Modulation of mammalian dendritic GABAA receptor function by the kinetics of Cl− and HCO3− transport
Kevin J. Staley,
Departments of Neurology and Pediatrics, Medical Center, Denver, CO 80262, USA
Corresponding author K. J. Staley: Departments of Neurology and Pediatrics, University of Colorado Health Sciences Center, Box B182, 4200 East 9th Avenue, Denver, CO 80262, USA. Email: firstname.lastname@example.org
1During prolonged activation of dendritic GABAA receptors, the postsynaptic membrane response changes from hyperpolarization to depolarization. One explanation for the change in direction of the response is that opposing HCO3− and Cl− fluxes through the GABAA ionophore diminish the electrochemical gradient driving the hyperpolarizing Cl− flux, so that the depolarizing HCO3− flux dominates. Here we demonstrate that the necessary conditions for this mechanism are present in rat hippocampal CA1 pyramidal cell dendrites.
2Prolonged GABAA receptor activation in low-HCO3− media decreased the driving force for dendritic but not somatic Cl− currents. Prolonged GABAA receptor activation in low-Cl− media containing physiological HCO3− concentrations did not degrade the driving force for dendritic or somatic HCO3− gradients.
3Dendritic Cl− transport was measured in three ways: from the rate of recovery of GABAA receptor-mediated currents between paired dendritic GABA applications, from the rate of recovery between paired synaptic GABAA receptor-mediated currents, and from the predicted vs. actual increase in synaptic GABAA receptor-mediated currents at progressively more positive test potentials. These experiments yielded estimates of the maximum transport rate (vmax) for Cl− transport of 5 to 7 mmol l−1 s−1, and indicated that vmax could be exceeded by GABAA receptor-mediated Cl− influx.
4The affinity of the Cl− transporter was calculated in experiments in which the reversal potential for Cl− (ECl) was measured from the GABAA reversal potential in low-HCO3− media during Cl− loading from the recording electrode solution. The calculated KD was 15 mM.
5Using a standard model of membrane potential, these conditions are demonstrated to be sufficient to produce the experimentally observed, activity-dependent GABAA depolarizing response in pyramidal cell dendrites.
The direction of current through the GABAA ionophore is determined by the relative permeabilities and transmembrane electrochemical gradients of Cl− and HCO3− (reviewed in Alvarez-Leefmans, 1990 and Kaila, 1994). Pyramidal neurons utilize outwardly directed KCl cotransport so that at resting membrane potential (RMP), the electrochemical gradient for Cl− favours hyperpolarizing Cl− influx (Misgeld et al. 1986; Thompson & Gähwiler, 1989a; Thompson, 1994). In contrast, HCO3− flux through the GABAA ionophore is outward and depolarizing, because the HCO3− reversal potential (EHCO3) is 50 mV more positive than RMP (Kaila, 1994; Staley et al. 1995). GABAA receptor-mediated net anionic flux is inward and hyperpolarizing initially, since both the GABAA channel permeability and the concentration of Cl− at the GABAA ionophore entrance are 4-5 times that of HCO3− (Bormann et al. 1987). However, the bidirectional flow of anions through the GABAA receptor results in a short-circuit condition such that even when the membrane potential has settled at the GABAA reversal potential (EGABA), large Cl− and HCO3− currents continue to flow, because the membrane potential cannot reach either ECl or EHCO3. If the GABAA current is larger than the local anionic transport capacity, a significant shift in the concentrations of the permeant ions can occur (Ballanyi & Grafe, 1985; Huguenard & Alger, 1986; Kaila et al. 1989; Thompson & Gähwiler, 1989a). This shift will occur more rapidly in structures such as dendrites with a high ratio of GABAA receptors to volume (Qian & Sejnowski, 1990).
