Shift from depolarizing to hyperpolarizing glycine action in rat auditory neurones is due to age-dependent Cl⁻ regulation

Abstract

1. The inhibitory neurotransmitter glycine can elicit depolarizing responses in immature neurones. We investigated the changes in glycine responses and their ionic mechanism in developing neurones of the rat lateral superior olive (LSO), an auditory brainstem nucleus involved in sound localization.

2. Whole-cell and gramicidin perforated-patch recordings were performed from visually identified LSO neurones in brain slices and glycine was pressure applied for 3–100 ms to the soma. Glycine-evoked currents were reversibly blocked by strychnine. They were mostly monophasic, but biphasic responses occurred in ∼30% of P8-11 neurones in perforated-patch recordings.

3. In whole-cell recordings from P2-11 neurones, the reversal potential of glycine-evoked currents (E_Gly) was determined by the transmembranous Cl⁻ gradient and corresponded closely to the Nernst potential for Cl⁻, regardless of age. This indicates that Cl⁻ is the principle ion permeating glycine receptors, but is also consistent with a low relative (10–20%) permeability for HCO₃⁻. The Cl⁻ gradient also determined the polarity and amplitude of glycine-evoked membrane potential changes.

4. Leaving the native intracellular [Cl⁻] undisturbed with gramicidin perforated-patch recordings, we found a highly significant, age-dependent change of E_Gly from −46.8 ± 1.8 mV (P1-4, n = 28) to −67.6 ± 3.3 mV (P5-8, n = 10) to −82.2 ± 4.1 mV (P9–11, n = 18). The majority of P1–4 neurones were depolarized by glycine (∼80%) and spikes were evoked in ∼30%. In contrast, P9–11 neurones were hyperpolarized.

5. In perforated-patch recordings, E_Gly was influenced by the voltage protocol and the glycine application interval; it could be shifted in the positive and negative direction. For a given application interval, these shifts were always larger in P1–4 than in P8–11 neurones, pointing to less effective Cl⁻ regulation mechanisms in younger neurones.

6. Furosemide (frusemide), a blocker of cation-Cl⁻ cotransporters, reversibly shifted E_Gly in the negative direction in P2–4 neurones, yet in the positive direction in P8–10 neurones, suggesting the blockade of net inward and net outward Cl⁻ transporters, respectively.

7. Taken together, age-dependent changes in active Cl⁻ regulation are likely to cause the developmental shift from depolarizing to hyperpolarizing glycine responses. A high intracellular [Cl⁻] is generated in neonatal LSO neurones which decreases during maturation.

The mechanisms underlying the development and maturation of inhibitory synapses in the central nervous system (CNS) are still poorly understood. This is mainly due to the fact that inhibitory neurones are mostly interneurones which are diffusely distributed within a brain region. In the adult CNS, fast synaptic inhibition is mediated by the release of GABA or glycine and subsequent activation of GABA_A receptors (GABA_ARs) or glycine receptors (GlyRs). These receptors are ion channels permeable to Cl⁻ and, to a much lesser extent, to other anions such as bicarbonate (HCO₃⁻) (Bormann et al. 1987). A flux of Cl⁻ into neurones causes a hyperpolarization of the membrane potential and, hence, inhibition of neuronal activity. However, GABA and glycine depolarize neurones in many CNS areas during embryonic and neonatal development, e.g. in the hippocampus (Ben-Ari et al. 1989; Cherubini et al. 1990), cerebral cortex (Luhmann & Prince, 1991; Lo Turco et al. 1995), hypothalamus (Chen et al. 1996), brainstem (Kandler & Friauf, 1995b; Singer et al. 1998) and spinal cord (Wu et al. 1992; Reichling et al. 1994). The depolarizing action of GABA and glycine can be excitatory, i.e. generate spike activity and induce Ca²⁺ influx, probably via the opening of voltage-gated Ca²⁺ channels (VGCCs) (Reichling et al. 1994; Leinekugel et al. 1995; Obrietan & van den Pol, 1995; Owens et al. 1996; Flint et al. 1998). It has been suggested that the transiently depolarizing action of inhibitory transmitters and the resulting influx of Ca²⁺ may play a role in inhibitory synapse maturation and plasticity, similar to the situation seen at excitatory synapses (McLean et al. 1996; Ben-Ari et al. 1997). There is evidence that depolarizing GABA_A responses in developing neurones are caused by an efflux of Cl⁻ which is attributable to a high internal Cl⁻ concentration ([Cl⁻]_i), and it has been suggested that the high [Cl⁻]_i is generated by an active inward Cl⁻ transport involving transporter molecules (Rohrbough & Spitzer, 1996; Owens et al. 1996).

We use the lateral superior olive (LSO), a nucleus within the mammalian auditory brainstem, as a model system to study the development of glycinergic synapses. The LSO is a binaural nucleus that contributes to sound localization by processing interaural intensity differences (for review see Helfert et al. 1991). Adult LSO neurones receive excitatory, glutamatergic input from the ipsilateral ear via the ventral cochlear nucleus and inhibitory, glycinergic input from the contralateral ear via the medial nucleus of the trapezoid body (MNTB). The LSO is of particular interest for studying the development of inhibitory connections, because the input from the MNTB is exclusively inhibitory and tonotopically ordered (for review see Helfert et al. 1991). During development, there is a refinement of the MNTB- LSO projection which appears to be activity dependent and involves pruning of the afferent axon terminals of MNTB neurones (Sanes & Siverls, 1991; Sanes & Takacs, 1993) as well as of the postsynaptic dendritic trees of LSO neurones (Sanes & Chokshi, 1992; Sanes et al. 1992; Rietzel & Friauf, 1998). In the rat, glycinergic afferents reach their targets within the LSO on embryonic day (E) 18 (Kandler & Friauf, 1993), i.e. 4 days before birth, but the onset of hearing is not until postnatal day (P) 12. Glycinergic transmission from the MNTB to the LSO starts at E18 and is depolarizing until the end of the first postnatal week, after which it becomes hyperpolarizing (Kandler & Friauf, 1995b). A depolarizing action of glycine has also been demonstrated in perforated-patch recordings from the rat LSO using long bath applications of glycine (Backus et al. 1998). However, no difference was found in glycine-evoked responses between the first and the second postnatal week, but a new, HCO₃⁻-dependent biphasic response type was described. In gerbils, where auditory development appears to follow a similar time course to that in rats, depolarizing glycinergic transmission from MNTB to LSO was not described initially (Sanes, 1993), but recent data from whole-cell patch-clamp recordings suggest that GABAergic and glycinergic postsynaptic currents in the medial, but not the lateral, part of the LSO have reversal potentials more positive than the resting membrane potential (V_m) during the first two postnatal weeks (Kotak et al. 1998). Thus, the developmental changes in glycine action in the LSO are still controversially discussed.

Here, we studied the development and ionic mechanisms of glycine-evoked currents of LSO neurones in brain slices from neonatal rats employing the patch-clamp technique. We found that the transmembranous Cl⁻ gradient of developing neurones determines the polarity of their glycine response. In gramicidin perforated-patch recordings with intact [Cl⁻]_i (Abe et al. 1994; Kyrozis & Reichling, 1995), a negative shift in the reversal potential of glycine-evoked responses (E_Gly) occurs. Accordingly, many neonatal neurones were depolarized by glycine, changing to hyperpolarization in the second postnatal week. LSO neurones actively regulate their [Cl⁻]_i in an age-dependent manner using furosemide-sensitive Cl⁻ transporters: in neonatal neurones, inward Cl⁻ transport generates a high [Cl⁻]_i whereas in older neurones, outward Cl⁻ transport keeps [Cl⁻]_i low. Some of these data have been presented in abstract form (Ehrlich et al. 1998).

