A Dynamic Role for GABA Receptors on the Firing Pattern of Midbrain Dopaminergic Neurons

Abstract

Dopaminergic neurons are subject to a significant background GABAergic input in vivo. The presence of this GABAergic background might be expected to inhibit dopaminergic neuron firing. However, dopaminergic neurons are not all silent but instead fire in single-spiking and burst firing modes. Here we present evidence that phasic changes in the tonic activity of GABAergic afferents are a potential extrinsic mechanism that triggers bursts and pauses in dopaminergic neurons. We find that spontaneous single-spiking is more sensitive to activation of GABA receptors than phasic N-methyl-d-aspartate (NMDA)-mediated burst firing in rat slices (P15–P31). Because tonic activation of GABA_A receptors has previously been shown to suppress burst firing in vivo, our results suggest that the activity patterns seen in vivo are the result of a balance between excitatory and inhibitory conductances that interact with the intrinsic pacemaking currents observed in slices. Using the dynamic clamp technique, we applied balanced, constant NMDA and GABA_A receptor conductances into dopaminergic neurons in slices. Bursts could be produced by disinhibition (phasic removal of the GABA_A receptor conductance), and these bursts had a higher frequency than bursts produced by the same NMDA receptor conductance alone. Phasic increases in the GABA_A receptor conductance evoked pauses in firing. In contrast to NMDA receptor, application of constant AMPA and GABA_A receptor conductances caused the cell to go into depolarization block. These results support a bidirectional mechanism by which GABAergic inputs, in balance with NMDA receptor–mediated excitatory inputs, control the firing pattern of dopaminergic neurons.

INTRODUCTION

Dopamine released by the activity of midbrain dopaminergic neurons plays an important role in Parkinson's disease (Blandini et al. 2000), reinforcement learning (Schultz 1998), and schizophrenia (Grace 1991). At least 70% of all inputs onto dopaminergic neurons are GABAergic, the majority of which arise from the striatum, globus pallidus (GP) and substantia nigra pars reticulata (SNr) (Bolam and Smith 1990; reviewed in Tepper and Lee 2007). Because spontaneously active neurons in the GP and SNr fire at rates as high as 50 and 60 Hz, respectively (Celada et al. 1999; Deniau et al. 1978; Guyenet and Aghajanian 1978; Kita and Kitai 1991), the activity of dopaminergic neurons is subject to a large GABAergic input. Dopaminergic neurons under the influence of this GABAergic inhibition would be expected to be mostly silent. Although some dopaminergic neurons are silent (Grace et al. 2007; but see Dai and Tepper 1998), many of them are spontaneously active in anesthetized or in awake, behaving animals (Freeman et al. 1985; Hyland et al. 2002; Kiyatkin and Rebec 1998; Schultz et al. 1997). They typically fire single spikes with varying degrees of regularity and can generate high-frequency bursts by a mechanism activating N-methyl-d-aspartate (NMDA) receptors (Chergui et al. 1993; Deister et al. 2009; Overton and Clark 1992, 1997; Zweifel et al. 2009). Dopamine is released either tonically or phasically depending whether the neuron is in a single-spiking or burst-firing mode (Goto et al. 2007; Grace and Bunney 1984a,b; Wilson et al. 1977). Does tonic GABAergic input simply hinder these firing modes or could it play a more integral role in their generation?

Dopaminergic neurons receive a combination of tonic inhibitory and excitatory inputs in vivo. Dopaminergic neurons are bombarded by chloride-mediated inhibitory postsynaptic potentials (IPSPs) in vivo (Grace and Bunney 1985). Local application of GABA_A receptor (GABA_AR) antagonists shifts the firing pattern of dopaminergic neurons from a single-spike mode into a burst-firing mode, whereas application of GABA_B receptor antagonists regularizes the firing pattern (Brazhnik et al. 2008; Engberg 1993; Paladini and Tepper 1999). These results were interpreted to mean that tonic activation of GABA_A receptors inhibits burst firing and that the action of GABA_B receptors is mostly presynaptic. The shift into the burst-firing mode suggests that dopaminergic neurons are also subject to tonic excitation. Local application of NMDA receptor (NMDAR) antagonists, but not an AMPA receptor (AMPAR) antagonist, significantly reduced burst firing (Chergui et al. 1993; Overton and Clark 1992), suggesting that the majority of the tonic excitation driving bursts is NMDAR mediated.

Here we show that spontaneous, single-spike firing is more sensitive to activation of both GABA_A and GABA_B receptors than phasic, NMDAR-mediated burst firing evoked by either iontophoresis or dynamic clamp. This suggests that the activity seen in dopaminergic neurons in vivo is the result of a balance between excitatory and inhibitory conductances. Single spiking continued after application of balanced, constant NMDAR/GABA_AR, but not AMPAR/GABA_AR, conductances by dynamic clamp. Applying tonic NMDAR and GABA_AR conductances, we show that phasic removal of the GABA_AR conductance causes burst firing and phasic increases in the GABA_AR conductance causes a pause in firing. Bursts generated by disinhibition have a higher frequency than bursts generated by identical excitation alone. Our results provide evidence for an extrinsic mechanism by which phasic changes in GABAergic drive can generate bursts and pauses in firing in midbrain dopaminergic neurons.

METHODS

Slice preparation and recordings

Electrophysiological experiments were performed on slices obtained from Sprague-Dawley rats (Charles River Laboratories) 15–31 days of age. Although dopaminergic neurons recorded in vivo do show changes in firing pattern with age (Tepper et al. 1990), we found no difference among neurons from 15- or 31-day-old animals during in vitro recordings. All experimental procedures were approved by the University of Texas at San Antonio Institutional Animal Care and Use Committee. Rats were anesthetized with ketamine/xyzaline and decapitated, and the brains were rapidly removed and cooled. Horizontal slices (240 μm) were cut using a vibrating microtome (Microm HM 650V) in oxygenated, cold artificial cerebrospinal fluid (ACSF) containing (in mM) 110 cholineCl, 2.5 KCl, 1.25 NaH₂PO₄, 7 MgCl₂, 0.5 CaCl₂, 10 dextrose, 25 NaHCO₃, 1.3 ascorbic acid, and 2.4 sodium pyruvate. Slices were transferred to an incubation chamber containing (in mM) 126 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 4 MgCl₂, 2 CaCl₂, 10 dextrose, 25 NaHCO₃, 1.3 ascorbic acid, 2.4 sodium pyruvate, and 0.05 glutathione. Slices were incubated at 32°C for at least 1 h before recording and kept at room temperature thereafter. Before recording, a slice was held in a submerged chamber filled with ACSF similar to the incubating solution except that 2 mM MgCl₂ was used, and glutathione was not added. The slice was superfused at a rate of 2 ml/min by a gravity feed system and heated to 32–34°C with an inline heater.

SNc neurons were visualized with a gradient contrast imaging system. Perforated-patch or whole cell recordings were made from presumed SNc dopaminergic neurons. Perforated patch recordings were made with the whole cell internal described below or an internal solution containing (in mM) 140 KMeSO₄, 0.2 EGTA, 7 NaCl, and 10 HEPES. The antibiotic gramicidin A or D (dissolved in DMSO, 100 μg/ml of internal solution) was used to maintain natural intracellular Cl⁻ levels (Kyrozis and Reichling 1995) and to reduce run-down. Accidental break-in was determined by a large, instantaneous jump in spike height. Cells in which accidental break-in occurred were rejected. Whole cell recordings were made with an internal solution containing (in mM) 138 K-gluconate, 10 HEPES, 2 MgCl₂, 0.2 EGTA, 0.0001 CaCl2, 4 Na-ATP, and 0.4 Na-GTP. All internal solutions were adjusted to a pH of 7.3 using 1 M KOH and an osmolarity of 270–275 mOsm. Recordings were acquired with a Multiclamp 700B and digitized (Instrutech) under command of the AxographX software program.

Dopaminergic neurons were identified by a slow, spontaneous firing rate (usually 1–4 Hz), a prominent spike afterhyperpolarization, and a large I_h current on passage of a hyperpolarizing voltage step. The presence of a large mGluR1-mediated hyperpolarizing response to iontophoresis of glutamate after a burst (Morikawa et al. 2003) was also used to identify dopaminergic neurons (Marino et al. 2001).

Dynamic clamp

Dynamic clamp experiments were conducted in whole cell mode as previously described (Deister et al. 2009). The equations used to calculate the applied current are

$$I_{\text{NMDA}}=-g_{\text{NMDA}}\times \left(1/\left\{1+\left(\left[\text{Mg}\right]\text{/3.57}\right)\times e^{\left(-V_{\text{m}}\times 0.062\right)}\right\}\right)\times \left(V_{\text{m}}-E_{\text{NMDA}}\right)$$

$$I_{\text{GABAA}}=-g_{\text{GABAA}}\times \left(V_{\text{m}}-E_{\text{GABAA}}\right)$$

$$I_{\text{GABAB}}=-g_{\text{GABAB}}\times \left(V_{\text{m}}-E_{\text{GABAB}}\right)$$

where [Mg] = 1.5 mM, E_NMDA = 0 mV, E_GABAA = −60 mV (−63 mV was the mean reversal potential for GABA_A receptors in perforated patch; Gulásci et al. 2003), and E_GABAB = −100 mV unless otherwise stated. An AMPAR-mediated current was created from the I_NMDA equation by setting [Mg] to 0. All recordings were done with a balanced bridge in continuous current clamp (Bridge Mode). A junction potential of −6 mV was corrected on-line in these experiments.

Evoked responses

Iontophoresis was chosen over electrical stimulation of the slice to evoke phasic bursts that did not contain a GABA-mediated component, which would be concomitantly activated on electrical stimulation. Iontophoresis was also chosen over bath application of NMDA (Komendantov et al. 2004; Paladini et al. 1999b) to isolate the effects of GABA receptor activation on single-spiking and burst firing. Bursts were generated by iontophoresis of glutamate (typically onto dendrites 50 μm from the recording electrode; 50–250 ms pulses, holding +1 to 10 nA, ejection −50 to −300 nA; Dagan ION-100) as previously described (Deister et al. 2009). The iontophoretic pipette contained 1 M glutamate at pH 7–9. All experiments involving iontophoresis were performed in the presence of the AMPA receptor blockers, NBQX, or GYKI-52466.

Drugs

All drugs were applied to the slice via superfusion. Isoguvacine hydrochloride (1–100 μM), (R)-baclofen (0.01–10 μM), picrotoxin (100 μM), NBQX (25 μM), GYKI-52466 (20–50 μM), and CGP-55845 (1–2 μM) were purchased from Tocris.

Data analysis

Action potentials were detected using a derivative threshold in AxographX (1–5 V/s), and changes in spiking frequency were measured. Analysis was done with Mathematica 7 (Wolfram Research). A burst was defined as a series of the action potentials that occurred within 1 s after the onset of the iontophoretic pulse. Bursts evoked by dynamic clamp were analyzed for the entire time window that the NMDA conductance was on. In the case of bursts evoked by disinhibition, the window was defined as the period in which the GABA_AR conductance was phasically turned off. Maximum burst frequency was determined as the reciprocal of the minimum interspike interval (ISI) for all spikes in the burst. Mean burst frequency was determined as the reciprocal of the mean intraburst ISI. For spontaneous single spiking, mean frequency was determined as the reciprocal of the mean ISI measured over a time window of 20 s in which no current was injected to the cell. Average membrane potential was calculated as the mean of all sampled voltages, including spikes.

Statistics

In several experiments, inhibition of single-spike or burst firing frequency by GABA receptor activation was measured. Because one of the factors (GABA_AR activation) was quantitative rather than categorical, ANOVA was not appropriate. Instead, regression analysis was used to fit each curve. The data were fit with a quadratic model. This was implemented in SAS (SAS Institute) using the equation

$$y_{\text{i}}={\tau }_{\text{i}}+{\beta }_{1}x+{\left(\tau {\beta }_1\right)}_{i}x+{\beta }_2{x}^2+{\left(\tau {\beta }_2\right)}_i{x}^2+\epsilon$$

where

y_i = dependent variable representing inhibited firing frequency in terms of percent of control, τ_i = effect of firing frequency type (Figs. 1 and 2, B and E: maximum and mean burst frequency and mean single spiking frequency) or receptor type (Fig. 2D: GABA_A and GABA_B), β₁, β₂, (τβ₂)_i, (τβ₁)_i = regression coefficients, x = continuous independent variable for GABA_AR activation in terms of concentration of agonist (μM) or conductance applied (nS), and ε = residuals, ε ∼ N(0,σ²).

Fig. 1.

Single-spiking is more sensitive to suppression by GABA_A receptor (GABA_AR) activation than N-methyl-d-aspartate receptor (NMDAR)-mediated burst firing. A: a phasic burst (15 Hz maximum burst frequency) was evoked in a perforated patch recording of an identified dopaminergic neuron by iontophoresis of glutamate (Glu; black bars beneath trace, top) in 25 μM NBQX. The GABA_AR agonist isoguvacine (40 μM) was applied to the bath (middle). Single-spike firing was completely suppressed while the burst frequency was decreased by 62%. The effect of isoguvacine was reversed by subsequent application of the chloride channel blocker, picrotoxin (100 μM, bottom). B: summarized data show the percent inhibition of maximum burst frequency (filled circles), mean burst frequency (gray circles), and mean spontaneous firing frequency (empty circles) from control frequencies after bath application of isoguvacine (1, 10, 40, and 100 μM). Numbers in parentheses indicate n for each concentration. C: application of a 40 nS NMDAR conductance using dynamic clamp elicits a burst (20 Hz maximum) in a whole cell somatic recording of an identified dopaminergic neuron (top). Application of a 2.8 nS GABA_AR conductance (E_GABAA = −60 mV) suppresses single-spike activity while the burst remains (18 Hz maximum; bottom). Above each voltage trace is the truncated current applied from the dynamic clamp (downward represents an outward current). D: summarized data (n = 4) show the percent of inhibition of maximum burst frequency (filled circles), mean burst frequency (gray circles), and mean spontaneous firing frequency (empty circles) to application of a series of GABA_AR conductances in dynamic clamp (as in C).

Fig. 2.

Single-spike firing is more sensitive to GABA_BR activation than NMDAR-mediated burst firing. A: a phasic burst was evoked in a perforated patch recording of an identified dopaminergic neuron by iontophoresis of glutamate (Glu; black bars beneath trace; top) in 25 μM NBQX. Bath application of 1 μM baclofen abolished single-spike activity and reduced the burst firing frequency. This effect was reversible on removal of baclofen [return to ACSF]. Spike heights are truncated. B: summarized data for the percent of inhibition of maximum burst frequency (filled circles), mean burst frequency (gray circles), and mean spontaneous firing frequency (empty circles) from control after bath application of baclofen (0.01, 0.1, 1, and 10 μM). Numbers in parentheses represent number of cells. C: a representative recording showing inhibition of single-spike and burst firing by application of a 1.2 nS GABA_BR conductance (E_GABAB = −100 mV). The total current (truncated) applied by the dynamic clamp is shown above each voltage trace. D: summarized data (n = 4) show the effectiveness of GABA_AR- and GABA_BR-mediated inhibition in suppressing maximum burst frequency (filled squares and circles, respectively) and mean spontaneous firing frequency (empty squares and circles, respectively). Conductances were applied into the same group of cells.

For dynamic clamp experiments in which a range of GABA_AR conductances were applied to the same cell, an additional random effect c_k was added to y_i to account for within-cell effects.

After each curve was fit, multiple comparisons of firing or receptor type (τ_i) given GABA_AR conductance were made using the Scheffe adjustment (P < 0.05). These comparisons permit us to compare, for example, whether the inhibition of single-spike firing was statistically different from the inhibition of burst firing at a specific GABA_AR conductance.

For all other statistical tests, Prism (Graphpad Software) was used.

All effects are given in terms of means ± SE unless stated otherwise. In all experiments presented here, statistical significance is considered at P < 0.05.

RESULTS

GABA_A receptor activation

We first studied the inhibition of spontaneous single-spiking and NMDAR-mediated burst firing by activation of GABA_A receptors. Identified dopaminergic cells fired spontaneously (typically 1–4 Hz) in a regular, single-spike firing pattern during perforated patch recordings. Bursts of action potentials (mean, 22 Hz maximum burst frequency; similar to in vivo, e.g., Grace and Bunney 1984a) were evoked every 30 s by iontophoretic application of glutamate onto the recorded neuron in the presence of an AMPA receptor antagonist, NBQX (25 μM), or GYKI-52466 (20–50 μM). We applied the GABA_AR agonist, isoguvacine (1, 10, 40, or 100 μM), to the bath and measured its effect on single-spike and burst firing frequency (Fig. 1). Regression analysis of the concentration-response data showed that isoguvacine inhibited single-spike firing at significantly lower concentrations than burst firing (P < 0.05, 15–75 μM isoguvacine, regression analysis). Application of 100 μM isoguvacine abolished both single-spike and burst firing (Fig. 1B). The addition of the GABA_AR antagonist, picrotoxin (100 μM), reversed the effects of isoguvacine (Fig. 1A; P < 0.05, repeated-measures ANOVA with Tukey's multiple comparison test; control single-spike firing frequency: 2 ± 0.57 Hz, 40 μM isoguvacine: 0 ± 0 Hz, picrotoxin: 1.9 ± 0.68 Hz, n = 5). Spontaneous and burst firing frequencies in the presence of picrotoxin were not significantly different from control (P > 0.05; repeated-measures ANOVA with Tukey's multiple comparison test). These results suggest that spontaneous, single-spiking is more sensitive to GABA_AR activation than burst firing.

We obtained similar results by applying a range of GABA_AR conductances (0–30 nS) into cells using dynamic clamp (Fig. 1, C and D; Robinson and Kawai 1993; Sharp et al. 1993). Bursts (20.6 Hz maximum burst frequency on average) were evoked by application of a moderate NMDAR conductance (35 nS; range, 20–40 nS). These conductances were high enough to cause sustained high-frequency firing but were generally not high enough to cause the cell to go into a state of depolarization block by the end of the conductance pulse (which typically occurred at conductances >60 nS; data not shown). Again, single-spiking was more sensitive to GABA_AR activation than burst firing (P < 0.05, 0–24 nS, regression analysis). Both single-spike and burst firing were abolished in all cells at a GABA_AR conductance of 30 nS. The average GABA_AR conductance at which single-spike firing was suppressed was 25% of the conductance needed to suppress burst firing (Fig. 1D; single-spike 4.8 ± 2.1 nS; burst, 19.5 ± 4.5 nS; n = 4). Together, these results show that in the presence of a tonic GABA_AR conductance, single-spike firing is more sensitive to GABA_AR activation than burst firing.

GABA_B receptor activation

Tonic activity in GABAergic afferents may also activate postsynaptic GABA_B receptors on midbrain dopaminergic neurons. Cells were recorded in perforated patch and NMDAR-mediated bursting was evoked by iontophoresis at regular intervals as above. The GABA_BR receptor agonist, baclofen (0.01, 0.1, 1, or 10 μM), suppressed single-spike and burst firing in a dose-dependent manner (Fig. 2). Single-spike firing was inhibited at lower concentrations than burst firing (P < 0.05, 0.6–2 μM baclofen, regression analysis).

Inhibition of single-spike and burst firing by application of a GABA_BR conductance was also investigated in dynamic clamp (Fig. 2, C and D; E_GABAB = −100 mV). Bursts were evoked by application of a NMDAR conductance as before (Fig. 1 C–D). There was a significant difference in the regressions of the single-spike inhibition curves and both of the burst firing inhibition curves (P < 0.05, 0–8 nS, regression analysis).

We also tested whether GABA_BR activation was more effective at suppressing firing than GABA_AR activation. The use of dynamic clamp allowed for direct comparison of these two receptor types in the same sample (from Fig. 1D). We found that there was a significant difference between GABA_AR and GABA_BR conductances for inhibition of burst firing (P < 0.05, 0–24 nS, regression analysis) and single-spike firing (P < 0.05, 0–7.2 nS, regression analysis).

Together these results extend our findings with GABA_AR to show that with activation of either GABA_A or GABA_B receptors single-spike firing will be abolished before the suppression of burst firing mediated by the activation of an NMDAR conductance. These results further show that GABA_B receptors are more effective at suppressing both single-spike and burst firing than GABA_AR. Because the conductances used in Fig. 2D were equal, this indicates that GABA_BR are more effective because of the more hyperpolarized reversal potential for potassium.

Bursting by disinhibition

Previous studies have suggested that bursts are suppressed by tonic activation of GABA_AR in vivo (Brazhnik et al. 2008; Paladini and Tepper 1999). However, here we show that single-spike firing is more sensitive than burst firing to activation of either GABA_A or GABA_B receptors (Figs. 1 and 2). Therefore tonic activation of GABA receptors in vivo would be expected to suppress single spiking, causing dopaminergic neurons to be silent. However, single spiking is seen in vivo (Grace and Bunney 1984b; Paladini and Tepper 1999; Wilson et al. 1977). This suggests that the firing pattern recorded in vivo is caused by a combination of excitatory and inhibitory conductances in addition to its pacemaking currents. These results also suggest that phasic removal of this inhibitory conductance can evoke a burst and that a phasic increase in this inhibitory conductance can evoke a pause in firing.

A control burst was evoked by application of a constant NMDAR conductance in whole cell recordings, as shown above (Figs. 1C and 2C). The same NMDAR conductance was applied in the presence of a constant GABA_AR conductance (E_GABAA = −60 mV; −63 mV measured in perforated patch recordings in Gulácsi et al. 2003) sufficient to counteract the NMDAR conductance-mediated burst and return single-spike firing back to control levels (Fig. 3, A and B; P > 0.05; control mean ISI, 0.55 ± 0.11 s; balanced ISI, 0.67 ± 0.18 s; paired t-test, n = 10). Overall, the ratio of NMDAR/GABA_AR conductance used was 3.2 ± 0.4 (n = 7). In similar experiments using an AMPAR conductance instead of NMDAR, background firing could not be restored. Instead, the cell quickly went into depolarization block (Fig. 3D; 4/4 cells). Block in the AMPA/GABAA configuration occurred at much smaller conductances than those used in the balanced NMDAR/GABA_AR configuration (typical NMDAR/GABA_AR conductance: 25/8nS; AMPA/GABA_AR conductances were <4 nS each).

Fig. 3.

Bursts and pauses can be generated by phasic changes in input. A: in a whole cell recording, balanced GABA_AR and NMDAR conductances were applied into a dopaminergic neuron. A disinhibition burst could be evoked by phasic removal of the GABA_AR conductance. A pause in firing could be evoked by phasic removal of the NMDAR conductance (A) or by a phasic increase in the GABA_AR conductance (C, increase of 2 nS). The total current applied by the dynamic clamp is shown above each voltage trace. The total current in C has been truncated. B: single-spiking frequency in the presence of tonic NMDAR and GABA_AR (Balanced) conductances was not significantly different from spontaneous, single-spiking (Control; P > 0.05, paired t-test). Horizontal lines indicate the means for the sample. D: single spiking is not maintained with application of balanced GABA_AR and AMPAR conductances.

In the NMDAR/GABA_AR balanced configuration, a burst of action potentials could be evoked by briefly turning off the GABA_AR conductance (Fig. 3A). A pause in firing was evoked by transiently turning off the NMDAR conductance (Fig. 3A) or by increasing the GABA_AR conductance (Fig. 3C). Bursts produced by disinhibition had a greater frequency than control bursts with an identical NMDAR conductance (Fig. 4, A and B; P < 0.05; control maximum burst frequency, 22.2 ± 4.9 Hz; disinhibition, 27.8 ± 5.4 Hz; paired t-test, n = 10; 1st 5 spikes mean burst frequency; control, 13.8 ± 1.6 Hz; disinhibition, 20.0 ± 3.4 Hz; paired t-test, n = 10). Both disinhibition and excitation-only bursts displayed prominent spike frequency adaptation (Fig. 4A). The increase in burst firing frequency was sustained over the first eight spikes of the burst (Fig. 4C). This suggests that the mechanism by which disinhibition bursts are faster occurs on the order of hundreds of milliseconds.

Fig. 4.

Disinhibition bursts have a greater frequency than phasic NMDA bursts of the same conductance. A: representative example of a burst produced by phasic activation of NMDA receptors (Control, top trace; g_GABAA = 0; 18 Hz maximum) and by disinhibition (bottom trace; removal of g_GABAA = 5.6 nS; 29 Hz maximum). The total current applied by the dynamic clamp is shown above each voltage trace. In both cases (control and disinhibition), an NMDA conductance of 20 nS was used. B: summary data show a significant increase in maximum (I) and mean (II) burst frequencies of disinhibition bursts compared with control bursts. Horizontal lines indicate the means for the sample. C: summary data show that the interspike interval (ISI) for disinhibition bursts was significantly shorter than control bursts for the 1st 7 ISIs (* = P < 0.05).