Thus one possible mechanism for the depolarizing response is an activity-dependent alteration of the dendritic EGABA (Staley et al. 1995): during intense GABAA receptor activation, Cl− on both sides of the membrane comes to equilibrium so that ECl is driven towards the membrane potential and Cl− flux decreases (Wong & Watkins, 1982; Ballanyi & Grafe, 1985; Huguenard & Alger, 1986; Thompson & Gähwiler, 1989a). If HCO3− does not come to equilibrium, the remaining GABAA current would consist primarily of depolarizing HCO3− efflux. Here we demonstrate the validity of key assumptions of this model: (1) the rate at which Cl− can be transported out of the neuron is fixed and saturable; (2) GABAA receptor-mediated Cl− influx can exceed dendritic transport capacity; (3) the driving force for transmembrane HCO3− flux is only minimally affected during maximal GABAA receptor activation; and (4) Cl− accumulation sufficient to alter ECl occurs in the dendrites but not the soma during intense GABAA receptor activation.
In accordance with the guidelines of our institutional animal welfare committee, following pentobarbitone anaesthesia (60 mg kg−1i.p.), adult Sprague-Dawley rats were decapitated and hemibrain slices were cut in the coronal plane. Whole-cell recordings from the CA1 subfield of the hippocampi of these slices were performed as described previously (Staley, 1994). HCO3−-buffered artificial cerebrospinal fluid (ACSF) saturated with 95 % O2-5 % CO2 (pH 7.4) contained (mM): 126 NaCl, 2.5 KCl, 26 NaHCO3, 2 CaCl2, 2 MgCl2, 1.25 NaH2PO4 and 10 glucose. Nominally HCO3−-free ACSF was saturated with 100 % O2, 26 mM Hepes acid replaced NaHCO3 and pH was titrated to 7.4 at 35°C with NaOH. Recording pipette solution contained (mM): 123 K+, 2 MgCl2, 8 Na+ (gluconate and Cl− salts were combined as necessary to produce the desired Cl− concentration, usually 10 mM), 1 K2EGTA, 4 K2ATP, 0.3 Na2GTP and 16 KHCO3; pH was 7.2 when solution was saturated with 95 % O2-5 % CO2. In HCO3−-free experiments, the electrodes were buffered with 10 mM Hepes instead, and the pH titrated to 7.2 with KOH. In low-Cl− experiments, 1-4 mM Cl− was kept in the media to stabilize the Ag/AgCl2 electrodes, and the balance of the extracellular Cl− was replaced with isethionate. Where noted, caesium, which is a substrate for the outward cation-Cl− transporter (Aickin et al. 1982), was used as the cation in electrode solutions for tetanic stimulation experiments in which somatic and proximal dendritic GABAA receptors were blocked (Fig. 5) in order to improve the voltage clamp of the distal dendrites (Staley & Mody, 1992).
Recordings were performed with an Axopatch-1D amplifier, and were digitized at 1-5 kHz using routines written in AxoBasic (Axon Instruments). Only recordings with stable access resistances < 15 MΩ were analysed. Approximately 80 % of the electrode access resistance was compensated in voltage clamp recordings. Junction potentials were corrected, and saline-bridged earths used. Low-frequency synaptic activation of dendritic GABAA receptors was accomplished using single 20-80 μs stimuli via bipolar tungsten electrodes in stratum moleculare (s. moleculare); high-frequency activation utilized trains of 10-40 of the same stimuli delivered at 200 Hz.
Synaptic GABAA receptor activity was isolated using the glutamate receptor blockers 6,7-dinitro-quinoxaline-2,3(1H,4H)-dione (DNQX, 20 μm) and dl-amino-5-phosphonovaleric acid (APV, 50 μm), and the GABAB receptor blockers CGP-35348 (100 μm) and CGP-55845A (1 μm) (Ciba-Geigy). In some experiments, GABAA receptors were selectively and non-synaptically activated by pressure application of 50-250 μm muscimol or GABA in ACSF using pulses of 5-20 ms × 10 p.s.i. through a 1 μm diameter whole-cell pipette. GABAB receptors were blocked when GABA was applied. Tetrodotoxin and pentobarbitone were applied by bath. When measuring dendritic Cl− transport, somatic GABAA receptors were blocked, where noted, by local application of 100 μm picrotoxin to stratum pyramidale (s. pyramidale).
