Control of a Depolarizing GABAergic Input in an Auditory Coincidence Detection Circuit

Abstract

Neurons in the chicken nucleus laminaris (NL), the third-order auditory neurons that detect the interaural time differences that enable animals to localize sounds in the horizontal plane, receive glutamatergic excitation from the cochlear nucleus magnocellularis (NM) and GABAergic inhibition from the ipsilateral superior olivary nucleus. Here, we study metabotropic glutamate receptor (mGluR)- and GABA_B receptor (GABA_BR)-mediated modulation of synaptic transmission in NL neurons. Gramicidin-perforated recordings from acute brain stem slice preparations showed that the reversal potential of the GABAergic responses in NL neurons was more depolarized than the spike threshold. Activation of the GABAergic input produced a mix of inhibitory and excitatory actions in NL neurons. The inhibitory action is known to be critical in improving the acuity of temporal processing of sounds. The excitatory action, however, would reduce the phase locking fidelity of NL neurons in response to their excitatory inputs from the NM. We show that activation of presynaptic mGluRs or GABA_BRs by either exogenous agonists or synaptically released neurotransmitters reduced the GABAergic responses, preventing the excitatory action of GABA while leaving the inhibitory action intact. Unlike most CNS synapses, the glutamatergic transmission in the NL was not modulated by either mGluRs or GABA_BRs, indicating that fixed (nonmodulatory) excitatory inputs to the NL may be optimal for coincidence detection. This study contributes to our understanding of how selective neuromodulation is achieved to suit a particular function of neuronal circuits in the brain.

INTRODUCTION

Sound localization depends on two major cues, interaural time difference (ITD) and interaural intensity difference (IID), which are encoded by distinct neuronal circuits in the central auditory system. The ITD coding circuits formed among lower auditory brain stem nuclei in mammals and birds are similar, and the ITD coding process relies on faithful transfer of phase-locked excitation (i.e., neuronal firing to a specific phase of a sinusoidal signal) through various auditory brain stem nuclei. Bushy cells in the anteroventral cochlear nucleus (AVCN, in mammals) or the nucleus magnocellularis (NM, in birds) receive phase-locked excitation from the auditory nerve. In turn, these cells send bilaterally segregated phase-locked excitation to neurons in the medial superior olive (MSO, in mammals) or the nucleus laminaris (NL, in birds), where neurons perform coincidence detection of the inputs from the two ears and transform ITD into a place code (firing of neurons at a certain location within the nucleus). This allows the animals to localize sounds in the horizontal plane.

Excitation, however, is not the sole player for precise time coding and ITD processing. In mammals, sharply timed glycinergic inputs that originate from the medial and lateral nuclei of the trapezoid body set up the temporal sensitivity of MSO neurons to the physiologically relevant ITDs (Grothe 2003; McAlpine 2005; but also see Joris and Yin 2007; Zhou et al. 2005). Elimination of the inhibition shifts the ITD curve and produces ambiguity in the ITD coding (Brand et al. 2002; Pecka et al. 2008). In birds, neurons in the superior olivary nucleus (SON), which are driven by excitatory inputs from the NL and the nucleus angularis (NA), send feedback GABAergic inputs to the NA, NM, and NL (Burger et al. 2005a) (Fig. 1A). The GABA_A receptor (GABA_AR)-mediated depolarizing inhibition improves ITD coding by enhancing neuronal phase locking fidelity in the NM (Monsivais et al. 2000; Yang et al. 1999), providing a gain control for the excitatory inputs to the NL (Burger et al. 2005a; Dasika et al. 2005; Peña et al. 1996; Viete et al. 1997) and sharpening the coincidence detection window in NL neurons (Funabiki et al. 1998; Kuba et al. 2002a). Lesioning the SON results in diminished acuity of ITD coding (Nishino et al. 2008).

FIG. 1.

Experimental preparations. A: a schematic diagram showing the time-coding circuits in the avian auditory brain stem and the stimulation and recording locations. On entering the brain stem, the auditory nerve bifurcates. One branch innervates the cochlear nucleus magnocellularis (NM) and the other innervates the cochlear nucleus angularis (NA). Neurons in the NM send segregated bilateral excitatory projections to the nucleus laminaris (NL). Stimulation electrode 1 (stim 1) was placed in the ipsilateral NM to activate the dorsal excitatory input to the NL. Stim 2 was placed on fibers ventral and medial to the NL to activate the excitatory input from the contralateral NM. Neurons in the NL also receive GABAergic inputs originating primarily from the ipsilateral superior olivary nucleus (SON). Stim 3 was placed on SON fibers lateral to the NL to activate the GABAergic pathway. B: morphology of an NL neuron shown by biocytin staining after physiological recordings. The dorsal and ventral dendrites show similar dendritic configurations. The arrow indicates the axon coming off the cell body, and the asterisks indicate 3 neighboring NL neurons.

Interestingly, the effects of the inhibitory inputs are dynamic. Depending on the extent of activation of the GABAergic inputs under in vitro conditions, the excitability of neurons in the NM and the NL can be enhanced or reduced or remain unchanged when GABA is exogenously applied (Brückner and Hyson 1998; Hyson 2005). Regulation of the inhibition may therefore play pivotal roles in ITD coding. Under in vivo conditions, the strength of the inhibition is primarily determined by two factors: the excitatory driving force impinging on SON neurons and the modulation of GABA release onto targeted cells. The excitatory inputs to SON neurons originate from the NL and the NA and determine the level of spiking activity of SON neurons; the stronger the excitatory inputs, the higher the spiking activity of SON neurons (Burger et al. 2005a; Dasika et al. 2005; Peña et al. 1996; Viete et al. 1997). This study examines whether and how the inhibitory and the excitatory synaptic transmission in the NL are regulated by G protein–coupled receptors (GPCRs) that are activated by GABA or glutamate, the two known native neurotransmitters in the NL. Mechanisms underlying the modulation and its physiological role in ITD coding are also investigated.

METHODS

Slice preparation and in vitro whole cell recordings

Fertilized chicken eggs were purchased from Purdue University (West Lafayette, IN). Eggs were incubated using an RX2 Auto Turner (Lyon Electric, Chula Vista, CA) from embryo day 1 (E1) to E18, and a Clearview Brooder (Lyon Electric) from E19 to E21. Brain stem slices (250–300 μm in thickness) were prepared from E18–E21 chicken embryos and early hatchlings (P1), as described previously (Monsivais et al. 2000), with modification of the components of the artificial cerebrospinal fluid (ACSF) used for dissecting and cutting the brain tissue. The modified ACSF, which is a glycerol-based solution (Ye et al. 2006), contained (in mM) 250 glycerol, 3 KCl, 1.2 KH₂PO₄, 20 NaHCO₃, 3 HEPES, 1.2 CaCl₂, 5 MgCl₂, and 10 dextrose, pH 7.4, when gassed with 95% O₂-5% CO₂. The procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at Northeastern Ohio Universities Colleges of Medicine and Pharmacy and are in accordance with National Institutes of Health policies on animal use. Slices were incubated at 34–36°C for >1 h in normal ACSF containing (in mM) 130 NaCl, 26 NaHCO₃, 3 KCl, 3 CaCl₂, 1 MgCl₂, 1.25 NaH₂PO₄, and 10 dextrose. ACSF was constantly gassed with 95% O₂-5% CO₂ (pH 7.4). For recording, slices were transferred to a 0.5-ml chamber mounted on a Zeiss Axioskop 2 FS Plus microscope (Zeiss) with a ×40 water-immersion objective and infrared, differential interference contrast optics. The chamber was continuously superfused with ACSF (1–2 ml/min) by gravity. The microscope was positioned on the top center of an Isolator CleanTop II and housed inside a Type II Faraday cage (Technical Manufacturing, Peabody, MA). Recordings were performed at 34–36°C, controlled by a single channel temperature controller TC324B (Warner Instruments, Hamden, CT).

Patch pipettes were drawn on an Electrode Puller PP-830 (Narishige) to 1- to 2-μm tip diameter using borosilicate glass Micropipets (ID of 0.86 mm, OD of 1.60 mm; VWR Scientific, Seattle, WA). The electrodes had resistances between 3 and 7 MΩ when filled with a solution containing (in mM) 105 K-gluconate, 35 KCl, 5 EGTA, 10 HEPES, 1 MgCl₂, 4 ATP-Mg, and 0.3 GTP-Na, with pH of 7.2 (adjusted with KOH) and osmolarity between 280 and 290 mOsm. The Cl⁻ concentration (37 mM) in the internal solution approximated the physiological Cl⁻ concentration (36.3 mM in average, calculated based on data in Fig. 2) in NL neurons. Placement of the electrodes was controlled by a motorized micromanipulator MP-225 (Sutter Instruments, Novato, CA). The liquid junction potential was 10 mV, calculated using a software package by Barry (1994), and data were corrected accordingly.

FIG. 2.

The reversal potential (E_GABA) of GABA_A receptor (GABA_AR)-mediated inhibitory postsynaptic currents (IPSCs) in the NL is depolarizing. IPSCs were evoked with a train of synaptic stimulation (10 pulses at 100 Hz or 20 pulses at 200 Hz) and recorded at varying holding potentials that were changed in a 10-mV step. A: IPSCs recorded under the gramicidin-perforated patch configuration. The arrowhead indicates where the amplitude of the currents was measured to obtain the data plotted in B. B: the amplitude of IPSCs was plotted against the holding potentials. A linear regression was fitted, and E_GABA (the membrane potential at which the IPSC was 0) was −41 mV for this cell. C and D: the conventional whole cell recording using the same internal solution in another cell showed that the currents reversed at ∼1.6 mV, which is reasonably close to the calculated equilibrium potential for Cl⁻ (E_Cl, 0.4 mV). E: the average E_GABA was −36 mV (n = 6), which was significantly more depolarized than the average resting membrane potential (RMP) of all the cells recorded in this study (−59 mV, n = 112; P < 0.001, unpaired t-test). E_GABA is also more positive (by 8–13 mV) than the spike threshold of NL neurons (−49 to −44 mV, the range indicated by the double arrow; Kuba et al. 2002a). F: under the whole cell current-clamp mode and in the presence of ionotropic glutamate receptor blockers DNQX and APV, activation of the GABAergic pathway with a train of synaptic stimulation (5 pulses at 10 Hz) induced spikes. The RMP of this cell was −64 mV. Stimulus artifacts are truncated (A and C) or blanked (F) for clarity. In this and subsequent figures where current-clamp data are presented, recordings were performed without any somatic current injections. In the figures, we show means ± SE. Means and SD are reported in the text.

The voltage- and current-clamp experiments were performed with an AxoPatch 200B and an AxoClamp 2B amplifier, respectively (Molecular Devices, Union City, CA). In each voltage-clamp recording, series resistance was compensated by ∼80%, and cells were clamped at a membrane potential of −60 mV unless indicated otherwise. Before each synaptic stimulation protocol was applied, a 5-mV hyperpolarizing command (duration of 5 ms) was given to monitor series resistance and input resistance during the experiment. Data were low-pass filtered at 3–10 kHz and digitized using a Data Acquisition Interface ITC-18 (Instrutech, Great Neck, NY) at 20 kHz. Recording protocols were written and run using the acquisition and analysis software Axograph, version 4.9 (Molecular Devices, Union City, CA).

All chemicals and drugs were obtained from Sigma (St. Louis, MO) except for (±)-1-aminocyclopentane-trans-1,3-dicarboxylic acid (tACPD), 3,5-dihydroxyphenylglycine (3,5-DHPG), (2S,2′R,3′R)-2-(2′,3′dicarboxycyclopropyl)glycine (DCG-IV), l-(+)-2-amino-4-phosphonobutyric acid (L-AP4), 3-[[(3,4-dichlorophenyl)methyl]amino] propyl] diethoxymethyl)phosphinic acid (CGP 52432), (2S)-2-amino-2-[(1S,2S)-2-carboxycycloprop-1-yl]-3-(xanth-9-yl) propanoic acid (LY341495), (RS)-α-cyclopropyl-4-phosphonophenylglycine (CPPG), dl-threo-β-benzyloxyaspartic acid (dl-TBOA), and (3S)-3-[[3-[[4-(trifluoromethyl)benzoyl]amino]phenyl]methoxy]-l-aspartic acid (TFB-TBOA), which were obtained from Tocris (Ballwin, MO). All drugs were bath-applied except for muscimol, which was applied with pressure ejection (puff application) by using a multichannel picospritzer (General Valve, Fairfield, NJ). Muscimol (10 μM) was prepared in ACSF containing DNQX (50 μM) and APV (100 μM), antagonists for ionotropic glutamate receptors [AMPA and N-methyl-d-aspartate (NMDA) receptors, respectively]. Puff electrodes were prepared using the same pulling methods as producing recording electrodes except that the puff electrodes had larger tip diameter (2–5 μm). The puff electrode was placed above and lateral to the cell at a distance of 50–100 μm. Positive pressure (30–70 kPa, duration of 200 ms) was used to eject the muscimol-containing solution.

Synaptic stimulation and recordings of synaptic responses

Extracellular synaptic stimulation was performed using concentric bipolar electrodes with a tip core diameter of 127 μm (World Precision Instruments, Sarasota, FL). Neurons in the NL receive GABAergic inhibitory inputs from the ipsilateral SON (Burger et al. 2005a; Lachica et al. 1994; Yang et al. 1999). To activate the GABAergic pathway, the stimulation electrode was placed using a Micromanipulator NMN-25 (Narishige) in the area immediately lateral to the NL where the ipsilateral SON fibers travel to innervate the NL (Fig. 1A). To activate both the GABAergic and the glutamatergic pathways, the stimulation electrode was placed in an area dorsal and lateral to the NL, where fibers from the ipsilateral NM and SON fibers are mixed. Square electric pulses of 200-μs duration were delivered through a Stimulator A320RC (World Precision Instruments). The standard stimulus was a train stimulation (10 Hz, 5 pulses) at the intensity of 0.1–2.5 mA (average of ∼0.5 mA). Optimal stimulus parameters were selected for each cell to give rise to the maximal postsynaptic currents or potentials.

