Direction of action of presynaptic GABA_A receptors is highly dependent on the level of receptor activation

Abstract

Many synapses, including parallel fiber synapses in the cerebellum, express presynaptic GABA_A receptors. However, reports of the functional consequences of presynaptic GABA_A receptor activation are variable across synapses, from inhibition to enhancement of transmitter release. We find that presynaptic GABA_A receptor function is bidirectional at parallel fiber synapses depending on GABA concentration and modulation of GABA_A receptors in mice. Activation of GABA_A receptors by low GABA concentrations enhances glutamate release, whereas activation of receptors by higher GABA concentrations inhibits release. Furthermore, blocking GABA_B receptors reduces GABA_A receptor currents and shifts presynaptic responses toward greater enhancement of release across a wide range of GABA concentrations. Conversely, enhancing GABA_A receptor currents with ethanol or neurosteroids shifts responses toward greater inhibition of release. The ability of presynaptic GABA_A receptors to enhance or inhibit transmitter release at the same synapse depending on activity level provides a new mechanism for fine control of synaptic transmission by GABA and may explain conflicting reports of presynaptic GABA_A receptor function across synapses.

NEW & NOTEWORTHY GABA_A receptors are widely expressed at presynaptic terminals in the central nervous system. However, previous reports have produced conflicting results on the function of these receptors at different synapses. We show that presynaptic GABA_A receptor function is strongly dependent on the level of receptor activation. Low levels of receptor activation enhance transmitter release, whereas higher levels of activation inhibit release at the same synapses. This provides a novel mechanism by which presynaptic GABA_A receptors fine-tune synaptic transmission.

INTRODUCTION

Neurotransmitter receptors expressed in the presynaptic membrane have proven to be widespread and important modulators of synaptic transmission. Activation of presynaptic GABA_A receptors (GABA_ARs) is one of the earliest observed forms of synaptic modulation, beginning with primary afferent depolarization in the spinal cord (Andersen et al. 1964; Eccles et al. 1963). In recent years there has been a renewed interest in these receptors as they continue to be found at an expanding range of synapses throughout the central nervous system (Soiza-Reilly et al. 2013; Wang et al. 2019; Weisz et al. 2016; Zorrilla de San Martin et al. 2017). However, there continue to be widely divergent reports on the primary function of presynaptic GABA_ARs. While it is generally agreed that activation of these receptors depolarizes the presynaptic terminal (Price and Trussell 2006; Pugh and Jahr 2011; Ruiz et al. 2003; Zhang and Jackson 1995), studies are split on whether this depolarization reduces transmitter release (Eccles et al. 1963; Nicoll and Alger 1979; Wang et al. 2019; Zhang and Jackson 1995) or enhances transmitter release (Pouzat and Marty 1999; Pugh and Jahr 2011; Ruiz et al. 2010; Stell et al. 2007; Turecek and Trussell 2002; Zorrilla de San Martin et al. 2017). Currently, it is not clear why GABA_ARs inhibit transmitter release at some synapses and increase release at others. GABA_AR function is regulated by a wide diversity of modulators, including GABA_B receptors (Connelly et al. 2013; Tao et al. 2013), ethanol (Palmer and Hoffer 1990; Valenzuela and Jotty 2015), neurosteroids (Majewska et al. 1986), barbiturates (Nicoll et al. 1975), and benzodiazepines (Costa and Guidotti 1979), raising the possibility that modulation of presynaptic GABA_ARs may influence whether they enhance or inhibit synaptic transmission.

Parallel fiber synapses in the cerebellum are perhaps the most heavily studied site of presynaptic GABA_AR function in recent years (Antflick and Hampson 2012; Berglund et al. 2016; Dellal et al. 2012; Howell and Pugh 2016; Pugh and Jahr 2011, 2013; Santhakumar et al. 2013; Stell 2011; Stell et al. 2007). These studies have resulted in broad agreement that presynaptic GABA_ARs enhance glutamate release at this synapse. However, even at this synapse there have been indications of inhibitory actions of GABA_ARs, particularly during periods of very high or prolonged GABA application (Berglund et al. 2016; Dellal et al. 2012; Stell 2011). This suggests that presynaptic GABA_ARs may have differing effects on transmitter release depending on the properties of the synapse or degree of activation (Trigo et al. 2008).

