative explanations for the effects of low micromolar concentrations of CCh on evoked IPSC amplitude and on the late phases of multiple pulse depression seem markedly less plausible. In this regard, it is worth noting that a very recent study has now used immunohistochemical techniques in combination with electron microscopy to provide clear anatomical evidence for the existence of axonal GABA_A receptors on GABAergic neurons in another system. Specifically, Trigo et al. (2010) have demonstrated expression of α1-containing axonal GABA_A receptors on molecular layer interneurons in rat cerebellum on postnatal day 12.

A third major conclusion is that, in the hilus, CCh-mediated facilitation of spontaneous IPSCs is mechanistically related, via changes in ambient GABA, to CCh-mediated inhibition of evoked IPSCs. This is an important conclusion, even independent of the particular subtype of GABA receptors involved. It is also a conclusion that has not yet been reached in other systems with apparently divergent effects of CCh on evoked and spontaneous synaptic transmission. In fact, in most other such cases, a mechanistic explanation for the opposing effects of CCh is either lacking or preliminary. For example, in the lateral spiriform nucleus of the chick, opposing effects of mAChR agonists on spontaneous and evoked synaptic transmission are pharmacologically distinguishable, but appear to both be dependent on either M1 or M3 receptors (Guo & Chiappinelli, 2001). In this case, the authors reasonably speculate that there may be an inhibitory effect of presynaptic M1 receptors. However, in most mammalian systems (including in the hippocampus), M1 receptors are typically expressed postsynaptically, and are generally coupled to the excitatory G_q/11 (Caulfield, 1993; Caulfield & Birdsall, 1998). Similarly, a recent study has revealed the same paradoxical effects of muscarinic agonists on spontaneous and evoked synaptic transmission in the entorhinal cortex; however, in this case both effects were blocked by 4-DAMP, a relatively selective M3 antagonist (Xiao et al. 2009). Although the authors noted a small muscarine-induced decrease in action potential amplitude in entorhinal interneurons, overall the mechanism through which an M3-coupled receptor could effectively inhibit evoked GABAergic transmission remained unclear. Finally, area CA1 of the hippocampus is probably the best-studied system where opposing effects of CCh on spontaneous and evoked inhibitory synaptic transmission are well documented (Pitler & Alger, 1992; Behrends & ten Bruggencate, 1993). Interestingly, in that system, there is strong evidence that the opposing effects of CCh are likely to be mechanistically distinct (Fukudome et al. 2004), and there is no direct evidence to date suggesting an essential involvement of high-affinity GABAergic receptors.

A fourth and final main conclusion of this study is already apparent above, but is worthy of particular consideration. In fact, in our view, one of the single most intriguing aspects of this study is the observation that the specific mechanism of feedback inhibition observed in inhibitory afferents to hilar mossy cells is in fact activity dependent. In summary, one mechanism is activated by bath application of the GAT-1 antagonist NO-711, is also activated by 500 nm CCh (which produces mild increases in spontaneous IPSC frequency and integrated area), contributes to paired pulse depression of unitary or near-unitary IPSCs, but does not have a significant impact on multiple pulse depression produced by longer trains of stimuli. This mechanism is also completely eliminated by CGP and as such is very likely to depend on activation of presynaptic GABA_B autoreceptors. By contrast, the other mechanism is not activated by blocking GABA uptake when spontaneous IPSC frequency is low, is not activated by 500 nm CCh (which produces mild increases in spontaneous IPSC frequency and integrated area), but is activated in a predictable manner by low micromolar concentrations of CCh that produce robust increases in spontaneous IPSC frequency, amplitude, area and integrated area. This mechanism has minimal impact on paired pulse depression of unitary of near-unitary IPSCs but clearly impacts development of multiple pulse depression during longer trains of stimuli. As stated above, we believe this mechanism is likely to depend critically on activation of presynaptic or axonal ionotropic GABA_A autoreceptors. However, regardless of mechanism, to our knowledge these findings represent the first demonstration of an activity-dependent switch between distinct forms of feedback inhibition in these or other central synapses. We believe that further exploration of the activity dependence of feedback inhibitory systems in the dentate and throughout the CNS may provide new insight into the regulation of cortical excitability in epilepsy and related disorders, and that it may also reveal potential for the development of novel therapeutic strategies that selectively target activity-dependent mechanisms.

Glossary

Abbreviations

CCh

carbachol

mAChR

muscarinic acetylcholine receptor

Author contributions

Both authors contributed to the experimental design, data analysis and presentation of this manuscript, and approved the final version. The vast majority of the experimental work was done by C.L.

Supplemental material

Supplemental Figure 1

Supplemental Figure 2

Supplemental Figure 3

Supplemental Figure 4

Supplemental Figure 5

Supplemental Figure 6

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors