Mixed excitatory and inhibitory GABA-mediated transmission in chick cochlear nucleus

Abstract

1. Neurons of the chick nucleus magnocellularis (NM) receive depolarizing GABAergic input from the superior olivary nucleus (SON). We examined the response to exogenous GABA or to stimulation of GABAergic fibres in order to identify the ionic basis of GABAergic synaptic transmission and its physiological implications.

2. Reversal potentials of GABA responses (E_GABA) were determined exclusively by the Cl⁻ gradient, measured using whole-cell recording. With gramicidin-perforated patch recording, E_GABA was -25 ± 5 mV (mean ± S.D.), and was stable between embryonic day 17 and post-hatch day 10. With normal intracellular Cl⁻, GABA depolarized neurons by 12 mV.

3. In current clamp, repetitive activation of the GABAergic axons reduced the probability of spiking in response to simultaneous stimulation of excitatory axons. However, IPSPs could themselves elicit action potentials, and facilitation of IPSPs by repetitive activation could lead to a characteristic pattern of spiking.

4. These data indicate that IPSPs with reversal potentials positive to spike threshold may have dual functions, depending on the context of their activation.

In circuits that encode interaural timing of auditory signals, a variety of morphological and physiological adaptations ensure preservation of information through sequential synaptic levels (Trussell, 1999). Given the need for faithful transmission of signals, it is perhaps surprising that inhibitory innervation, by GABAergic and glycinergic fibres, is a major feature of the lower auditory pathways (Wenthold et al. 1986; Code et al. 1989). How might inhibitory synapses alter acoustic signalling? Activation of GABA_A receptors can either depolarize or hyperpolarize neurons, depending on the driving force, V_rest - E_GABA. While hyperpolarizing GABA responses may shunt excitatory currents and bring the potential away from spike threshold, depolarizing ones both shunt and cause Na⁺ channel inactivation (Staley & Mody, 1992; Golding et al. 1995; Zhang & Jackson, 1995). Shunting also shortens the membrane time constant, leading to briefer synaptic potentials and more precise temporal summation (Funabiki et al. 1998). Therefore, understanding the role of GABA in auditory processing requires determining directly how excitatory and inhibitory synapses interact.

We examined the actions of GABA and GABAergic IPSPs in neurons of the chick NM, to help define the biophysical actions of GABA on auditory neurons. Due to the simpler anatomical layout and well-described developmental time course, chick NM is a useful model for the initial level of auditory processing. NM neurons are functionally homologous to spherical bushy cells of the mammalian cochlear nucleus, and contain the first central synapses of the timing pathway (Trussell, 1999). Each cell is innervated by two to three auditory nerve fibres (Parks, 1981); interspersed among these glutamatergic inputs are boutons with GABA immunoreactivity (Code et al. 1989). GABAergic inputs originate largely from SON, with possible contributions from interneurons scattered just outside NM (Lachica et al. 1994; Monsivais et al. 2000). SON inputs to NM may be activated in response to acoustic stimuli, probably via collaterals from nucleus angularis and nucleus laminaris (Lachica et al. 1994). Hyson et al. (1995) showed that exogenous GABA depolarizes neurons in chick NM, even in animals that were mature with respect to their auditory system. We report that the depolarizing nature of GABA in NM is attributed to a remarkably elevated intracellular Cl⁻ concentration, yielding a suprathreshold reversal potential, which persisted even in relatively mature animals. GABAergic IPSPs provided both powerful inhibition and a transient excitatory signal.

METHODS

Brainstem slices (250 μm thick) were taken from chicks between embryonic day 17 and post-hatch day 10 (E17-P10). Use of animals followed procedures approved by the Oregon Health Sciences University Institutional Animal Care and Use Committee. Hatchling chicks, used for the experiments of Fig. 3, were anaesthetized with halothane vapour prior to being killed by decapitation. All extracellular solutions contained (mm): 140 NaCl, 5 KCl, 3 CaCl₂, 1 MgCl₂, 10 Hepes and 20 glucose, pH 7.4 with NaOH. Slices were continuously perfused with oxygenated solution at 35 ± 1 °C and at a rate of 2 ml min⁻¹.

Figure 3.

Gramicidin-perforated patch recordings

A, recordings on the same cell with 133 mm Cl⁻ in recording pipette, E_GABA = −25 mV with patch intact (left) and near 0 mV 7 min after break-in (right). Holding potentials were set at 10 mV intervals. B, GABA (10 μm)-induced depolarization on another cell recorded with a gramicidin patch. Bath application of 10 mm TEA enhanced the depolarization by almost twofold. In A and B, the duration of GABA application is indicated with horizontal bars. Occasional spontaneous synaptic currents are visible in the records. C, during a 15 day developmental period E_GABA (▴) and membrane potential (X), both measured with gramicidin patch, were stable. Linear fits yielded r = 0.24, P > 0.2 for E_GABA, and r = 0.16, P > 0.5 for membrane potential.

Whole-cell recordings

Patch-clamp recordings were made on visually identified NM neurons (Zhang & Trussell, 1994). For cell-attached recordings, pipettes contained (mm): 140 KCl, 3 CaCl₂, 1 MgCl₂ and 10 Hepes (pH 7.3 with KOH). For experiments measuring E_GABA with different [Cl⁻]_i, the pipettes contained (mm): X CsCl, (140 - X) CsMeSO₃, 1 MgCl₂, 10 Hepes and 5 Cs₄-BAPTA (pH 7.3 with CsOH; X = 10, 20, 40, or 140 mm). For current-clamp experiments, pipettes contained (mm): 92 potassium gluconate, 58 KCl, 1 MgCl₂, 10 Hepes and 4 K₄-BAPTA (pH 7.3 with KOH). For experiments recording synaptic currents, pipettes contained (mm): 100 CsCl, 10 TEACl, 5 MgCl₂, 4 MgATP, 0.4 Na₂GTP, 14 Tris phosphocreatine, 10 QX-314Cl (Alomone Labs), 10 Hepes and 4 Cs₄-BAPTA (pH 7.3 with CsOH). For conventional whole-cell recordings, series resistance was 3-8 MΩ before compensation by 90 %. After filtering at 5-10 kHz, the stimulus waveform and current signal were digitized and stored on computer and analysed with Clampfit 8 (Axon Instruments), Origin 6 (Microcal Software) and Microsoft Excel 2000. All reported voltages were corrected for junction potentials.

E_GABA was measured using a series of voltage steps (protocol I) or a voltage ramp (protocol II). In protocol I, the depolarizing and hyperpolarizing voltage steps were interleaved. The voltage ramps had slopes of less than 500 mV s⁻¹ and E_GABA was estimated as the voltage intercept of traces with or without GABA.

Glutamatergic and/or GABAergic synaptic responses were elicited by extracellular stimulation through a glass pipette (tip diameter: 5-10 μm, intensity: 0.5-2 μA, duration: 100-200 μs) positioned ≈40 μm from the cell body in the dorsolateral direction. AP-5, 6,7-dinitro-quinoxaline-2,3-dione (DNQX), GYKI 52466, bicuculline methiodide and SR 95531 were obtained from both RBI and Tocris Cookson. GABA was pressure ejected onto the recorded neuron through another patch pipette positioned ≈30 μm from soma.

Gramicidin experiments

Gramicidin D (2.5 mg; Sigma) was dissolved in 10 μl anhydrous DMSO, vortexed, aliquoted and kept at -20 °C for up to a week. The pipette solution for gramicidin experiments contained (mm): 120 potassium gluconate, 5 sodium gluconate, 3 CaCl₂, 3.5 MgCl₂, 30 Hepes and 0.4 Lucifer yellow (pH 7.3 with KOH). Prior to experiments, 1 μl of gramicidin stock was added into 1 ml of pipette solution and vortexed. Gigaseals were formed with little or no suction to avoid break-in. Perforation began within 5-10 min, developing steadily for the following 10-20 min. Typical series resistance was 10-15 MΩ before compensation of 90 %. The intactness of the patch (estimated by exclusion of dye from the cell) was monitored every few minutes. In some experiments, gluconate was replaced with Cl⁻, with no effect on E_GABA, confirming that patches remained unruptured.

We observed a -2 to −6 mV jump in resting potential at the instant the perforated patch was broken, probably reflecting a Donnan potential across the perforated patch. Indeed, a potential of −4 mV was predicted using the method of Horn & Marty (1988), and hence all potentials in gramicidin experiments were corrected by this value.

Statistics

All statistical analysis was done with Student's paired and unpaired t test. All data are shown as means ± s.d.

RESULTS

Anion permeability of NM GABA receptor channels

We first examined the permeability of GABA receptor channels. A standard concentration of 10 μm GABA was used, which produced currents of < 5 nA when the cell was within ± 30 mV of E_GABA. E_GABA was measured both using the interpolated zero-current value in I-V relations based on data such as those in Fig. 1A, and from leak-subtracted currents obtained using a voltage-ramp command (Fig. 1B). Figure 1C shows that E_GABA varied linearly with log[Cl⁻]_pipette (r = 0.99, P < 0.001), changing 53 mV for a 10-fold change in [Cl⁻]_i, close to the Nernst prediction for a Cl⁻-selective channel. GABA_A receptors have a higher permeability to nitrate than to Cl⁻ (Bormann et al. 1987). In NM, E_GABA was 15 ± 3 mV (n = 7) when the pipettes were filled with 130 mm CsNO₃ plus 10 mm CsCl. The Goldman-Hodgkin-Katz relation yielded an NO₃⁻/Cl⁻ permeability ratio of 2.05, similar to estimates made on spinal neurons (2.1; see Bormann et al. 1987).

Figure 1.

Relationship between E_GABA and [Cl⁻]_i

A and B, protocols utilized for E_GABA measurement in whole-cell recording. A, protocol I: the cell was held steadily at different potentials (10 mV interval) for at least 10 s before 10 μm GABA was applied by pressure ejection. B, protocol II: a voltage-ramp command was delivered before and during the application of GABA (inset); the range of the voltage ramp was from -60 to −30 mV and the duration was 75 ms. In A and B, horizontal bars over the traces indicate the duration of GABA application. Data in each panel are from different cells and with different [Cl⁻]_i, as indicated. C, summary of E_GABA in relation to [Cl⁻]_i. Number of cells for each data point is shown. Linear fit to data yields a slope of 53 mV per decade, consistent with predictions of the Nernst equation.

Depolarizing GABA responses due to elevated [Cl⁻]_i

As GABA-evoked currents are mediated by Cl⁻, yet are depolarizing in microelectrode recordings (Hyson et al. 1995), we examined GABA's action without disturbing intracellular Cl⁻ concentration. We first used a technique described by Zhang & Jackson (1995) (Fig. 2A). Cell-attached recordings were made using a high [K⁺] pipette solution. Spontaneous outward-going single-channel currents were observed (Fig. 2B) which reversed when the pipette potential (V_pipette) equalled the resting potential. Given symmetrical K⁺ on either side of the patch, these events are probably due to K⁺ channel openings. By determining the reversal potential of the channels with and without GABA, we could see if GABA depolarizes the cell and by how much. In Fig. 2B, the pipette was held at −20 mV, corresponding to a patch potential of about −50 mV, assuming a resting membrane potential of −70 mV. Under these conditions, the spontaneous channel events were outward. When 50 μm GABA was pressure ejected onto the neuron for 5 s, the amplitude of single-channel currents declined, indicating a reduction in driving force and a depolarization of the neuron. Figure 2C shows for the same neuron the average amplitude of channel events measured just before and during GABA application versus V_pipette. Application of GABA to this neuron produced an 8 mV depolarization. On average GABA (50 μm) shifted the membrane potential from -63 ± 8 to -51 ± 9 mV (n = 8; paired t test, P < 0.02), a difference of +12 mV.

Figure 2.

GABA-induced depolarization monitored with cell-attached recording

A, experimental protocol. B, examples of single K⁺ channel currents before, during and after the application of 50 μm GABA for 5 s. At the V_pipette of −20 mV the currents were outward. Note the reduction in current amplitude with GABA application. C, I-V plot for single K⁺ channel currents recorded in the same cell, with or without GABA. For each data point 100-2862 events were averaged in control solutions and 12-151 events were averaged in the presence of GABA. Membrane potential is measured by extrapolating a linear fit of data points to zero current.

We next used gramicidin-perforated patch recording to measure E_GABA with endogenous [Cl⁻]_i intact (Kyrozis & Reichling, 1995; Kakazu et al. 1999). Figure 3A (left panel) shows a typical perforated patch recording, in which the reversal potential of the response to 10 μm GABA was −25 mV. In Fig. 3A (right) a few minutes after patch rupture, E_GABA was −1 mV, as expected after dialysis of the intracellular Cl⁻ by the pipette solution. As the bath did not contain HCO₃⁻, the E_GABA with gramicidin is not due to the contribution of E_HCO3 (Staley et al. 1995), but must reflect an elevated E_Cl. E_GABA was positive to spike threshold, based on measurements of the response to current steps (-39 ± 2 mV, n = 4, data not shown; see Zhang & Trussell, 1994). In current-clamp mode, 10-20 μm GABA depolarized the cells by 11 ± 3 mV (n = 4; Fig. 3B), similar to the 12 mV depolarization reported above using single-channel recording. In a bath containing 10 mm TEA, the depolarization by GABA was larger (Fig. 3B), indicating that activation of K⁺ currents limits the effects of GABA (Monsivais et al. 2000).

GABA is depolarizing in the developing brain, but generally hyperpolarizing in mature neurons, due to a change in E_Cl (Cherubini et al. 1991). Using gramicidin patches we therefore examined the age dependence of E_GABA on chicks ranging from embryos 4 days before hatching (E17) to hatchlings up to 10 days old (P10). Figure 3C shows that E_GABA did not change with development over a period extending from 2 days before the ear canals open (at E19) to 12 days after (Saunders et al. 1973). The mean E_GABA over these ages was -25 ± 5 mV (n = 27). Given the maturity of the auditory system at these ages, it is therefore likely that a depolarized E_GABA is a normal component of NM in vivo. The mean resting potential of these cells was -66 ± 3 mV (n = 15) and showed no systematic change over this developmental period.

Excitatory effects of GABAergic transmission

To assess the functional consequences of GABAergic transmission, experiments were performed on GABAergic synaptic responses in which the recording pipettes were filled with 60 mm Cl⁻ (E_GABA = −25 mV), mimicking the level found in intact cells. IPSCs isolated in voltage clamp had submillisecond rise times, decayed with a time constant of about 25 ms (see Lu & Trussell, 2000), and were blocked by 94 ± 3 % (n = 7) and 94 ± 4 % (n = 6) with 10 μm bicuculline or SR95531, respectively. IPSPs measured in current clamp and elicited at 1 Hz peaked rapidly (< 1 ms) and had a biphasic decay (Fig. 4A, upper panel). The slow phase (24 ± 10 ms time constant, n = 13) is similar to the decay time of the IPSC (Lu & Trussell, 2000), suggesting that the major part of the voltage decay is determined by the decay time of the synaptic conductance. The initial, fast part of the decay was a strongly damped oscillation, reflecting active properties of the membrane (Zhang & Trussell, 1994; Monsivais et al. 2000). In particular, depolarization of NM neurons rapidly activates a low-threshold K⁺ conductance, resulting in a transient repolarization (Zhang & Trussell, 1994).

Figure 4.