To understand how GABA_ARs inhibit transmitter release under some conditions but enhance release in others, we systematically investigated presynaptic GABA_AR function at parallel fiber synapses in the cerebellum over a range of physiological conditions. We found that the direction of GABA_AR effects on transmission is strongly dependent on GABA concentration and modulation of GABA_ARs. Contrary to previous studies, we found that transmitter release was enhanced over only a very narrow range of relatively low GABA concentrations, whereas higher (but likely still physiological) GABA concentrations inhibited release. We then tested whether modulators of presynaptic GABA_ARs could shift the balance of enhancement vs. inhibition of transmitter release. We found that increasing GABA_AR activity through application of ethanol and the neurosteroid allopregnanolone shifts GABA_AR responses toward greater inhibition of transmitter release across almost all GABA concentrations tested. In addition, GABA_BR antagonists reduce GABA_AR activity, resulting in greater enhancement of transmitter release across a wide range of GABA concentrations.

METHODS

Animals.

All experimental procedures involving animals were approved by the Institutional Animal Care and Use Committee at the University of Texas Health Science Center San Antonio. Male and female C57BL/6 mice (Charles River, Wilmington, MA) 14–30 days old were used for all experiments. Animals were kept on a 12:12-h light-dark cycle with ad libitum access to food and water.

Slice preparation.

Male and female mice were deeply anesthetized with isoflurane, and the cerebellum was rapidly dissected and placed in ice-cold oxygenated artificial cerebrospinal fluid (ACSF) containing the following (in mM): 119 NaCl, 26.2 NaHCO₃, 2.5 KCl, 1.0 NaH₂PO₄, 11 glucose, 2 CaCl₂, and 1.3 MgCl₂. Parasagittal slices (300 µm) were cut from the vermis of the cerebellum using a vibratome (Leica Biosystems, Buffalo Grove, IL) and then incubated at 34°C for 30 min before being transferred to the microscope slice chamber for recording.

Electrophysiology experiments.

During recording, slices were superfused with ACSF at a flow rate of ~2 ml/min. Electrophysiological currents were recorded with a Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA), filtered at 5 kHz, and digitized at 50 kHz. Data were collected using pCLAMP software (Molecular Devices). For GABA uncaging experiments, 10 ml of ACSF containing 60 μM RuBi-GABA (Tocris, Bristol, UK) were recirculated using a fluid pump (Cole-Parmer, Vernon Hills, IL). This concentration of RuBi-GABA (60 μM) had no effect on the membrane resistance (P = 0.75, n = 4) or the amplitude of excitatory synaptic responses (P = 0.59, n = 4) in cerebellar stellate cells. GABA was uncaged by a brief (5 ms) illumination from a 473-nm laser light source (PSU-III-LED; Opto Engine, Midvale, UT) over a range of intensities (0.28–315 µW) at 30-s inter-sweep intervals to allow for clearance of GABA between sweeps. The uncaging laser intensity was measured using a photometer placed under the objective of the microscope. The theoretical maximum GABA concentration produced by photolytic uncaging is 60 μM (the concentration of RuBi-GABA in the bath solution). The uncaging laser intensity was adjusted over an ~1,000-fold range, suggesting the uncaged GABA concentrations varied roughly from tens or hundreds of nanomolar to tens of micromolar. Where indicated, ACSF contained one or more of the following: 3 µM CGP-55845 (CGP; Abcam, Cambridge, MA); 200 µM picrotoxin (PTX; Abcam); 1 μM tertiapin-Q (TPQ; Tocris); and 20 mM ethanol (Fisher, Hampton, NH).

For experiments on somatodendritic GABA currents, whole cell patch-clamp recordings were made from granule cells, with the following internal solution (in mM): 135 CsCl, 10 HEPES, 4 MgCl₂, 5 EGTA, 4 Na-ATP, 0.5 Na-GTP, and 2 QX-314. The pH was adjusted to 7.2–7.4 using CsOH, and the osmolarity was 280–300 mosmol/l.

For experiments on presynaptic GABA_A receptor function, putative stellate cells were identified in the outer third of the molecular layer and whole cell patch-clamp recordings were made with the following internal solution (in mM): 137 K-gluconate, 2 KCl, 4 MgCl₂, 10 HEPES, 5 EGTA, 4 Na-ATP, and 0.5 Na-GTP. The pH was adjusted to 7.2–7.4 using KOH and, the osmolarity was 280–300 mosmol/l. The postsynaptic membrane potential was held near the chloride reversal potential to minimize postsynaptic GABA currents following GABA uncaging. However, small, long-lasting GABA_AR-mediated currents were sometimes still observed immediately following the laser pulse as a slight change in the baseline holding current. Excitatory postsynaptic currents (EPSCs) were evoked by stimulation of parallel fibers with two pulses (100 μs, 10–60 V) 20 ms apart through a patch pipette filled with ACSF. GABA was uncaged by a laser pulse 50 ms before EPSCs were evoked, to activate presynaptic GABA_ARs as described previously (Howell and Pugh 2016). Uncaging and control sweeps were interleaved. Asynchronous events were identified by eye. Only synaptic events occurring more than 5 ms after the onset of the second stimulus artifact (in a pair of stimuli delivered at 50Hz) were counted as asynchronous events.