The mechanism by which disinhibition bursts are faster than control bursts may lie in differences in the availability of spiking currents just before the removal of the GABA conductance (Fig. 5). To test this, we constructed phase plots of both spontaneous single-spiking (Fig. 5A, black) and balanced conductance single-spiking modes (Fig. 5A, red). There was a significant decrease in maximum dV_m/dt (Fig. 5BI; P < 0.05; control 67.8 ± 8.3, balanced 44.4 ± 2.9, paired t-test, n = 7). There was also a significant increase in minimum dV_m/dt (Fig. 5BII; P < 0.05; control −41.0 ± 4.9, balanced −29.7 ± 3.6, paired t-test, n = 7). Balanced spiking had a more depolarized threshold than control spiking (Fig. 5BIII; P < 0.05; control −34.4 ± 1.3 mV, balanced −32.0 ± 1.5 mV, paired t-test, n = 7). This suggests that an increase in sodium channel availability is not the cause of the increased burst frequency of disinhibition bursts.

Fig. 5.

Disinhibition bursts may be faster because of depolarization. A: phase plots of spontaneous single-spiking (black) and balanced single-spiking (red). Threshold was defined as the voltage in which there was a significant break (4 × SD) of dV_m/dt from baseline (shown in inset for control) and is shown on the control and balanced phase plots by black and red arrows, respectively. B: there was a significant decrease in maximum dV_m/dt (I) during balanced spiking. There was also a significant increase in minimum dV_m/dt (II) and spiking threshold (III) during balanced spiking. C: overlay of all-points histograms for sampled membrane potentials during spontaneous firing (Control, black bars) and balance between excitation and inhibition (Balanced, white bars). Black and white arrows represent the average membrane potential for Control and Balanced, respectively. The histograms were generated from 9 s of single spike firing in each configuration. Ten superimposed single spikes (aligned at the peak) are shown as an inset for Control (black) and Balanced configurations (red). Scale bars for inset are 10 mV (ordinate) and 100 ms (abscissa). Dashed line in inset shows −40 mV. D: average membrane potential for balanced spiking was more depolarized than control spiking. E: overlay of control (black) and disinhibition (red) bursts (I; same bursts as shown in Fig. 4A). The total current applied by the dynamic clamp is shown above each voltage trace. Summary data show a significant decrease in the time to the peak of the 1st action potential (II) from application of an NMDAR conductance (Control, black) or removal of a GABA_AR conductance (Disinhibition, red). Arrow indicates the time of conductance change. Horizontal lines indicate means for the sample in each panel.

Histograms of control and balanced spiking showed that the average membrane potential was significantly more depolarized for balanced than control spiking (Fig. 5, C and D, also Fig. 5A; example histogram taken from cell shown in Fig. 4A; P < 0.05; control −46 ± 1.0 mV, balanced −39 ± 1.5 ms, paired t-test, n = 10). There was also a significant decrease in the latency from the time of conductance change to the peak of the first action potential in the burst (Fig. 5E; P < 0.05; control 58 ± 16 ms, disinhibition 27 ± 9 ms, paired t-test, n = 10). This suggests that a current that inactivates with depolarization (e.g., an A-type potassium current) may be the cause of the increased burst frequency of disinhibition bursts.

DISCUSSION

High conductance state of the dopaminergic neuron

Dopaminergic neurons recorded in vivo display a variety of firing patterns: silent, regular single-spiking, irregular single-spiking, and bursty (reviewed in Tepper and Lee 2007). Dopaminergic neurons recorded in slices, however, fire only in the regular, single-spiking mode. This difference in firing patterns of dopaminergic neurons recorded in slices and in vivo is often ascribed to the loss of afferent input in slices (Overton and Clark 1997). This suggests that single-spiking in vivo is generated by an in vitro–like pacemaking mechanism in which spikes can be advanced or delayed because of afferent input (hereafter referred to as the pacemaker mechanism). Under such a scheme, bursts are caused by phasic excitatory inputs and pauses in firing are caused by phasic inhibitory inputs. Such a mechanism may underlie the reward-related responses of dopaminergic neurons described by Schultz (1998). However, additional mechanisms for generating phasic bursts and pauses are likely for dopaminergic neurons in the high conductance state (reviewed in Destexhe et al. 2003).

There is much evidence that suggests the presence of a tonic inhibitory drive onto dopaminergic neurons. Dopaminergic neurons are bombarded by chloride-mediated IPSPs in vivo (Grace and Bunney 1985). These IPSPs are most likely caused by spontaneously active afferent cells in the globus pallidus (GP) or substantia nigra pars reticulata (SNr), which fire at ∼50 Hz (Celada et al. 1999; Deniau et al. 1978; Guyenet and Aghajanian 1978; Kita and Kitai 1991). The presence of this tonic inhibitory drive can also be shown by the local application of GABA antagonists in vivo. Paladini and Tepper (1999); and Brazhnik et al. (2008) found that local application of a variety of GABA_A antagonists by pressure ejection onto the recorded neuron shifted the firing pattern of the dopaminergic neuron from a single-spiking mode to a bursting one, indicating that tonic GABA_AR activation suppresses burst firing.