METHODS

Brain slice preparation

Young Sprague-Dawley rats (P1-11) were used. Animals were housed in our facility and treated in compliance with the current German animal protection law. All protocols were approved by the local animal care and use committee (RP Darmstadt, Germany). Rat pups were killed by decapitation and their brains were dissected in a chilled (4°C) solution containing (mm): NaHCO₃, 25; KCl, 2.5; NaH₂PO₄, 1.25; MgCl₂, 1; CaCl₂, 2; D-glucose, 260; sodium pyruvate, 2; myo-inositol, 3; kynurenic acid, 1 (continuously bubbled with 95 % O₂-5 % CO₂, pH 7.4). Coronal sections (300 μm) of the auditory brainstem were obtained using a VT-1000 vibratome (Leica, Bensheim, Germany). Slices were transferred to extracellular solution containing (mm): NaCl, 125; NaHCO₃, 25; KCl, 2.5; NaH₂PO₄, 1.25; MgCl₂, 1; CaCl₂, 2; D-glucose, 10; sodium pyruvate, 2; myo-inositol, 3; ascorbic acid, 0.4 (continuously bubbled with 95 % O₂-5 % CO₂, pH 7.4), incubated at 36°C for 60 min, and stored at room temperature (20-25°C) until recording.

Electrophysiological recording

For whole-cell recordings, patch pipettes were pulled from filamented borosilicate glass capillaries (GB150F-8P, Science Products, Hofheim, Germany) on a horizontal puller (P-87, Sutter Instruments, Novato, CA, USA) and had resistances of 4–8 MΩ when filled with intracellular solution. We used pipette solutions with four different Cl⁻ concentrations ([Cl⁻]_p). One solution (132 mm Cl⁻) contained (mm): KCl, 130; Hepes, 10; EGTA, 5; MgCl₂, 1; Na₂ATP, 2; and Na₂GTP, 0.3 (pH 7.2 with KOH). In two other solutions (32 or 12 mm Cl⁻), 130 mm KCl was substituted by 100 mm potassium gluconate with 30 mm KCl, and by 120 mm potassium gluconate with 10 mm KCl, respectively. The fourth solution (4.2 mm Cl⁻) contained (mm): potassium gluconate, 135; Hepes, 10; EGTA, 1; MgCl₂, 2; CaCl₂, 0.1; and Na₂ATP, 2 (pH 7.2 with KOH). For perforated-patch recordings, patch pipettes were pulled from unfilamented borosilicate glass capillaries (GB150-8P, Science Products), frontfilled with gramicidin-free pipette solution (132 mm Cl⁻) for 45-60 s and then backfilled with the same solution that additionally contained 30-80 μg ml⁻¹ gramicidin. A stock solution of gramicidin (10 mg ml⁻¹) was prepared in dimethyl sulfoxide (DMSO) and was then diluted such that the pipette solution never contained more than 0.8 % DMSO.

Slices were transferred to a recording chamber and continually perfused with extracellular solution at room temperature and a rate of 3–4 ml min⁻¹. LSO neurones were visualized with DIC-infrared optics using a ×40, 0.75 NA water-immersion objective on an Axioscope FS microscope (Zeiss, Germany) equipped with a video camera system (Hamamatsu, Japan). For whole-cell recordings, a gigaseal was established and the membrane patch ruptured to gain low resistance access to the cell (series resistance: 17 ± 1 MΩ). For gramicidin perforated-patch recordings, the progress of perforation was monitored after formation of a gigaseal until the series resistance had stabilized to ≤30 MΩ (after 10-30 min). Recordings were made with an Axopatch-1D amplifier (Axon Instruments, Foster City, CA, USA) and signals were filtered at 1–10 kHz using the amplifier's 4-pole Bessel filter. Data were digitized at 2–5 times the filter frequency via a Digidata 1200 interface (Axon Instruments) and stored on a personal computer using pCLAMP software (Axon Instruments). Series resistance compensation was set to 70-80 % with a lag of 80-100 μs.

Drugs and drug application

Kynurenic acid, EGTA, Na₂ATP and Na₂GTP were obtained from Fluka (Neu-Ulm, Germany). Ascorbic acid, potassium gluconate, strychnine, glycine, gramicidin and furosemide were purchased from Sigma (Deisenhofen, Germany) and tetrodotoxin (TTX) was from RBI Biotrend (Cologne, Germany). All other chemicals were obtained from Merck (Darmstadt, Germany). Stock solutions of glycine and strychnine were stored at −20°C and diluted to working concentrations in extracellular solution before use. Furosemide (1 mm) was directly dissolved in extracellular solution before use.

Glycine (1 mm) was pressure applied via a wider-tip (opening diameter μ2-5 μm) patch pipette mounted on a multi-channel Picospritzer (General Valve Corp., Fairfield, NJ, USA). The application pipette was positioned over the soma of the recorded neurone. Glycine was applied at a low pressure (μ3 p.s.i.) to avoid movement of the tissue. The duration of the application pulses (3–100 ms, mostly ≤30 ms) was adjusted so that a well-measurable current response was evoked. All other drugs were applied via the bath.

Voltage protocols

The holding potential (V_h) was −70 mV between voltage steps and different experimental protocols. To determine E_Gly, we used three voltage protocols (‘random’, ‘up’, and ‘down’) that differed in the order in which the voltage steps were given. In the ‘random’ protocol, voltage steps were applied in randomized order, in the ‘up’ protocol in ascending order and in the ‘down’ protocol in descending order. In all protocols, voltage steps lasted 2–3 s and glycine was applied 500 ms after the onset of the step, when fast voltage-activated currents are negligible. Step voltages (V_s) ranged between −90 and +20 mV in whole-cell recordings and between −100 and −10 mV in perforated-patch recordings. The interval between the voltage steps was varied between 5 and 50 s. Long intervals (30 and 50 s) were used in young neurones (≤P4) to avoid shifts in E_Gly and to unambiguously determine E_Gly, shorter intervals (5 and 10 s) were used in experiments in which the susceptibility of neurones to shifts in E_Gly was investigated.

Data analysis

Data analysis was performed using Igor (WaveMetrics Inc., Lake Oswego, OR, USA) or Origin (Microcal Inc., Northampton, MA, USA) software packages. Current amplitudes (I) were measured as the difference between mean holding current and mean maximally evoked current in two time windows (5–8 ms duration), one positioned before the pulse and the second positioned at the peak response. To generate current-voltage relations (I-V relations), V_s values were corrected for the voltage errors due to incomplete series resistance compensation. However, when experimental protocols are described, uncorrected V_s values are given. We did not correct for liquid junction potentials (LJPs) for the following reasons. In whole-cell recordings, we applied high positive pressure to the pipette while penetrating the slice and found that the buildup of the LJP was very slow, i.e. much slower than the time needed for seal formation. Secondly, the V_m of LSO neurones measured with four different pipette solutions (132 mm Cl⁻: −59.4 ± 1.7 mV; 32 mm Cl⁻: −57.5 ± 1.3 mV; 12 mm Cl⁻: −58 ± 1.0 mV; 4.2 mm Cl⁻: −57.8 ± 2.5 mV) did not differ from each other (ANOVA, P = 0.8) and were comparable to V_m values obtained in sharp electrode recordings (Kandler & Friauf, 1995a), where LJP problems do not exist. In perforated-patch recordings, with little or no positive pressure on the pipette, the LJP amounted to μ3 mV and was therefore neglected. I-V relations were well fitted by a linear regression in most cases, and E_Gly was determined as the voltage at zero current of the regression line or the extrapolation of this line. In a few cases, E_Gly was determined by interpolation of the data points in the I-V relation. Grouped data are given as means ±s.e.m., and n is the number of cells. The current and voltage traces shown in the figures represent single events; likewise, values in the I-V relations represent single measurements. Statistical analysis between experimental groups was performed using either Student's two-tailed t test or analysis of variance (ANOVA). If applicable, Tukey's honestly significant difference (HSD) post hoc test was used in combination with ANOVA. The P values are given for each test, and significant differences are marked in the figures; *P≤ 0.05, **P≤ 0.01, ***P≤ 0.001.