Fiber volleys from parallel fibers were measured using a standard patch-clamp electrode filled with ACSF inserted into the molecular layer of transverse cerebellar slices. Parallel fibers were stimulated by a second patch electrode filled with ACSF inserted in the molecular layer ~150–250 μm from the recording electrode. On alternating sweeps, GABA was uncaged 50 ms before parallel fiber stimulation by a 5-ms laser pulse centered on either the recording or stimulating electrode (results were similar at both uncaging locations). The stimulus intensity was kept relatively low to avoid saturating the fiber volley response. At the beginning of each experiment we confirmed that increasing the stimulus intensity increased the fiber volley amplitude.

Stimulus artifacts have been digitally removed in all figures.

Data analysis and statistical analysis.

Data were analyzed in IgorPro (Wavemetrics, Lake Oswego, OR) using the Neuromatic toolkit (Rothman and Silver 2018) and custom macros. Statistical significance was determined using two-tailed paired Student’s t-tests in Excel (Microsoft, Redmond, WA) and one- and two-way repeated-measures ANOVAs using SigmaStat 4.0 (Systat Software, San Jose, CA). Statistical values of P ≤ 0.05 were considered significant. Data are means ± SE.

RESULTS

To investigate presynaptic GABA_AR function, we made whole cell patch-clamp recordings from stellate cells in acute cerebellar slices and evoked EPSCs by electrical stimulation of parallel fibers. On alternating sweeps, RuBi-GABA was uncaged by a brief (5 ms) laser flash in the molecular layer 50 ms before parallel fiber stimulation to activate presynaptic GABA_ARs (Fig. 1A). We have shown previously that this protocol activates GABA_ARs with minimal GABA_BR activation at the time of synaptic stimulation (Howell and Pugh 2016). To measure GABA_AR effects on transmission across a range of activity levels, we systematically modulated the laser power used to uncage RuBi-GABA from 0.28 to 315 μW. We used a relatively low RuBi-GABA concentration (60 μM) in the bath solution to mimic the likely range of GABA concentrations resulting from spillover transmission (Barbour 2001; Barbour and Häusser 1997; Dzubay and Jahr 1999; Rusakov and Kullmann 1998). We first tested the effectiveness of modulating the laser power to control the uncaged GABA concentration by recording from cerebellar granule cells and focusing the laser on the soma. We found that the amplitude of the uncaging current varied directly with the power of the uncaging laser (Fig. 1B, inset). From these data we constructed a dose-response curve that is well fit by the Hill equation, indicating modulation of the laser intensity effectively alters the uncaged GABA concentration over the range of laser intensities used (0.28–850 μW).

Fig. 1.

Activation of presynaptic GABA_A receptors has bidirectional effects on synaptic currents. A: diagram of GABA uncaging (laser) and parallel fiber stimulation (stim) protocol. Representative synaptic responses in control (black) and following GABA uncaging (red) are shown. B: dose-response relationship of GABA_A receptor-mediated currents (I) at the granule cell body and GABA uncaging laser intensity. Inset: example traces of GABA uncaging-evoked currents at each laser intensity. C: representative average traces of excitatory postsynaptic current (EPSCs) from control sweeps (black) and GABA uncaging sweeps (red) for each uncaging laser power. D: change in evoked EPSC amplitudes recorded in stellate cells following GABA uncaging at each laser intensity compared with interleaved control sweeps without uncaging in control artificial cerebrospinal fluid (circles; n = 8–12) or in the presence of the GABA_A receptor antagonist picrotoxin (PTX; triangles; n = 7). All traces are the average of 10 sweeps. *P ≤ 0.05.

Contrary to previous studies that showed parallel fiber GABA_ARs increase transmission (Pugh and Jahr 2011; Stell et al. 2007), we found that GABA_ARs increase EPSC amplitudes over only a very narrow range of GABA concentrations. Specifically, we observed enhancement of EPSCs at a laser power of 14.2 μW (148.1 ± 16.4% of control, n = 11, P = 0.004; Fig. 1, C and D), and all other laser powers tested either produced no change (power < 14.2 μW, n = 9–12) or decreased EPSC amplitudes (power > 14.2 μW, n = 8–11; for laser intensity of 315 μW: 56.1 ± 9.6% of control, n = 9, P = 0.005; Fig. 1, C and D). As expected from previous studies (Pugh and Jahr 2011; Stell et al. 2007), the enhancement of EPSCs at 14.2 μW was blocked by bath application of PTX (14.2 μW: 91.7 ± 3.3% of control, P = 0.05, n = 7). However, the inhibition of EPSCs observed at higher laser intensities was also blocked by PTX (315 μW: 100.2 ± 7.8% of control, P = 0.63, n = 7 for all laser intensities; Fig. 1D), indicating that this inhibition is also mediated by GABA_ARs and not by GABA_BRs or activation of G protein-coupled inwardly rectifying potassium (GIRK) channels. Interestingly, the uncaging laser power that produced an enhancement of EPSCs (14.2 μW) produced a current in granule cells only approximately one-third the maximal current (Fig. 1B), suggesting presynaptic GABA_ARs may only enhance transmitter release when partially activated. Collectively, these data suggest that presynaptic GABA_ARs may frequently inhibit release depending on the GABA concentration reaching the receptors.