Evoked inhibitory postsynaptic currents (IPSCs) were recorded in the presence of DNQX (50 μM) and APV (100 μM), antagonists for ionotropic glutamate receptors. The standard train of synaptic stimulation was repeated 6–12 times (once every 5 or 10 s) under each experimental condition. The raw traces were averaged off-line, and the peak values of IPSCs were measured. The average of the peak values of the IPSCs was considered as one data point, representing the averaged IPSC under the experimental conditions. TTX (1 μM), a sodium channel blocker, was added when recording miniature IPSCs (mIPSCs). These methods have been established in our previous studies (Lu 2007; Lu et al. 2005). Evoked excitatory postsynaptic currents (EPSCs) were elicited with electrical shocks delivered to either the ipsilateral NM or the fibers originating from the contralateral NM and innervating the ventral neuropils of the NL (Fig. 1A). EPSCs were recorded in the presence of GABA_A receptor blocker SR-95531 (gabazine, 10 μM) or bicuculline (20 μM). The morphology of some recorded cells was shown by adapting a method by Hamam and Kennedy (2003). Biocytin (0.1–0.5%) (Sigma) was added to the internal solution used for whole cell recordings. After physiological recordings, brain slices were fixed overnight in 4% paraformaldehyde. Biocytin was visualized with avidin-biotin peroxidase and diaminobenzidine (Fig. 1B).

Gramicidin-perforated recordings

A fresh stock solution of gramicidin was prepared using DMSO (5 mg/ml) before each experiment. The final concentration of gramicidin in the recording internal solution was ∼25 μg/ml, prepared fresh every 90 min. The internal solution contained (in mM) 140 KCl, 5 EGTA, 10 HEPES, 1 MgCl₂, and 0.5 CaCl₂, with pH of 7.2 (adjusted with KOH) and osmolarity around 290 mOsm. The liquid junction potential was 5 mV, and data were corrected accordingly. Under voltage-clamp mode, a gigaohm seal was obtained, and series resistance (Rs) was monitored continuously by applying a 5-mV hyperpolarizing test pulse once every 10 s. In 15–30 mins Rs decreased to <50 MΩ, at which time recordings were started. During the process of membrane perforation, whole cell break-ins can be noticed by an abrupt drop in Rs, and subsequently, the reversal potential of GABA_ARs (E_GABA) measured under this condition is close to the equilibrium potential for Cl⁻ (E_Cl = 0.4 mV), calculated with the Nernst Equation based on the Cl⁻ concentration inside the recording electrode (143 mM) and outside the cell (141 mM) at 35°C. Following perforated patch recordings, whole cell configuration was obtained to confirm that the measurements give rise to the correct E_Cl. Summed IPSCs were evoked with high-frequency stimulation (10 pulses at 100 Hz or 20 pulses at 200 Hz), and obtained at different holding potentials in the range of –100 to +30 mV, with an increment or decrement step of 10 mV. The peak current amplitude was plotted against the holding potentials. A line fitting was performed, and the intercept voltage value (where I = 0) was defined as the E_GABA. To minimize a potential accumulation of intracellular Cl⁻ and thus a shift of E_GABA in the positive direction (Frech et al. 1999), we ran the protocols multiple times whenever possible and alternated the direction of varying the holding potentials. An average E_GABA was obtained in such cases.

Statistical analyses were performed using Excel (Microsoft, Redmond, WA) and Statview (Abacus Concepts, Berkeley, CA), and graphs were constructed in Igor (Wavemetrics, Lake Oswego, OR). Means and SD are reported in the text (n in parentheses indicates number of cells), and means and SE are shown in figures. The presentation of SE in figures is primarily for the purpose of clear illustration of the means. ANOVA post hoc Fisher's test was used for statistical analyses unless indicated otherwise.

RESULTS

Reversal potential for somatic GABA currents (E_GABA) in NL neurons is depolarizing

The polarity of a synaptic input is key to understanding its impact on neuronal functions. In contrast to MSO neurons (the mammalian homolog of NL), where the glycinergic inputs are hyperpolarizing (Smith et al. 2000), the GABAergic inputs to NL neurons in birds are depolarizing in late embryos (Funabiki et al. 1998), and they remain so during maturation into hatchlings (Hyson et al. 1995). The E_GABA in NL neurons, however, remains unknown. E_GABA defines the mechanisms whereby activation of the GABAergic pathway affects neuronal activity. The combination of the resting membrane potential (RMP), the spike threshold, and E_GABA determines whether the GABAergic input is purely inhibitory and/or can be excitatory. Thus we used the gramicidin-perforated patch recording technique (Kyrozis and Reichling 1995) to measure E_GABA in NL neurons. Inhibitory postsynaptic currents (IPSCs) in response to single or low-frequency synaptic stimulations were largely variable in amplitude. We therefore used a high-frequency train stimulation (10 pulses at 100 Hz or 20 pulses at 200 Hz) to activate the GABAergic pathway to the NL. Summed IPSCs were recorded at varying holding potentials in the presence of DNQX (50 μM) and APV (100 μM), antagonists for ionotropic glutamate receptors (Fig. 2). The IPSCs were completely blocked by gabazine (10 μM, data not shown), a selective antagonist for GABA_ARs. GABA_BRs may also be activated by synaptically released GABA. Activated presynaptic GABA_BRs could reduce the amplitude of the IPSCs (as shown in Fig. 7); however, E_GABA would not be affected by this effect because E_GABA is determined purely by the chloride concentrations inside and outside the cells. Activated postsynaptic GABA_BRs are generally coupled to activation of K⁺ channels, giving rise to membrane hyperpolarization (Calver et al. 2002). However, the effect is slow, and even if it exists in NL neurons, it would not contribute significantly to the fast GABA_AR-mediated currents we recorded. Other effects of postsynaptic GABA_BRs, such as modulation of Ca²⁺ channels, are also common. Elimination of the fast IPSCs by gabazine does not necessarily rule out a functional expression of GABA_BRs at the postsynaptic site.

The results show that E_GABA in NL neurons was −36 ± 10 mV (n = 6), which was significantly more depolarized than the average RMP of all the cells recorded in this study (−59 ± 4 mV, n = 112; P < 0.001, unpaired t-test). E_GABA was also more positive (by 8–13 mV) than the spike threshold of NL neurons (−49 to −44 mV, Kuba et al. 2002a); thus activation of the GABAergic pathway to the NL induced spikes (Fig. 2F). The depolarized E_GABA is unlikely to be a developmental phenomenon (see discussion). Spikes induced by GABA, should they occur in vivo, would reduce phase-locking fidelity and directly disturb coincidence detection in the NL. We show below that presynaptic metabotropic receptors activated by either exogenous agonists or synaptically released neurotransmitters control the strength of the GABAergic inputs and effectively prevent GABA-induced spikes.

