Tonic GABA_A Conductance Favors Spike-Timing-Dependent over Theta-Burst-Induced Long-Term Potentiation in the Hippocampus

Abstract

Synaptic plasticity is triggered by different patterns of network activity. Here, we investigated how LTP in CA3-CA1 synapses induced by different stimulation patterns is affected by tonic GABA_A conductances in rat hippocampal slices. Spike-timing-dependent LTP was induced by pairing Schaffer collateral stimulation with antidromic stimulation of CA1 pyramidal neurons. Theta-burst-induced LTP was induced by theta-burst stimulation of Schaffer collaterals. We mimicked increased tonic GABA_A conductance by bath application of 30 μm GABA. Surprisingly, tonic GABA_A conductance selectively suppressed theta-burst-induced LTP but not spike-timing-dependent LTP. We combined whole-cell patch-clamp electrophysiology, two-photon Ca²⁺ imaging, glutamate uncaging, and mathematical modeling to dissect the mechanisms underlying these differential effects of tonic GABA_A conductance. We found that Ca²⁺ transients during pairing of an action potential with an EPSP were less sensitive to tonic GABA_A conductance-induced shunting inhibition than Ca²⁺ transients induced by EPSP burst. Our results may explain how different forms of memory are affected by increasing tonic GABA_A conductances under physiological or pathologic conditions, as well as under the influence of substances that target extrasynaptic GABA_A receptors (e.g., neurosteroids, sedatives, antiepileptic drugs, and alcohol).

SIGNIFICANCE STATEMENT Brain activity is associated with neuronal firing and synaptic signaling among neurons. Synaptic plasticity represents a mechanism for learning and memory. However, some neurotransmitters that escape the synaptic cleft or are released by astrocytes can target extrasynaptic receptors. Extrasynaptic GABA_A receptors mediate tonic conductances that reduce the excitability of neurons by shunting. This results in the decreased ability for neurons to fire action potentials, but when action potentials are successfully triggered, tonic conductances are unable to reduce them significantly. As such, tonic GABA_A conductances have minimal effects on spike-timing-dependent synaptic plasticity while strongly attenuating the plasticity evoked by EPSP bursts. Our findings shed light on how changes in tonic conductances can selectively affect different forms of learning and memory.

Introduction

Tonic conductances mediated by GABA_A receptors have received significant attention over the last two decades (Brickley et al., 1996; Semyanov et al., 2004; Farrant and Nusser, 2005; Glykys and Mody, 2007a; Brickley and Mody, 2012). Often referred to as tonic current or tonic inhibition, this is thought to be a mechanism that decreases the excitability of specific cell populations in the brain (Semyanov, 2003; Semyanov et al., 2003; Ade et al., 2008; Vardya et al., 2008; Urban-Ciecko et al., 2010). The tonic GABA_A conductance is set by concentrations of ambient GABA, the expression of high-affinity GABA_A receptors, and the presence of endogenous and exogenous modulators of these receptors (Semyanov et al., 2003; Stell et al., 2003; Scimemi et al., 2005; Glykys and Mody, 2007b; Song et al., 2013). Endogenous modulators include neurosteroids, the concentrations of which change in the brain during puberty, pregnancy, and the ovarian cycle (Herd et al., 2007; Maguire and Mody, 2007, 2008). The corresponding changes in the tonic GABA_A conductance can promote stress, anxiety, and depression (Maguire and Mody, 2007, 2008; Shen et al., 2007; MacKenzie and Maguire, 2013). A socially relevant exogenous factor that augments the tonic GABA_A conductance is ethanol (Wei et al., 2004; Hanchar et al., 2005). Hence, acute ethanol intake impairs synaptic plasticity, learning, and memory (McCool, 2011). In contrast, chronic ethanol abuse downregulates the GABAergic system (Davies, 2003).

Indeed, the tonic GABA_A conductance is involved in far more complex neuronal computations than would be expected for a simple inhibitory action. Tonic conductances modulate neuronal gain during synaptic excitation in small-size cells (e.g., cerebellar granule cells) and neuronal offset in larger cells (e.g., hippocampal pyramidal neurons) (Mitchell and Silver, 2003; Semyanov et al., 2004; Pavlov et al., 2009). The tonic GABA_A-conductance mediated decrease in cellular input resistance reduces both the membrane time constant and the membrane length constant (Rall, 1969; Jack and Redman, 1971). These constants influence the shape of EPSPs, their integration, and EPSP-spike coupling in hippocampal interneurons and pyramidal cells (Wlodarczyk et al., 2013; Pavlov et al., 2014). In addition, the tonic GABA_A conductance modulates synaptic plasticity and brain rhythms (Mann and Mody, 2010; Shen et al., 2010; Pavlov et al., 2014).

Moreover, the tonic activation of GABA_A receptors is not always inhibitory. Immature neurons have depolarizing reversal potentials for GABA (E_GABA), which become hyperpolarizing in the adult brain (Ben-Ari, 2002). Nevertheless, several types of mature neurons maintain depolarizing E_GABA: for example, striatal projection neurons (Bracci and Panzeri, 2006), hippocampal granule cells (Chiang et al., 2012), suprachiasmatic nucleus neurons (Choi et al., 2008), vasopressin-secreting hypothalamic neurons (Haam et al., 2012), and interneurons of the hippocampus and amygdala (Banke and McBain, 2006; Woodruff et al., 2006). Low levels of tonic GABA_A conductance excite hippocampal interneurons by exerting small depolarizations that recruit voltage-dependent membrane conductances (Song et al., 2011). A high level of tonic GABA_A conductance inhibits these neurons by a shunting effect. The tonic activation of presynaptic GABA_A receptors depolarizes hippocampal mossy fiber boutons and increases synaptic release probability (Ruiz et al., 2010).

Thus, tonic GABA_A conductances influence cell excitability and the integration of synaptic inputs depending on cell size, E_GABA, and the magnitude of this conductance. Acute increases in tonic conductance inhibit the LTP induced by high-frequency stimulation in the hippocampus (Whissell et al., 2013). However, until now, it has remained unclear whether tonic GABA_A conductances differentially affect synaptic plasticity which triggered by different neuronal network dynamics. To address this issue, we compared the effect of tonic GABA_A conductances on spike-timing-dependent LTP (stLTP) induced by pairing synaptic input activation and postsynaptic cell spiking and theta-burst-induced LTP (tbLTP) induced by theta-bursts of presynaptic cell firing.

Materials and Methods

Hippocampal slice preparation

Transverse hippocampal slices were prepared from 3- to 5-week-old Sprague Dawley male rats in accordance with RIKEN regulations. Animals were anesthetized with 2-bromo-2-chloro-1,1,1-trifluoroethane (halothane) and decapitated. The brain was exposed and cooled with an ice-cold solution containing the following (in mM): 75 sucrose, 87 NaCl, 2.5 KCl, 0.5 CaCl₂, 1.25 NaH₂PO₄, 7 MgCl₂, 25 NaHCO₃, 1 Na-ascorbate, and 25 D-glucose. Hippocampi from both hemispheres were isolated and placed in an agar block. Transverse slices (350-400 µm) were prepared with a vibrating microtome (Microm HM 650V, Thermo Fisher Scientific) and left to recover for 20 min at 34°C and then for 40 min in an interface chamber with storage solution containing the following (in mM): 127 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 2 MgCl₂, 1 CaCl₂, 25 NaHCO₃, and 11 D-glucose. The slices were then transferred to the recording chamber and continuously perfused with a recording solution containing the following (in mM): 127 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 1 MgCl₂, 2 CaCl₂, 25 NaHCO₃, and 11 D-glucose at 34°C. All solutions were saturated with 95% O₂ and 5% CO₂. Osmolarity was adjusted to 295 ± 5 mOsm; 5 μm 3-[[(3,4-dichlorophenyl)methyl]amino]propyl] diethoxymethyl) phosphonic acid (CGP52432) and 400 μm (S)-α- methyl-4-carboxyphenylglycine (S-MCPG) were routinely added to the solution to block GABA_B and metabotropic glutamate receptors, respectively. Cells were visually identified under infrared DIC using an Olympus BX-61 microscope (Olympus).

Electrophysiology

A glass electrode, with a resistance of 3-5 MΩ, filled with the extracellular solution, was placed in stratum radiatum for field potential recordings (see Fig. 1a). Synaptic responses were evoked by stimulation via two bipolar stainless-steel electrodes (FHC). Theta-burst stimulation of Schaffer collaterals (SCs) was performed with the electrode placed in the stratum radiatum >200 μm from the recording site to induce tbLTP. Theta-burst stimulation consisted of 10 bursts with a 200 ms interburst interval; each burst consisted of 4 pulses with a duration of 0.2 ms at a frequency of 100 Hz. stLTP was induced with one electrode placed in the stratum radiatum to trigger glutamate release from SC, and another electrode placed in the stratum oriens to trigger antidromic (AD) action potentials (APs) in postsynaptic CA1 pyramidal neurons (AD stimulation). SC stimulation was followed by a 10 ms pause before AD stimulation, and this paired stimulation was repeated every 6 seconds for 10 min. Stimulus strength was adjusted to 30%-50% of the maximal amplitude of fEPSPs.

Figure 1.

Tonic GABA_A conductances have no effect on stLTP but suppress tbLTP. a, Left, A schematic of a hippocampal slice showing the location of stimulating and recording electrodes. The recording electrode was placed in CA1 stratum radiatum (Rec), the first stimulating electrode on Shaffer collaterals (SC stim.), the second electrode in stratum oriens for AD stimulation (AD stim.). Right, AD stimulation induced an antidromic population spike, whereas SC stimulation induced fEPSP. stLTP was induced by the pairing of CS and AD stimulations; tbLTP was induced by bursts of SC stimulation. b, Average fEPSPs 30-40 min after stLTP induction in control (gray), in the presence of GABA (green), and in the presence of APV (orange). Black represents fEPSPs before stLTP induction. c, The time course of averaged and normalized fEPSP slope in control (gray), in the presence of GABA (green), and in the presence of APV (orange). The zero time point is the beginning of the stLTP induction protocol (gray bar). d, The summary of data showing the average magnitude of stLTP 30-40 min after induction. Gray represents control (Ctrl). Green represents GABA. Orange represents APV. e, Average fEPSPs 30-40 min after tbLTP induction in control (gray), in the presence of GABA (green), and in the presence of APV (orange). Black represents fEPSPs before tbLTP induction. f, The time course of averaged and normalized fEPSP slope in control (gray), in the presence of GABA (green), and in the presence of APV (orange). The zero time point is the beginning of the tbLTP induction protocol (gray bar). g, The summary of data showing the average magnitude of tbLTP 30-40 min after induction. Gray represents control (Ctrl). Green represents GABA. Orange represents APV. h, The time course of averaged and normalized fEPSP slope showing the effect of two sequential stLTP protocols followed by HFS. i, The summary of data showing the average magnitude of stLTPs and HFS-induced LTP 5-10 min after induction. j, The time course of averaged and normalized fEPSP slope showing the effect of two sequential tbLTP protocols followed by HFS. k, The summary of data showing the average magnitude of tbLTPs and HFS-induced LTP 5-10 min after induction. Data are mean ± SEM. NS, Not significant (p > 0.05). *p < 0.05; **p < 0.01; ***p < 0.001; two-sample t-test.

Whole-cell recordings in CA1 pyramidal neurons were obtained using patch electrodes with a resistance of 3-5 MΩ, filled with a solution containing the following (in mM): 130 KCH₃SO₃, 8 NaCl, 10 HEPES, 10 Na₂-phosphocreatine, 4 Na₂ATP, 0.4 NaGTP, 3 L-ascorbic acid, pH adjusted to 7.2 with KOH (osmolarity to 290 mOsm). The recording solution also contained the morphologic tracer AlexaFluor-594 (50 μm, red channel, R) and the Ca²⁺-sensitive dye Fluo-4F (250 μm, green channel, G) for stLTP, and Fluo-5F (300 μm, green channel, G) for tbLTP. Baseline Ca²⁺ transients were significantly different for the two types of stimulation; thus, different dyes were selected to ensure that the response was within the dynamic range of each dye.

APs were induced by somatic current injections (2 ms, 1-2 nA). The membrane potential was kept at −70 mV for all recordings to ensure experimental conditions were consistent. The threshold of APs was calculated from phase portraits as the membrane potential (V_m) when ΔV_m/Δt was 20 mV/ms. Input resistance was calculated from the voltage response to 500 ms current injection in a stepwise fashion from −50 to 50 pA in 10 pA steps (the I–V curves). The series resistance was typically < 20 MΩ, and data were discarded if this changed by >20% during the recording. The series resistance was compensated for by the “bridge balance” function in current-clamp mode.

Recordings were amplified with a patch-clamp amplifier, Multiclamp 700B (Molecular Devices), and digitized at 10 kHz with a NI PCI-6221 card (National Instrument). The data were visualized and stored with WinWCP software (supplied free of charge to academic users by John Dempster, University of Strathclyde).

Two-photon imaging

Cells were filled with dye for at least 20 min before imaging to ensure dye equilibration. Two-photon Ca²⁺ imaging was performed with a two-scanner FV1000-MPE laser-scanning microscope (Olympus) equipped with a mode-locked (<140 fs pulse width) tunable 720-930 nm laser Chameleon XR (Coherent). Both dyes were excited at a wavelength of 830 nm, and their fluorescence was chromatically separated and detected with two independent photomultipliers. We used the bright AlexaFluor-594 emission to identify oblique apical dendrites (∼150 mm from the soma) and their spines. Line-scan imaging was performed to record Ca²⁺ signals in the dendritic shaft and from one to three spines. Imaging was synchronized with electrophysiological recordings. At the end of each recording, we tested that the Ca²⁺ transients were below the dye saturation level by prolonged somatic depolarization, causing cell firing and Ca²⁺ buildup in neurons. The changes in baseline Ca²⁺ levels were monitored as the ratio between baseline Fluo-4 or Fluo-5F and AlexaFluor-594 fluorescence (G_baseline/R_baseline). If this ratio increased during the experiment by >20%, the recordings were discarded. The dark noise of the photomultipliers was collected during each recording when the laser shutter was closed.

Glutamate uncaging

Glutamate concentration in the synaptic cleft can reach ∼1 mM and decays with a time constant of 1.2 ms (Clements et al., 1992). Glutamate uncaging cannot precisely mimic this concentration profile. First, it is impossible to uncage glutamate only in the synaptic cleft without activating extrasynaptic NMDA receptors (NMDARs). Second, to obtain 1 mM of glutamate, 5–10 mM of the caged compound should be added to the bath or locally applied. This can potentially cause osmotic stress. Here we used a bath application of 400 μm 4-methoxy-7-nitroindolinyl-caged-L-glutamate, which did not produce significant osmolarity changes. Then we adjusted the duration and intensity of laser stimulation to mimic synaptically induced EPSPs of 2–3 mV [uncaging induced (u) EPSP]. Single-photon uncaging was conducted using 5–10 ms laser pulses (405 nm diode laser; FV5-LD405; Olympus) with the “point scan” mode in Fluoview software (Olympus). Uncaging spots were typically positioned at the edge of the spine heads of imaged dendrites. The Ca²⁺ transients were measured in response to (1) a single backpropagating AP (bAP) induced by somatic current injection; (2) bAP/uEPSP pairing in which a single bAP was followed by uncaging at a single dendritic spine with a 20 ms delay (the strength of uncaging was adjusted to trigger 2-3 mV uEPSP); and (3) a uEPSP burst in which uncaging was simultaneously performed at four dendritic spines 4 times at 100 Hz (the strength of uncaging was adjusted to trigger two APs).

Drugs and chemicals

All drugs were kept frozen at −20°C in 100-200 μl 1000× concentration aliquots (stock solutions). CGP52432, S-MCPG, GABA, D-APV, picrotoxin, and 4-methoxy-7-nitroindolinyl-caged L-glutamate were purchased from Tocris Cookson. AlexaFluor-594, Fluo-4F, and Fluo-5F were obtained from Invitrogen.

Data analysis

Electrophysiological data were analyzed with WinWCP and Clampfit (Molecular Devices). Imaging data were analyzed using FluoView (Olympus), ImageJ (a public domain Java image processing program by Wayne Rasband), and custom software written in LabView (National Instruments). Statistical analysis was performed using Excel (Microsoft) and Origin 8 (OriginLab). Ca²⁺ transients were represented as ΔG/R: (G_peak – G_baseline)/(R_baseline – R_{dark noise}); the baseline Ca²⁺ level was estimated as G/R: (G_baseline – G_{dark noise})/(R_baseline – R_{dark noise}), where G is the Fluo-4F/Fluo-5F fluorescence and R is AlexaFluor-594 fluorescence. G_baseline and R_baseline are the average fluorescence levels 50-100 ms before stimulation. G_peak is the mean fluorescence 30-40 ms following stimulation. G_{dark noise} and R_{dark noise} are the dark currents of the corresponding photomultipliers. For illustration purposes, single traces were processed by a 5 point moving average, and then four or five sequential traces were averaged. The statistical significance was tested using a paired-sample, two-sample, or one-sample t-test when appropriate. Data are presented as mean ± SEM; and n indicates the number of recordings.

Mathematical modeling

Simulations were performed in the NEURON 7.1 simulation environment (Hines and Carnevale, 1997). A biophysically detailed CA1 hippocampal pyramidal cell model was modified from Poirazi et al. (2003). The implemented membrane mechanisms included location-dependent R_m (membrane resistance) and R_a (axial resistance). The model incorporated sodium, delay rectifier-, A-type, M-type, Ca²⁺-activated potassium, and h-type conductances, as well as L-, R-, and T-type voltage-dependent Ca²⁺ channels (Poirazi et al., 2003). The detailed conductance densities and their subcellular distribution properties are provided in the supplementary information of Poirazi et al. (2003). The excitatory synaptic conductance was composed of AMPA receptors (AMPARs) and NMDARs (gAMPA and gNMDA): gAMPA was represented by a double exponential function with τ_rise of 1 ms τ_decay of 12 ms (Otmakhova et al., 2002); gNMDA was implemented as in Kampa et al. (2004). Six excitatory synapses were placed at six different apical oblique dendrites (apical dendrite 5, 8, 10, 18, 113, and 118) within the stratum radiatum. gAMPA and gNMDA were set to generate 10 pA somatic excitatory postsynaptic currents with a physiological NMDA/AMPA charge ratio in each excitatory synapse (Otmakhova et al., 2002). The tonic GABA_A conductance (gGABA) with an outward-rectifying property was adapted from Pavlov et al. (2009). gGABA density was distributed homogeneously along the cell membrane and was set so as to increase the membrane conductance by 25% (1.7 mS/cm²) with E_GABA = −75 mV to mimic experimental conditions. The initial resting membrane potential of simulated neurons was set to −70 mV, and the simulation temperature was 34°C.

Membrane potential and intracellular Ca²⁺ concentrations were simulated during bAP/EPSP pairing, and EPSP burst. For bAP/EPSP pairing, activation of six excitatory synapses was followed by somatic current injection (1.7 nA, 2 ms) to trigger a single AP with a 20 ms delay; for the EPSP burst, six excitatory synapses were activated 4 times at 100 Hz; to mimic D-APV and picrotoxin applications, gNMDA and gGABA were set to zero, respectively.

Results

Tonic GABA_A conductance has no effect on stLTP but suppresses tbLTP

We recorded fEPSPs from the CA1 stratum radiatum of rat hippocampal slices in response to extracellular stimulation of SCs (Fig. 1a). Two experimental protocols were used to trigger LTP in CA3-CA1 synapses: the stLTP protocol (paired synaptic and AD stimulation repeated 100 times within 10 min) and tbLTP protocol (four synaptic stimulations at 100 Hz repeated 10 times with a 200 ms interval). Each type of LTP was entirely blocked by 50 μm D-APV (APV), an NMDAR antagonist (Fig. 1b–g). An increase in tonic GABA_A conductance was simulated by the bath application of 30 μm GABA. Because of efficient uptake, we assumed the acting concentration of GABA to be in the low micromolar range expected under physiological conditions in vivo. Nevertheless, because GABAergic synapses are located on the dendritic shaft, they have fewer astrocyte processes in their vicinity and are relatively exposed to ambient GABA (Gavrilov et al., 2018). Thus, the bath application of GABA can recruit both extrasynaptic and synaptic GABA_A receptors.

GABA increased the membrane conductance of CA1 pyramidal neurons to 122 ± 6% of baseline (n = 5, p = 0.02, one-sample t-test) but did not significantly affect stLTP (fEPSP slope: 121 ± 6% of baseline in control, 30 min after stLTP induction, n = 8; 125 ± 9% of baseline in the presence of GABA, n = 7, p = 0.753, two-sample t-test; Fig. 1b–d). However, it reduced tbLTP by approximately half (fEPSP slope: 161 ± 13% of baseline in control, 30 min after tbLTP, n = 7; 130 ± 7% of baseline in the presence of GABA, n = 8, p = 0.041, two-sample t-test; Fig. 1e–g). These results suggest that, although both types of LTP are NMDAR-dependent, they are differentially affected by tonic GABA_A conductances.

An alternative explanation to the differential effect of tonic GABA_A conductances on two types of LTP is that stLTP is saturated with the induction protocol, which we used. When LTP is saturated, it may become insensitive to any kind of modulation. To test this possibility, we repeated the stLTP protocol twice and then applied high-frequency stimulation (HFS, 100 Hz for 1 s). The second stLTP protocol increased the magnitude of LTP, which was further increased by HFS (fEPSP slope: 132 ± 5% of baseline after first stLTP protocol, 158 ± 13% after second stLTP protocol, p = 0.03 for difference; 256 ± 30% after HFS, p = 0.0012 for the difference with second stLTP, n = 8, paired-sample t-test; Fig. 1h,i). Then we performed a similar experiment for tbLTP. The second tbLTP protocol also increased the magnitude of LTP. Subsequent HFS produced an increase in LTP similar to HFS in the stLTP experiment (fEPSP slope: 152 ± 13% of baseline after first tbLTP protocol, 176 ± 16% after second tbLTP protocol, p = 0.0003 for difference; 258 ± 31% after HFS, p = 0.002 for the difference with second stLTP, n = 8, paired-sample t-test; Fig. 1j,k). These results indicate that neither stLTP nor tbLTP was saturated. A similar magnitude of LTP induced by HFS suggests that slices were in similar conditions in both experiments.

Tonic GABA_A conductance does not affect amplitude and kinetics of somatic AP

To understand the mechanism underlying the differential effects of tonic GABA_A conductances on LTP, we first tested whether tonic GABA_A conductance directly modulates somatic AP amplitude and kinetics. Previous reports have suggested that the activation of extrasynaptic GABA_A receptors can influence the AP waveform (Szemes et al., 2013; Xia et al., 2014). Because stLTP depends on voltage-dependent removal of the Mg²⁺-block of NMDARs by bAP, changes in its waveform might lead to changes in the effectiveness of stLTP induction. However, we did not observe a significant change in amplitude, half-width, or threshold of somatic APs in CA1 pyramidal neurons following GABA application (Table 1; Fig. 2). This result is consistent with our finding that stLTD was not affected by tonic GABA_A conductance. Indeed, APs are regenerative events, and their amplitude depends on the activation of Na⁺ and K⁺ conductances, which are considerably larger than the tonic GABA_A conductance (Hausser et al., 2001).

Table 1.

Tonic GABA_A conductances do not affect AP properties in the soma of CA1 pyramidal neurons^a

	Control	GABA	Paired-sample t-test (p)
AP amplitude	107.8 ± 5.6 mV	108.7 ± 6.7 mV	0.50
AP half-width	1.47 ± 0.15 ms	1.49 ± 0.15 ms	0.19
AP threshold	−54.1 ± 4.0 mV	−54.6 ± 4.3 mV	0.45

^aData are mean ± SEM; n = 7.

Figure 2.

Tonic GABA_A conductances do not affect the properties of APs. a, Sample APs in control (Ctrl, gray) and in the presence of GABA (green). b, Phase plots of APs do not differ in control (gray) and in the presence of GABA (green). The threshold of APs was calculated as the membrane potential (V_m) at AP velocity ΔV_m/Δt = 20 mV/ms. The plot was low-pass filtered at the values of V_m below the AP threshold for clarity. c-e, The summary of data are shown for the amplitude (c), half-width (HW, d), and threshold (e) of APs in control (gray) and in the presence of GABA (green). Data are mean ± SEM. NS, Not significant (p > 0.05, paired-sample t-test).

Tonic GABA_A conductance differentially affects responses to bAP/EPSP pairing and EPSP burst in model pyramidal neurons

The properties of somatic APs are determined by the high density of Na⁺ and K⁺ channels in the nearby triggering zone. When the AP propagates along a dendrite, the density of voltage-dependent conductances decreases; and as such, bAPs can be influenced by tonic GABA_A-mediated shunting as nonregenerative events (Golding et al., 2001; Groen et al., 2014; Brunner and Szabadics, 2016; but see Bereshpolova et al., 2007). To address the effects of tonic GABA_A conductances on bAPs in dendritic spines and thin dendrites, we used a previously suggested mathematical model (Poirazi et al., 2003). Six synapses were simulated at proximal regions of apical oblique dendrites to induce a single multisynaptic EPSP (Fig. 3a). A somatic current injection (1.7 nA, 2 ms) was simulated to trigger a bAP. The pairing of an EPSP with a bAP was used to mimic a single stimulation in the stLTP protocol. The somatic AP was triggered with a 10 ms delay, matching the bAP's arrival with the peak of the EPSP. The bAP alone induced much smaller Ca²⁺ transients than the EPSP (Fig. 3b). The pairing of a bAP with an EPSP induced a supralinear increase in the magnitude of the Ca²⁺ transient (Nevian and Sakmann, 2004). Tonic GABA_A conductance (1.7 mS/cm²) decreased the area under the curve (AUC) of the ΔV_m induced by the bAP/EPSP pairing by approximately one-fourth (Table 2; Fig. 3c); and consequently, the amplitude of the Ca²⁺ transient was reduced by a similar degree (Table 2; Fig. 3d,e). The subsequent blockade of NMDARs nearly abolished the remaining Ca²⁺ transient (Table 2, GABA + APV; Fig. 3d,f). Notably, after introducing the tonic GABA_A conductance, the blockade of NMDARs had no further effect on the AUC of ΔV_m induced by the bAP/EPSP pairing (Table 2; Fig. 3c). However, the blockade of NMDARs without a tonic GABA_A conductance reduced the AUC of ΔV_m (Table 2; Fig. 4a). This NMDAR blockade also suppressed the Ca²⁺ transient produced by the bAP/EPSP pairing (Table 2; Fig. 4b,c). Thus, most Ca²⁺ entering the spine during bAP/EPSP pairing is mediated by NMDARs. Approximately one-fourth of this Ca²⁺ transient and ΔV_m was blocked by a tonic GABA_A conductance.

Figure 3.

Tonic GABA_A conductances differentially affect responses to bAP/EPSP pairing and EPSP burst in the model pyramidal neuron. a, The synapse location (green dots) on the simulated pyramidal neuron. Two electrodes indicate where V_m was obtained: red, on soma; blue, on a second-order dendritic branch. The somatic electrode also indicates where the current injection used to elicit bAP occurred. b, The Ca²⁺ transients in the dendrite (top row); V_m in the dendrite (middle row) and the soma (bottom row). Left column, EPSP. Middle column, bAP. Right column, bAP/EPSP pairing. [Ca²⁺]_i, Intracellular Ca²⁺ concentration. c, The ΔV_m to bAP/EPSP pairing in control (Ctrl, gray), on an increase in the tonic GABA_A conductance (GABA, green), and subsequent blockade of NMDARs (GABA + APV, orange). d, The Ca²⁺ transient in response to bAP/EPSP pairing in control (gray), and on an increase in the tonic GABA_A conductance (green), and subsequent blockade of NMDARs (orange). e, The summary of data on the effect of tonic GABA_A conductances (GABA) on the amplitude of the Ca²⁺ transient induced by bAP/EPSP pairing (empty bar) and EPSP burst (filled bar). f, The summary of data on the effect of subsequent NMDAR blockade (GABA + APV) on the amplitude of the Ca²⁺ transient induced by bAP/EPSP pairing (empty bar) and EPSP burst (filled bar). g, The ΔV_m to EPSP burst in control (gray), and on an increase in the tonic GABA_A conductance (green), and subsequent blockade of NMDARs (orange). h, The Ca²⁺ transient in response to EPSP burst in control (gray), and on an increase in the tonic GABA_A conductance (green), and subsequent blockade of NMDARs (orange).

Table 2.

Tonic GABA_A conductances reduce the ΔV_m and Ca²⁺ transient induced by EPSP burst to a greater extent than those induced by bAP/EPSP pairing in the model neuron^a

	GABA	APV	GABA + APV
bAP/EPSP pairing
Dendrite ΔVm AUC	73.3 ± 0.4% (6)	87.5 ± 1.2% (6)	73.3 ± 0.7% (6)
Peak Δ[Ca2+]i	76.7 ± 1.8% (6)	6.8 ± 0.7% (6)	5.6 ± 0.5% (6)
EPSP burst
Dendrite ΔVm AUC	48.8 ± 1.1% (6)	66.2 ± 0.2% (6)	44.6 ± 0.9% (6)
Peak Δ[Ca2+]i	53.8 ± 0.9% (6)	7.6 ± 0.6% (6)	0.0 ± 0.0% (6)

^aData are mean ± SEM (n). The responses were normalized to control. GABA, the introduction of the tonic GABA_A conductance; APV, removal of NMDAR conductance; GABA + APV, both.

Figure 4.