Although it has been shown previously that presynaptic GABA_ARs increase EPSC amplitudes by increasing vesicle release probability (Pugh and Jahr 2011; Stell et al. 2007), it is not clear whether the inhibition of EPSCs seen at higher uncaging laser intensities is due to pre- or postsynaptic mechanisms. The increase of EPSC amplitudes following GABA uncaging at 14.2 μW power was associated with a decrease in the paired-pulse ratio (3.2 ± 0.4 vs. 2.1 ± 0.2, n = 10, P = 0.02; Fig. 2, A and B) and failure rate (0.16 ± 0.07 vs. 0.09 ± 0.04, n = 9, P = 0.07), consistent with an increase in vesicle release probability. At higher uncaging laser intensities (34–167μW), the paired-pulse ratio was significantly increased (P = 0.03; Fig. 2, A and B) following GABA uncaging, and there was a trend toward increased failure rate (P = 0.17), suggesting a decrease in vesicle release. However, this does not rule out additional inhibition of EPSCs due to postsynaptic mechanisms. To test this we measured the amplitude of asynchronous EPSCs that result from fusion of single vesicles following the synchronous EPSCs (Atluri and Regehr 1998; Diamond and Jahr 1995). We found no change in asynchronous EPSC amplitudes between control and GABA uncaging sweeps (33–167 μW: 58.9 ± 3.4 vs. 53.7 ± 3.3 pA, n = 39, P = 0.12), even though the synchronous EPSC amplitude was reduced (Figs. 1D and 2C). The lack of change in asynchronous EPSCs suggests the reduction in the synchronous EPSC is not due to postsynaptic shunting through GABA_ARs or other postsynaptic mechanisms. Together, these data indicate that both the enhancement of EPSCs at low uncaging powers and the inhibition of EPSCs at higher uncaging powers are due to changes in the presynaptic vesicle release.

Fig. 2.

Mechanism of presynaptic GABA_A receptor inhibition of synaptic currents. A: representative traces of excitatory postsynaptic currents (EPSCs) from control sweeps (black) or GABA uncaging sweeps (red) using an uncaging laser power of 3.7 (top), 14 (middle), or 67 μW (bottom). GABA uncaging EPSCs have been normalized to the peak of the first control EPSC to show the change in paired-pulse ratio (PPR). B: average change in PPR (PPR GABA/PPR control) following GABA uncaging at low (3.7 μW), mid (14 μW), and high (34–167 μW) laser intensities. C: representative traces (10 overlaid sweeps) showing asynchronous synaptic events in control (black) and following GABA uncaging (red). D: average number (left) and amplitude (right) of asynchronous events during control or GABA uncaging sweeps for uncaging laser powers from 67 to 315 μW. E: representative traces of fiber volleys from control sweeps (black) or GABA uncaging sweeps (red). F: change in evoked EPSC amplitudes following GABA uncaging at each laser intensity in the presence of the G protein-coupled inwardly rectifying potassium (GIRK) channel blocker tertiapin Q (n = 6–7). The data from standard artificial cerebrospinal fluid are replotted from Fig. 1 (gray) for comparison. All traces are the average of 10 sweeps. *P ≤ 0.05; **P ≤ 0.01. asynch, Asynchronous; cnt, control.