Activation of GABA_BRs suppresses IPSCs in NL neurons via presynaptic action

The GABA_BR functions as an autoreceptor, modulating GABA release through a use-dependent feedback mechanism (Ulrich and Bettler 2007). Neurons in the NL receive inhibitory inputs primarily from the ipsilateral SON. Moreover, the expression of GABA_BRs in these neurons is very high (Burger et al. 2005b). To test the hypothesis that GABA_BRs modulate GABAergic transmission in the NL, we examined the effects of specific GABA_BR agonist baclofen on the IPSCs evoked with the standard train stimulation (5 pulses at 10 Hz). Baclofen (100 μM) significantly reduced IPSCs to 27 ± 20% of the control in NL neurons (Fig. 3, A and B, solid bars; P < 0.001; n = 7; control: −341 ± 179 pA; baclofen: −98 ± 86 pA; wash: −246 ± 202 pA). This effect is independent of the occurrence of synaptic failures because baclofen produced similar inhibition of IPSCs when synaptic failures were excluded in the calculation (Fig. 3B, open bars). Baclofen also significantly increased the failure rate of GABAergic transmission (Fig. 3C; P < 0.01; n = 7; control: 5 ± 9%; baclofen: 40 ± 27%; wash: 12 ± 18%; note the large variations in failure rates among cells, and the low failure rate in the example cell shown in Fig. 3A), suggesting presynaptic actions of GABA_BRs. The amplitude of the IPSCs partially recovered after washout of baclofen.

FIG. 3.

Activation of GABA_BRs by baclofen reduces IPSCs via a presynaptic action. A: averaged IPSCs obtained before, during, and after baclofen application. There were few synaptic failures in this example cell, so they were not observed in the averaged IPSCs. B and C: baclofen significantly reduces the amplitude of IPSCs, independent of the occurrence of synaptic failures (solid bars: including failures; open bars: excluding failures). C: baclofen increases the failure rate of synaptic transmission (n = 7). D: the effects of baclofen on miniature IPSCs (mIPSCs) of NL neurons. D₁: representative mIPSCs recorded in 1 cell. Baclofen significantly reduces the frequency of mIPSCs (D₂, P < 0.001, n = 6), without affecting the amplitude (D₃). Cumulative distribution of interevent interval (IEI) and amplitude of mIPSCs: 2,597 and 843 events for control (thin line) and baclofen (thick line), respectively. E: baclofen does not produce any noticeable effects on I-muscimol (n = 5), the currents evoked by puff application of GABA_AR agonist muscimol (10 μM), excluding the possibility of a postsynaptic action of GABA_BRs. Inset: 2 superimposed current traces obtained before and during baclofen (thicker line) application. F: analysis of the coefficient of variation (CV) of the IPSCs recorded under baclofen or a GABA_AR antagonist gabazine (1–2.5 μM). Partial block of GABA_ARs by gabazine reduces the amplitude of IPSCs but leaves the 1/CV² unaffected. Baclofen reduces the amplitude and decreases the 1/CV² of IPSCs (n = 7), suggesting a presynaptic action. In this and subsequent figures, *, **, and *** indicate P < 0.05, P < 0.01, and P < 0.001, respectively (ANOVA post hoc Fisher's test unless indicated otherwise). Cells were held at −60 mV for the voltage-clamp experiments.

Besides a strong expression in postsynaptic NL neurons, GABA_BRs are expressed presynaptically in the GABAergic terminals that impinge on NL neurons (Burger et al. 2005b), and they can modulate synaptic transmission through either presynaptic or postsynaptic mechanisms (Calver et al. 2002). To determine the action loci of GABA_BRs in modulating GABAergic transmission in the NL, we examined the effects of baclofen on the mIPSCs, the postsynaptic currents evoked by puff application of a GABA_AR agonist muscimol (I-muscimol), the coefficient of variation (CV) of IPSCs, and the paired-pulse ratio (PPR) of IPSCs, all of which may indicate the action loci of GABA_BRs. Modulation of the frequency, but not the amplitude, of mIPSCs implies a presynaptic mechanism. Conversely, modulation of the amplitude, but not the frequency, implies a postsynaptic mechanism. Finally, modulation of both implies a mechanism involving both pre- and postsynaptic elements of the synapse (Chu and Moenter 2005; Giustizieri et al. 2005). Unchanging parameters, however, do not necessarily exclude presynaptic modulation, because some drugs may influence transmission by inhibiting voltage-gated Ca²⁺ channels (VGCCs) on presynaptic terminals, without affecting Ca²⁺- and action potential–independent release of neurotransmitters (Gereau and Conn 1995; Lu 2007).

The results indicate a presynaptic action of GABA_BRs (Fig. 3, D–F). Baclofen significantly reduced the frequency of mIPSCs to 38 ± 21% of the control (Fig. 3D₂; P < 0.001, n = 6), without affecting the amplitude of mIPSCs (Fig. 3D₃; −58 ± 20, −58 ± 20, and −59 ± 24 pA for control, baclofen, and wash, respectively). Baclofen also did not produce any noticeable effects on I-muscimol (i.e., the postsynaptic current in response to puff application of muscimol; Fig. 3E; n = 5). Because muscimol (10 μM) directly activates postsynaptic GABA_ARs, bypassing presynaptic elements for synaptic transmission, the results excluded the possibility of a postsynaptic action of GABA_BRs on GABA_AR-mediated responses. Analysis of the CV can also be indicative of the action loci of drug effects. An increase in CV (and thus a decrease in 1/CV² of postsynaptic currents/potentials) indicates a presynaptic mechanism (Faber and Korn 1991; Tzounopoulos et al. 2007). We validated this analysis by examining the effects of the GABA_AR antagonist gabazine on the IPSCs of NL neurons. Partial block of GABA_ARs by gabazine (1–2.5 μM) reduced the amplitude of IPSCs (22 ± 17% of the control) but left 1/CV² unaffected (98 ± 72% of the control, n = 7; Fig. 3F), confirming a postsynaptic action of gabazine. Conversely, baclofen reduced the IPSC amplitude to 27 ± 20% and decreased the 1/CV² of the IPSCs to 40% of the control (n = 7), suggesting a presynaptic action of baclofen. Surprisingly, the PPR, measured at an interval of 100 or 50 ms, remained unchanged during baclofen application (control: 1.21 ± 0.31; baclofen: 1.31 ± 0.48; paired t-test, P > 0.05, n = 6). This discrepancy is possibly caused by the large variation in the amplitudes of IPSCs among trials, suggesting that PPR might not be a reliable indicator at this synapse to distinguish pre- from postsynaptic actions of GPCRs.

