INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Spontaneous activity is a property of many, perhaps all, networks in the developing nervous system. It comprises recurring episodes, containing one or many cycles of discharge, that are synchronized throughout the neurons of the active network (for review see O'Donovan 1999). This activity has been implicated in several aspects of network development (for review see Katz and Shatz 1996), but its genesis is not entirely understood. Recent modeling studies have argued that spontaneous activity can be produced by excitatory networks with activity-dependent synaptic depression (Senn et al. 1996; Tabak et al. 1999, 2000). In such models, the network is recruited through the positive-feedback excitation of recurrent synaptic connections. Once active, the network engages activity-dependent depression, which reduces functional network connectivity below a level that can sustain activity. The network then progressively recovers, allowing another episode of activity to occur (for reviews see O'Donovan and Chub 1997; O'Donovan et al. 1998).

In the chick embryo spinal cord, it is known that an episode of activity is followed by a prolonged depression of synaptic transmission in many spinal pathways (Fedirchuk et al. 1999), but the cellular mechanisms of this depression are still unknown. Ventrally located GABAergic neurons are an important source of synaptic drive to motoneurons during spontaneous episodes and have been implicated in the patterning of flexor and extensor motoneuron activity (Sernagor et al. 1995). In addition, it has been shown that GABAergic networks are capable of supporting spontaneous activity in the embryonic day 10 to 12 (E10-12) cord following excitatory amino acid blockade (Chub and O'Donovan 1998a) or in the E4-5 cord following cholinergic blockade (Milner and Landmesser 1999). Given the importance of GABAergic networks in early spinal cord development, in this paper we have characterized GABAergic transmission and its modulation by spontaneous activity. Preliminary reports of this work have been published in abstracts (Chub and O'Donovan 1995, 1998b, 1999).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Experiments were performed on the isolated spinal cord of E10-11 White Leghorn chicken embryos. The dissection procedures have been described in detail previously (Landmesser and O'Donovan 1984; O'Donovan 1989). In short, the lumbosacral cord was isolated and cooled to 10-15°C in Tyrode's solution containing (in mM) 139 NaCl, 5 KCl, 17 NaHCO₃, 1 MgCl₂, 3 CaCl₂, and 12 glucose, equilibrated with 95% O₂-5% CO₂ to pH 7.3-7.4. An isolated section of the lumbosacral cord (LS1-LS7) was then transferred to a recording chamber and superfused at room temperature (20-22°C). Recordings were made after heating the bath solution to 28°C. During both the dissection and the recording, the isolated cord was continuously superfused with Tyrode's solution equilibrated with O₂/CO₂.

Recordings were made from ventral horn neurons located in spinal segments LS1-LS3 that contain a significant proportion of GABAergic neurons at E10-12 (Antal et al. 1994). The whole cell recording ("blind patch" technique) (see Blanton et al. 1989) was used as modified for the isolated spinal cord of the chick embryo (Chub and O'Donovan 1998a; Sernagor and O'Donovan 1991). Micropipettes were pulled from thin-walled glass capillaries (TW 100-3, World Precision Instruments) in two stages using a Brown-Flaming puller (P-80/PC, Sutter Instrument). Tip resistances with intracellular solution were between 4 and 6 M (measured in the extracellular solution).

Initial current-clamp experiments were performed using an intracellular solution that contained (in mM) 10 NaCl, 130 K-gluconate, 10 HEPES, 1.1 EGTA, 1 MgCl₂, 0.1 CaCl₂, 1 Na₂ATP, and standard extracellular (Tyrode's) solution. For the next set of the experiments, using the Gramicidin-perforated patch, we used an intracellular solution containing (in mM) 140 KCl, 10 NaCl, 10 HEPES, and 30 sucrose. Gramicidin D ([Dubos], Sigma) was dissolved in methanol (2 mg in 100 µl) and added to the pipette solution at final concentration of 50 µg/ml. The tip of the micropipette was back-filled with gramicidin-free patch solution. Sealing, perforating (up to 30-40 min), and identification of the neurons were all made in standard extracellular (Tyrode's) solution. For measurements of E_GABAa (perforated patch configuration, voltage-ramp protocol) we added blockers of various voltage-activated and GABA_b conductances to the extracellular solution which then contained (in mM) 139 NaCl, 5 KCl, 17 NaHCO₃, 1 MgCl₂, 1 MnCl₂, 12 glucose, 10 tetraethylammonium acetate (TEA), 5 CsCl, 0.001 tetrodotoxin citrate (TTX), and 0.1 2-hydroxysaclofen (2-HS), pH 7.3-7.4.

Miniature synaptic postsynaptic currents (mPSCs) were recorded from cells under whole cell voltage clamp (at V_hold = 70 mV) that exhibited a good seal resistance (>5-7 G) and a high-input resistance (0.6-1.2 G). For these measurements, we used an intracellular solution containing (in mM) 10 NaCl, 94 K-gluconate, 36 KCl, 10 HEPES, 1.1 EGTA, 1 MgCl₂, 0.1 CaCl₂, 1 Na₂ATP, and 0.5 Na-GTP. For long-duration measurements of the agonist-evoked I_GABAa (in whole cell configuration), the ATP concentration was increased to 5 mM. In addition, to reduce the possible confounding effects of the intracellular Ca²⁺ elevation that accompanies spontaneous episodes (O'Donovan et al. 1994) and that may lead to changes of I_GABAa because of Ca-dependent receptor phosphorylation (Mozrzymas and Cherubini 1998), we used an intracellular solution with a high concentration of the Ca²⁺ buffer [11 mM bis-(o-aminophenoxy)-N,N,N',N'-tetraacetic acid (BAPTA) instead of 1.1 mM EGTA]. Thus for these experiments, the whole cell solution contained (in mM) 84 K-gluconate, 46 KCl, 10 HEPES, 11 BAPTA, 1 MgCl₂, 0.1 CaCl₂, 5 Na₂ATP, and 0.5 Na-GTP. In the final set of the experiments, we determined the effects of spontaneous episodes on E_GABAa. In these experiments, the voltage-dependent channel blockers could not be added to the bath solution because this prevented the generation of episodes. Therefore we added to the intracellular solution (containing BAPTA) the following channel blockers: 10 mM QX-314 (lidocaine, N-ethyl bromide quaternary salt), 10 mM TEA, and 0.1 mM verapamil [(±)-verapamil, methoxy-HCl]. All intracellular solutions were adjusted to pH 7.2 with KOH.

Membrane potential was recorded with an Axoclamp 2A amplifier, current with an Axopatch 200B amplifier (Axon Instruments) and ventral root activity by using tight-fitting plastic suction electrodes coupled to DAM70 DC amplifiers (World Precision Instruments). Signals were filtered and amplified (DC  2 or 5 kHz, Cyber Amp 380), digitized (DigiData 1200), and stored on the hard disk of an IBM-compatible computer using AxoScope 1, Clampex 6, or 7 software (Axon Instruments). E_GABAa was measured by applying a quasi-stationary voltage-ramp command generated by Clampex 6 and analyzed off-line with Clampfit 6 software. The amplitude and inter-event intervals of mPSCs were measured off-line with Mini Analysis Program 3.0.1 (Jaejin Software) and analyzed with Origin 4.1 (Microcal Software), Sigma Plot 3.0, and SigmaStat 2.0 (Jandel). Access resistance (series resistance) was not compensated. Typically, it was not more than 40 M (in perforated patch recording) and 25 M in whole cell configuration that would give an error in voltage command of up to 2.7 mV at V_hold = 30 mV (perforated patch) and up to 2.5 mV at V_hold = 0 mV for whole cell recording.

Local application of drugs was performed as described earlier (Sernagor et al. 1995). In brief, drugs were dissolved in Tyrode's solution and then pressure-applied (10-15 psi) from a micropipette similar to that used for whole cell recording. The injection micropipette was first positioned inside the ventral horn, and then the recording electrode was introduced. A whole cell recording was made from a cell located as close as possible to the injection side. The drugs for local and bath application were prepared as 10-mM stocks in distilled water and diluted to the final concentration in the extracellular solution immediately before use or made up fresh and used immediately. Drugs were obtained from Sigma (St. Louis, MO) with exception of TTX, 2-HS, 6-cyano-7-nitroquinoxaline-2,3-dione disodium (CNQX), and bicuculline-methiodide (BIC), which were obtained from RBI (Research Biochemicals International, Natick, MA).

Liquid junction potentials were measured as described by Neher (1992) with 3 M KCl agar bridges, well equilibrated (2-3 days) in Tyrode's solution. They were not more than 3-4.5 mV for our intracellular solutions. Data were not corrected these errors. Results are expressed as means ± SE.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Inter-episode changes of membrane potential and the hyperpolarizing effect of bicuculline during the inter-episode interval

Whole cell recordings were made from ventral horn neurons in spontaneously active isolated spinal cords of the chick embryo (E10-11). Spontaneous activity appeared as repetitive, rhythmic episodes nearly 60 s in duration with 10 to 15-min inter-episode intervals. Long-lasting recordings (up to 1 h in current-clamp configuration) revealed that spinal neurons were transiently hyperpolarized after an episode, and this recovered as a slow depolarization until the next episode. Figure 1 shows examples of the membrane potential trajectory of two ventral spinal neurons during, and in the interval between, spontaneous episodes. Approximately 1-2 min after an episode, the membrane became hyperpolarized by 6 to 10 mV (6.5 ± 1.2 mV, mean ± SE, 8 neurons) and then slowly depolarized (0.5-1 mV/min) until the next episode. In some cells, the inter-episode depolarization reached threshold and triggered several spikes (Fig. 1, dashed arrow). Otherwise, neurons did not fire between episodes.

View larger version (27K): [in this window] [in a new window]   Fig. 1. Whole cell recordings from 2 ventral horn neurons (top cell located in LS1 and the bottom cell in LS3) show the membrane potential trajectory between spontaneous episodes. The records were in current-clamp configuration without current injection. In both neurons, the membrane potential hyperpolarized after the episodes, and it slowly recovered as a progressively developing depolarization during the inter-episode interval. Arrows indicate the occurrence of spontaneous episodes. Note that 2-3 min before a spontaneous episode (record on the top) the membrane potential reached threshold (about 40 mV), and the cell started to fire. The 2 parallel dashed lines show the range of the post-episode hyperpolarization. The membrane potential calibration is on the left of the records. The spikes are truncated.

Earlier preliminary observations (Chub and O'Donovan 1995) had suggested that the inter-episode depolarization was affected by the GABAa antagonist bicuculline. This was consistent with indirect evidence in the chick spinal cord that GABA was depolarizing (Sernagor et al. 1995). Therefore to test the dependence of the inter-episode membrane potential on GABAergic conductances, we pressure-applied the GABAa antagonist bicuculline (100 µM, n = 7) at various times after spontaneously occurring episodes (20 s, 4 and 8 min) and measured the accompanying changes in membrane potential (V_m) and input resistance (R_m). The drug was most effective when it was applied 8 min after an episode, close to the time the next episode would have occurred (the inter-episode interval was about 10 min; Fig. 2A). At this time, it hyperpolarized the membrane by 6.4 ± 0.8 mV and increased R_m to 112.6 ± 2.6% (7 neurons). By contrast, we observed no change in membrane potential when the drug was applied 20 s after the episode (n = 5, 4 neurons). Application of the antagonist 4 min after an episode, approximately midway through the inter-episode interval, resulted in a smaller membrane hyperpolarization than that at 8 min (Fig. 2B, middle and left).