Effects of GABA on NMDAR-independent Ca²⁺ transients in the model neuron. a, The change in V_m in response to bAP/EPSP pairing in control (Ctrl, black), and on the blockade of NMDARs (APV, orange) and subsequent increase of the tonic GABA_A conductance (GABA + APV, green). b, The Ca²⁺ transient in response to bAP/EPSP pairing in control (black), and on the blockade of NMDARs (orange) and subsequent increase of the tonic GABA_A conductance (green). c, The summary of data on the effect of the NMDAR blockade (APV, orange) and subsequent increase in the tonic GABA_A conductance (GABA + APV, green) on the amplitude of the Ca²⁺ transient induced by bAP/EPSP pairing. d, The summary of data on the effect of the NMDAR blockade (APV, orange) and subsequent increase in the tonic GABA_A conductance (GABA + APV, green) on the amplitude of the Ca²⁺ transient induced by EPSP burst. e, The change in V_m in response to EPSPs burst in control (Ctrl, black), and on blockade of NMDARs (APV, orange) and subsequent increase in the tonic GABA_A conductance (GABA + APV, green). f, The Ca²⁺ transient in response to EPSP burst in control (black), and on blockade of NMDARs (orange) and subsequent increase of the tonic GABA_A conductance (green).

Next, we simulated an EPSP burst similar to that in the tbLTP protocol: four EPSPs at 100 Hz were simulated at the same six synapses. The resulting EPSP burst triggered two somatic APs and a large Ca²⁺ transient in stimulated dendritic segments (Fig. 3g,h). The tonic GABA_A conductance reduced the AUC of ΔV_m induced by EPSP burst by approximately half, bringing it below the threshold for AP generation (Table 2; Fig. 3g). Consequently, the Ca²⁺ transient was proportionally reduced (Table 2; Fig. 3f,h). Further blockade of NMDARs had a small effect on the AUC but completely abolished the Ca²⁺ transient (Table 2, GABA + APV; Fig. 3f–h). NMDAR blockade without a tonic GABA_A conductance abolished one of the two APs and thereby reduced the AUC (Table 2; Fig. 4d). This blockade also largely suppressed the Ca²⁺ transient (Table 2; Fig. 4e,f). Thus, the model of CA1 pyramidal neuron proposed by Poirazi et al. (2003) suggests that tonic GABA_A conductances should have a more profound effect on the ΔV_m and Ca²⁺ transient induced by EPSP burst than on those induced by bAP/EPSP pairing.

Tonic GABA_A conductance differentially affects responses to bAP/uEPSP pairing and uEPSP burst in CA1 pyramidal neuron

To confirm the model's prediction, we performed two-photon Ca²⁺ imaging in dendritic spines (Fig. 5a). The dendritic Ca²⁺ transients (ΔG/R) induced by bAPs alone were not significantly affected by GABA and subsequent 100 μm picrotoxin (GABA_A antagonist) application (GABA: 100.2 ± 8.2% of control, n = 9, p = 0.98, one-sample t-test; picrotoxin: 96.0 ± 8.3% of control, n = 9, one-sample t-test). This finding suggests that the tonic GABA_A conductance did not affect bAPs in imaged dendrites. This, however, does not rule out its action on more distal dendritic branches. bAP/EPSP pairing was mimicked by uEPSP, followed in 20 ms by the injection of depolarizing current through the patch pipette (Fig. 5b). EPSP burst was mimicked by repeated glutamate uncaging (4 times at 100 Hz) at four synapses and further referred to as uEPSP burst (Fig. 5c). The application of GABA had a small effect on the somatic ΔV_m induced by bAP/uEPSP pairing that was not statistically significant (ΔV_m AUC: 94 ± 2% of control, n = 7, p = 0.07, one-sample t-test; Fig. 5d,e). Indeed, ΔV_m induced by the bAP/uEPSP pairing was dominated by somatic AP that was insensitive to tonic GABA_A conductances. Nevertheless, the Ca²⁺ transient induced by the bAP/uEPSP pairing was significantly reduced (peak ΔG/R: 86 ± 4% of control, n = 6, p = 0.01, one-sample t-test; Fig. 5d,f).

Figure 5.

Tonic GABA_A conductances differentially affect responses to bAP/uEPSP pairing and uEPSP burst in CA1 pyramidal neurons. a, Representative image of a CA1 pyramidal neuron filled with the morphologic tracer AlexaFluor-594 (50 μm). White box represents a second-order dendritic branch used for Ca²⁺ imaging. b, Left, Zoomed-in dendritic branch with a line-scan trajectory (white line) and an uncaging spot (Unc., white dot). Middle, The bAP/uEPSP pairing protocol (uncaging dot and current injection step) and corresponding V_m change in the soma. Right, A line-scan image in the Fluo-4F channel showing Ca²⁺ transients in dendritic spines and the shaft crossed by the line scan. c, Left, Zoomed-in dendritic branch with a line-scan trajectory (white line) and four uncaging spots (Unc., white dots). Middle, The uEPSP burst protocol (four uncaging episodes at 100 Hz) and corresponding V_m change in the soma. Right, A line-scan image in the Fluo-5F channel showing Ca²⁺ transients in dendritic spines and the shaft crossed by the line scan. d, The ΔV_m in the soma (left column) and the Ca²⁺ transients in dendritic spines (ΔG/R, right column) triggered by bAP/uEPSP pairing (top row) and uEPSP burst (bottom row). Gray represents Control (Ctrl). Green represents GABA application. Orange represents Subsequent APV application. e, The summary of data on the effect of GABA (green) and subsequent APV application (orange) on the AUC of the ΔV_m to bAP/uEPSP pairing (empty symbols) and uEPSP burst (filled symbols). f, The summary of data on the effect of GABA (green) and subsequent APV application (orange) on the Ca²⁺ transients induced by bAP/uEPSP pairing (empty symbols) and uEPSP bursts (filled symbols). Data are mean ± SEM. NS, Not significant (p > 0.05). *p < 0.05; **p < 0.01; two-sample t-test.