Activation of mGluRs suppresses IPSCs via presynaptic actions

Neurons in the NL receive glutamatergic inputs from the NM. The firing rate of NM neurons is among the highest in the brain, with a spontaneous firing rate of 50–100 Hz, and even higher rates when sound stimuli are present (Fukui et al. 2006; Warchol and Dallos 1990). High firing rates are also observed in NL neurons, with a spontaneous firing rate of 30–80 Hz and a sound-evoked firing rate of 200–400 Hz (Nishino et al. 2008). Glutamate can activate a variety of mGluRs, as well as ionotropic receptors. To date, eight members of mGluRs have been identified, which are further divided into three groups (group I: mGluR1 and 5; group II: mGluR2 and 3; and group III: mGluR4, 6, 7, and 8) based on their homology, pharmacology, and signal transduction pathways (Kew and Kemp 2005). Nonspecific and specific agonists for each group have been developed. For example, group I mGluRs can be activated by the agonist 3,5-DHPG at an EC₅₀ of 6–60 μM. DCG-IV selectively activates group II mGluRs at an EC₅₀ of 0.1–0.3 μM. Finally, L-AP4 activates mGluR4, 6, and 8 at an EC₅₀ <1 μM, but a much higher concentration (>100 μM) is needed to activate mGluR7 (Cartmell and Schoepp 2000). Each of the three groups of mGluRs, especially presynaptic group II and III mGluRs, can modulate synaptic transmission as auto- or heteroreceptors (Cartmell and Schoepp 2000).

Morphological results indicate that the excitatory inputs to the NL synapse primarily on dendrites, with some inputs on the cell bodies (Parks et al. 1983). The GABAergic inputs to the NL also spread over both the cell bodies and the dendrites of NL neurons (Code et al. 1989; Parks et al. 1983). It is conceivable that some glutamatergic and GABAergic terminals are close, and thus activation of mGluRs on GABAergic terminals by synaptically released glutamate would be possible. Hence, we predicted that the GABAergic transmission in the NL was also modulated by mGluRs. Our results described below confirmed our prediction (Fig. 4). We activated mGluRs by bath applying a nonspecific mGluR agonist tACPD (100 μM) or group-specific mGluR agonists (3,5-DHPG, DCG-IV, and L-AP4 for groups I, II, and III, respectively). The amplitude of the IPSCs was significantly reduced by tACPD, independent of the occurrence of synaptic failures (Fig. 4E, P < 0.001, n = 6). The failure rate was not affected by the application of tACPD (control: 2 ± 5%, tACPD: 20 ± 25%, P = 0.067; Fig. 4F). Because tACPD activates both group I and II mGluRs and does not activate group III mGluRs effectively, we further tested the effects of specific and potent agonists for each group. At a concentration ≥3 times higher than their EC₅₀, DCG-IV (2 μM, n = 8) and L-AP4 (10 μM, n = 6) significantly reduced the amplitude of IPSCs (Fig. 4, G and I). The failure rate was increased significantly from 3 ± 5 to 29 ± 22% by DCG-IV. Application of L-AP4 did not affect the failure rate (control: 1 ± 1%; L-AP4, 14 ± 7%; P = 0.058). In addition, 3,5-DHPG (200 μM) did not have any significant effects on either the IPSC amplitude or the failure rate (n = 7). Based on these results, we conclude that groups II and III, but not group I, mGluRs are involved in modulating GABA release onto the NL.

FIG. 4.

GABAergic transmission in NL neurons is modulated by group II and III but not group I mGluRs. A–D: averaged IPSCs obtained under conditions of control, mGluR agonists, and wash. (±)-1-Aminocyclopentane-trans-1,3-dicarboxylic acid (tACPD) is a nonspecific mGluR agonist, and (2S,2′R,3′R)-2-(2′,3′dicarboxycyclopropyl)glycine (DCG-IV), l-(+)-2-amino-4-phosphonobutyric acid (L-AP4), and 3,5-dihydroxyphenylglycine (3,5-DHPG) are specific agonists for groups II, III, and I, respectively. Synaptic failures were obvious during application of tACPD. IPSC traces from 4 different cells. E and F: tACPD significantly reduces the amplitude of IPSCs (solid bars: including failures; open bars: excluding failures), without affecting the failure rate (n = 6). G and H: DCG-IV (n = 8) significantly reduces the amplitude of IPSCs and increases the failure rate. I and J: L-AP4 (n = 6) significantly reduces the amplitude of IPSCs, without affecting the failure rate. K and L: 3,5-DHPG (n = 7) does not affect either parameter.

To further examine the action loci of group II and III mGluRs on GABAergic transmission, we performed the same experiments as shown in Fig. 3, D–F, using mGluR agonists. The results suggested presynaptic actions of group II and III mGluRs (Fig. 5). DCG-IV (2 μM) produced a small (18%) but statistically significant reduction in the frequency of mIPSCs (Fig. 5A₂; P < 0.05, n = 6), without affecting the amplitude (Fig. 5A₃; control: −59 ± 5 pA; DCG-IV: −64 ± 13 pA; wash: −64 ± 8 pA), had no reliably noticeable effects on I-muscimol (Fig. 5B; n = 5), and reduced the 1/CV² of IPSCs (Fig. 5C; n = 8). L-AP4 (10 μM) affected neither the frequency nor the amplitude of mIPSCs (Fig. 5D; n = 8), rendering the conclusion of the action loci based on this analysis uncertain. However, L-AP4 did not have any noticeable effects on I-muscimol (Fig. 5E; n = 5), excluding the possibility of a postsynaptic action. Furthermore, the CV analyses for the effects of L-AP4 on IPSCs (n = 6) also supported a presynaptic action (Fig. 5F). Similar to baclofen, DCG-IV and L-AP4 did not alter PPR (control for DCG-IV: 1.74 ± 0.49; DCG-IV: 1.79 ± 0.59; control for L-AP4: 1.66 ± 0.29; L-AP4: 1.66 ± 0.44; paired t-test, P > 0.05, n = 5 for each drug), providing further support that PPR might not be a reliable indicator at this synapse that can be used to distinguish prefrom postsynaptic action of GPCRs.

FIG. 5.

The combination of 3 lines of indirect evidence suggests presynaptic actions of group II and III mGluRs. A: DCG-IV produces a small but consistent reduction in the frequency of mIPSCs (P < 0.05, n = 6), without affecting the amplitude (2,841 and 2,486 events for control and DCG-IV, respectively). B: DCG-IV does not produce any noticeable effects on I-muscimol (n = 5). C: CV analyses for the effects of DCG-IV on IPSCs (n = 8). D: L-AP4 does not affect the frequency and the amplitude of mIPSCs (n = 8, 3,856 and 3,908 events for control and L-AP4, respectively). E: L-AP4 has no noticeable effects on I-muscimol (n = 5). F: CV analyses for the effects of L-AP4 on IPSCs (n = 6). See Fig. 3 for detailed legend. For clarity, cumulative fractions for mIPSC frequency (A₂ and D₂) <0.90 are not shown, and the portions of the 2 curves that are not shown overlap each other closely.