Our results, showing that NMDA-mediated burst firing is suppressed at greater levels of both GABA_A and GABA_B receptor activation than single spiking, also suggest that dopaminergic neurons are tonically inhibited in vivo. However, if single spiking in vivo is generated by a pacemaker mechanism, our results suggest that both single-spiking and bursting would be suppressed and there would be no tonic level of dopamine in efferent structures. However, many dopaminergic neurons recorded in vivo are not silent (Grace and Bunney 1984a,b; Paladini and Tepper 1999; Paladini et al. 1999a; Wilson et al. 1977), and tonic levels of dopamine are observed at target loci (Floresco et al. 2003; Gonon 1988). Therefore our results suggest that dopaminergic neurons receive tonic excitatory input.

Local application of NMDAR antagonists but not an AMPAR antagonist significantly reduce spontaneous burst firing (Chergui et al. 1993; Overton and Clark 1992; see also Charlety et al. 1991), suggesting that tonic excitation is present and is NMDAR mediated. Similar results were obtained with genetic NMDAR inactivation (Zweifel et al. 2009), and lesion of the subthalamic nucleus (STN), a spontaneously active glutamatergic afferent, regularized the firing pattern of dopaminergic cells (Smith and Grace 1992). Incomplete antagonism of NMDAR activation may explain why the strongly bursting cells of Chergui et al. (1993) do not become silent, as would be expected for a dopaminergic neuron in the high conductance state.

Are the GABA_AR and NMDAR conductances applied here with dynamic clamp comparable to the conductances activated by tonic GABAergic synaptic input in vivo? Using a GABA_A receptor time constant of 6 ms (τ; Brancucci et al. 2004), a single channel conductance of 5–30 pS (Δg; Guyon et al. 1999; Macdonald and Olsen 1994), and a 50 Hz input rate (F), we calculate that 887–5,330 GABA_A receptors must be activated by globus pallidus and substantia nigra pars reticulata inputs to achieve the 8 nS steady-state conductance (g_ss) used in the balanced configuration in Fig. 1D (g_ss = Δg/(1 − e^{[−1/(τFn)]}); Wilson et al. 2004). Assuming 12 receptors per synapse [∼5 immunogold particles per synapse (Fujiyama et al. 2002) times 2.5 GABA_A receptors per gold particle (Nusser et al. 1997)], this corresponds to a minimum ∼1–8% of the total GABAergic synapses onto an SNc neuron (∼5,600 GABAergic synapses total; Henny et al. 2009). Similarly, we calculate that 2,300 NMDA receptors from 348 synapses must be activated for the NMDAR conductance used in the balanced configuration (g_ss = 25 nS; τ = 43 ms, Schilström et al. 2006; Δg = 50 pS, Edmonds et al. 1995; F = 5 Hz for subthalamic nucleus, Chergui et al. 1994; 2.9 immunogold particles per synapse, Chatha et al. 2000; assuming 2.3 channels per synapse, similar to AMPA, Nusser 1999). This corresponds to ∼15% of the available glutamateric synapses (∼30% of synapses are VGluT²⁺, Henny et al. 2009). Thus the NMDAR/GABA_AR conductances used in this study are easily approachable in the intact network. These results also suggest that mechanisms that affect the properties of NMDAR and GABA_AR (e.g., conductance, desensitization, or kinetics) may also have significant effects on the balance achieved by the dopaminergic neuron and hence influence firing pattern.

We propose that many dopaminergic neurons in vivo may be in the high conductance state generated by tonic GABA_AR and NMDAR conductances. Activation of both AMPA and GABA_A receptors decrease the input resistance of the cell and shunt the pacemaker and burst-generating currents (Fig. 6A). Thus the negative slope region of the IV curve, created by voltage-gated sodium and calcium channels, disappears. Similar results are expected for all other excitatory conductances such as nicotinic acetylcholine receptors. However, activation of NMDAR is unique because its negative slope conductance produces an increase in input resistance (Koch 1999). This negative slope conductance is caused by the magnesium sensitivity of the NMDA receptor. The increase in input resistance with activation of NMDA receptors counteracts the decrease in input resistance caused by activation of GABA_A receptors, and the negative slope region of the IV curve persists (Fig. 6B). This explains why spikes can still be produced in the balanced NMDA/GABA_A configuration. Differences in input resistance measurements in vivo (31 MΩ; Grace and Bunney 1983) and in slices (∼20 0MΩ; 70–250 MΩ, Kita et al. 1986; 135 MΩ, Grace and Onn 1989; 384 MΩ, Paladini et al. 1999b) suggest that 27.3 nS of the total input conductance in vivo (32.3 nS measured in vivo − 5 nS measured in slices = 27.3 nS) are caused by tonic NMDA and GABA_A conductances. We found that cells went into block at AMPA/GABA_A conductances around 4 nS and thus cannot account for the 27.3 nS difference in input conductance. The 27.3 nS input conductance difference in the NMDA/GABA_A configuration may be underestimated because of the increase in input resistance with NMDAR activation. The drop in input resistance caused by impalement with sharp electrode recordings will overestimate this difference in input conductance.

Fig. 6.

The high conductance state of the dopaminergic neuron. The effect of tonic AMPA/GABA_A (A) and NMDA/GABA_A (B) conductances on the steady-state IV curve of a fictive dopaminergic neuron. Note that the negative slope region between −65 and −50 mV persists at much higher NMDA/GABA_A conductances than AMPA/GABA_A conductances. The thick black line represents the control condition where g_AMPA, g_NMDA, and g_GABAA are set to 0. The thin lines represent a 5 nS increase in g_AMPA (A) or g_NMDA (B) and a 1.6 nS increase in g_GABAA. Dashed arrows represent displacements of the IV curve as these conductances are increased.