Previous studies have found that presynaptic GABA_ARs at parallel fiber synapses enhance transmitter release by depolarizing the presynaptic terminal and activating voltage-gated calcium channels (Pugh and Jahr 2011; Stell et al. 2007). What is the mechanism producing inhibition of EPSCs at higher uncaging laser intensities? Is it also dependent on presynaptic depolarization? To address this question, we analyzed the frequency of asynchronous synaptic events in control and GABA uncaging sweeps. The frequency of asynchronous synaptic events is directly related to subthreshold depolarization (Christie et al. 2011) and bulk calcium (Atluri and Regehr 1998; Sun et al. 2007) in the presynaptic bouton. We found that the frequency of asynchronous events was increased following GABA uncaging compared with control sweeps at high uncaging laser powers (67–315 μW; P = 0.002, n = 28; Fig. 2, C and D), consistent with depolarization of parallel fiber boutons and elevated bulk calcium following GABA uncaging. This raises the question of how the synchronous EPSC amplitude is reduced but the number of asynchronous events is increased. A simple change in vesicular release probability generally modulates synchronous and asynchronous release together (Atluri and Regehr 1998; Christie et al. 2011; Howell and Pugh 2016). One possibility is that the synchronous EPSC is reduced by a decrease in parallel fiber excitability (possibly due to sodium channel inactivation) such that fewer axons are brought to threshold with each stimulus. To test this, we measured the fiber volley amplitude following stimulation of parallel fibers in the molecular layer during control and GABA uncaging sweeps. We observed a slight increase in fiber volley amplitude following GABA uncaging at high laser power (167 μW: 106.8 ± 3.2% of control, n = 7, P = 0.005; Fig. 2E), consistent with earlier results (Dellal et al. 2012). This suggests more, not fewer, parallel fibers are activated with each stimulus in the presence of GABA, ruling out reduced parallel fiber excitability as an explanation for the reduced synchronous EPSC amplitude. Instead, our data are more consistent with an excessive depolarization of the presynaptic bouton and inactivation of voltage-gated calcium (Catterall et al. 2013; Forsythe et al. 1998; Xu and Wu 2005) or sodium channels (Leão et al. 2005; Ohura and Kamiya 2018; Zhang and Jackson 1995). The observation that fiber volley amplitude is not reduced following GABA uncaging suggests that action potential initiation and propagation are normal; however, it remains possible that sodium channel inactivation reduces action potential amplitude (Jang et al. 2005; Trigo et al. 2008; Zhang and Jackson 1995). Inactivation of either calcium or sodium channels due to subthreshold depolarization could result in increased global calcium but reduced action potential-evoked calcium influx, a mechanism that would explain the decrease in the synchronous EPSC but increase in the frequency of asynchronous EPSCs.

We then tested if excessive depolarization of parallel fibers results in inhibition of release by artificially depolarizing the resting membrane potential in parallel fibers. We choose to do this by blocking GIRK channels with TPQ, which, unlike block of voltage-gated potassium channels, is expected to depolarize the axons with little direct effect on the action potential waveform. Previous work has shown that activation of GABA_B or adenosine receptors reduces the excitability of parallel fibers (Dittman and Regehr 1996), suggesting GIRK channels are expressed in parallel fibers and can modulate their excitability. In the presence of TPQ, the frequency of spontaneous synaptic events was increased (control: 0.21 ± 0.04 Hz, n = 12; TPQ: 0.52 ± 0.19 Hz, n = 7, P = 0.05), consistent with a depolarization of parallel fibers by TPQ. With GIRK channels blocked we observed greater inhibition of EPSCs (34–167 μW: 49.5 ± 8.5% of control, P < 0.001) following GABA uncaging and did not observe enhancement of EPSCs at any laser power (Fig. 2F; data from Fig. 1D plotted for comparison). The increased inhibition observed in TPQ is consistent with excessive depolarization reducing transmitter release, possibly through inactivation of calcium channels.

These data suggest that presynaptic GABA_ARs may inhibit or enhance transmitter release depending on the magnitude of the GABA_AR-mediated depolarization of the presynaptic bouton. This raises the possibility that positive or negative modulators of GABA_ARs can also modulate GABA_AR effects on transmitter release. Previous studies have shown that GABA_BR activity enhances currents through extrasynaptic (likely δ-subunit containing) GABA_ARs in cerebellar granule cells and other cell types that exhibit a tonic GABA_AR current (Connelly et al. 2013; Tao et al. 2013); however, the mechanism of this modulation is not currently known. Although the exact subunit expression of GABA_ARs expressed at parallel fibers is not known, they may also include δ-subunits and can be broadly classified as extrasynaptic receptors (Berglund et al. 2016; Dellal et al. 2012; Santhakumar et al. 2013). Furthermore, our previous studies found that blocking GABA_BRs reduces presynaptic GABA_AR-mediated increase in transmitter release (Howell and Pugh 2016), suggesting parallel fiber GABA_ARs may also be enhanced by GABA_BR activity. Consistent with these earlier studies, in the presence of the GABA_BR antagonist CGP, we did not observe enhancement of EPSCs at an uncaging laser intensity of 14.2 μW as we did in control conditions (Fig. 3, A and C). Rather, the uncaging dose-response curve was shifted to the right, resulting in enhancement of EPSC amplitudes over a wide range of higher uncaging laser intensities (67–167 μW: 141.2 ± 12.4% of control, n = 37, P = 0.004; Fig. 3A; data from Fig. 1D plotted for comparison). Bath application of CGP alone did not alter the amplitude of control EPSC amplitudes (100.4 ± 15.8 vs. 114.7 ± 20.5 pA, n = 15, P = 0.2) or paired-pulse ratio (2.53 ± 0.28 vs. 2.29 ± 0.19, n = 15, P = 0.32). The range of laser intensities producing enhancement of EPSCs is likely broadened in CGP by saturation of GABA_ARs or saturation of RuBi-GABA uncaging (Fig. 1B). These data suggest that blocking GABA_BRs reduces the presynaptic GABA_AR current and shifts the response to GABA uncaging toward enhancement of EPSCs.