View larger version (42K): [in this window] [in a new window]   Fig. 2. The effect of the GABAa antagonist bicuculline on the membrane potential and input resistance of ventral horn neurons depends on the time after an episode when the drug is applied. A: local application of bicuculline (BIC; time of application indicated by bar over the record) 8 min after a spontaneous episode, just before another episode would have occurred, hyperpolarized the neuron (9 mV) and increased its input resistance (+13%). B: the effect of bicuculline on the membrane potential progressively increased with time after an episode. In this experiment, bicuculline was applied to the neuron at 3 different times (20 s, 4 and 8 min) after a spontaneous episode. When applied 20 s after the episode, the drug had no detectable effect on the neuron's membrane potential or the input resistance. Application at 4 and 8 min hyperpolarized the neuron and increased its input resistance. In both records 100 µM bicuculline was pressure ejected for 15 s (indicated by horizontal bars) from a micropipette located near the cell. Membrane potential and input resistance were measured in whole cell current-clamp configuration. Hyperpolarizing current (20-50 pA, 250-300 ms, 1.5 Hz) was injected into the cell for input resistance measurement.

These findings suggested that the slow inter-episode depolarization were caused, in part, by endogenous GABA. Consistent with a role for GABA in the inter-episode depolarization, we found that low concentrations of bicuculline (5 µM; 3 experiments) strongly and reversibly lengthened the inter-episode interval from 11.8 ± 1 min to 51.3 ± 4.7 min. Following wash out of the drug, the intervals partially recovered to 25.1 ± 5.2 min (2 h wash out).

Depolarizing GABAa equilibrium potentials measured with gramicidin-perforated-patch recording

To confirm the depolarizing nature of GABA on ventral spinal neurons, we used perforated-patch recording with gramicidin D as the voltage-insensitive cation ionophore (Hladky and Haydon 1984; Kyrozis and Reichling 1995; Reichling et al. 1994). In the gramicidin-perforated-patch configuration, the resting membrane potential (V_m) of spinal neurons varied between 50 and 70 mV (60 ± 3 mV, n = 10). Neurons were distinguished from glial cells by their ability to fire action potentials.

When a neuron was identified, voltage-dependent channel blockers and 2-HS were included in the Tyrode's solution to block voltage-activated and GABAb conductances, respectively (see METHODS). GABAb conductances were blocked because we wanted to establish the current-voltage relationship (I-V) and the equilibrium potential for GABAa receptor activation (E_GABAa). This extracellular solution also blocked spontaneous episodes. The I-V of the GABA-evoked currents (I_GABAa) was determined by applying a slow (80 mV/s) triangular voltage ramp to the neuron (Fig. 3A, top). Local application of 100 µM GABA at holding potential (V_hold) 50 mV evoked a long-lasting inward current (Fig. 3A, bottom) ranging from 58 to 276 pA (106 ± 37 pA, 6 neurons). For the neuron shown in Fig. 3A, V_m was 50 mV (estimated from the zero current crossing under control conditions), and E_GABAa was 30 mV (estimated from the intersection of the I-V curves generated under control conditions and during GABA application; Fig. 3B). I_GABAa was linear in the range of holding potentials from 60 to 0 mV (Fig. 3C, as determined by subtraction of the control current from the current recorded in the presence of GABA). The mean E_GABAa, determined from six perforated-patch recordings, was 29.1 ± 2.9 mV and varied from 37 to 18 mV (Fig. 3E). It should be noted that inclusion of the channel blockers in the external solution decreased the resting potential of the neuron by 5-10 mV, but E_GABAa was always about 20 mV more positive than the resting V_m.

View larger version (24K): [in this window] [in a new window]   Fig. 3. Gramicidin-perforated-patch recording revealed that the GABAa equilibrium potential is depolarizing in ventral horn neurons. A: an example of the membrane currents induced by a triangular voltage ramp (inset on the top) before and during local application of 100 µM GABA (horizontal bar on the bottom) at a holding potential (Vhold) of 50 mV. To facilitate estimating the resting membrane potential at zero current, the ramp command had 10-mV hyperpolarization from Vhold = 50 mV. Voltage-dependent and GABAb conductances were eliminated by bath application of the appropriate blockers (see METHODS). B: current-voltage (I-V) relation obtained from the record shown in A before (control) and during GABA application. The resting membrane potential was determined from the 0 crossing point of the control curve and was approximately 50 mV. The GABAa equilibrium potential (EGABAa), measured at the intersection of the 2 curves, was 32 mV. C: the GABAA current (IGABAa) was computed as the difference between the 2 curves (GABA - control) shown in B. Note the nearly linear I-V relation of IGABAa from 50 to 0 mV. D: in another neuron IGABAa was measured in gramicidin-perforated-patch configuration as in C, and then the neurons patched membrane was ruptured to obtain the whole cell configuration. The voltage ramps and GABA application were repeated to obtain measurements of the I-V relation in the whole cell configuration with symmetrical intracellular and extracellular Cl solutions. Note the shift of EGABAa from 21 mV in the gramicidin-perforated-patch recording to 0 mV in the whole cell configuration, consistent with Cl as the charge carrier for IGABAa. E: bar graph showing EGABAa determined by gramicidin-perforated-patch recording in 6 neurons (inset on the left indicates the mean ± SE for the 6 cells).

In five neurons, the recording was converted into the whole cell configuration by applying gentle suction. Under these conditions, I_GABAa was measured no earlier than 3 min after membrane rupture to allow adequate time for somal dialysis. In the example shown in Fig. 3D, E_GABAa was 21 mV in the gramicidin-perforated-patch and +1 mV in the whole cell configuration. Measured in five neurons, the average E_GABAa in the whole cell configuration, with 150 mM Cl in the pipette solution, was +2.1 ± 2.8 mV, a value close to the chloride equilibrium potential (0.5 mV) predicted by the Nernst equation. These results suggest that Cl ions are the main charge carrier for the I_GABAa and allow us to estimate the physiological intracellular [Cl] as 50 mM (from the mean E_GABAa measured using the gramicidin-perforated patch).