RESULTS

Characterization of glycine-evoked currents in LSO neurones

We investigated the properties and development of glycine-evoked responses in LSO neurones of rats between P1 and P11. We applied glycine focally to the soma, because here the membrane potential and the intracellular milieu can be better controlled through the patch pipette than at more distant locations. Glycine applications were short in an attempt to mimic synaptic actions. Glycine responses were most probably generated by the recorded neurone rather than by network activity, because they were not altered by bath application of 1 μm TTX (n = 2, not shown) and they readily disappeared when the application pipette was moved away from the recorded cell. In whole-cell recordings with a pipette chloride concentration ([Cl⁻]_p) of 132 or 32 mm, glycine-evoked currents were reversibly blocked by bath application of 0.5 μm strychnine (98.7 ± 0.7 % block, n = 3, Fig. 1A) or 2 μm strychnine (95.3 ± 1.5 % block, n = 2).

Figure 1. Characterization of glycine-evoked currents in LSO neurones.

A, glycine-evoked currents were reversibly blocked by bath application of strychnine. Glycine-evoked currents obtained in a whole-cell recording of a P3 neurone with 32 mm[Cl⁻]_p and V_s =−70 mV before, during and after application of 0.5 μm strychnine. The glycine response recovered to 68 % of control after 20 min washout. B, E_Gly depended on the recording configuration. The upper panel shows glycine-evoked currents recorded in a P7 neurone in gramicidin perforated-patch configuration (left) and, after membrane rupture, in whole-cell configuration (right), at V_s =−50 and −70 mV. The corresponding I-V relations in the lower panel illustrate that E_Gly changed from −63.7 mV in perforated-patch recording (•, V_s =−80 to −20 mV, ‘up’ protocol) to −6.2 mV in the subsequent whole-cell recording (○, V_s =−80 to −10 mV, ‘up’ protocol). During whole-cell recording, the [Cl⁻] in the pipette and bath solution were about equimolar (132 vs. 133.5 mm). C, biphasic glycine-evoked currents were seen in perforated-patch recordings of some P8-11 neurones. The upper panel shows biphasic currents recorded from a P10 neurone at V_s =−90, −80 and −60 mV; glycine was applied for 20 ms. The same neurone displayed monophasic currents in response to briefer applications (5 ms, inset). Scale bars are identical in the main panel and inset. Dotted lines mark the time when current amplitudes were measured to determine E_Gly of phase 1 (−90.1 mV) and phase 2 (−75.3 mV). The lower panel shows a summary of all 9 neurones displaying biphasic glycine responses; ○, results from the neurone shown in the upper panel. E_Gly of phase 1 was always more negative than that of phase 2 (−80.6 ± 2.3 vs.−69.2 ± 3.6 mV).

When perforated-patch recordings were compared with whole-cell recordings obtained from the same neurone, the importance of [Cl⁻]_i in determining the size, polarity and reversal potential of glycine responses was demonstrated. This is shown in Fig. 1B, where a P7 neurone was initially recorded in perforated-patch configuration and E_Gly was −63.7 mV. When the membrane was ruptured underneath the patch, E_Gly changed very rapidly to −6.2 mV, close to the reversal potential for Cl⁻ (E_Cl) predicted by the Nernst equation (E_Cl =−0.3 mV for 132 mm[Cl⁻]_p and 133.5 mm external [Cl⁻]). Similar results were obtained in all 21 neurones tested where E_Gly changed to −2.7 ± 1.4 mV after membrane rupture. Because of this rapid change, it was easy to discriminate between a perforated-patch recording and an unwanted whole-cell recording occurring through a spontaneous membrane rupture.

Monophasic currents comprised the majority of glycine-evoked responses (see e.g. Figs 1A and B, 2A and B, and 4A-C). They were observed in all whole-cell recordings regardless of age (n = 44), all perforated-patch recordings between P1 and P7 (n = 39) and most of the perforated-patch recordings between P8 and P11 (n = 22 of 31). Additionally, we found biphasic glycine-evoked currents in about a third of perforated-patch recordings in P8-11 neurones (n = 9 of 31, Fig. 1C), consisting of an early and a later phase (phase 1 and phase 2) that had a different E_Gly. Reducing the duration of the glycine application by a factor of four, e.g. from 20 to 5 ms, resulted in monophasic currents (Fig. 1C, inset) that had the same E_Gly as phase 1 of the biphasic currents (n = 2). Phase 1 always reversed at a more negative potential than phase 2 (−80.6 ± 2.3 and −69.2 ± 3.6 mV, respectively, n = 9, Fig. 1C). Biphasic glycine responses appeared to consist of a fast, initial phase with a similar E_Gly and, possibly, the same ionic basis as monophasic responses, and a second phase ‘blending’ into the first. We did not investigate the ionic mechanisms of biphasic responses (see Discussion for a short account on biphasic responses), yet concentrated on monophasic responses in the following.

Figure 2. E_Gly corresponds closely to the Nernst potential for Cl⁻.

Glycine-evoked currents were recorded in whole-cell configuration with 12, 32 or 132 mm[Cl⁻]_p. E_Gly was determined and compared with E_Cl values predicted by the Nernst equation. A and B, the upper panels show glycine-evoked currents recorded in a P10 and a P9 neurone with 12 and 132 mm[Cl⁻]_p, respectively. The lower panels depict the corresponding I-V relations, with E_Gly amounting to −58.1 and −1.7 mV. V_s =−90 to −20 mV (‘up’ protocol) in A and V_s =−60 to +20 mV (‘random’ protocol) in B. C, semilogarithmic plot of E_Gly±s.e.m.versus[Cl⁻]_p; pooled data from two age groups are shown (P2-7, □, and P8-11, ▪, n in parentheses). In P2-7 neurones, E_Gly was 3.1 ± 1.9, −25.3 ± 2.2 and −59.6 ± 2.3 mV for 132, 32 and 12 mm[Cl⁻]_p, respectively. The slope of the linear fit to these data points (upper thin line) was +58.4 mV per 10-fold increase in [Cl⁻]_i. In P8-11 neurones, E_Gly was −0.1 ± 2.4, −33.7 ± 2.7 and −67.5 ± 4.6 mV for 132, 32 and 12 mm[Cl⁻]_p, respectively. The slope of the linear fit to these data points (lower thin line) was +60.9 mV per 10-fold increase in [Cl⁻]_i. These data for both age groups agree well with values predicted by the Nernst equation for Cl⁻ (thick line). Dashed and dotted curves represent the mixed ion reversal potentials calculated from the Goldman-Hodgkin-Katz equation using relative HCO₃⁻ permeabilities that are 11 and 33 % of the Cl⁻ permeability, respectively.

Figure 4. E_Gly changes during development.

E_Gly was determined from LSO neurones of different ages recorded in gramicidin perforated-patch configuration. A-C, the upper panels show examples of glycine-evoked currents obtained at different step potentials in a P3, a P7 and a P10 neurone. The lower panels depict the corresponding I-V relations, where E_Gly amounted to −38.4 mV (P3), −62.4 mV (P7) and −79.7 mV (P10). The current families and I-V relations shown here were obtained with the ‘up’ voltage protocol. In all cases, care was taken to unambiguously determine E_Gly (see Methods for details). V_s =−70 to −20 mV in A, −80 to −20 mV in B and −90 to −40 mV in C. D, E_Gly values from LSO neurones (n = 57, P1-11) were plotted against age; open circles mark results from the three neurones shown in A-C. An age-related negative shift of E_Gly was seen. E, grouped data show a highly significant shift of E_Gly from −46.8 ± 1.8 mV at P1-4 (n = 28) to −67.6 ± 3.3 mV at P5-8 (n = 10) to −82.2 ± 4.1 mV at P9-11 (n = 18, ▪, ANOVA, P < 0.001). A parallel shift in V_m was not observed (□); a significant difference was found only between P1-4 and P9-11 (ANOVA and Tukey's HSD post hoc test, P < 0.05). Average V_m values were −58.3 ± 0.8 mV at P1-4 (n = 24), −63.0 ± 2.4 mV at P5-8 (n = 9) and −62.7 ± 1.4 mV at P9-11 (n = 15).