Activation of GABA_BRs or mGluRs eliminates GABA-induced spikes and regulates NL excitability

To test the hypothesis that the dual modulation of the GABAergic transmission by GABA_BRs and mGluRs prevents the excitatory action of GABA, we performed whole cell current-clamp experiments in which the effects of baclofen and tACPD on GABA-induced spikes were examined. From the values of E_GABA (ranging from −49 to −23 mV) and the known Cl⁻ concentration in the ACSF (141 mM), we calculated the corresponding intracellular Cl⁻ concentration (22–59 mM) using the Nernst equation. We then included 47 mM Cl⁻ inside the recording pipette (which is likely to be within the physiological range, but has a higher probability of generating GABA-induced spikes than the normal internal solution containing 37 mM Cl⁻), and used the current-clamp configuration to test the effects of activating GABA_BRs or mGluRs on action potentials (APs) generated in NL neurons by stimulation of the GABAergic pathway (Fig. 6). Glutamatergic responses were blocked with DNQX (50 μM) and APV (100 μM). The GABA-induced spike probability varied widely among cells. Activation of either GABA_BRs or mGluRs almost completely eliminated GABA-induced spikes. Moreover, baclofen and tACPD failed to change the intrinsic firing properties, including the threshold currents, in response to prolonged somatic current injections in NL neurons (n = 7, data not shown), suggesting that both postsynaptic GABA_BRs and mGluRs do not change NL excitability. These results indicate that modulation of the GABAergic transmission by presynaptic GABA_BRs and mGluRs effectively prevents the excitatory action of GABA in the NL.

FIG. 6.

Activation of GABA_BRs or mGluRs almost completely eliminates GABA-induced spikes in NL neurons and regulates NL excitability. A: baclofen reduces the amplitude of GABA-mediated inhibitory postsynaptic potentials (IPSPs) and eliminates GABA-induced spikes. Inset: 1 IPSP and 1 spike (indicated by #) at an enlarged time scale. B: tACPD has similar effects. C: baclofen (100 μM) almost completely eliminates GABA-induced spikes (n = 9). AP, action potential. D: tACPD (100 μM) produces similar effects (n = 7). E: regulation of the excitability of NL neurons. Somatic current injections (50 Hz, 50 pulses) at a level slightly above the threshold current were used to cause spikes and some subthreshold responses. Synaptic stimulation (100 Hz, 20 pulses) of SON fibers activates the GABAergic pathway, resulting in a membrane depolarization of ∼15 mV in amplitude, and a suppression of spike generation during and after the SON stimulation (2 spikes indicated by #). Baclofen (n = 4) reduces the inhibitory effects of SON stimulation, allowing more spikes to be generated after SON stimulation. Application of tACPD (100 μM, n = 3), or of both baclofen and tACPD (n = 3), produces similar results (data not shown). Stimulus artifacts are blanked for clarity. DNQX (50 μM) and APV (100 μM) were present in all recordings.

Neurons in the NL respond to excitatory inputs from the NM in a phase-locked manner, i.e., cells fire spikes that correspond to a certain phase of a sinusoidal sound wave (Köppl and Carr 2008; Nishino et al. 2008). We mimicked such responses using current injections into the cell bodies of NL neurons and examined the effects of GABA_BRs and/or mGluRs. Figure 6E shows that somatic current injections (50 Hz, 50 pulses) at a level of slightly above the threshold current elicited spikes as well as some subthreshold responses in an NL neuron. Activation of the GABAergic pathway by a train stimulation (100 Hz, 20 pulses) caused a GABA_AR-mediated membrane depolarization and inhibited spike generation, consistent with previous observations in the NL (Yang et al. 1999). Such depolarizing GABAergic inputs can exert a strong inhibitory effect on cell excitability via a combination of shunting inhibition, inactivation of voltage-gated Na⁺ channels, and activation of low voltage-gated K⁺ channels (Monsivais and Rubel 2001). GABA_BRs and/or mGluRs regulated the dynamics of the inhibitory effects and therefore the cellular excitability of NL neurons. Baclofen (n = 4) reduced the inhibitory effect on NL excitability, allowing more spikes to be generated after the SON stimulation. This is similar to a previous observation, in which noradrenaline reduced a deopolarization plateau elicited by synaptic stimulation in neurons of the medial nucleus of trapezoid body, thus causing more spike firing (Leao and von Gersdorff 2002). Application of tACPD (100 μM, n = 3), or of both baclofen and tACPD (n = 3) produced similar results (data not shown). Therefore the firing properties of NL neurons are regulated by the GABAergic inputs via GABA_ARs, and the dynamics of the inhibitory strength are fine-tuned by mGluRs and GABA_BRs.

Endogenous activity of GABA_BRs and mGluRs suppresses GABA release at NL

Although activation of presynaptic GABA_BRs and mGluRs by exogenous agonists reduced GABA release onto the NL, a more important question related to physiological functions of these receptors is whether they can be activated by synaptically released neurotransmitters. To determine whether endogenous GABA and glutamate modulate GABAergic transmission in the NL in vitro, we examined the effects on IPSCs of a GABA_BR antagonist (CGP 52432, 10 μM) and a cocktail of mGluR antagonists (4 μM LY341495 plus 10 μM CPPG) at their saturating concentrations that completely block GABA_BRs (Lanza et al. 1993) and the majority of mGluRs (Schoepp et al. 1999), respectively. Temporally summed IPSCs were evoked by a train of synaptic stimulation at a frequency of 100 or 200 Hz (20 pulses). The stimulation frequencies approximate the discharge rates of NM and SON neurons in vivo (Fukui et el. 2006; Nishino et al. 2008). Figure 7A shows averaged IPSCs obtained in one NL neuron under conditions of control, CGP 52432 (10 μM), and washout. CGP 52432 increased the amplitude of IPSCs, and a nearly complete recovery of the responses was observed after washout. Pooled data show that CGP 52432 significantly increased the amplitude of IPSCs to 164 ± 49% of the control (Fig. 7B; P < 0.01, n = 6), and after washout, the amplitude returned to 103 ± 39% of the control. The IPSC amplitude was −757 ± 546, −1,064 ± 538, and −823 ± 464 pA for control, CGP 52432, and washout, respectively.

FIG. 7.

Endogenous GABA and glutamate suppress the GABAergic transmission in NL neurons. A: averaged IPSCs elicited by a train of synaptic stimulation (100 Hz, 20 pulses) under conditions of control, GABA_BR antagonist 3-[[(3,4-dichlorophenyl)methyl]amino]propyl] diethoxymethyl)phosphinic acid (CGP 52432; 10 μM), and washout. B: CGP 52432 significantly increases the IPSC amplitude (P < 0.01, n = 6) measured at 20 ms after the last stimulus pulse. C: to study endogenous activity of presynaptic mGluRs in modulating GABA release in the NL, the stimulation electrode was placed in an area dorsal and lateral to the NL to activate both the GABAergic and the glutamatergic pathways. Under control conditions, both excitatory postsynaptic currents (EPSCs) and IPSCs were elicited in response to the same electrical stimulus. Antagonists for ionotropic glutamate receptor (50 μM DNQX plus 100 μM APV) eliminated the EPSCs. D: averaged IPSCs under conditions of control, mGluR antagonists (2S)-2-amino-2-[(1S,2S)-2-carboxycycloprop-1-yl]-3-(xanth-9-yl) propanoic acid (LY341495; 4 μM) plus (RS)-α-cyclopropyl-4-phosphonophenylglycine (CPPG; 10 μM), and washout. E: LY341495 and CPPG significantly increase the IPSC amplitude (P < 0.001, n = 6). Stimulus artifacts in A and D are blanked for clarity.