In agreement with the model's prediction, GABA reduced the number of APs induced by the uEPSP burst from 2 ± 0 to 1.1 ± 0.3 (n = 6, p = 0.031, paired-sample t-test, Fig. 5d); and as such, both the ΔV_m and the Ca²⁺ transient were significantly decreased (ΔV_m AUC: 76 ± 6% of control, n = 6, p = 0.007, one-sample t-test; ΔG/R: 71 ± 3% of control, n = 6, p < 0.001, one-sample t-test; Fig. 5d–f). Consistent with our hypothesis, this decrease was significantly larger than the decrease observed with the bAP/uEPSP pairing (ΔV_m AUC: p = 0.013, ΔG/R: p = 0.014, two-sample t-test; Fig. 5d–f)

The subsequent blockade of NMDARs with APV did not further decrease the ΔV_m in either case (ΔV_m AUC in GABA + APV: 95 ± 3% of control, n = 7 for bAP/uEPSP pairing; 76 ± 5% of control, n = 6 for uEPSP burst; Fig. 5d,e). However, the Ca²⁺ transients were further reduced to a similar level in both cases (ΔG/R in GABA + APV: 43 ± 4% of control, n = 6 for bAP/uEPSP pairing; 41 ± 4% of control, n = 6 for uEPSP burst; p = 0.77, two-sample t-test for difference between stimulations; Fig. 5d,f). These results demonstrate that the Ca²⁺ transients induced both by bAP/uEPSP pairing and by uEPSP burst in dendritic spines are partially mediated by NMDARs and are sensitive to tonic GABA_A conductance. However, it is not clear to what extent tonic GABA_A conductances affect the NMDAR-dependent and NMDAR-independent components of Ca²⁺ transients. To address this issue, we reversed the order of drug application: first, we applied APV and then GABA. APV had a small, non statistically significant, effect on the ΔV_m to bAP/uEPSP pairing (ΔV_m AUC: 94 ± 4% of control, n = 7; p = 0.14, one-sample t-test) but significantly reduced the ΔV_m to uEPSP burst (ΔV_m AUC: 76 ± 7% of control, n = 8; p = 0.008, one-sample t-test; p = 0.03, two-sample t-test for difference between stimulations; Fig. 6). Nevertheless, NMDAR blockade reduced Ca²⁺ transients in both cases by approximately half (ΔG/R: 46 ± 4% of control, n = 6 for bAP/uEPSP pairing; 53 ± 6% of control, n = 6 for uEPSP burst; p = 0.285, two-sample t-test for difference between stimulations; Fig. 6). Subsequent GABA application had a small effect on the ΔV_m (ΔV_m AUC: 84 ± 4% of control, n = 6 for bAP/uEPSP pairing; 73 ± 4% of control, n = 6 for uEPSP burst; p = 0.086, two-sample t-test for difference between stimulations; Fig. 6b) and Ca²⁺ transients in both cases (to 39 ± 4%, n = 6, for bAP/uEPSP pairing and 38 ± 3%, n = 6, for uEPSP burst; p = 0.865, two-sample t-test for difference between stimulations; Fig. 6c). Notably, NMDAR-independent Ca²⁺ transients were equal for both types of stimulation before and after the application of GABA. This finding suggests that the effect of tonic GABA_A conductance on different forms of LTP is due to its action on NMDAR-mediated Ca²⁺ transients.

Figure 6.

Tonic GABA_A conductances have small effects on NMDAR-independent responses to bAP/uEPSP pairing and uEPSP burst in CA1 pyramidal neurons. a, The ΔV_m in the soma (left column) and the Ca²⁺ transients in dendritic spines (ΔG/R, right column) triggered by bAP/uEPSP pairing (top row) and uEPSP burst (bottom row). Gray represents Control (Ctrl). Orange represents APV application. Green represents Subsequent GABA application. b, The summary of data on the effect of APV (orange) and subsequent GABA application (green) on the AUC of ΔV_m to bAP/uEPSP pairing (empty symbols) and uEPSP burst (filled symbols). c, The summary of data on the effect of APV (orange) and subsequent GABA application (green) on the Ca²⁺ transients induced by bAP/uEPSP pairing (empty symbols) and uEPSP burst (filled symbols). Data are mean ± SEM. NS, Not significant (p > 0.05). *p < 0.05, two-sample t-test.

Discussion

We show that tonic GABA_A conductance suppresses tbLTP, but not stLTP. Both forms of plasticity require NMDARs, but these NMDARs are recruited by different mechanisms. At baseline, NMDARs are blocked by Mg²⁺ (Nowak et al., 1984). Depolarization of the postsynaptic neuron removes this block and enables ion movement through glutamate-bound NMDARs (Wu et al., 2012). While synaptic activity serves as a source of glutamate in both forms of plasticity, the manner in which the Mg²⁺ block is removed differs. In the case of tbLTP, the EPSP burst propagates to the soma and triggers APs, which propagate back and sum with the burst. The resulting depolarization unblocks NMDARs. Tonic GABA_A conductances suppress the EPSP burst by shunting. The diminished EPSP burst is unable to trigger APs and recruits fewer NMDARs. Thus, tbLTP is reduced. In the case of stLTP, the Mg²⁺ block of NMDARs is removed by bAPs. The AP is a regenerative event mediated by significantly larger conductances than those of tonic GABA_A or the underlying EPSP (Hausser et al., 2001). Indeed, we did not detect any significant effect of tonic GABA_A conductances on AP parameters (threshold, amplitude, half-width) recorded in the soma. However, this finding does not reflect the effect of tonic GABA_A conductances on bAPs in the dendrite. Dendrites have a lower density of Na⁺ and K⁺ channels than the axonal trigger zone, which determines the shape of the somatic AP (Lorincz and Nusser, 2010; Kole and Stuart, 2012; Kirizs et al., 2014). Nevertheless, the bAP-induced Ca²⁺ transients recorded with two-photon imaging were not significantly affected by GABA application. As a result, Ca²⁺ entry through NMDARs during bAP/EPSP pairing was decreased by tonic GABA_A conductance significantly less than during EPSP burst. Consequently, no significant decrease in stLTP magnitude was detected, in contrast to tbLTP. We also demonstrated that the NMDAR-independent component of Ca²⁺ transients was similarly affected by GABA in the bAP/uEPSP pairing and uEPSP burst, suggesting that the observed effect is mediated by a GABA-dependent decrease of NMDAR-dependent Ca²⁺ transients during uEPSP burst stimulation.tbLTP can be linked to cell firing during theta rhythms, which is important for the encoding and retrieval of space- and time-related information (Hyman et al., 2003; Hasselmo and Stern, 2014; Larson and Munkacsy, 2015). stLTP can occur when the neuron receives synaptic inputs from many presynaptic cells and needs to choose which inputs are more relevant (Markram et al., 2012). Although the relevance of stLTP as a general model for synaptic plasticity is sometimes put into question (Lisman and Spruston, 2005, 2010), this phenomenon has been demonstrated in different species and brain regions in vitro and in vivo (Sjöström et al., 2008; Shulz and Jacob, 2010; Feldman, 2012; Gambino and Holtmaat, 2012; Jones et al., 2017). stLTP also has a broad appeal in computational neuroscience (Abbott and Nelson, 2000; Senn et al., 2001; Graupner and Brunel, 2007; Clopath et al., 2010; Vignoud et al., 2018). Our findings that stLTP is more resistant to tonic GABA_A conductances than tbLTP may provide further insight into how the activity-dependent accumulation of ambient GABA can affect brain computations, learning, and memory. These findings can also be considered in the context of physiological/pathologic processes that control the magnitude of tonic GABA_A conductance: densities and properties of extrasynaptic GABA_A receptors, GABA release and clearance.