Dual excitatory/inhibitory effect of IPSPs

A, subthreshold (upper panel) and suprathreshold (lower panel) depolarizing IPSPs stimulated at 1 Hz in two cells. Inset shows full height of spikes. B, facilitation of IPSPs induced by trains of 20 stimuli at 200 Hz at an interval of 500 ms in another cell where low-frequency IPSPs were subthreshold. Extent of facilitation in IPSC is shown in upper traces. IPSP trains (lower traces) reliably induced a single spike after 2 trials (lower panels, spikes marked with an asterisk). C, simultaneously activated EPSPs and IPSPs reveal inhibition. Left panels show trains of 20 stimuli at 300 Hz (indicated with small arrows at the bottom) applied to both glutamatergic and GABAergic axons, resulting in a 20 mV depolarizing plateau and few spikes. After blocking GABAergic synapses with 10 μm SR95531 (right panel), more and larger spikes were generated while the plateau decreased to about 10 mV. D, summary of the inhibitory effect of IPSP on plateau depolarization and spike generation in 7 cells. Spike probability was measured between stimulus nos 5 and 20, from 5-10 trials of trains at 200 or 300 Hz. ** Data illustrated in C. All recordings were made with 60 mm[Cl⁻]_i. In A and B, glutamatergic synaptic responses were blocked with 100 μm AP-5, 20 μm DNQX and 20 μm GYKI 52466.

Low-frequency IPSPs varied widely in amplitude, consistent with the quantal variation measured under voltage clamp (Fig. 4A, upper panel and see Lu & Trussell, 2000). In 11 of 20 neurons, IPSPs were occasionally large enough to elicit single action potentials (Fig. 4A, lower panel). Spontaneous IPSPs were never suprathreshold (12 cells, data not shown). This, plus the fact that larger IPSPs were likely to be due to synchronous activation of multiple fibres (Lu & Trussell, 2000), suggests that background or low-frequency activity of GABAergic fibres is not likely to induce action potentials routinely.

However, IPSPs in NM exhibit strong facilitation by high-frequency trains of activity (Lu & Trussell, 2000). For example, a train of 20 stimuli at 100-300 Hz led to a fourfold increase in peak IPSC amplitude, when tested by a train of stimuli 0.4 s after the conditioning train was terminated (Fig. 4B; see Lu & Trussell, 2000). As this facilitation could, in principle, enhance the reliability of spiking induced by IPSPs, we examined the effects of series of IPSP trains in current clamp, again with 60 mm[Cl⁻]_i. In the first two trains of 20 IPSPs, all IPSPs were subthreshold, even though they summated to a level roughly 3-4 times higher than the amplitude of the first IPSP in the train. However, the first IPSP in subsequent trains was consistently suprathreshold (Fig. 4B). This behaviour was observed in 4 of 9 neurons in which no spikes occurred with low-frequency shocks. Spikes never occurred later in the train. Rather, IPSPs late in the train became more irregular, due to the onset of desynchronization of vesicle release (Lu & Trussell, 2000). It is likely, therefore, that desynchronization, plus activation of strong outward K⁺ currents, prevents IPSPs from generating trains of spikes. These results indicate that repeated trains of IPSPs could be excitatory, signalling the onset of bursts of activity in GABAergic fibres.

Inhibition by depolarizing IPSPs

Monsivais et al. (2000) showed that stimulation of GABAergic inputs to NM could prevent current steps from inducing spikes. To examine how excitatory and inhibitory synaptic inputs interact, we delivered stimuli (from a single stimulus pipette) that activated GABAergic and glutamatergic fibres simultaneously at 200-300 Hz for 20 stimuli. We could then test the role of GABA by repeating the stimulus trial in the presence of a GABA receptor antagonist, SR95531 (10 μm). Action potentials quickly fell subthreshold during the trains in the absence of the receptor antagonist (Fig. 4C). Here, without GABA_A receptor antagonists, the plateau depolarization induced by the high-frequency train was 10−30 mV, varying presumably from differential recruitment of GABAergic axons. The spiking probability (P_S) late in the trains also varied widely (Fig. 4C). SR 95531 was then applied on the same group of neurons and its effects determined on both P_S and the plateau depolarization during the trains. As shown in Fig. 4C and D, blocking GABA_A receptors resulted in a decreased plateau depolarization during trains but an increased P_S for the last 15 stimuli. With GABA_A receptors blocked, the 200 or 300 Hz train produced a smaller plateau depolarization (P < 0.001) and increased P_S (P < 0.05; Fig. 4C and D). An increase in P_S with the application of SR 95531 suggests that although E_GABA is positive to threshold, GABA is largely inhibitory during high-frequency trains. It is unlikely that GABA_B receptors contribute to these effects, as baclofen, a GABA_B receptor agonist, produces a net enhancement of EPSP amplitude in NM (Brenowitz et al. 1998).