To study endogenous activity of presynaptic mGluRs in modulating GABA release in the NL, the stimulation electrode was placed in an area dorsal and lateral to the NL to activate both the GABAergic and glutamatergic pathways. Concurrent activation of GABAergic and glutamatergic pathways to NL neurons was observed in response to the same single stimulus (Fig. 7C). To enhance the chance of detecting endogenous mGluR activity in brain slices, blockade of glutamate uptake systems was used to increase glutamate accumulation in the synaptic cleft and thus possible activation of mGluRs on GABAergic terminals by glutamate spillover (Shimamoto et al. 1998; Tsukada et al. 2005). Indeed, with perfusion of dl-TBOA (50 μM) and TFB-TBOA (10 μM), both of which are blockers of glutamate transporters, we observed endogenous activity of mGluRs in suppression of IPSCs. Figure 7D shows representative IPSCs obtained in one NL neuron. Pooled data show that LY341495 plus CPPG significantly increased the amplitude of IPSC to 130 ± 8% of the control (Fig. 7E; P < 0.001, n = 6), and after washout, the amplitude returned to 81 ± 9% of the control. The IPSC amplitude was −713 ± 569, −907 ± 692, and −416 ± 249 pA for control, LY341495 plus CPPG, and washout, respectively.

GABA_BRs and mGluRs do not modulate glutamatergic transmission in NL neurons

NL neurons receive segregated dorsal versus ventral glutamatergic inputs from the ipsilateral and contralateral NM, respectively. The bipolar dendrites of one NL neuron are approximately symmetrical in morphology in terms of dendritic length, area, and general branching complexity. Synaptic responses (EPSCs) are approximately symmetrical in terms of amplitude and kinetics (Funabiki et al. 1998). Therefore the number of excitatory inputs from the two NMs and subunit compositions for ionotropic glutamate receptors in NL neurons are expected to be about the same between the dorsal and the ventral inputs. These morphological and physiological properties imply that the two segregated excitatory inputs to the NL may have equal synaptic strength at the same distance along the dorsal and ventral dendrites. We hypothesized that the excitatory inputs to the NL were subject to modulation by presynaptic GABA_BRs and mGluRs, and the strength of modulation was equal between the two segregated inputs from ipsilateral and contralateral NM. We stimulated either the dorsal or the ventral glutamatergic pathway and recorded EPSCs in the presence of bicuculline (20 μM) or gabazine (10 μM), which are antagonists for GABA_ARs. EPSCs were relatively stable in amplitude among trials in response to a repeated single stimulus. To our surprise, the results did not support our hypothesis (Fig. 8). Neither baclofen (100–200 μM, n = 7) nor tACPD (100 μM, n = 5) produced significant changes in EPSCs (Fig. 8, A, B, D, and E; P > 0.05, paired t-test). The amplitude and the profile of synaptic depression of EPSCs evoked by a train stimulation (100 Hz, 10 pulses) were also not affected (Fig. 8, C and F). Because tACPD does not effectively activate group III mGluRs, we tested group III agonist L-AP4 (10 μM) and also found no effects (data not shown). This lack of modulation of EPSCs in the NL is unlikely because of a saturated activation of endogenous GABA_BRs and mGluRs, because endogenous activity of these receptors is hardly evoked with low-frequency stimulation in slice preparations, let alone being saturated. Although the glutamatergic transmission was not subject to modulation by GABA_BRs and mGluRs, an activity-dependent modulation of EPSCs in the form of synaptic depression was obvious (Fig. 8, C and F), which has been shown to improve coincidence detection in NL neurons (Kuba et al. 2002b).

FIG. 8.

Neither GABA_BRs nor mGluRs modulate the glutamatergic transmission in NL neurons. A: baclofen does not affect the EPSCs of an NL neuron. B: pooled data show that baclofen has no effects on EPSC amplitude (P > 0.05, paired t-test, n = 7). C: baclofen has little effect on EPSCs elicited with a train of synaptic stimulation (100 Hz, 10 pulses). D--F: tACPD has similar effects (P > 0.05, paired t-test, n = 5). The GABA_AR blocker gabazine (10 μM) was present in all recordings.

DISCUSSION

The major findings of this study are as follows: 1) the GABAergic input to the NL is depolarizing and can be both inhibitory and excitatory; 2) presynaptic GABA_BRs and mGluRs control the synaptic strength of the GABAergic input, preventing generation of GABA-induced spikes; and 3) neither GABA_BRs nor mGluRs modulate the excitatory inputs to the NL. To the best of our knowledge, this is the first evidence of a glutamatergic synapse in the vertebrate CNS that is not subject to modulation by either autoreceptors or GABA_B heteroreceptors. We focus on the depolarized E_GABA in NL neurons, the selective modulation of the GABAergic but not the glutamatergic inputs, and the functional significance in ITD coding.

GABAergic input to NL neurons is depolarizing

The polarity of a synaptic input is key to understanding its effects on cell activity. Both GABAergic and glycinergic responses are depolarizing and excitatory during early development and become hyperpolarizing and inhibitory after maturation (Ben-Ari 2002; Kandler et al. 2002; Owens and Kriegstein 2002; Sanes and Friauf 2000). It is not uncommon, however, that GABA and glycine mediate depolarizing and excitatory responses in the adult brain (Marty and Llano 2005), including in the central auditory system (Awatramani et al. 2005; Golding and Oertel 1996; Price and Trussell 2006; Turecek and Trussell 2001). Such a depolarization can enhance or reduce firing depending on the physiological status of the cells and is critically involved in synaptic plasticity (Tzounopoulos et al. 2004), indicating that the effects mediated by traditionally believed “inhibitory” transmitters may be dramatically regulated to suit their specific functions. For two main reasons, the observation of a depolarizing GABAergic input to the NL is unlikely a developmental phenomenon. First, depolarizing responses mediated by GABA have been shown in chicken hatchlings up to posthatching day 14 (Hyson et al. 1995). Chickens are precocial animals; they are able to hear before they hatch (Saunders et al. 1973). By the time of hatching, hearing threshold (Saunders et al. 1973) and behavior (Gray and Rubel 1985) including sound localization are fairly mature. Second, both intrinsic and synaptic neuronal properties of NL neurons show adult-like features by E18 (Gao and Lu 2008). By this age, the E_GABA in SON and nonauditory neurons in the same preparations is already hyperpolarizing (Monsivais and Rubel 2001).

The depolarizing GABAergic input to the NL improves coincidence detection of excitatory inputs by decreasing the duration of excitatory postsynaptic potentials and thus sharpening the time window for convergent excitatory inputs to coincide (Funabiki et al. 1998). Through a combination of shunting, inactivation of sodium channels, and activation of low-threshold potassium channels, such a depolarizing inhibition is more potent than a conventional hyperpolarizing one (Monsivais and Rubel 2001). Because E_GABA was more positive than the spike threshold, however, activation of the GABAergic input could generate spikes in our in vitro conditions (Figs. 2 and 6). The GABAergic inputs to the NL can therefore produce a mix of inhibitory and excitatory actions, as observed in the NM (Lu and Trussell 2001). GABA-induced spikes are unlikely to phase-lock to the excitatory inputs from the NM (Lachica et al. 1994) and are therefore potentially disadvantageous for phase locking fidelity and for ITD processing in NL neurons. This necessitates the fine control of the GABAergic strength and prevention of the excitatory action of GABA.