Quantal and persistent GABAa currents during the inter-episode interval

mPSCs were recorded in whole cell voltage-clamp configuration using an electrode chloride concentration (48.2 mM, E_Cl = 30 mV) close to the intracellular [Cl] estimated from the perforated-patch recordings. The mPSCs were observed in ventral horn neurons but not in glial cells and were analyzed only during the inter-episode interval (not during spontaneous episodes). The mPSCs ranged from 4 to 190 pA and had variable kinetics with an average rise time of 3.5 ms (measured between 10 and 90% of peak amplitude) and a slower, approximately monoexponential decay,  = 26.1 ms (V_hold = 70 mV, 3 neurons, 154 events). These parameters were in general similar to the parameters of mPSCs recorded in cultured chick spinal neurons (O'Brien and Fischbach 1986) and from other types of embryonic spinal neuron (Ali et al. 2000; Gao et al. 1998; Rohrbough and Spitzer 1999).

To establish whether generation of the mPSCs required action potentials, we determined the effects of TTX on the amplitude and frequency of the events. We found that the mPSCs persisted following bath application of 1 µM TTX (3 experiments; Fig. 4), which also abolished spontaneous episodes and action potentials (data not shown). To establish the effects of TTX on the mPSCs, we measured their frequency and amplitude during a 2-min period before and 5-8 min after application of the TTX. All measurements were done at a V_hold = 70 mV. The mean amplitude of the mPSCs was 13.3 pA in control conditions (476 events), and this decreased to 11.5 pA after 5-8 min in TTX (458 events). The corresponding averaged inter-event intervals were 743.5 ms in control and 771.5 ms in TTX. Both the amplitude and frequency of mPSCs in TTX were not statistically significantly different from the controls (ANOVA on Ranks, Dunn's Method; compared in 3 neurons). Representative examples of averaged mPSCs are illustrated in Fig. 4B and are shown superimposed (obtained during 2 1-min time period in control and in TTX). These findings suggest that the great majority of mPSCs occurring during the inter-episode interval arise from action-potential-independent transmitter release. However, just before an episode, some neurons do start to fire so that some of the mPSCs occurring at this time will arise from action potentials.

View larger version (26K): [in this window] [in a new window]   Fig. 4. During the inter-episode interval, miniature postsynaptic currents are largely action potential independent. A: bath application of TTX does not block the miniature postsynaptic currents (mPSCs) recorded during the inter-episode interval from a ventral horn neuron. Control and 1 µM TTX records (obtained 5-8 min after drug application) are from the same cell. B: comparison of the mPSCs waveforms averaged over a 1-min period under control conditions and in the presence of TTX (from the cell shown in A). The averaged and superimposed mPSCs are shown on the right. C: a longer section (50 s) recorded from another cell, showing mPSCs under control conditions and the presence 1 µM TTX. Five successive 10-s traces are superimposed for each condition. D: cumulative counts of the mPSCs amplitude (left) and the inter-event interval (right) distribution in control, and 5-8 min after application of TTX. The cumulative-frequency distributions were obtained from 3 sets of 2-min measurements for each condition, which were recorded from 3 cells. Note that frequency and amplitude of the mPSCs was not changed significantly by TTX, suggesting that most mPSCs are action potential independent. All records are from neurons at Vhold = 70 mV with an electrode [Cl] = 48.2 mM, which was close to level estimated from the gramicidin-perforated-patch recordings.

To establish the transmitters responsible for generation of the mPSCs, we bath-applied receptor antagonists in the presence of 1 µM TTX. The blockers were added in the following sequence: for the first 15 min we added the glutamate antagonists 2-amino-5-phosphonopentanoic acid (AP5; 50 µM) and CNQX (10 µM). Then we added the GABAa antagonist bicuculline (BIC; 25 µM) for an additional 15 min, followed for a further 15 min by the addition of the glycine antagonist strychnine (10 µM). The AP5 and CNQX always reduced, but never completely abolished, the mPSCs (Fig. 5A). When bicuculline was added to the glutamate antagonists, the remaining mPSCs were completely abolished in four of seven cells. In the other three cells, strychnine abolished all of the remaining spontaneous events (Fig. 5A). The mean frequency of mPSCs measured in the presence of AP-5 and CNQX decreased to 45 ± 5.9% from 100% of the control. When bicuculline was added to the excitatory blockers, the mPSCs frequency decreased to 3% from the control level (Fig. 5B; 2-min periods were used for measurement in 7 neurons). These data suggest that glutamate and GABA are the major transmitters contributing to the generation of mPSCs and allow us to estimate that approximately 42% of the total mPSCs are mediated by action-potential-independent quantal release of GABA. The amplitude of the GABAergic mPSCs was measured in four neurons, which did not have glycinergic mPSCs. In these cells (recorded in the presence of AP5 and CNQX), subsequent application of bicuculline completely abolished the mPSCs demonstrating that they originated from GABAergic transmission. The amplitude of the GABAergic miniature currents ranged from 6 to 31 pA with a median value of 8 pA (571 events from 4 neurons; each cell was separately analyzed and then averaged).

View larger version (17K): [in this window] [in a new window]   Fig. 5. Both mPSCs and tonic GABAa currents are present in ventral horn neurons. A: comparison of mPSCs in the presence of TTX (control) and during sequential application of various transmitter antagonists. First the glutamate antagonists 2-amino-5-phosphonopentanoic acid (AP5; 50 µM) and 6-cyano-7-nitroquinoxaline-2,3-dione disodium (CNQX; 50, 10 µM) were applied, then the GABAa antagonist bicuculline (BIC; 25 µM) was added, and finally the glycine antagonist strychnine (STR; 10 µM) was added to the other drugs. Then all of the drugs were washed out (washout). Note that bicuculline reduced the membrane noise (thickness of traces) and it blocked most of the mPSCs after it was added to the glutamate antagonists. B: bar graph showing the frequency (top axis) and percentage change in the frequency (bottom axis) of mPSCs under control conditions, after addition of the antagonists and following washout (mean ± SE). The data were generated from measurements of the mPSC frequency over a 2-min interval, and averaged from 7 cells. C: bath application of bicuculline (25 µM), in the presence of AP5, CNQX, and TTX blocks a persistent GABAergic current (bars indicate the antagonist application). Note that when 10 µM STR was added to the other drugs, the remaining mPSCs are completely blocked, but that the baseline current was unchanged. The 2 dashed lines delineate the amplitude of the tonic current (15 pA), which was blocked by bicuculline, and the arrow indicates its inward direction. All currents were recorded at a Vhold = 70 mV.