Tonic GABA_A conductance is mediated by multiple and plastic GABA_A receptors (Scimemi et al., 2005). The α5-subunit-containing GABA_A receptor is a receptor subtype that contributes to the tonic inhibition of CA1 pyramidal neurons (Caraiscos et al., 2004). Notably, the pharmacological blockade of α5GABA_A receptors enhances LTP in the CA1 region of the ventral hippocampus (Pofantis and Papatheodoropoulos, 2014). This receptor subtype also raises the LTP induction threshold in the dorsal hippocampus through predominantly nonsynaptic mechanisms and regulates certain forms of memory in vivo (Martin et al., 2010).

The change in tonic conductance magnitude due to modifications to receptor composition/density has been reported during development and in pathologic conditions (Brickley and Mody, 2012). Indeed, both increased tonic conductance and attenuated LTP were reported in animal models of epilepsy (Scimemi et al., 2005; Plata et al., 2018). In Alzheimer's disease, elevated concentrations of extracellular GABA occur due to GABA production and release by astrocytes (Wu et al., 2014), and the increased tonic GABA_A conductance suppresses LTP in the hippocampal dentate gyrus in this disease.

Our results suggest that stLTP is insensitive to tonic GABA_A conductance. However, it does not rule out that this form of plasticity remains under the control of other forms of GABAergic inhibition: for example, feedforward or feedback IPSPs curtailing EPSPs. The pairing of a single EPSP with a single AP, which we used in this study, is efficient in juvenile but not young adult mice or rats (Meredith et al., 2003). However, stLTP can still occur in response to EPSP pairing with a postsynaptic burst of APs. The blockade of GABA_A receptors restores the ability of a single AP–EPSP pairing to induce LTP in adult rats. Therefore, this effect may be attributed to the maturation of the GABAergic system.

The transient emergence of α4βδ GABA_A receptors on the spines of CA1 pyramidal neurons reduces activation of NMDARs and prevents tbLTP at puberty in mice (Shen et al., 2010). This effect of perisynaptic inhibition is more profound than the reduction of tbLTP mediated by tonic conductance reported in this study. Our work was conducted on prepubertal animals, where tonic conductance is mediated by GABA_A receptors located on the dendritic shaft and soma. How the activation of perisynaptic α4βδ GABA_A receptors would affect stLTP in puberty requires further investigation, although we speculate that stLTP may also be affected at puberty. Although spines are relatively electrically isolated and are weakly affected by dendritic conductance, the positioning of GABA_A receptors on their surface may be a game-changer.

Ambient GABA concentrations build up due to GABA spillover and, thus, reflect neuronal activity (Glykys and Mody, 2007b; Song et al., 2013). Neuronal activity can also lead to an increase in extracellular GABA through astrocytes (Heja et al., 2012; Unichenko et al., 2013; Kirischuk et al., 2016). Synaptically released glutamate is taken up by astrocytic transporters along with Na⁺ in a 1:3 ratio. Intracellular Na⁺ accumulation then reverses Na⁺-dependent GABA transporters, which start to move GABA to the extracellular space.

A high level of neuronal activity is also required for tbLTP induction (Larson and Munkacsy, 2015). In many synapses, induction of LTP can lead to excessive excitability of the brain, seizures, and excitotoxicity. However, an accompanying increase in extracellular GABA reduces both the excitability and magnitude of tbLTP. Thus, the activity-dependent elevation of ambient GABA can serve as a protective mechanism to maintain balanced levels of brain excitation. This phenomenon is reminiscent of the homeostatic downregulation of individual cell excitability by the upregulation of h-channels following LTP induction (Fan et al., 2005; Wu et al., 2012). On the other hand, stLTP does not require a high rate of presynaptic firing but depends on the temporally correlated occurrence of synaptic inputs and postsynaptic APs. Thus, the induction of stLTP does not have to be associated with significant activity-dependent accumulation of extracellular GABA. Indeed, tonic GABA_A conductance can be beneficial for stLTP, as stLTP is sensitive to spike jitter (Cui et al., 2018). Tonic GABA_A conductance can reduce spike jitter by decreasing the membrane time constant (Wlodarczyk et al., 2013). In addition, we addressed only postsynaptic mechanisms by which tonic GABA_A conductance can have a differential effect on LTP. In our model and glutamate uncaging experiments, we assumed synaptic release probability of 1, which means that each presynaptic AP triggers the vesicular release. Indeed, vesicular release probability in CA3-CA1 synapse varies from 0 to 1 (Hanse and Gustafsson, 2001). During burst stimulation, release probability can increase because of presynaptic facilitation and decrease because of depletion of immediately releasable vesicles. The activity-dependent facilitation occurs because of residual Ca²⁺ accumulation after each presynaptic AP and because of depolarization of presynaptic terminal by local elevation of extracellular K⁺ (Shih et al., 2013). Most of the extracellular K⁺ leaks to the synaptic cleft through postsynaptic NMDARs. When tonic GABA_A conductance decreases activation of NMDARs, it also decreases NMDAR-mediated K⁺ efflux and, hence, presynaptic facilitation. Reduced presynaptic facilitation will further decrease postsynaptic Ca²⁺ transients during burst stimulation and contribute to a more profound reduction of tbLTP.

In conclusion, brain states and activity that increase the tonic GABA_A conductance suppress rate coding (tbLTP) but not temporal coding (stLTP) in the hippocampus. This phenomenon may have important implications for overall brain computations, learning, and memory.