Current-voltage plots were obtained by fitting the GABAA receptor-mediated currents to the Goldman-Hodgkin-Katz constant field equation, rewritten as:
where g is the GABAA conductance, V is the membrane potential, F is Faraday's constant (96 487 C mol−1), R is the gas constant (8.315 J mol−1 K−1) and T is temperature (K). Ao, the extracellular concentration of permeant anions, was calculated as:
where [Cl−]o and [HCO3−]o are the extracellular concentrations of Cl− and HCO3−, respectively, and PermHCO3 and PermCl are the permeabilities for the individual ion species HCO3− and Cl− through the GABAA channel. V0 represents the membrane potential corresponding to the zero-current condition for the constant field current equation:
(Hodgkin & Katz, 1949), where [Cl−]i and [HCO3−]i are the cytoplasmic concentrations of Cl− and HCO3−, respectively. Charge transfer by GABAA receptor-mediated post-synaptic currents (PSCs) was calculated by numerical integration of the PSC waveform. Time-averaged currents were obtained by dividing this integral by the duration of the current; 500 ms time intervals were used for averaging.
Kinetics of cation-Cl− transport
In these experiments, the change in direction and amplitude of the transmembrane Cl− current was assumed to be due to alterations in [Cl−]i and ECl as a consequence of GABAA receptor-mediated Cl− flux. Under the experimental conditions used in this study, we did not find evidence for substantial alterations in ECl or EK based on interneuron network activity (Kaila et al. 1997; Fig. 1A). Cytoplasmic Cl− loading and extrusion via voltage-dependent Cl− channels (Staley, 1994; Smith et al. 1995; Fig. 1B) were minimized by the selection of holding potentials: holding potentials positive to -25 mV were not used to avoid activation of ClC-3 chloride channels (Smith et al. 1995) and holding potentials negative to ECl were not used to avoid activation of ClC-2 channels (Staley, 1994; Staley et al. 1996).
We used the following three methods to characterize neuronal Cl− transport.
1. When the dendritic transport rate was calculated from the rate of recovery of the outward currents evoked by exogenous GABA application (Figs 3 and 4), the resting ECl was calculated from the I-V relationship of the currents evoked at 20 s intervals between GABA applications, using eqn (3). In experiments in which both Cl− and HCO3− were present (Fig. 4), we used a value of 0.25 for the HCO3−:Cl− GABAA permeability ratio (Bormann et al. 1987; Kaila, 1994). The GABAA conductance was calculated from the slope of the I-V plot at potentials near ECl. [Cl−]i at the time of repeat GABA application was estimated from the change in EGABA necessary to account for the initial current resulting from the repeat GABA application (eqn (1)). [Cl−]i was then calculated from EGABA using eqn (3). Data for the rate of change of [Cl−]i were fitted to a monoexponential curve using least-squares fit, where t= 0 was the time at which the initial evoked current decayed to 0 or when the test potential was changed to a value near ECl.
2. Large-amplitude dendritic Cl− currents were elicited either by brief tetanic stimuli (10-20 stimuli at 200 Hz) during block of GABAB and ionotropic glutamate receptors in s. moleculare or by applicaton of GABA to s. moleculare. These Cl− currents did not increase in amplitude at depolarized membrane potentials as predicted based on the slope of the I-V plot at potentials close to ECl (Figs 3C and 5B). The largest PSC evoked at depolarized potentials that could be predicted from the slope of the I-V curve near ECl was assumed to represent the maximum Cl− transport rate at the steady-state [Cl−]i, i.e. this PSC equalled the largest Cl− influx rate at which the Cl− driving force did not deteriorate as a consequence of Cl− accumulation due to Cl− influx exceeding the Cl− transport rate. The difference in charge transferred by the PSCs at depolarized membrane potentials vs. the charge transfer expected based on extrapolation from the slope of the I-V curve near ECl was used to estimate the charge lost due to Cl− accumulation. The GABAA conductance was obtained from the slope of the I-V curve near ECl. The shift in ECl was calculated as (GABAA conductance)−1× lost current (= lost charge/PSC duration). The volume of the dendrites into which the GABAA currents flowed was calculated as Cl− influx (measured as the charge transferred by PSCs at depolarized potentials, converted from coulombs to mM)/change in [Cl−]i necessary to account for the calculated change in ECl (eqn (3)). This allowed calculation of maximum Cl− transport per unit volume.