DISCUSSION

Depolarizing GABAergic transmission is characteristic of embryonic or young postnatal mammalian brain (Cherubini et al. 1991; Kandler & Friauf, 1995; Kotak et al. 1998), and is thought to be an outcome of a high expression of Na⁺-K⁺-Cl⁻ cotransporters, which elevate intracellular Cl⁻, and a reduced expression of K⁺-Cl⁻ cotransporters, which reduce intracellular Cl⁻ (Kakazu et al. 1999). In mammalian auditory brainstem, depolarizing GABA and glycine responses turn hyperpolarizing by the end of the first week after birth (Kandler & Friauf, 1995; Kotak et al. 1998; Ehrlich et al. 1999; Kakazu et al. 1999), at least several days before the onset of hearing. In chicks, the ear canals clear of fluids several days before hatching, and a newly hatched chick hears with nearly adult thresholds and spectrum (Saunders et al. 1973; Gray & Rubel, 1985). This fact, plus the complete absence of a systematic change in E_GABA by up to 10 days post-hatch, suggests that a depolarizing GABAergic synapse is a key component of mature NM.

Depolarizing inhibition

Synaptic currents whose reversal potential is negative to action potential threshold are necessarily inhibitory. However, even when the reversal potential is above threshold, IPSPs may still be inhibitory if their action is to decrease the effectiveness of excitatory signals without themselves generating excitation. A classic example is the primary afferent depolarization, in which depolarizing GABA responses suppress excitatory transmitter release (Eccles et al. 1962; Kennedy et al. 1980; Edwards, 1990). The depolarizing IPSPs in NM accomplish this apparently through a shunting mechanism which is enhanced by the large voltage-gated K⁺ conductances of these neurons (Monsivais et al. 2000). Although K⁺ currents oppose the depolarization produced by GABA (Fig. 3B), they markedly lower input resistance and thereby increase the inhibitory effect of GABA on EPSPs. Moreover, depolarization will quickly lead to inactivation of Na⁺ conductance, thus effectively raising spike threshold (Zhang & Jackson, 1995). Indeed, Koyano et al. (1996) showed that 50 % of somatic Na⁺ current in NM was inactivated at −58 mV. Thus, GABA should produce a significant reduction in available Na⁺ channels.

Excitatory actions of GABA

The ability of IPSPs in NM to evoke action potentials seems counterproductive. It is not likely that such excitation could provide well-timed firing of NM neurons, at least in relation to the phase of sound or of firing by other inputs. SON, the major source of GABA in NM, is probably driven by neurons of nucleus angularis (Lachica et al. 1994; Monsivais et al. 2000), a cochlear nucleus involved in level coding that does not exhibit strong phase-locking to sound stimuli (Warchol & Dallos, 1990). Irregular activity of SON could therefore produce firing in NM, but would not be likely to generate much more activity than would spontaneous activity of auditory nerve fibres. However, we have shown here that NM could be excited by serial bursts of activity from GABAergic fibres. During such bursts of activity, IPSCs are initially synchronized, but then transmitter release degrades into steady asynchronous activity (Lu & Trussell, 2000). Suprathreshold responses in NM were therefore concentrated early in the burst of IPSPs, before the depolarization became ‘smoothed’ by asynchronous release. The functional significance of transient excitation by GABA is unclear, and will require an understanding of what natural signals drive SON activity, and what is the best-frequency relationship between nucleus angularis, SON and the targets of the SON.