Fig. 3.

Modulation of presynaptic GABA_A receptor function. A: change in evoked excitatory postsynaptic current (EPSC) amplitudes recorded in stellate cells following GABA uncaging at each laser intensity in the presence of CGP-55845 (+CPG; black; n = 7–11). B: change in evoked EPSC amplitudes recorded in stellate cells following GABA uncaging at each laser intensity in the presence of ethanol (+EtOH; blue; n = 6) or allopregnanolone (+Allo; green; n = 4–6). The data from standard artificial cerebrospinal fluid are replotted from Fig. 1 (gray) in A and B for comparison. C: representative EPSCs recorded in control sweeps (black) or following GABA uncaging with laser power of 14.25 μW (red) for each pharmacological condition. All traces are the average of 10 sweeps.

Previous studies have shown that extrasynaptic GABA_ARs are particularly sensitive to enhancement by ethanol (Diaz and Valenzuela 2016; Hanchar et al. 2005; Santhakumar et al. 2007; but see Borghese and Harris 2007) or neurosteroids (Abramian et al. 2014; Stell et al. 2003). We therefore used ethanol and allopregnanolone to test how increasing presynaptic GABA_AR activity alters effects on synaptic transmission. In the presence of 20 mM ethanol, GABA uncaging inhibited EPSCs at all uncaging laser powers except the lowest power tested (3.7–315 μW: n = 6, P = 0.009–0.057), shifting the dose-response curve to the left (Fig. 3B; data from Fig. 1D plotted for comparison). Ethanol alone did not increase fiber volley amplitudes (n = 6, P = 0.53), indicating the shift toward greater inhibition is not due to depolarization of parallel fibers by ethanol itself. Similarly, application of allopregnanolone resulted in significant inhibition of EPSC amplitudes following GABA uncaging across almost all laser powers tested (14.2–315 μW: n = 6, P = 0.0002–0.05; 167 μW: P = 0.36; Fig. 3B). We did not observe enhancement of EPSC amplitudes at any laser power in the presence of ethanol or allopregnanolone. This is likely due to the narrow range of GABA concentrations that produce enhancement and the relatively large (~10-fold) gaps in laser intensities tested at the low range (0.28 vs. 3.7 vs. 14.25 μW). These results demonstrate that the effects of GABA_ARs on transmitter release are highly sensitive to GABA_AR modulators, shifting responses to primarily enhance or inhibit release across a range of GABA concentrations.

DISCUSSION

In this study we show that presynaptic GABA_ARs can enhance or inhibit transmitter release depending on the GABA concentration. In vivo, these receptors are likely exposed to GABA concentrations ranging from hundreds of nanomolar to tens of micromolar from either ambient GABA (Berglund et al. 2016; Bright et al. 2011; Santhakumar et al. 2006) or spillover transmission from neighboring inhibitory synapses (Barbour 2001; Barbour and Häusser 1997; Dzubay and Jahr 1999; Rusakov and Kullmann 1998). The GABA concentrations tested in this study likely span a similar range (see methods), suggesting both enhancement and inhibition of transmitter release could occur over physiological GABA concentrations.

Mechanisms of GABA-mediated enhancement and inhibition of vesicle release.

Previous studies have shown that activation of presynaptic GABA_ARs in parallel fibers depolarizes the bouton (Dellal et al. 2012; Stell 2011; Stell et al. 2007), activating voltage-gated calcium channels (Pugh and Jahr 2011) and enhancing vesicle release. In the present study we find that higher concentrations of GABA can also inhibit vesicle release. We argue that the switch from enhancement to inhibition is dependent on the amplitude of the GABA_AR-mediated depolarization and inactivation of voltage-gated calcium and/or sodium channels (see Fig. 4) based on the following evidence: 1) The uncaging laser power producing enhancement of EPSCs in control solutions (14.25 μW) results in a GABA current only approximately one-third of the maximal current in granule cell body, suggesting higher laser powers can produce substantially larger GABA_AR currents and depolarization in parallel fiber boutons. 2) The number of asynchronous synaptic events was significantly increased at high uncaging laser powers (Fig. 2, C and D), but not at lower powers (14.25 μW; 1.03 vs. 1.22 events/sweep, P = 0.12, n = 12), suggesting greater depolarization and bulk calcium in the boutons at uncaging laser intensities that produce inhibition of release. 3) Tonic depolarization of parallel fibers by blockade of GIRK channels results in greater inhibition of vesicle release (Fig. 2F). One possible caveat to this mechanism is that it requires depolarization and/or elevated calcium in the bouton for at least 50 ms (the interval between GABA uncaging and synaptic stimulation). In our experimental system this is not an issue because the GABA concentration likely remains elevated in the slice for at least 50 ms, maintaining GABA_AR activation and presynaptic depolarization. This is evident in the relatively slow-decaying GABA_AR-mediated currents in Fig. 1B. In a previous study we found that enhancement of EPSCs by GABA was the same with the use of a 20- or 50-ms interval between uncaging and synaptic stimulation (Howell and Pugh 2016), suggesting a prolonged depolarization is not necessary.