3. The neuronal cytoplasm was dialysed with an isotonic high-Cl− solution in the whole-cell electrode (Fig. 7). Under these conditions, the neuronal Cl− concentration increases to a steady state that is determined by the influx of Cl− (determined by the electrode Cl− concentration and rate of diffusion) and the rate of neuronal Cl− efflux (Staley, 1994). The neuronal cytoplasmic Cl− concentration was estimated from the reversal potential of the GABAA receptor-mediated PSCs in HCO3−-free media. The maximum cation-Cl− cotransport rate is a function of the transport capacity and the gradients for K+ and Cl− (Läuger, 1987). When either K+ or Cl− gradients are held constant, the rate of transport of the other ion as a function of its cytoplasmic concentration can be characterized using the Michaelis-Menten kinetic model (Tas et al. 1987; Gasbjerg & Brahm, 1991; Lauf et al. 1992). If spontaneous GABAA receptor fluxes are neglected, then at steady state, the rate of Cl− influx (from the electrode) equals the rate of efflux via transport (v). From the diffusion equation, Cl− influx is proportional to the difference between the electrode Cl− concentration ([Cl−]E) and the cytoplasmic Cl− concentration ([Cl−]i), so that v= ([Cl−]E - [Cl−]i)DEC, where DEC is the diffusion coefficient that characterizes diffusion of Cl− from the electrode solution to the cytoplasm for a given concentration gradient. The maximum Cl− efflux rate and the Cl− affinity of the cation-Cl− cotransporter (for a specified K+ gradient) were derived from the intercepts of the Lineweaver-Burke equation:
where KD is the neuronal Cl− concentration at which the extrusion rate is half-maximal and vmax is the maximum rate of Cl− extrusion. Data for Lineweaver-Burke plots were fitted using a least-squares algorithm. GABA responses were evoked shortly after whole-cell recording was established, and repeated until the calculated reversal potential reached a steady value; this required between 10 and 30 min (Staley, 1994). In these experiments, Cl− diffusion from the electrode will set up a somatic-dendritic Cl− gradient. Although proximal stimulation was used to elicit primarily somatic GABA responses, we cannot determine the subcellular location of the activated GABA receptors, thus our measurements represent a weighted mean of both somatic and dendritic concentrations.
Membrane potential model of the depolarizing GABAA response
We utilized a standard model of membrane potential (Finkelstein & Mauro, 1977) in which the membrane potential is considered to be generated by a group of electrochemical potentials, each representing one ionic species, in series with time-varying conductances (Fig. 8B). The currents flowing through the conductances were considered to flow into a structure of specified volume. HCO3− was assumed to be replaced as rapidly as it flowed out, but Cl− could accumulate or decrease in this volume depending on the amplitude of the Cl− current vs. the Cl− pump velocity. Cationic conductances were lumped into a time-invariant leak conductance. Voltage-dependent conductances and membrane capacitance were not included in the model.
The equation governing the membrane potential is derived from the principal that the sum of all membrane currents In equals the sum of the products of all conductances and the respective driving forces (Finkelstein & Mauro, 1977):
so that after rearranging,
where gn and V0n are the conductance and reversal potential for each ionic species, here restricted to ICl, IHCO3, and a cationic Irest.
The constant field equation was used for GABAA currents, using intracellular anion concentrations rather than V0 as in eqn (1), and setting g=gGABA× fractional ionic permeability. Cl− permeability was assumed to be 0.8, and HCO3− permeability was 0.2 (sum of fractional permeabilities, 1.0; HCO3−:Cl− permeability ratio, 0.25; Kaila, 1994).
[Cl−]i was calculated as:
where F is Faraday's constant, vol is the volume of the structure into which the GABAA current flows, v is the pump veolcity and dt is the time increment.
The pump velocity (v) was set according to the Michaelis-Menten model (eqn (4)). KD and vmax were taken from the data; KD was 15 mM and vmax was ∼5 mmol l−1 s−1. [Cl−]i(0) in eqn (9) is the steady-state [Cl−]i. This was assumed to be 2 mmol l−1 and was achieved by including a Cl− leak of 0.3 mmol l−1 s−1, in which case the Cl− pump reached a steady-state [Cl−]i of 2 mM. The source of this Cl− leak could include the electrode Cl− as well as Cl− components of the input conductance and spontaneous GABAA receptor-mediated activity.