We also found that bath application of bicuculline (in the presence of TTX and glutamatergic receptor antagonists) reduced the membrane noise (see Fig. 5A) and blocked an inwardly directed, tonic I_GABAa (Fig. 5C). The amplitude of the tonic I_GABAa was measured as shown in Fig. 5C and ranged from 4 to 18 pA with a mean of 9.6 ± 2.8 pA (n = 5). It is unlikely that the tonic I_GABAa arises because of summation of high-frequency quantal GABAergic release, because individual GABAergic mPSCs were clearly separated one from another and often completely decayed before the next event (Fig. 5A).

Activity-dependent modulation of mPSCs

The mPSCs recorded at V_hold = 70 mV were superimposed on a slow inward current (Fig. 6, dashed line on the top) that was presumably responsible for the slow inter-episode depolarization we recorded under current clamp (Fig. 1). The slow inward current was voltage dependent: at V_hold = 70 mV its peak value was 15-35 pA, but at V_hold = 30 mV (close to E_GABAa) it was greatly reduced (data not shown). These observations are consistent with the idea that slow inter-episode depolarization is mediated by persistent activation of the GABAa receptors.

View larger version (35K): [in this window] [in a new window]   Fig. 6. Spontaneous episodes depress both mPSCs and the persistent current. A: 20-min record, similar to those shown in Fig. 1, but recorded under voltage clamp at a holding potential of 70 mV. The dashed line on the top of the record provides a baseline reference for the persistent inward current, which progressively increased during the inter-episode interval. Synaptic currents accompanying the spontaneous episode (arrows) are off-scale. The rectangles at the beginning of the top trace illustrate 30-s segments before and after an episode, which are expanded in the bottom panels to show the synaptic events clearly. Asterisks (bottom left) show events with multiple peaks, which usually occurred before an episode. B and C: cumulative distributions show the changes of mPSCs amplitudes and inter-event intervals before and after spontaneous episodes. The data for the cumulative distributions were obtained from 5 neurons (30-s measurement periods) before and after spontaneous episodes.

We also found that mPSCs changed systematically during the inter-episode interval. Before a spontaneous episode, the amplitude of the mPSCs was highest, and they often comprised multi-peak events (Fig. 6A, bottom left, asterisks). After a spontaneous episode, the mPSCs amplitude declined (Fig. 6A, bottom right) and then progressively recovered throughout the inter-episode interval. We quantified these changes by comparing the mPSCs (V_hold = 70 mV) during 30-s intervals (indicated by the rectangles in the top panel of Fig. 6A) before and after spontaneous episodes. The cumulative distributions show that mPSC amplitude distribution is shifted to the left after an episode (Fig. 6B). We also found that the mPSC intervals increased after an episode, and this is illustrated by the right hand shift of their cumulative distribution (Fig. 6C). Following an episode, the mPSCs mean amplitude decreased from 15.9 pA (1,904 events) to 11.3 pA (1,627 events), and the mPSCs interval increased from 156 ms (1,894 events) to 184 ms (1,617 events). The post-episode mPSC measurements were significantly different from the control, pre-episode mPSC measurements (P < 0.001, Mann-Whitney rank sum test; 10 measurements, 5 neurons). The percentage change of the mean mPSCs amplitude (29%) was greater than that for the mPSCs interval (+18%).

Post-episode depression of GABAa-mediated current

The time-dependent effects of locally applied bicuculline on the inter-episode depolarization, coupled with the post-episode depression of the mPSCs amplitude, raised the possibility that the postsynaptic response to GABA might be depressed after an episode and then recover during the inter-episode interval. To test this hypothesis, we measured the GABAa currents evoked by local application of the GABAa agonist isoguvacine (100 µM) or GABA (100 µM) onto a recorded neuron, before and after spontaneous episodes. To avoid initiation of an episode by the local application of the agonist, the duration and pressure of the puff were routinely adjusted for each cell and thereafter kept constant during the measurements on that cell. For these, and all following measurements, we used an intracellular solution (see METHODS) with a high concentration of ATP to minimize the time-dependent rundown of the agonist-evoked I_GABAa (Kapur et al. 1999).

We found that the amplitude of the isoguvacine-evoked current measured at V_hold = 60 mV was transiently depressed 30 s after an episode and partly recovered by 2 min 30 s (Fig. 7A). To quantify this post-episode depression, we compared the amplitude of the currents evoked 40-50 s before and 30-40 s after an episode. The mean amplitude of the GABA-induced current decreased from a pre-episode value of 63.9 ± 10.2 pA to 49.6 ± 9.4 pA after the episode (n = 17, 11 neurons). Similarly, the isoguvacine-evoked current decreased from 194.5 ± 27 pA to 153.3 ± 20.5 pA after the episode (n = 5, 2 neurons). Both of these decreases were statistically significant (P < 0.001, paired t-test). We found no change in the normalized time course of the current decay (not shown) suggesting that the post-episode decrease of the evoked current was probably not a result of GABAa receptor desensitization.

View larger version (34K): [in this window] [in a new window]   Fig. 7. The amplitude of GABAa agonist-evoked currents is transiently reduced after an episode. A: isoguvacine-evoked currents recorded under voltage clamp (Vhold = 60 mV) in a ventral horn neuron, before and after the occurrence of a spontaneous episode. Isoguvacine (100 µM) was puff-applied for 5 s as indicated by the bars above the record. The bottom panel shows the isoguvacine-induced currents adjusted to the same baseline level and at expanded time scale. Note that amplitude of the GABAa currents has partly recovered 2 min 30 s after the episode. B: control record to show that the depression of the isoguvacine currents does not depend on voltage clamping the cell during the episode (see text for details). In this experiment the GABAa currents (application of 100 µM isoguvacine, before and after episode) were measured under voltage clamp (Vhold = 60 mV), but during the episode the recording was switched to current clamp (without current injection). Note the 2 calibrations: the membrane potential calibration immediately follows the episode, and the current calibration is displayed at the end of the record. The isoguvacine-evoked currents before and after the episode are shown superimposed to the right of the record.