E_Gly is determined by the transmembranous Cl⁻ gradient and corresponds closely to the Nernst potential for Cl⁻

Depolarizing GABA and glycine responses in neurones have been attributed to a number of mechanisms, e.g. an increase in the conductance to anions such as HCO₃⁻ (Staley et al. 1995; Perkins & Wong, 1996; Backus et al. 1998) or to cations (Andersen et al. 1980). Because it was suggested that a high [Cl⁻]_i is responsible for GABA-induced depolarizations in immature neurones (Rohrbough & Spitzer, 1996; Owens et al. 1996), we investigated the role of Cl⁻ in glycine-evoked responses during the first and second post-natal week, when a change from glycinergic depolarization to hyperpolarization occurs in the LSO (Kandler & Friauf, 1995b). In whole-cell recordings, we biased the [Cl⁻]_i to that of the pipette, using three different [Cl⁻]_p of 12, 32 and 132 mm, while the external [Cl⁻] was always 133.5 mm. Results of such experiments are illustrated in Fig. 2A and B, demonstrating that two neurones at about the same age, yet recorded with different [Cl⁻]_p, displayed a different E_Gly. As shown in Fig. 2C, E_Gly determined from neurones in the first (P2-7) and the second (P8-11) postnatal week depended on the [Cl⁻]_p. Linear fits to the data yielded a change of +58.4 and +60.9 mV per 10-fold increase in [Cl⁻]_p for P2-7 and P8-11 neurones, respectively. These values correspond well to the values for E_Cl predicted by the Nernst equation (at 25°C, +59.2 mV change per 10-fold increase in [Cl⁻]_p). However, E_Gly in whole-cell recordings from younger neurones was slightly more positive than E_Cl. Taking a HCO₃⁻ permeability of GlyRs into account, we also calculated mixed reversal potentials on the basis of the Goldman-Hodgkin-Katz equation. We assumed HCO₃⁻ permeabilities of 11 % (Bormann et. al, 1987) and 33 % (arbitrarily chosen) relative to that of Cl⁻ (Fig. 2C; dashed and dotted curve, respectively). The external [HCO₃⁻] was 25 mm and the [HCO₃⁻]_i, as set to a constant level by the intracellular pH, was calculated to be 17.7 mm. From this, we estimate that the experimentally obtained E_Gly values are also consistent with a relative HCO₃⁻ permeability of GlyRs in the range of 10-20 %. Taken together, our results support the hypothesis that Cl⁻ is the principle ion mediating glycine responses in LSO neurones, and that the Cl⁻ permeability of GlyRs is barely, if at all, altered during development.

Polarity and amplitude of glycine responses are determined by [Cl⁻]_i

To elucidate the physiological relevance of differences in [Cl⁻]_i, we recorded glycine-evoked membrane potential changes while controlling the [Cl⁻]_i via the pipette. For all [Cl⁻]_p tested, neurones responded to glycine application as expected from the calculated values for E_Cl (and from the measured values for E_Gly). Glycine strongly depolarized neurones recorded with 132 mm[Cl⁻]_p, and spike activity occurred on top of the depolarization in all neurones tested (n = 7, P2-9). This spike activity ranged from three spikes generated within 30 ms to 17 spikes generated within 560 ms (Fig. 3A). All neurones recorded with 32 mm[Cl⁻]_p showed a less pronounced depolarization (n = 9, P3-11), and in four of these, 1–3 spikes were triggered (P3-8, Fig. 3B). When 12 mm[Cl⁻]_p was used, neurones were slightly depolarized (n = 2 of 9), showed no response (n = 3 of 9) or were slightly hyperpolarized (n = 4 of 9, P3-9, Fig. 3C). When the [Cl⁻]_p was reduced further to 4.2 mm, glycine hyperpolarized all neurones tested (n = 4, P2-3, Fig. 3D). The results show that the polarity and the amplitude of glycine responses are determined by the [Cl⁻]_i. They also suggest that, under physiological conditions, glycine may excite neurones by inducing spikes when E_Cl is above spike threshold.

Figure 3. [Cl⁻]_i determines the polarity and amplitude of glycine responses.

Examples of glycine-evoked membrane potential changes observed in four neurones recorded in whole-cell configuration with different [Cl⁻]_p. A, a P9 neurone recorded with 132 mm[Cl⁻]_p had a calculated E_Gly of −0.3 mV; it was strongly depolarized and a train of 17 spikes was evoked within 560 ms. B, a P8 neurone recorded with 32 mm[Cl⁻]_p had a calculated E_Gly of −36.7 mV; it was less depolarized than the neurone shown in A and a single spike was evoked. C, a P3 neurone recorded with 12 mm[Cl⁻]_p had a calculated E_Gly of −61.9 mV; it showed a very small hyperpolarization. D, a P3 neurone recorded with 4.2 mm[Cl⁻]_p had a calculated E_Gly of −88.9 mV; it was hyperpolarized. Spikes in A and B are truncated. Same time scale for A-D.

Perforated-patch recordings reveal a negative shift of E_Gly between P1 and P11

Previous studies have yielded controversial results concerning whether, when and where in the LSO a shift from depolarizing to hyperpolarizing glycine action occurs (Sanes, 1993; Kandler & Friauf, 1995b; Backus et al. 1998; Kotak et al. 1998). The controversy may be due to different experimental approaches. Since our whole-cell recordings demonstrated the importance of [Cl⁻]_i in determining properties of glycine-evoked currents, we had to record from LSO neurones with intact [Cl⁻]_i to correctly assess their E_Gly. We therefore used the gramicidin perforated-patch technique and characterized age-dependent changes in E_Gly. Since we had observed that short application intervals led to shifts in E_Gly (cf. Figs 6 and 7), sufficiently long application intervals were chosen to avoid this problem (see Methods). Three examples of glycine-evoked currents recorded from a total of 57 cells are shown in the upper panels of Fig. 4A-C. In early postnatal neurones (P1-4, Fig. 4A), E_Gly was generally more positive than at the turn of the first to the second postnatal week (P5-8, Fig. 4B) and it became even more negative in older neurones (P9-11, Fig. 4C). A plot of E_Gly against animal age illustrates the negative shift of E_Gly during development (Fig. 4D). The negative shift of E_Gly from −46.8 ± 1.8 mV at P1-4 (n = 28) to −67.6 ± 3.3 mV at P5-8 (n = 10) and to −82.2 ± 4.1 mV at P9-11 (n = 18) was highly significant (ANOVA, P < 0.001, Fig. 4E). A parallel and equally large shift in V_m was not seen. Values for V_m ranged from −58.3 ± 0.8 mV (P1-4, n = 24) to −63.0 ± 2.4 mV (P5-8, n = 9) to −62.7 ± 1.4 mV (P9-11, n = 15); a significant difference was found only between P1-4 and P9-11 (ANOVA and Tukey HSD post hoc test, P < 0.05). These results suggest that the relatively positive values for E_Gly in young neurones, compared to older neurones, are due to a higher [Cl⁻]_i and that [Cl⁻]_i decreases during development. Furthermore, the shift from an E_Gly that is more positive than V_m (at P1-4) to an E_Gly that is more negative than V_m (at P9-11) indicates that [Cl⁻]_i is actively and differentially regulated throughout development (otherwise E_Gly should be equal to V_m, independent of age).

Figure 6. E_Gly is influenced by voltage protocol and application interval.

Comparison of E_Gly determined in perforated-patch recordings with ‘up’ and ‘down’ protocols and various Δt values (see Methods for details). A and B, examples of glycine-evoked currents in a P2 and a P8 neurone (upper panels) and the resulting E_Gly (lower panels) obtained with an ‘up’ protocol (•), followed by a ‘down’ protocol (○), for Δt = 10 s. In the P2 neurone, E_Diff =E_Gly(up) - E_Gly(down) was 24.2 mV, whereas in the P8 neurone, E_Diff amounted to only 4.0 mV. V_s =−90 to −30 mV in A and −90 to −20 mV in B. C, group data show that E_Diff was significantly larger in P1-4 than in P8-11 neurones (14.1 ± 1.7 mV, n = 17 vs. 3.8 ± 1.8 mV, n = 12; Student's t test, P < 0.01). D, plot of E_Diff as a function of Δt demonstrates that shorter intervals increased whereas longer intervals decreased E_Diff in both P1-4 (□) and P8-11 (▪) neurones (n in parentheses). For a given Δt, E_Diff was always greater in P1-4 neurones than in P8-11 neurones.