Presynaptic GABA_BRs and mGluRs exert selective modulation on the GABAergic but not the glutamatergic inputs to NL

We present cellular mechanisms involving presynaptic GABA_BRs and mGluRs that control the synaptic strength of the GABAergic input to the NL. Neurons in the NL receive inhibitory inputs from the ipsilateral SON and have an abundance of GABA_BRs. These receptors are located both postsynaptically in NL neurons and presynaptically on the GABAergic terminals (Burger et al. 2005b). Presynaptic GABA_BR-mediated modulation of GABA release is a common feedback mechanism throughout the CNS (Ulrich and Bettler 2007). Our results provided physiological support for presynaptic GABA_BRs in the GABAergic terminals that impinge on NL neurons. Functions of postsynaptic GABA_BRs in NL neurons, however, remain to be determined. Similar to GABA_BRs, mGluRs, especially group II and III members, can regulate transmitter release via affecting presynaptic VGCCs (Cartmell and Schoepp 2000; Schoepp 2001). Although the action loci for these receptors are at the presynaptic terminal, the exact mechanisms underlying their modulation of GABA release in the NL seem to be different. We found that L-AP4 had no effects on mIPSCs, whereas baclofen and DCG-IV reduced the frequency without affecting the amplitude of mIPSCs. This discrepancy might be interpreted by differential effects of different metabotropic receptors on upstream and downstream cellular events of Ca²⁺ influx through VGCCs (Giustizieri et al. 2005; Lu 2007; Zheng and Johnson 2003). Group III mGluRs may cause presynaptic modulation of GABA release evoked only by neuronal activity, perhaps because of a separate modulation of two disparate pools of GABA vesicles (Sara et al. 2005). Because GABA_BRs and group II mGluR agonists affected both evoked IPSCs and mIPSCs, additional mechanisms for the modulation mediated by these receptors could exist (Kelly et al. 2009).

Although modulation of synaptic transmission by metabotropic receptors is found in many neuronal systems, the inputs that are modulated in a particular neuronal circuit, the type of metabotropic receptors involved, and the mechanisms underlying the modulation vary broadly among different systems. Neuromodulation mediated by GPCRs (specifically GABA_BRs and mGluRs), as well as the lack of neuromodulation, may play a pivotal role in ITD coding, as summarized in a schematic diagram (Fig. 9). Presynaptic GABA_BRs on the auditory nerve terminals modulate glutamate release in the NM, enhancing synaptic efficacy and rendering NM neurons capable of following high-frequency synaptic inputs (Brenowitz and Trussell 2001; Brenowitz et al. 1998; Otis and Trussell 1996). Interestingly, there is no autoregulation of glutamate release by mGluRs in the NM (Otis and Trussell 1996). Modulation of the GABAergic transmission in the NM by presynaptic GABA_BRs (Lu et al. 2005) and mGluRs (Lu 2007) prevents generation of GABA-induced spikes, helping to preserve fidelity of temporal information encoded by NM neurons and ensuring faithful transfer of phase-locked excitation to the NL. Although similar effects of these receptors on the GABAergic transmission occur in the NL, mGluR members that mediate the modulation of the GABAergic transmission in the NL are different from those in the NM, in that group II and III mGluRs are involved in the NL, whereas all three groups are involved in the NM (Lu 2007).

FIG. 9.

A schematic diagram showing neuromodulation mediated by GABA_BRs and mGluRs in the avian auditory brain stem. Presynaptic GABA_BRs, but not mGluRs, on the auditory nerve (8th n.) terminals modulate glutamate release onto the NM, enhancing synaptic efficacy and rendering NM neurons capable of following high-frequency synaptic inputs (Brenowitz and Trussell 2001; Brenowitz et al. 1998; Otis and Trussell 1996). The GABAergic transmission in the NM is subject to a dual modulation by presynaptic GABA_BRs (Lu et al. 2005) and mGluRs (Lu 2007). This study showed a selective modulation of the GABAergic but not the glutamatergic transmission in the NL by presynaptic GABA_BRs and mGluRs.

Functional significance in ITD coding

Because synaptically released GABA and glutamate activate GABA_BRs and mGluRs and modulate GABA release in the NL (Fig. 7), fine regulation of the synaptic strength of the GABAergic input bears functional significance. Neurons in the NL fire maximally to glutamatergic inputs converging from the ipsilateral and contralateral NM at the same time. The time window for coincidence detection needs to be very brief to code the small ITDs in the chick (Hyson 2005; Köppl and Carr 2008). The coincidence detection window measured under GABA_AR activation is sharper than that measured without inhibition, implying that coincidence detection in the NL is improved by synaptic inhibition via the ionotropic GABA receptors (Funabiki et al. 1998). These two measurements represent the two extreme situations: one without and the other with full strength of inhibition. The area between these two windows constitutes the dynamic range of coincidence detection, the extent of which is likely to be regulated by mGluRs and GABA_BRs. Given the high discharge rate of NM neurons in vivo (Fukui et al. 2006; Warchol and Dallos 1990), a tonic activation of mGluRs may exist and regulate the basal level of GABA release. Feedback control through GABA_BRs could limit excessive GABA release, thereby maintaining a homeostatic range.

It is unusual that the glutamatergic transmission in the NL is not affected by GABA_BRs and mGluRs, given the universal modulatory effects of these receptors on glutamate synapses in the CNS (Cartmell and Schoepp 2000; Ulrich and Bettler 2007). We speculate that this lack of modulation may uniquely fit the NL's function in coincidence detection. The morphological symmetry of bilateral dendrites of NL neurons implies physiological symmetry between the two segregated excitatory inputs, which may be optimal for coincidence detection. Neurons of the NL show similar thresholds to tone-burst stimulation presented to either ear (Köppl and Carr 2008; Rubel and Parks 1975), suggesting that they have approximately equal sensitivity to their segregated excitatory inputs. The importance of balanced excitatory inputs to coincidence detectors can also be inferred from computer modeling studies in which the same membrane parameters are assigned to the two excitatory inputs (Agmon-Snir et al. 1998; Dasika et al. 2007; Grau-Serrat et al. 2003). The lack of modulation of the excitatory inputs may be one key to preserving physiological symmetry, because modulation of one but not the other excitatory input or the differential degree of modulation of the two inputs would lead to imbalanced excitation. A differential driving force may bias the NL firing toward the stronger input, reducing the acuity for sound localization. Taken together, maintenance of the overall balance of excitation and inhibition in NL neurons may rely primarily on regulation of the inhibitory inputs to the NL, and the selective regulation of the GABAergic transmission by GABA_BRs and mGluRs in the NL may be crucial to the maintenance of fine temporal information in the auditory brain stem.

GRANTS

This work was supported by National Institute on Deafness and Other Communication Disorders Grant R01 DC-008984 to Y. Lu.