We were concerned that the post-episode depression of the evoked I_GABAa might be due, in part, to clamping the membrane potential during the episode. For this reason, in three control experiments, the voltage clamp was switched to current clamp during the episode (Fig. 7B), and the evoked I_GABAa was measured as before under voltage clamp. We found that the currents were still depressed after an episode indicating that the post-episode decrease was not due to clamping the synaptic drive or holding the membrane potential constant during episodes of rhythmic activity.

Changes of E_GABAa after episodes

Previous work in the chick embryo spinal cord has shown that ventral horn neurons receive a prolonged (up to 60 s) and strong GABAergic synaptic drive during an episode of activity (Chub and O'Donovan 1998a; Sernagor et al. 1995), raising the possibility that intracellular Cl redistribution might be responsible for the post-episode I_GABAa changes. If so, then experimental manipulation of the direction of Cl flux during the episode should be accompanied by corresponding changes in post-episode amplitude (increase or decrease) of the evoked I_GABAa. To change the direction of neuronal Cl flux, the cell was clamped at one of three different V_hold during the episode: 0 mV, in-flux of Cl; 60 mV, out-flux of Cl; 30 mV, E_Cl, minimal Cl flux. The isoguvacine-induced pre- and post-episode currents (Fig. 8) were measured at V_hold = 60 mV as described above. To reduce the possible confounding effect of intracellular Ca²⁺ elevation (during the episode), we used an electrode solution with a high concentration of the Ca²⁺ buffer (11 mM BAPTA instead of 1.1 mM EGTA, see METHODS).

View larger version (21K): [in this window] [in a new window]   Fig. 8. The change in the post-episode amplitude of isoguvacine-evoked currents depends on the neuron's holding potential during the episode. Examples of the isoguvacine-evoked current before and 30 s after episodes during which the neuron was voltage clamped at 1 of 3 different holding potentials (Vhold): 0 mV the GABAergic drive should result in Cl in-flux during the episode; 30 mV is close to EGABAa so that Cl flux during the episode should be minimal; and 60 mV the Cl flux should be outward during the episode. The isoguvacine-evoked current measurements were all made at the same Vhold = 60 mV. The time of occurrence of the episodes can be seen from the LS3 ventral root record shown on the top, and the timing of the isoguvacine puffs is shown on the bottom trace. The voltage protocol is shown above current record. Note that the holding potential for the last episode was 60 mV, the same as for measurements of all isoguvacine-evoked currents.

For episodes at V_hold = 60 mV (i.e., below E_Cl), the post-episode, isoguvacine-induced currents decreased as described previously (Fig. 7). By contrast, when the neuron was clamped during the episode at 0 mV (i.e., above E_Cl), the amplitude of the isoguvacine-induced currents increased by 18% (3 cells) after the episode. When the neurons were clamped at 30 mV (close to the estimated E_Cl), we observed a small 5% decrease (mean of the same 3 cells) in the amplitude of the evoked currents after the episode. These findings indicate that that the direction of the intracellular Cl flux during the episodes determines the amplitude and direction of the post-episode changes of the evoked GABAa current.

In the next set of experiments, we analyzed the changes of E_GABAa after an episode by measuring the reversal potentials of the isoguvacine-evoked currents. The E_GABAa was measured before and after episodes using the voltage-ramp protocol described for the perforated-patch recordings (Fig. 3). However, it was necessary to add channel blockers to the electrode solution to avoid abolishing episodes of activity (see METHODS). Under these conditions the I-V relation of the evoked I_GABAa was linear from 60 to 15 mV.

We found that E_GABAa became more negative when the neuron was clamped at 60 mV during the episode (Fig. 9, A and B). Under these conditions, E_GABAa shifted by 5 to 18 mV (10 ± 1.1 mV, n = 11, 6 neurons) after the episode. As expected, for cells at V_hold = 0 mV during the episode, E_GABAa shifted in the opposite direction (Fig. 9C) by +3 to +11 mV (+7 ± 1.1 mV, n = 7). When the cell was at V_hold = 30 mV (the assumed E_GABAa) during the episode, the episode-induced changes of E_GABAa were small (1.2 ± 0.5 mV, n = 4).

View larger version (27K): [in this window] [in a new window]   Fig. 9. The reversal potential of isoguvacine-evoked currents changes after an episode. A: currents generated by the injection of 2 triangular voltage ramps (from a Vhold = 60 mV) before and after an episode. The 1st ramp was used to calculate the I-V relationship for the cell under control conditions in the absence of applied drug. During the 2nd ramp, isoguvacine (100 µM) was applied to the cell for the time indicated by the bar. The same protocol was repeated 30 s after a spontaneous episode. The I-V curve for IGABAa was estimated by subtracting the I-V curves generated by the 2 ramps (isoguvacine-control). B: graph comparing the I-V curve for IGABAa before (continuous line) and after (broken line) an episode that was at Vhold = 60 mV. The I-V curve was shifted to the left after the episode, as EGABAa became more negative. Note that the slope of the I-V curve was reduced after the episode, suggesting that the GABAa conductance is decreased after an episode. C: similar graph to that shown in B, but in this case the cell was at Vhold = 0 mV during the episode. Under this condition, the I-V curve shifted to the right after the episode, as EGABAa became more positive. The data in the graphs were generated from the same cell illustrated in A.

These data allow us to estimate (from the Nernst equation) the post-episode changes in the intracellular [Cl] under the different holding potentials. When the neurons were at V_hold = 60 mV during the episode, the mean intracellular [Cl] fell by 15 mM at the end of the episode, whereas the opposite change (+15 mM) occurred if the cells were at V_hold = 0 mV during the episode. It should be noted that the use of whole cell configuration (rather than perforated-patch) in these experiments would inevitably tend to counteract the episode-induced changes of intracellular [Cl] because of dialysis. As a result, we may have underestimated the actual changes.