Figure 7. E_Gly can be shifted in the positive and negative direction by voltage protocols.

A, E_Gly(up) and E_Gly(down) deviate from the ‘real’E_Gly. The left panel shows an example of a P2 neurone where E_Gly(up) (•) and E_Gly(down) (○) are plotted for different Δt values (perforated-patch recording). For Δt = 50 s, E_Gly(up) and E_Gly(down) were almost identical (−50.3 and −49.2 mV, respectively) and the arithmetic mean was taken to represent ‘real’E_Gly (−49.8 mV). For shorter Δt values, E_Gly(up) shifted in the negative direction and E_Gly(down) in the positive direction when compared to ‘real’E_Gly. The right panel shows that the average deviation (E_Shift) for Δt = 10 s amounted to −4.3 ± 1.8 mV for the ‘up’ and +10.3 ± 0.7 mV for the ‘down’ protocol in all 12 P1-4 neurones investigated. B and C, ‘fading’, a reduction of glycine-evoked current amplitudes in response to repetitive glycine application, was prominent at P1-4, but not at P10 (perforated-patch recordings). The right panels show currents evoked by the first and the last of 20 repetitive glycine applications (Δt = 10 s) at a V_s that was either ≈20 mV positive to ‘real’E_Gly (top traces) or ≈20 mV negative to ‘real’E_Gly (bottom traces). The left panels depict the time course and magnitude of the reduction of normalized current amplitudes (Normalized I). The P2 neurone shown in B had a ‘real’E_Gly of −59.0 mV; the glycine-evoked current amplitudes decreased by 62 % at V_s =−40 mV and by 66 % at V_s =−80 mV. The P10 neurone shown in C had a ‘real’E_Gly of −80.0 mV, the current amplitude decreased by 19 % at V_s =−60 mV and by 14 % at V_s =−100 mV.

In P1-4 neurones, we found quite variable values for E_Gly, ranging from −21.0 to −68.0 mV (Fig. 4D). To investigate whether this variability was due to regional differences of E_Gly along the tonotopic axis of the LSO (cf. Kotak et al. 1998), we compared E_Gly values from P1-4 neurones in the medial (−46.1 ± 4.8 mV, n = 5), central (−45.6 ± 4.7 mV, n = 6) and lateral part (−55.3 ± 5.3 mV, n = 6) and found no significant difference (ANOVA, P = 0.32). However, when comparing two to three neurones in the same slice, E_Gly was indeed 10-20 mV more negative in the more lateral parts of the LSO in 4 of 6 slices investigated. Nonetheless, with the variability of E_Gly and the small number of cells investigated in each subregion, we have no conclusive evidence for a regional gradient of E_Gly in the rat LSO.

Glycine depolarizes the majority of P1-4 neurones and can be excitatory

To examine how the developmental shift in E_Gly affects the membrane potential of LSO neurones in response to glycine application, we performed voltage recordings in the gramicidin perforated-patch configuration. We especially wanted to see whether glycine can excite early postnatal neurones, i.e. elicit spike activity. Even though the average E_Gly was well above V_m in P1-4 neurones, some neurones had a very positive E_Gly, and a few others showed an E_Gly close to, or even below, −60 mV (Fig. 4D). We found this diversity being reflected in the neurones’ voltage responses: the vast majority of P1-4 neurones (n = 14 of 18) were depolarized by glycine. In some cases (n = 5 of 18), a train of 2–10 spikes was generated within 50-530 ms on top of a μ15-20 mV depolarization (Fig. 5Aa). In other cases (n = 9 of 18), the depolarization was below spike threshold (μ10 mV, Fig. 5Ab). Few of these P1-4 neurones (n = 4 of 18) showed no voltage change or a small hyperpolarization (μ5 mV) following glycine application (Fig. 5Ac). LSO neurones at P5-8 responded with a slight depolarization (μ5 mV, n = 1 of 5, not shown), with no change in membrane potential (n = 2 of 5, Fig. 5B a) or with a hyperpolarization of μ10 mV (n = 2 of 5, Fig. 5B b). In P9-11 neurones, glycine consistently induced a hyperpolarization of μ15 mV (n = 3, Fig. 5C). These results show that glycine-evoked responses in the LSO are transiently depolarizing during early postnatal development, and they may even be excitatory. Hyperpolarizing responses occur mainly in the second postnatal week, as expected from the age-dependent changes in E_Gly determined in voltage-clamp recordings.

Figure 5. Glycine responses change from depolarizing to hyperpolarizing during development.

Examples of glycine-evoked membrane potential changes recorded in gramicidin perforated-patch configuration in six LSO neurones of different age. A, P1-4 neurones exhibited diverse glycine responses, but most cells were depolarized. a, in a P3 neurone, a depolarization was evoked, giving rise to a train of 8 spikes within 530 ms (spikes are truncated). b, a P2 neurone was depolarized, but no spikes occurred. c, a P4 neurone was slightly hyperpolarized. B, the majority of P5-8 neurones showed no voltage change or a hyperpolarization. a, a P6 neurone showed almost no response. b, a P8 neurone was clearly hyperpolarized. C, P9-11 neurones were strongly hyperpolarized, as illustrated for a P10 neurone. Same time scale in A-C.

E_Gly can be shifted in the positive and negative direction depending on voltage protocol and application interval

During development, patterned repetitive synaptic activity has been demonstrated in several neuronal systems (for review see Katz & Shatz, 1996). It is conceivable that repetitive activity of Cl⁻ conductances can lead to short-term alterations of [Cl⁻]_i, and thus to changes in Cl⁻-mediated responses. Indeed, repetitive activation of GABA_ARs with brief pulses or single activation with a prolonged pulse results in a ‘fading’ of the activated currents due to a shift in E_Cl and in GABA reversal potential (E_GABA); this shift in E_GABA is frequency dependent (Huguenard & Alger, 1986; Akaike et al. 1987). These experiments showed that Cl⁻ regulation mechanisms can be challenged and the neuronal Cl⁻ gradient can be manipulated by GABA-evoked activity. To assess whether the same holds for Cl⁻ fluxes through activated GlyRs in LSO neurones, we varied the voltage protocol and glycine application interval (Δt) and investigated whether shifts in E_Gly occurred. ‘Up’ and ‘down’ voltage protocols were applied and Δt was varied; an ‘up’ protocol was immediately followed by a ‘down’ protocol. In P1-4 neurones with Δt = 10 s, E_Gly determined with the ‘down’ protocol (E_Gly(down)) was considerably more positive than E_Gly determined with the ‘up’ protocol (E_Gly(up)). In contrast, this effect was much smaller in P8-11 neurones. Examples of our observations are illustrated for a P2 and a P8 neurone in Fig. 6A and B, respectively. We calculated E_Diff =E_Gly(down) - E_Gly(up) to use it as a measure for shifts in E_Gly. Group data showed that E_Diff was significantly greater in P1-4 than in P8-11 neurones (14.1 ± 1.7 mV, n = 17 vs. 3.8 ± 1.8 mV, n = 12; Student's t test, P < 0.01; Fig. 6C). What could be the reason for the observed effects? We think that Cl⁻ regulation mechanisms are not effective enough to maintain a constant [Cl⁻]_i when Δt is as short as 10 s. As a consequence, E_Gly(up) and E_Gly(down) should deviate from the ‘real’ value for E_Gly in that E_Gly(up) is more negative and E_Gly(down) is more positive than the ‘real’E_Gly. We assume that E_Gly(up) is generated by Cl⁻ depletion, i.e. an efflux of Cl⁻ during the voltage steps more negative than E_Cl, which is sufficient to result in a considerable reduction of [Cl⁻]_i and, consequently, in a negative shift of E_Cl. Likewise, E_Gly(down) is generated by Cl⁻ loading, i.e. an influx of Cl⁻ during the voltage steps more positive than E_Cl which results in an increase in [Cl⁻]_i and, thus, a positive shift of E_Cl. Finally, the finding that E_Diff is larger in P1-4 than in P8-11 neurones may reflect less efficient Cl⁻ regulation mechanisms in the younger group. One way to test these ideas is to vary Δt in the ‘up’ and ‘down’ protocol: an increase in Δt should result in a smaller E_Diff, whereas a larger E_Diff should be observed when Δt is decreased. We analysed P1-4 and P8-11 neurones and found that E_Diff indeed depended on Δt in both groups (Fig. 6D). In the younger group, Δt had to be increased to 50 s to yield a small E_Diff of 3.5 ± 1.6 mV (n = 5). However, when Δt was decreased to 5 s, E_Diff increased to 20.3 ± 2.3 mV (n = 9). In the older group, E_Diff was negligible for Δt = 20 s (1.1 ± 1.2 mV, n = 3) but increased to 13.7 ± 3.2 mV when Δt was decreased to 5 s (n = 7). For a given Δt, E_Diff was always larger in the younger than in the older group. These results demonstrate that E_Gly can be shifted in both a frequency-dependent and an age-dependent manner.