An imbalance in tonic inputs may drive the cell into hyperpolarization or depolarization block in some proportion of dopaminergic neurons (Dai and Tepper 1998; Grace et al. 2007). For example, silent cells are cells whose membrane potential (−65 to −70 mV; Grace and Bunney 1984b) is near the chloride reversal potential (−68 mV; Grace and Bunney 1985). They represent the case where there is a tonic GABA_A conductance sufficient to shunt intrinsic pacemaker currents (Grace et al. 2007) but lack a strong enough NMDA conductance to return the cell to the spiking regime. Silent cells, like other shunting configurations, may still be able to produce spikes in response to input fluctuations.

A dopaminergic neuron firing mainly under the influence of a balance between excitatory and inhibitory inputs would be more sensitive to dynamic fluctuations in rate and gain of inputs within the basal ganglia network (Carvalho and Buonomano 2009; Destexhe et al. 2003; Vogels and Abbott 2009) and adopt a more irregular single-spike firing pattern, as normally seen in vivo (Paladini et al. 1999a).

Tipping the scales: generating pauses and bursts

NMDA receptors play an important role in the generation of bursts (Chergui et al. 1993, 1994; Deister et al. 2009; Overton and Clark 1992; Tong et al. 1996; Zweifel et al. 2009). The simplest mechanism by which bursts are generated in vivo is phasic activation of NMDA receptors. However, in the presence of tonic synaptic conductances, this requires that the NMDA receptor currents overcome a substantially reduced input resistance (≤90% from in vitro estimates) and an increased Mg²⁺ block on the NMDA receptor (Paladini and Tepper 1999). These limitations can be overcome by an alternative mechanism where bursting is triggered by disinhibition (Hong and Hikosaka 2008; Jhou et al. 2009; Matsumoto and Hikosaka 2009).

Here we show that not only can bursts be produced by disinhibition but that disinhibition bursts have a greater firing frequency than bursts evoked by NMDA receptors alone. The mechanism by which disinhibition bursts are faster than excitation-only bursts is probably caused by a more depolarized voltage in the balanced state compared with spontaneous, single-spiking (Fig. 5C). This depolarization is a result of the total current produced from the application of NMDA and GABA_A conductances with an effective synaptic reversal potential that is a conductance-weighted average of E_NMDA (0 mV) and E_GABAA (−60 mV). A depolarized membrane potential may inactivate A-type potassium channels (Gentet and Williams 2007; Khaliq and Bean 2008; Liss et al. 2001). The time to first spike measurement of Fig. 5EII is influenced by both spike-generating sodium channels and A-type potassium channels, which play an important role in the determination of the interspike interval. However, in control conditions, more sodium channels are available, which would overestimate the contribution of I_A. This suggests that A-type potassium channels are inactivated in the balanced configuration and may be responsible for the increase in burst frequency. Inactivation of A-type potassium channels would also increase single-spiking frequency (Liss et al. 2001). However, an increase is not seen here (Fig. 3B) because the procedure by which we obtain the balanced configuration was designed to keep the firing rate similar to the control, spontaneous firing rate. These results also suggest that disinhibition bursts may be more precisely timed to rewarding stimuli than bursts by NMDA receptor activation alone.

Recent studies have shown that synaptic plasticity may enhance synaptically triggered burst firing and therefore underlie the acquisition of the burst response to a reward-predicting stimulus (Harnett et al. 2009). Although changes in plasticity can be effected within minutes, changes in the balance between excitatory and inhibitory inputs are capable of modulating bursts quicker. Thus scaling of intraburst firing rate can occur according to the probability of reward associated with a stimulus within a single trial (Morris et al. 2004) or on successive trials of a learning task (Pan et al. 2005).

Phasic increases in GABAergic drive activate postsynaptic GABA_A and/or GABA_B receptors to induce a pause in firing. The length of the pause would be proportional to the increase in drive. GABA_B receptors would be expected to produce a longer pause than GABA_A receptors and may be recruited by synaptic spillover (Galvan et al. 2006). This suggests that the role of GABA_B receptors may not be only presynaptic. Pauses in firing may also be initiated or augmented by removal of excitation (Fig. 4A, e.g., the STN).

The firing pattern of dopaminergic neurons is a behaviorally important signal necessary for normal reward learning (Schultz 1998; Zweifel et al. 2009). Bursts and pauses in firing encode reward prediction error. Our results provide an extrinsic mechanism by which both reward-related responses could be evoked by phasic changes in GABAergic drive. These results suggest that the role of tonic GABAergic inhibition is not simply to suppress single-spiking and burst firing. On the contrary, the presence of tonic background inhbition allows for bidirectional changes in the firing rate and pattern in dopaminergic neurons. Thus GABAergic inhibition plays a fundamental role in the generation of the reward prediction error signal.

GRANTS

This work was supported by National Institutes of Health Grants MH-084494 to C. J. Lobb and MH-079276 and NS-060658 to C. A. Paladini.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).