Fig. 4.

Model of presynaptic GABA_A receptor (GABA_AR) action at parallel fiber synapses. Proposed action of presynaptic GABA_ARs when exposed to low (left) or high (right) GABA concentrations is shown. GABA_ARs depolarize the presynaptic terminal, activating voltage-gated calcium channels, which allow calcium influx (Ca²⁺) and potentiation of action potential (AP)-evoked transmitter release. When the GABA_AR-mediated current is increased (through a higher GABA concentration or exposure to positive modulators), the bouton experiences greater depolarization and calcium influx, resulting in reduced AP-evoked transmitter release, likely due to inactivation of calcium or sodium channels. Blocking GABA_BRs with CGP-55845 (+CGP) decreases the current through GABA_ARs, resulting in little or no GABA_AR effect at low GABA and enhancement of AP-evoked transmitter release at high GABA. +EtOH/Allo, presence of ethanol or allopregnanolone; ΔV_m, change in membrane potential.

Modulation by GABA_B receptors.

We find that blocking GABA_B receptors shifts the effects of presynaptic GABA_A receptors such that we observed enhancement of release across a wide range of higher GABA concentrations. Several mechanisms could explain this shift in GABA_A receptor behavior. Previous studies have shown that GABA_B receptor activity enhances currents through extrasynaptic (likely δ-subunit containing) GABA_A receptors in cerebellar granule cells and other neurons that exhibit a tonic GABA_A receptor current (Connelly et al. 2013; Tao et al. 2013). The mechanism of this enhancement is currently not understood but likely involves activation of PKA and phosphorylation of extrasynaptic GABA_A receptors (Connelly et al. 2013). If this is also true at parallel fiber boutons, then inhibition of GABA_B receptors would reduce the GABA_A receptor-mediated currents at each GABA concentration, resulting in smaller depolarizations of the bouton and favoring enhancement of release over inhibition. Another possibility is that blocking GABA_BRs shifts the chloride reversal potential (E_Cl). Wright et al. (2017) recently showed that activation of GABA_BRs in hippocampal neurons reduces the expression of the potassium-chloride cotransporter KCC2, which is largely responsible for extruding chloride from the cell. This raises the possibility that blockade of GABA_BRs at parallel fiber boutons increases KCC2 expression (and chloride extrusion from the bouton), resulting in a hyperpolarizing shift in E_Cl. This mechanism would also limit depolarization of the presynaptic bouton by GABA_AR activation and, hence, favor enhancement of release over inhibition. However, this would not explain the results using allopregnanolone or ethanol, because these compounds have been shown to have little or no effect on E_Cl (Rodgers-Neame et al. 1992; Shimura et al. 1996; Soldo et al. 1994; Ye et al. 2001). Either of these potential mechanisms would be consistent with our hypothesis that small GABA_A receptor-mediated depolarizations of the presynaptic membrane result in enhancement of release, whereas larger depolarizations result in inhibition. Alternatively, it is well established that GABA_BRs activate GIRK channels and inhibit voltage-gated calcium channels (Mintz and Bean 1993; Misgeld et al. 1995; Slesinger et al. 1997), either of which could possibly produce the shift in GABA enhancement/inhibition of EPSCs when GABA_BRs are blocked. We do not favor either of these potential mechanisms, because blocking GABA_BRs should reduce GIRK channel activity, depolarizing the boutons and shifting responses to GABA toward greater inhibition, as shown in Fig. 2F, not greater enhancement of EPSCs as seen in the presence of CGP (Fig. 3A). Likewise, blocking GABA_BRs should increase the activity of voltage-gated calcium channels, which is expected to shift enhancement of EPSCs toward lower GABA concentrations, not higher concentrations as seen in Fig. 3A. Regardless of the precise mechanism involved, these results demonstrate that activation of GABA_BRs can have a powerful influence on how GABA_ARs effect synaptic transmission.