In a next step, we determined how much E_Gly(up) and E_Gly(down) deviated from the ‘real’E_Gly in P1-4 neurones when Δt = 10 s. Values for the deviation will be referred to as E_Shift in the following. An example for a P2 neurone is shown in the left panel of Fig. 7A. Almost identical values for E_Gly(up) and E_Gly(down) were found for Δt = 50 s, and the arithmetic mean was therefore taken as the ‘real’E_Gly. However, when Δt = 10 s, E_Gly(up) was more negative and E_Gly(down) was more positive than the ‘real’E_Gly, resulting in an E_Shift of −4.8 and +12.4 mV, respectively. When the analysis included all 12 P1-4 neurones tested, the average E_Shift was −4.3 ± 1.8 mV for E_Gly(up) and +10.3 ± 0.7 mV for E_Gly(down) (Fig. 7A, right panel). The results demonstrate that E_Gly is shifted in the positive and negative direction in a frequency-dependent manner by glycine-evoked activity.

Further evidence that positive and negative shifts of E_Gly occur in an age-dependent manner was obtained in another series of experiments: glycine was applied repetitively (20 pulses with Δt = 10 s) at a V_s that was 20 mV more negative than the ‘real’E_Gly. After a waiting period of about 2 min, another 20 pulses were applied at a V_s that was 20 mV more positive than the ‘real’E_Gly. In both cases, a reduction of successive amplitudes of glycine-evoked currents (‘fading’) was observed in P1-4 neurones (Fig. 7B), amounting to 58 ± 6 % in inward and 60 ± 6 % in outward currents (n = 6). This demonstrates that E_Gly was shifted in the positive and negative direction. The ‘fading’ effect was larger for Δt = 5, as it amounted to 79 ± 10 % for inward currents and 83 ± 3 % for outward currents (n = 3, not shown), demonstrating frequency dependence. In two P10 neurones, in which Δt was 10 s, the ‘fading’ of glycine-evoked inward and outward currents amounted to 10 ± 4 % and 26 ± 6 %, respectively (Fig. 7C). Thus, the ‘fading’ effect was much smaller in P10 than in P1-4 neurones, again indicating age dependence.

Furosemide shifts E_Gly in the negative direction in young neurones yet in the positive direction in older neurones

Aside from the above experiments, we used a second approach to challenge Cl⁻ regulation mechanisms. We pharmacologically blocked Cl⁻ transporter molecules to assess their contribution in the homeostasis of [Cl⁻]_i and in shaping glycine responses in LSO neurones. The loop diuretic furosemide is a blocker of most known members of the cation-chloride cotransporter superfamily, including the electroneutral Na⁺-K⁺-2Cl⁻ and K⁺-Cl⁻ cotransporters (for review see Haas & Forbush, 1998). Furosemide has been shown to perturb Cl⁻ outward transport mechanisms in mature neurones (Misgeld et al. 1986; Thompson et al. 1988; Zhang et al. 1991) and to block Cl⁻ inward transport mechanisms in immature neurones (Owens et al. 1996). In perforated-patch recordings, we first determined the ‘real’E_Gly under control conditions (subsequently called control E_Gly) by adjusting Δt to avoid shifts in E_Gly due to Cl⁻ loading or depletion (see above). Subsequently, 1 mm furosemide was applied to the bath solution and changes in E_Gly were monitored; only neurones with reversible furosemide effects were included in the analysis. Furosemide caused a shift of the control E_Gly towards more negative values in P2-4 neurones. This is illustrated for a P4 neurone in Fig. 8A, where the shift occurring within 35 min amounted to −19.5 mV (from −48.5 to −68.0 mV). In contrast, a shift of the control E_Gly towards more positive values was observed in P8-10 neurones, as illustrated in Fig. 8B for a P10 neurone, where the shift occurring within 15 min amounted to +24.1 mV (from −88.6 to −64.5 mV). Data from 10 neurones are summarized in Fig. 8C, where the control E_Gly (filled circles) and the E_Gly under furosemide (open circles) are plotted and the corresponding values are connected by an arrow that indicates the direction and magnitude of the shift for each neurone. In P2-4 neurones, the control E_Gly was more positive than −60 mV and always shifted in the negative direction towards V_h =−70 mV (n = 5). In contrast, in P8-10 neurones, the initial E_Gly was more negative than −60 mV and always shifted in the positive direction towards, or even positive to, V_h =−70 mV (n = 5). In addition to shifting E_Gly, a second and independent effect of furosemide was to block the GlyR conductance directly (cf. Kumamoto & Murata, 1997). This can be seen as a reversible reduction of the slope of the I-V relation in young and older neurones (Fig. 8A and B, lower panels). We quantified this effect by comparing the slopes of the I-V relations before and during furosemide application and found a decrease of the GlyR conductance to 50 ± 6 and 66 ± 7 % of control values in P2-4 and P8-10 neurones, respectively (n = 5 in each group). The block of GlyRs was not significantly different in the younger and the older group (t test, P = 0.10). In summary, the observed, furosemide-induced shifts in E_Gly suggest the presence of net inward Cl⁻ transport mechanisms in young LSO neurones (P2-4), yet of net outward Cl⁻ transport mechanisms in older neurones (P8-10). This allows these neurones to generate a high [Cl⁻]_i or a low [Cl⁻]_i as a prerequisite for the depolarizing and hyperpolarizing action of glycine, respectively.

Figure 8. Furosemide causes a shift of E_Gly in the negative direction in young neurones, yet in the positive direction in older neurones.

Bath application of 1 mm furosemide caused shifts in E_Gly in perforated-patch recordings. A, example of glycine-evoked currents in a P4 neurone before (Control) and during application of furosemide for 35 min (Furosemide). E_Gly shifted from −48.5 mV (•) to −68.0 mV (○; lower panel). B, example of glycine-evoked currents in a P10 neurone before (Control) and during application of furosemide for 15 min (Furosemide). E_Gly shifted from −88.6 mV (•) to −64.5 mV (○; lower panel). The slope of the I-V relation was reduced in both the P4 and the P10 neurone to 44 and 58 % of control, respectively (lower panels in A and B). V_s =−70 to −20 mV in A, and −90 to −40 mV in B. C, summary of furosemide effects: the control E_Gly (•) is compared to E_Gly after furosemide (○) and the corresponding values are connected by an arrow that illustrates the direction and magnitude of the furosemide-induced shift. In P2-4 neurones, the control E_Gly was between −46.4 and −59.7 mV and furosemide always induced a shift in the negative direction (average: −15.9 ± 3.2 mV, n = 5), indicative of the blockade of a net inward Cl⁻ transport mechanism. In P8-10 neurones, the control E_Gly was between −64.4 and −88.6 mV and furosemide always induced a shift in the positive direction (average: +28.4 ± 4.8 mV, n = 5), indicative of the blockade of a net outward Cl⁻ transport mechanism. The dotted line marks V_h =−70 mV.