We also noticed that the I-V relationships of I_GABAa before and after the episode were not completely parallel when the cell was at V_hold = 60 mV during the episode (Fig. 9B). After the episode, the GABAergic conductance was lower than before the episode falling from 17.3 ± 2.9 to 13.4 ± 2.2 nS (n = 5, paired t-test P < 0.05; Fig. 10). By contrast, when the cell was at V_hold = 0 mV, the isoguvacine-evoked conductance was similar before and after the episode (13.4 ± 3.6 before, 12.9 ± 3.2 nS after; not statistically different at P < 0.05, Fig. 10). This finding raises the possibility that some voltage-dependent process in the postsynaptic cell is affecting the GABAa conductance (see DISCUSSION).

View larger version (31K): [in this window] [in a new window]   Fig. 10. Graph summarizing the percentage change of the GABAa current amplitude (I) and conductance (G) after an episode. The neurons were clamped at different holding potentials (shown on bottom) during episodes of the activity. The numbers above graph show the corresponding changes of the GABAa equilibrium potentials (EGABAa) in mV. Asterisks indicate the bars that are significantly different from the pre-episode level.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this study, we have demonstrated that evoked I_GABAa is depressed after an episode and recovers during the inter-episode interval. One of the important contributors to this depression is a change of intracellular [Cl] after an episode of spontaneous activity. We estimate that the intracellular [Cl] decreases from a pre-episode concentration nearly 50 mM to about 35 mM. This change is presumably due to the prolonged activation of action-potential-dependent GABAergic transmission during the episode (Chub and O'Donovan 1998a; Sernagor et al. 1995). In addition, our results establish that GABAa conductances are active during the inter-episode interval in E10-11 spinal neurons. These conductances appear to be activated by action-potential-independent GABAergic mPSCs and a tonic I_GABAa.

Depolarizing GABAergic transmission in the embryonic chick spinal cord

In the present study, we employed gramicidin-perforated-patch recording, which avoids dialysis-induced changes in the intracellular [Cl] and divalent cation concentrations (Hladky and Haydon 1984). Using this approach, we found that E_GABAa was distributed over a wide range (37 to 18 mV) with a mean of 30 mV. A similar distribution of E_GABAa was found recently in embryonic Xenopus spinal neurons using perforated-patch recordings (Rohrbough and Spitzer 1996) and in cell-attached recordings of GABAa channels on spinal neurons isolated from E15 rat spinal cords (Serafini et al. 1995). Depolarizing GABA responses have also been observed in other developing networks including the hippocampus, cortex, and retina (Ben-Ari et al. 1989; Fischer et al. 1998; Yuste and Katz 1991).

Although the E_GABAa is more positive than V_m at rest, activation of GABAergic conductances can be functionally inhibitory as well as functionally excitatory. In the E11 chick cord, flexor sartorius and extensor femorotibialis motoneurons both receive synchronous GABAergic, cholinergic, and glutamatergic inputs during spontaneous episodes, but generate an alternating pattern of discharge. The alternation arises because sartorius motoneuron discharge is interrupted in each cycle at the time of peak extensor discharge. The interruption in firing is mediated primarily by a shunting GABAergic conductance (O'Donovan 1989). Recently, it was proposed that the differences in the functional action of GABA may depend, in part, on whether the synaptic conductances are somatically or dendritically located (Sernagor et al. 1995).

Tonic and quantal GABA release during the inter-episode interval

Approximately 42% of the mPSCs we recorded during the inter-episode interval were mediated by quantal GABA release. The majority of these events were action potential independent because the mPSCs frequency and amplitude were not significantly changed after 5-8 min in the presence of TTX. Consistent with this conclusion, our current-clamp recordings showed that spinal neurons only fired during, and just before, spontaneous episodes.

We estimate that about 10 channels (range 8-40) are located postsynaptically at each GABAergic synapse. This number is derived from the mean amplitude of the GABAergic mPSCs recorded in presence of TTX, AP5, and CNQX (8 pA, see RESULTS) and recent data indicating that one quantum of GABA release can activate all the GABAa receptors located at a single synapse (Nusser et al. 1997). In addition, the calculation assumes a single GABAa channel conductance 26 pS (Yang and Zorumski 1989) and P_o,max ~ 0.8 for GABAergic mPSCs (Jones and Westbrook 1995; Newland et al. 1991). The estimated density of synaptic GABAa receptors is approximately one-half that reported for mature neurons that have 10-60 GABAa channels located at a single synapse (De Koninck and Mody 1994; Edwards et al. 1990; Salin and Prince 1996).

In addition to quantal release, we also found evidence for a tonic I_GABAa. We consider it unlikely that the tonic current arises from the summation of asynchronous mPSCs because individual GABAergic events (isolated under glutamate blockade) decayed without substantial fusion with each other. It seems likely that the tonic current is responsible, in part, for the membrane potential changes occurring during the inter-episode interval. In addition, several lines of evidence suggest that the slow inter-episode depolarization was the result of endogenous GABA acting on GABAa receptors. First, local-application bicuculline during the inter-episode interval hyperpolarized spinal neurons, consistent with depolarizing action of GABA. Second, the tonic inward current we recorded during the inter-episode interval under voltage clamp was largely abolished when the cell membrane potential was clamped near E_GABAa (30 mV). We also found that bath application of bicuculline abolished a tonic inward current in the presence of AP5, CNQX, and TTX. A tonic I_GABAa of the magnitude we measured would depolarize ventral horn neurons approximately 8 mV, consistent with the maximum amplitude of the bicuculline-induced hyperpolarization recorded under current clamp during the inter-episode interval.