The GABAA conductance was modelled using a two-term exponential function, using values of gmax= 40 nS, τ1= 0.05 s, τ2= 1 s:
Currents other than GABAA receptor-mediated Cl− and HCO3− currents were lumped together as a rest current:
The rest (input) conductance was set to 10 nS, corresponding to a 100 MΩ input resistance, and the resting membrane potential was set to -65 mV, i.e. V0rest= -65 mV.
When circular references were encountered (for instance, ICl depends on [Cl−]i and [Cl−]i changes according to ICl× dt), the value of [Cl−]i at the previous time point, T - dt, was used. The source code (Quickbasic 4.5 and Visual Basic 6.0, Microsoft) is available on request.
Stability of the Cl− and HCO3− gradients
To assess the stability of the dendritic Cl− and HCO3− gradients, we applied GABA (100 μm) directly to the distal dendrites of voltage-clamped CA1 pyramidal cells in media in which only Cl− or HCO3− was present in physiological concentration; the concentration of the other anion was minimized as described in Methods. GABAB conductances were blocked with 1 μm CGP 55845A. Paired GABA applications were utilized. The first GABA application was designed to collapse the anion gradient by evoking a large current at a test potential far from the reversal potential. To assess the impact of this current on the anion gradient, the membrane was then stepped to various test potentials and GABA was applied 2.5 s after the first application. The experiment was repeated with the first GABA application performed at a holding potential very close to the steady-state anionic reversal potential in order to minimize the initial anion flux. By comparing the anionic reversal potentials after large and small anion fluxes, effects due to the collapse of the anion gradient could be separated from other effects such as receptor desensitization (Huguenard & Alger, 1986; Thompson & Gähwiler, 1989a). Cl− gradients were assessed in nominally HCO3−-free ACSF containing 136.5 mM Cl− using a 10 mM Cl− electrode solution. For eight dendritic (s. moleculare) and four somatic (s. pyramidale) locations of GABA application in six CA1 pyramidal cells, the difference in the first dendritic Cl− current evoked at -30 vs. -70 mV was 646 ± 117 pA (Fig. 2A and B). The average shift in dendritic ECl due to the large Cl− current evoked at -30 mV was 12.6 ± 4 mV (Fig. 2C). When GABA was applied to the soma, the average difference in the first GABAA receptor-mediated Cl− currents elicited at the different holding potentials was larger (1180 ± 525 pA) than at the dendrites because the maximum response at the soma was larger. The corresponding shift in ECl at the soma was 1.5 ± 1 mV (Fig. 2D). HCO3− gradients were assessed using ACSF and electrode solutions containing physiological HCO3−/CO2 and 4 mM extracellular Cl−, and 1 mM electrode Cl−. For seven dendritic and two somatic locations in four cells, the average difference in the first dendritic HCO3− current evoked at -20 vs. -70 mV was 238 ± 91 pA, and the average shift in EHCO3 due to the first HCO3− current evoked at -70 mV was -0.5 ± 1 mV at the dendrites (Fig. 2E and F). At the soma, the average difference in the first HCO3− current was 304 pA, and the corresponding shift in EHCO3 was 0 mV. These data confirm that the dendritic Cl− gradient can collapse significantly during GABAA receptor activation (Ballanyi & Grafe, 1985; Huguenard & Alger, 1986; Thompson & Gähwiler, 1989a), and that dendritic and somatic HCO3− gradients and somatic Cl− gradients are stable under the same conditions.
The first GABA application produced HCO3− currents that were smaller than the corresponding Cl− currents, and the ratio of these maximum Cl− and HCO3− currents is consistent with the reported permeabilities of Cl− and HCO3− through the GABAA receptor ionophore (Bormann et al. 1987). While searching in low-Cl− media for the dendritic location at which GABA application resulted in the largest possible inward current at -70 mV, we never found a dendritic region with a higher amplitude current. This argues against the existence of a population of GABAA receptors with a uniquely high HCO3− or cationic permeability.