DISCUSSION

Three major results emerge from this study. First, by studying glycine-evoked responses during the development of the rat LSO, we found that glycine is generally depolarizing and may be excitatory during the first week, yet it is hyperpolarizing thereafter. Second, concentrating on the cellular mechanisms underlying the changes of glycine action, we found that, independent of age, glycine-evoked responses are mainly carried by Cl⁻ (>80 %). This means that glycinergic depolarizations in immature neurones are most likely to be mediated by Cl⁻ efflux while hyperpolarizations in older neurones are mediated by Cl⁻ influx. Finally, we present evidence that an age-dependent and furosemide-sensitive active regulation of [Cl⁻]_i underlies the changes in glycine action. The shift from depolarizing to hyperpolarizing action and its time course are consistent with our previous results obtained with sharp electrode recordings (Kandler & Friauf, 1995b), yet here we present the first evidence for the ionic basis and for active cellular regulation mechanisms.

Ionic basis of glycine-evoked responses in the LSO

In whole-cell recordings, the E_Gly of somatic glycine-evoked currents corresponded well to the E_Cl set by the experimental conditions. Although the values for E_Gly in both young and older neurones (P2-7 vs. P8-11) deviated slightly from those predicted by the Nernst equation, the slope of the linear regression was almost identical to that of the Nernst relation in both groups. This would not be the case if there were a prominent permeability of the GlyRs to other ions, e.g. anions such as HCO₃⁻, or cations such as Na⁺. The deviation of a mixed ion reversal potential from the E_Cl is always larger for low [Cl⁻]_i. For example, when [Cl⁻]_i is 12 mm and the HCO₃⁻ permeability is 33 % of the Cl⁻ permeability, the mixed ion reversal potential is μ9 mV more positive than E_Cl (Fig. 2C). Because this deviation becomes smaller for lower HCO₃⁻ permeabilities (e.g. 11 %, Bormann et al. 1987), we cannot exclude the possibility that a HCO₃⁻ flux occurs through GlyRs under our experimental conditions. However, especially for young neurones with high [Cl⁻]_i, it is extremely unlikely that other ions than Cl⁻ contributed significantly to monophasic currents through GlyRs and E_Gly. This implies that the glycine-evoked depolarizations were mediated by an efflux of Cl⁻. In fact, our whole-cell recordings showed that depolarizations consistently occurred when E_Cl > V_m. Our results are in line with other studies demonstrating that Cl⁻ efflux is primarily responsible for GABA_AR-mediated depolarizations in immature neurones (Rohrbough & Spitzer, 1996; Owens et al. 1996).

Interestingly, a considerable contribution of HCO₃⁻ ions to both GABA- and glycine-evoked responses has been reported in several studies (Staley et al. 1995; Perkins & Wong, 1996; Backus et al. 1998). However, in contrast to the above-mentioned monophasic and Cl⁻-mediated GABA and glycine responses, the responses were biphasic with an early hyperpolarizing and a late and long-lasting depolarizing component. Moreover, they were evoked by high-frequency stimulation or long applications of agonists and were thought to be mediated by the collapse of the opposing concentration gradients of Cl⁻ and HCO₃⁻ (Staley et al. 1995; Backus et al. 1998). Occasionally, we also found glycine responses with two phases, but we did not investigate their ionic basis, which would have required additional ion substitution experiments. Thus, it remains unclear whether and how HCO₃⁻ participated in the generation of these responses. Nevertheless, it is probable that the first phase was predominantly mediated by Cl⁻, because its E_Gly equalled that seen in the monophasic responses. Although biphasic glycine responses in LSO neurones have been described by Backus et al. (1998) and in the present study, we think that these responses differ from each other in the following aspects: (1) long bath applications vs. short focal applications were used, (2) responses lasted for minutes vs. seconds, and (3) during the second phase of the response, E_Gly was more positive vs. more negative than V_m. Taken together, we conclude that the monophasic responses, following short glycine applications, represent the predominant response type in LSO neurones and are generated predominantly by Cl⁻ flux while prolonged glycine applications may result in biphasic responses to which both Cl⁻ and HCO₃⁻ contribute.

Age-dependent changes in glycine responses and E_Gly

We observed an age-dependent change from depolarizing to hyperpolarizing glycine action between P5 and P8, which was accompanied by a corresponding negative shift in E_Gly. The data were obtained with gramicidin perforated-patch recordings and correspond well to our previous finding with sharp electrode recordings that a change in the polarity of the glycinergic transmission from MNTB to LSO occurs around the end of the first postnatal week (Kandler & Friauf, 1995b). In addition, we observed a variability of E_Gly and a diversity of glycine-evoked membrane potential changes among P1-4 neurones that was not clearly attributable to a medial to lateral gradient of E_Gly in the LSO. Since the rat LSO is not composed of a uniform population of neurones (Rietzel & Friauf, 1998), we rather think that some neurone types express more mature properties than others, e.g. regarding their complement of ion channels or ion transporter molecules. While in the neonatal gerbil LSO, sharp electrode recordings demonstrated hyperpolarizing PSPs already at P0 (Sanes, 1993), whole-cell patch-clamp recordings revealed that glycinergic PSCs reverse at V_m in the lateral part, yet positive to V_m in the medial part (Kotak et al. 1998). It is not clear why regional differences were seen in these whole-cell recordings which, most probably, disturb the native [Cl⁻]_i through a dialysis of the cellular content and bias the E_Cl. Together, the above results indicate that there may be species-specific regional and temporal differences in LSO development.

The depolarizing action of glycine in immature neurones appears to be a general principle, because it has also been demonstrated in other regions of the CNS. It has been observed in the spinal cord (Wu et al. 1992), the brainstem (Kandler & Friauf, 1995b; Backus et al. 1998; Singer et al. 1998), the hippocampus (Ito & Cherubini, 1991) and the cerebral cortex (Flint et al. 1998). Furthermore, a change from depolarizing to hyperpolarizing action of glycine and GABA appears to be a general principle as well. In rats, this change happens earlier in caudal than in more rostral parts of the CNS: around P0 in the spinal cord (Wu et al. 1992), after P3 in the brainstem (Singer et al. 1998), and from P7 to the middle of the second postnatal week in the cerebellum (Brickley et al. 1996), the hippocampus (Ben-Ari et al. 1989), and the cerebral cortex (Owens et al. 1996). The change in the LSO between P5 and P8 fits well into this general scheme (Kandler & Friauf, 1995b; present study).

Active regulation of [Cl⁻]_i as a mechanism for changes in glycine action

Our data suggest that the prerequisite for depolarizing and hyperpolarizing glycine responses is the generation of an E_Gly > V_m and an E_Gly < V_m, respectively, caused by the differential activity of Cl⁻ transporter molecules. Based on the Nernst equation and the mean values for E_Gly (−46.8 mV at P1-4; −82.2 mV at P8-11), we calculate that [Cl⁻]_i in LSO neurones is reduced from μ22 to μ5 mm during this period. If the Cl⁻ distribution were not actively regulated, E_Gly would equal V_m, and [Cl⁻]_i would hence amount to μ13 mm.

In order to interfere with the active Cl⁻ transport mechanisms, we used furosemide as a pharmacological blocker, even though this has two problems. First, furosemide non-competitively blocks GABA_ARs and GlyRs (Kumamoto & Murata, 1997). Our approach, however, to compare the I-V relations before, during and after washout of furosemide allowed us to discriminate between a block of GlyRs, seen as a decrease in the slope of the I-V relation, and the shift in E_Gly, seen as a slow and gradual shift of the I-V relation along the X-axis. The mean furosemide-induced block of GlyRs (50 % at P2-4, 44 % at P8-10) in our study is consistent with an IC₅₀ for furosemide of about 1 mm reported by Kumamoto & Murata (1997) in cultured septal neurones. The second problem associated with furosemide is that it inhibits most members of the cation- Cl⁻ cotransporter superfamily and more specific blockers are not available at present (Haas & Forbush, 1998). Therefore, furosemide cannot be used to distinguish between specific transporter molecules, but it may be applied to determine the net direction of Cl⁻ transport in a given neurone. In neonatal LSO neurones (P2-4), the furosemide-induced shift of E_Gly in the negative direction towards V_h most probably results from a block of a net inward Cl⁻ transport, leading to a passive distribution of Cl⁻. In contrast, in older neurones (P8-10) the shift of E_Gly in the positive direction towards or above V_h most probably results from the block of a net outward Cl⁻ transport. These data are consistent with previous reports on furosemide-induced shifts of E_Gly and E_GABA in immature and mature neurones (Misgeld et al. 1986; Thompson et al. 1988; Owens et al. 1996; Kakazu et al. 1998). Furthermore, our finding that E_Gly shifted to values more positive than V_h indicates the presence of inward Cl⁻ transport mechanisms which were not completely blocked by furosemide.