Our data also allow us to estimate the approximate number of active GABAa channels underlying the tonic current. This estimate assumes a single-channel GABAa conductance of 26 pS and probability of opening of 0.33, obtained from earlier studies of cultured chick spinal neurons (Yang and Zorumski 1989). Based on these values, we calculate that approximately 30 GABAa channels (range 12-57) would be responsible for generation of the persistent GABAa current. It is possible that these channels are primarily extrasynaptic because in postnatal day 4-28 (P4-P28) rat cerebellar granule cells, a tonic GABAa conductance, similar to the one we have identified, has been attributed to activation of extrasynaptic GABAa receptors (Brickley et al. 1996). In the cerebellum, one source of this extrasynaptic GABAa receptors activation appears to be spillover from quantally released GABA (Brickley et al. 1996; Rossi and Hamann 1998). It is also possible that the endogenous GABA is released nonquantally, perhaps leaking from neuronal or glial cells (Liu at al. 2000; Vesce et al. 1999).

Post-episode depression of GABAa current

After an episode, the amplitude of evoked GABAa current was transiently depressed and subsequently recovered during the inter-episode interval. Our data suggest that changes of intracellular [Cl] during the episode were largely responsible for this effect. First, the post-episode depression of isoguvacine-evoked currents was largely eliminated when the cell membrane potential was clamped close to the estimated E_GABAa (30 mV). Under these conditions intracellular [Cl] changes should be minimal. Second, post-episode depression could be reversed to potentiation if the Cl flux was changed from outward to inward by holding the membrane potential at 0 mV during the episode. Finally, consistent with the above arguments, we established that E_GABAa (and by inference E_Cl) changed after the episode in a manner that depended on the holding potential of the neuron during the episode. When the membrane potential was clamped at 60 mV, close to the resting membrane potential, E_GABAa decreased by approximately 10 mV after the episode. Conversely, when the membrane potential was clamped at 0 mV during the episode, E_GABAa increased by 7 mV. These changes will lead to corresponding changes in the driving force for GABAa-mediated synaptic currents.

After an episode, the amplitude of mPSCs fell and then recovered during the inter-episode interval. Although only 42% of spontaneous mPSCs were GABAergic, it seems reasonable to assume that the post-episode changes of E_GABAa will also contribute to their modulation during the inter-episode interval. Whether or not this is the only factor contributing to their modulation is unknown. For instance, when the cell was voltage clamped at 60 mV, we found that the GABAa conductance was greater before an episode than after it, and this could contribute to the increased mPSCs amplitude before the episode. We also detected a small decrease in the frequency of spontaneous mPSCs after an episode. While a reduction in frequency might indicate decreased presynaptic release, it may be that the frequency change is a detection issue, secondary to decreased mPSCs amplitude.

The post-episode reduction in the Cl driving force is also likely to account, in part, for the transient post-episode hyperpolarization we recorded in spinal neurons. When bicuculline was applied shortly after an episode (20 s), it did not change the membrane potential, consistent with the post-episode reduction of E_GABAa. By contrast, application of bicuculline 4-8 min after the episode hyperpolarized the cell resting potential, which would be expected as E_GABAa recovered and became more positive. However, the changes of E_GABAa are unlikely to be the only mechanism responsible for the inter-episode membrane potential changes. Bicuculline applied shortly after an episode produced no effect on the membrane potential even though isoguvacine or GABA puff-applied at this time was capable of generating a significant current. One explanation for this apparent discrepancy would be that the extracellular concentration of GABA was lowest after the episode and gradually increased during the inter-episode interval.

Concluding remarks

In previous work, it was shown that GABAergic ventral root-evoked potentials are transiently depressed after an episode and recover during the inter-episode interval (Fedirchuk et al. 1999). The depression in this and other pathways has been implicated in the genesis of spontaneous episodes by developing networks in the chick cord (Tabak et al. 1999, 2000). The results of the present work suggest that chloride redistribution may be one factor contributing to the post-episode depression of ventral root-evoked synaptic potentials. However, two factors complicate determining the magnitude of the episode-induced changes of intracellular [Cl] under physiological conditions. First, the outward chloride flux during an episode may be higher in cells that are voltage clamped at 60 mV (as in most of the present experiments) than when the membrane potential during the episode is allowed to vary. To address this concern, we examined the post-episode depression of the evoked I_GABAa when the voltage clamp was switched to current clamp during the episode (Fig. 7B). Under this condition, we still observed a post-episode depression in the amplitude of the isoguvacine-evoked current. The second concern arises because it was not possible to employ the gramicidin-patch recording to assess E_GABAa before and after an episode. This was because GABA_b and voltage-dependent channel blockers (necessary to isolate I_GABAa) had to be added to the electrode because they prevented episode generation when added extracellularly. This precluded the use of the gramicidin patch. As a result, the actual chloride changes induced by the episode may have been underestimated because they would have been opposed by electrode dialysis. One way to address these problems in future experiments may be the use of noninvasive chloride imaging (Verkman 1990) to assess the episode-induced changes of intracellular chloride. Furthermore, if chloride redistribution is a significant factor in the post-episode depression of synaptic transmission in GABAergic pathways, then blocking inwardly directed chloride pumps should affect this modulation. Although the regulation of intracellular [Cl] in developing neurons is not well understood, it is likely that inwardly directed Cl transporters are important in maintaining an elevated intracellular [Cl] (Rivera et al. 1999; Sun and Murali 1999). In Xenopus spinal neurons the elevated Cl concentration of Rohon-Beard cells depends on a bumetanide-sensitive Na⁺-dependent Cl co-transporter (Rohrbough and Spitzer 1996).

Experimental manipulation of intracellular chloride may also provide insight into the recovery of spontaneous activity that occurs in E10-12 cords following glutamatergic blockade (Chub and O'Donovan 1998a) and in E4-5 cords following cholinergic blockade (Milner and Landmesser 1999). At both ages, the reappearance of spontaneous activity depends on functional GABAergic networks. In E10-12 embryos, the resumption of spontaneous activity was accompanied by an increase in the amplitude of spontaneous, presumably GABAergic synaptic currents and potentials (Chub and O'Donovan 1998a). If these increases are mediated by a progressive increase in intracellular chloride concentration due to the action of inwardly directed chloride pumps, then we would predict that the recovery should be abolished if the pumps are blocked.