The rate of recovery of the dendritic transmembrane Cl− gradient
The rate at which the dendritic Cl− gradient recovered was assessed using the protocol described above, except that the timing of the second GABAA application was varied, and the amplitude of the second GABAA current was plotted as a function of the time elapsed since the end of the first GABAA current. The amplitude of the second GABAA current was used to estimate ECl and [Cl−]i. Recordings were performed in nominally HCO3−-free media, and the electrode solutions contained 10 mM Cl−.
A single GABA application at -85 mV produced an outward current, but inward currents were produced when GABA was applied at -85 mV within 2 s of a preceding GABA application at a test potential of -30 mV. At larger time intervals the second GABA application produced outward currents at -85 mV (Fig. 3A). The steady-state ECl in these experiments, obtained from GABA applications at 0.05 Hz, was -98 ± 5 mV (n= 7; Fig. 3B and C). To account for the inward currents evoked 500 ms after the voltage step, ECl would need to shift 28 ± 4.7 mV (n= 7); this corresponds to an increase in the calculated dendritic [Cl−]i from 4.1 ± 0.9 mM at steady state to 10.8 ± 1.8 mM (n= 7). The [Cl−]i was calculated in this manner for each time interval between the loading and test GABAA currents. As shown in Fig. 3D, the rate of recovery of [Cl−]i following the loading current was monoexponential, with a time constant of 3.3 ± 0.2 s (n= 7). The maximum rate of decrease in [Cl−]i, calculated from the initial 100 ms of the fitted exponential function, was 6.1 ± 2.7 mmol l−1 s−1 (n= 7).
Desensitization of the dendritic GABAA receptors during the first GABAA application would reduce the conductance evoked by the second GABA application at short time intervals more than those at longer intervals; this would lead to an underestimate of the initial change in ECl and also the rate at which ECl recovers to the steady-state value. As in the previous experiments, to provide an estimate of the degree to which desensitization may alter our estimate of the activity-dependent change in ECl and the subsequent recovery, a smaller initial GABAA receptor-mediated current was evoked using the same GABA application at -85 mV rather than -30 mV (Fig. 3A, lower traces). If the amplitudes of these currents are assumed to be limited by receptor desensitization, then the GABAA receptor-mediated conductances evoked after the -30 mV initial current can be corrected by a factor proportional to the fractional recovery of the amplitude of the corresponding current evoked after the -85 mV initial GABAA current. This correction would increase by approximately 2 the estimated Cl− accumulation due to the initial current at -30 mV, as well as the rate at which [Cl−]i recovers (Fig. 3D, ○). However, the initial current evoked at -85 mV is likely to increase [Cl−]i to some degree (compare with Fig. 6B), and even a 2 mM increase would account for the change in amplitude of the subsequent currents.
These data indicate that intense activation of dendritic GABAA receptors produces currents sufficient to increase the dendritic Cl− concentration such that ECl is shifted far enough to explain the depolarizing response. To establish whether the net anionic current would change direction as predicted by our estimates of the shift in ECl, we performed voltage clamp experiments at potentials close to RMP using media containing physiological concentrations of HCO3− and Cl−.
CA1 pyramidal neurons were voltage clamped at -55 mV. The current evoked by dendritic application of 100 μm GABA consisted of a brief initial outward current followed by an inward current (Fig. 4A). As in Fig. 3, when the interval between GABA applications was decreased, the initial current was inward, rather than outward; the longer the interval between GABA applications, the more outward the initial current became, reaching a minimum by 10-20 s (Fig. 4A). We considered the large inward currents and the change in direction of the initial current at short application intervals to be due to collapse of the dendritic Cl− gradient in the face of ongoing HCO3− efflux; the outward current was re-established at long application intervals due to recovery of the Cl− gradient. As in Fig. 3, the resting EGABA was established from the initial amplitudes of currents evoked at different test potentials at 0.05 Hz (Fig. 4B and C). We then estimated the shift from the resting ECl and [Cl−]i necessary to explain the change in direction of the GABA currents, assuming EHCO3 remained constant. [Cl−]i following dendritic GABA application is plotted as a function of time in Fig. 4D. [Cl−]i decreased exponentially from the peak concentration at the end of the current evoked by GABA application to the steady-state value with a time constant of 2.3 ± 0.5 s (n= 3) and an average initial rate of decrease of 3.3 ± 0.8 mmol l−1 s−1. The calculated Cl− recovery rate was slower in these experiments compared to those shown in Fig. 3. This is probably due to innaccuracies related to the calculation of ECl and [Cl−]i from mixed anionic currents: we assumed here that the steady-state [Cl−]i= the electrode Cl− concentration, but in the experiments with Cl− as the only permeant anion, [Cl−]i was substantially less than the electrode Cl− concentration due to ongoing Cl− transport (see Fig. 7).