Recently, it was shown that the expression of two members of the cation-Cl⁻ cotransporter superfamily is developmentally regulated in the nervous system. One of them, the Na⁺-K⁺-2Cl⁻ cotransporter BSC2, mediates inward Cl⁻ transport and is strongly expressed during the first postnatal week in neurones of the hippocampus, cerebral cortex, cerebellum and brainstem, while it diminishes from P14 to adulthood (Plotkin et al. 1997). The second member, the neurone-specific K⁺-Cl⁻ cotransporter KCC2, is thought to serve as the major Cl⁻ extrusion mechanism in the adult CNS (Payne et al. 1996); its expression is upregulated during postnatal development of the hippocampus (Rivera et al. 1999). Conclusive evidence for a role of KCC2 in participating in the age-dependent change from depolarizing to hyperpolarizing GABA action in the hippocampus was obtained by blocking KCC2 expression with antisense mRNA (Rivera et al. 1999). When these experiments were performed at a time when GABA responses were already hyperpolarizing, responses became slightly depolarizing in 25 % of the treated neurones, indicating that Cl⁻ did not become passively distributed. We assume that this finding and our observation of a furosemide-induced shift of E_Gly to values more positive than V_h are similar phenomena, suggesting that various inward and outward Cl⁻ transporters are present in a given neurone and that their relative activity determines the net Cl⁻ transport. It will be interesting to see whether BSC2 and KCC2 show a complementary temporal expression pattern in the LSO which would make them good candidates for participating in the Cl⁻ regulation that underlies the change in glycine responses.

Glycine-induced activity-dependent shifts in E_Cl

Not only the up- and downregulation of functional Cl⁻ transporter molecules results in changes of E_Gly, but short-term shifts in E_Gly can also be induced in an activity-dependent manner. When we applied glycine repetitively at sufficiently short time intervals, E_Gly was shifted in the positive and negative directions (depending on whether inward or outward Cl⁻ currents were activated, cf. Figs 6 and 7). These shifts were probably due to Cl⁻ loading or Cl⁻ depletion of neurones. Our observations are consistent with, and extend, our previous finding of shifts in E_Gly due to intracellular Cl⁻ accumulation during prolonged GlyR activation in combination with depolarizing steps (Backus et al. 1998). Similar shifts in E_Cl and E_GABA have been seen in adult neurones following brief repetitive, or a single prolonged, activation of GABA_ARs (Huguenard & Alger, 1986; Akaike et al. 1987). Thus, a strong activation of Cl⁻ flux through GABA_ARs and GlyRs can lead to considerable changes of [Cl⁻]_i that cannot be counteracted immediately by Cl⁻ transporters, resulting in short-term shifts in E_Cl (and consequently E_GABA and E_Gly). It is unclear whether short-term shifts in E_Cl occur in LSO neurones under natural stimulation conditions. However, spontaneous spike activity has been described in the vertebrate auditory system prior to the onset of hearing (for review see Friauf & Lohmann, 1999). Although being of relatively low frequency (1 Hz), it is sustained and occurs in rhythmic bursts and may therefore be sufficient to induce shifts in E_Cl.

Other studies have linked short-term changes in a neurone's E_Cl to the activation of voltage-dependent, inwardly rectifying Cl⁻ currents (Staley, 1994; Staley et al. 1996). These currents are mediated by ClC-2 channels and activated at membrane potentials negative to E_Cl. Their activity causes shifts of E_Cl towards the given V_h, or shifts of V_m towards a more positive E_Cl when the neurone is loaded with high [Cl⁻] via the patch pipette. For several reasons, we think that a significant contribution of ClC-2 channels to the short-term shifts in E_Gly can be ruled out in LSO neurones. First, in whole-cell recordings with a high [Cl⁻]_p, we did not see any changes of V_m in the positive direction. Second, in perforated-patch recordings from young neurones, where E_Gly > V_m, we did not see spontaneous changes of E_Gly in the negative direction. Third, we observed shifts in E_Gly in the positive and negative directions and not exclusively in the negative direction when V_h < E_Gly. Finally, under normal conditions, i.e. with undisturbed intracellular milieu, ClC-2 activity is downregulated (Staley, 1994). Therefore, we may not have seen ClC-2 channel activity in our perforated-patch recordings.

In the LSO, short-term shifts in E_Gly were larger in neonatal (P1-4) than in older animals, and considerable shifts were seen at a stimulation frequency as low as 0.1 Hz in the neonatal group (based on the Nernst equation, shifts in E_Gly from −50 mV to −40 or to −55 mV correspond to changes in [Cl⁻]_i from μ19 mm to μ28 and to μ16 mm, respectively). The age dependency of the shifts can be explained by compromised Cl⁻ transport mechanisms and/or the small volume of the younger neurones, both of which providing favourable conditions for larger changes in [Cl⁻]_i. The resulting changes in the amplitude of glycine-evoked responses may be viewed as a short-term, activity-dependent modulation of glycinergic transmission. Taken together, our data suggest that Cl⁻ transporter molecules participate in the homeostasis of [Cl⁻]_i in the LSO, but Cl⁻ regulation can be put out of action by strong activation of GlyRs, particularly in immature neurones.

Possible physiological relevance of glycinergic depolarization

The fact that glycine depolarizes neonatal LSO neurones and can induce spikes is consistent with the action of GABA and glycine on immature neurones described in other CNS regions. The depolarizing action of GABA and glycine can activate VGCCs, resulting in increases of [Ca²⁺]_i (Reichling et al. 1994; Leinekugel et al. 1995; Obrietan & van den Pol, 1995; Lo Turco et al. 1995; Flint et al. 1998). Such increases may be involved in mediating the trophic effects of GABA on neuronal precursors and immature neurones and play a role in synapse maturation and plasticity (for reviews see Cherubini et al. 1991; Ben-Ari et al. 1997). For example, activity-dependent plasticity at GABAergic synapses in neonatal hippocampus was observed only if GABAergic PSPs were depolarizing (McLean et al. 1996). In light of the above observations, it is possible that glycinergic depolarization may play a similar role in mediating trophic and plastic events during LSO development. VGCCs are present in neonatal rat LSO neurones (Schmanns & Friauf, 1994), and glycine application can cause rises in [Ca²⁺]_i in neonatal gerbil and rat LSO neurones (Brückner et al. 1998; Kullmann & Kandler, 1999). That calcium plays a critical role in the neonatal LSO was shown in organotypic cultures, where KCl-induced Ca²⁺ influx via L-type Ca²⁺ channels is necessary for the survival of LSO neurones and the maintenance of the MNTB-LSO projection (Lohmann et al. 1998). In vivo, the activation of VGCCs may be achieved by spontaneous activity originating in the cochlea before hearing onset. Spontaneous activity of cochlear origin has been identified in embryonic chickens (Lippe, 1994), but at present, it is unknown whether it occurs in the LSO of neonatal rats and how glycinergic and glutamatergic inputs may contribute to this activity.

In summary, glycinergic depolarizations in neonatal LSO neurones may provide a basis for activity-dependent plasticity, such as synaptic gain changes and remodelling of synaptic connections. The fact that neonatal neurones are more sensitive to shifts in E_Gly may render glycinergic inputs more susceptible to short-term modifications than later on when Cl⁻ regulation is more efficient to maintain a relatively constant [Cl⁻]_i.

Note added in proof

After the submission of this paper, Y. Kakazu, N. Akaike, S. Komiyama & J. Nabekura have published a related study in which Cl⁻ regulation mechanisms were addressed in acutely isolated, dissociated LSO neurones (Journal of Neuroscience19, 2843-2851 (1999)). The results of both studies are very similar in that they indicate that developmental changes of Cl⁻ transporters alter the [Cl⁻]_i and are responsible for the shift from depolarizing to hyperpolarizing glycine action.