Synapse specificity of GABA_A receptor modulation.

These data suggest that presynaptic GABA_AR function may vary widely across synapses depending on the GABA concentration and modulation of receptors. For example, synapses which coexpress GABA_A and GABA_BRs may be more prone to GABA_AR-mediated inhibition of release compared with synapses with GABA_ARs alone. In addition, modulators of GABA_AR function, such as neurosteroids, may be present at some synapses and not others, or change over time. For example, changes in progesterone-derived neuroactive steroids during the menstrual cycle and pregnancy are linked to changes in tonic GABA_AR currents (Maguire et al. 2005; Stell et al. 2003). However, our results suggest that changes in these steroids may also shift or reverse the effects of presynaptic GABA_AR on vesicle release. These factors suggest that presynaptic GABA_AR function is very synapse and time specific, explaining the conflicting results of presynaptic GABA_AR function across synapses. Furthermore, modulation of presynaptic GABA_ARs by ethanol may play a role in the behavioral effects of alcohol consumption (Santhakumar et al. 2013), and modulation by GABA_BRs could contribute to the effects of GABA_BR agonists used clinically.

At parallel fiber synapses, previous reports have suggested that presynaptic GABA_ARs enhance transmitter release (Dellal et al. 2012; Pugh and Jahr 2011; Stell et al. 2007). However, in the present work we find that GABA_ARs inhibit release over the majority of GABA concentrations tested, with only a very narrow range of GABA concentrations enhancing release (Fig. 1D). This discrepancy is likely due to the use of GABA_BR antagonists or GABA_AR-specific agonists in previous studies to isolate GABA_AR effects. In the absence of GABA_BR activity, GABA_ARs are shifted toward much greater enhancement of release. Our results suggest that under physiological conditions (GABA_BRs unblocked), the functional consequence of presynaptic GABA_AR function may be inhibition of release more often than enhancement of release. This result highlights the need for careful attention to experimental conditions when studying the function of presynaptic GABA_ARs. It will be interesting in the future to examine whether other synapses with presynaptic GABA_ARs that are thought to enhance transmitter release, such as mossy fiber boutons (Ruiz et al. 2010), Shaffer collaterals (Jang et al. 2006), and calyx of Held (Turecek and Trussell 2002), also show inhibition of release when GABA_BRs are not blocked.

Physiological consequences.

Many theories of cerebellar function hypothesize that pauses in Purkinje cell firing are a critical signal for downstream plasticity (Pugh and Raman 2006) and behavioral output (Jirenhed and Hesslow 2016; Medina et al. 2000). Pauses in Purkinje cell activity are produced by both reduced excitatory parallel fiber input (Ito 1993; Steuber et al. 2007) and increased inhibition from molecular layer interneurons (Barmack and Yakhnitsa 2008; Mathews et al. 2012). Presynaptic GABA_ARs are well positioned to contribute to pauses in Purkinje cell firing. During periods of low interneuron activity, spillover of GABA from single synapses or ambient GABA (Berglund et al. 2016) may enhance glutamate release from parallel fibers, but during periods of elevated inhibition, spillover and/or pooling of GABA from multiple inhibitory synapses may instead inhibit glutamate release through increased activation of GABA_ARs. The net effect of this process is that periods of high inhibition are complimented with decreased glutamate release from parallel fibers, resulting in a more robust pause in Purkinje cell firing.

Parallel fiber synapses also express presynaptic GABA_BRs, which inhibit transmitter release by reducing the activity of voltage-gated calcium channels (Dittman and Regehr 1996 1997). We have shown that presynaptic GABA_BR expression is downregulated at parallel fiber-stellate cell synapses following 4-Hz parallel fiber stimulation (Orts-Del’Immagine and Pugh 2018). This change not only removes GABA_BR-mediated inhibition of transmitter release but also removes enhancement of GABA_AR currents, shifting presynaptic GABA_AR function toward increasing transmitter release across a wide range of GABA concentrations (Fig. 3B). The result of these complimentary actions will be significantly greater synaptic transmission at parallel fibers exposed to 4-Hz activity. This is a possible mechanism by which information flow through different parallel fiber pathways and susceptibility to long-term synaptic plasticity could be regulated.

GRANTS

This work was funded by National Institutes of Health Grant R01 NS092809.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

J.R.P. conceived and designed research; S.N.K., W.-C.W., and Y.Y. performed experiments; S.N.K., W.-C.W., Y.Y., and J.R.P. analyzed data; S.N.K. and J.R.P. interpreted results of experiments; J.R.P. prepared figures; J.R.P. drafted manuscript; S.N.K., W.-C.W., and J.R.P. edited and revised manuscript; S.N.K., W.-C.W., Y.Y., and J.R.P. approved final version of manuscript.