Synaptic anion fluxes
The experiments illustrated in Figs 3 and 4 provide an estimate of the rate of recovery of dendritic [Cl−]i following large Cl− influx. However, the Cl− currents used to increase [Cl−]i were induced by application of exogenous GABA. To determine whether Cl− currents resulting from synaptic activity could exceed transport capacity and thereby increase [Cl−]i, we elicited pharmacologically isolated GABAA postsynaptic currents by brief tetanic stimulation in the s. moleculare in nominally HCO3−-free media (Fig. 5A). The amplitude of the dendritic GABAA PSCs was increased by clamping the membrane at more depolarized test potentials. In this way the amplitude of the GABAA receptor-mediated current could be increased without altering the amount of GABA released at the activated synapses, which was important in order to avoid interpretational difficulties related to possible differences in the ionic permeabilities of synaptic vs. extrasynaptic GABAA receptors, or receptor desensitization. For large GABAA receptor-mediated conductances, as the test potential is made more positive, Cl− influx will increase to the point where it exceeds the vmax of Cl− transport. This will lead to Cl− accumulation during large PSCs with a consequent shift in ECl and decrease in the driving force. Thus at more positive test potentials, the GABAA current should be smaller than predicted based on the GABAA conductance and the initial driving force, so that the I-V relationship would become more shallow at test potentials substantially more positive than the resting ECl. The current amplitude at the point where the slope of the I-V curve decreases equals the maximum Cl− transport rate at the resting [Cl−]i.
In the initial series of synaptic Cl− loading experiments, the effect of Cl− accumulation on the GABAA PSCs was quite variable (n= 3). This variation was decreased substantially when GABAA receptors at the soma and proximal dendrites were blocked by local pressure application of 100 μm picrotoxin, suggesting (1) that tetanic stimulation was activating axons that were presynaptic to both dendritic and somatic GABAA receptors (Freund & Buzsaki, 1996), and (2) that large somatic GABAA currents were more stable than the dendritic currents (Fig. 2). When proximal GABAA receptors were blocked, the amplitude of PSCs evoked at depolarized membrane potentials by intense synaptic activation of dendritic GABAA receptors was smaller than predicted based on extrapolation from PSCs evoked at test potentials closer to resting ECl (decrease in actual vs. predicted charge transfer at -30 mV = -29 ± 8 % when resting ECl= -70 mV; n= 4; Fig. 5B; compare with similar effect demonstrated in Fig. 3C). In contrast, smaller PSCs evoked concurrently using single stimuli did increase as expected at the same test potentials, so that the depression of the large GABA currents was unlikely to be due to voltage-dependent alterations in GABAA receptor function or deterioration of the pre- or postsynaptic neurons. Further, when the experiment was repeated in bicarbonate-buffered, low-Cl− media to test the stability of the HCO3− transmembrane gradient, the HCO3− currents elicited by tetanic stimuli changed with the test potential as predicted by extrapolation from the currents elicited by single stimuli (maximum difference = 4 ± 9 %, n= 3; Fig. 5C-E). The simplest way to explain the nonlinear increase in the amplitude of the large vs. small PSCs carried by Cl− at depolarized test potentials is that at high current densities, Cl− influx exceeded the maximum Cl− cotransport rate, resulting in accumulation of [Cl−]i and a deterioration in the driving force and therefore the Cl− current. In contrast, the PSC carried by HCO3− shows no evidence of alteration in the transmembrane HCO3− gradient.
We obtained the GABAA conductance from the slope of the I-V relationship near ECl. The amount of Cl− influx was estimated by integrating the PSC waveform, and the shift in the Cl− driving force at depolarized test potentials was estimated from the difference between the extrapolated and the actual current amplitude (see Methods). Using these estimates, several parameters can be derived from the experiment shown in Fig. 5A and B, including the time-averaged depolarizing shift in ECl (+7.7 ± 1.2 mV; n= 4), the corresponding increase in [Cl−]i (4 ± 0.6 mM), the volume of the dendrites activated by the large IPSC (4 × 10−13± 1.5 × 10−13 l, which for a sense of scale corresponds to a 500 μm length of dendrite 1 μm in diameter), and the maximal rate of dendritic Cl− cotransport, 7 ± 1.5 mmol l−1 s−1. This value, which is not affected by receptor desensitization, is quite close to the estimate of 6.1 mmol l−1 s−1 obtained from the experiments shown in Fig. 3. However, there are several limitations to this experiment: first, only a few tetanic stimuli could be delivered before the postsynaptic response degraded. This limited the number of test potentials at which we could elicit tetanic responses, and thus the accuracy with which we could determine the maximum transport rate. The degradation after tetanic stimulation was presumably due to injury of nerve terminals arising from the high current densities required to elicit a large GABA release when ionotropic glutamate receptors were blocked. To control for this effect (a) smaller PSCs were evoked using single stimuli to assess the stability of the response, (b) tetanic responses at positive potentials were evoked before evoking the responses near ECl, and (c) tetanic stimuli were applied at 10 min intervals to avoid alterations in GABA release or Cl− transport due to changes in [K+]o (Kaila et al. 1997). The second experimental limitation was the accuracy with which the distal dendritic membrane potential could be controlled during the tetanic stimulation at positive test potentials. For the experiments shown in Fig. 5, to improve control of the dendritic membrane potential, Cs+ rather than K+ had to be used in the electrode solution (see Methods). Finally, tetanic stimulation releases a variety of transmitters and modulators besides GABA. For instance, glutamate release will activate both metabotropic glutamate receptors and electrogenic glutamate reuptake.
To avoid these experimental limitations, we repeated the protocol shown in Fig. 3, using single synaptic activation rather than tetanic stimulation or exogenous GABA application. Recordings were performed using the same low-HCO3− media as in Fig. 3. Ionotropic glutamate receptors and GABAB receptors were blocked, and pentobarbitone (50 μm) was added to the bath to enhance the GABAA receptor-mediated postsynaptic conductance (Alger & Nicoll, 1982; Thallman, 1988). Under these conditions, large (80 V for 80 μs) single stimuli in the distal dendritic layer produced enough Cl− loading at -30 mV to decrease the amplitude of subsequent GABAA receptor-mediated currents elicited by the same stimuli at -80 mV (Fig. 6). The average initial rate of decrease of dendritic Cl− concentration was calculated as for Fig. 3, and in these experiments was 2.8 ± 0.2 mmol l−1 s−1 (n= 6). Comparison of Fig. 6D with Fig. 3D indicates that the lower rate of decrease of Cl− appears to be related to the smaller increase in [Cl−]i in these experiments, suggesting that the GABAA current evoked by electrical stimulation was distributed more widely among the dendrites than the current evoked by local GABA application. Thus in these experiments it is less likely that we were measuring the vmax of the transporter.
Calculation of the dendritic transport rate from the decrease in the amplitude of the Cl− current evoked by GABA application at depolarized potentials relative to that predicted from the slope of the I-V curve near ECl (Fig. 3C) in recordings utilizing K+ electrode solutions led to a value of 5.4 ± 2.3 mmol l−1 s−1 (n= 7, same cells and experiments as shown in Fig. 3). This suggests that the value of 6.1 mmol l−1 s−1 calculated from the rate of recovery of the dendritic Cl− currents after exogenous GABA application, and the value of 7 mmol l−1 s−1 calculated from the tetanic stimulation experiments are reasonable estimates for the vmax of dendritic KCl cotransport, and that factors such as receptor desensitization (Fig. 3D) do not substantially reduce the accuracy of these calculated values.
Cl− affinity of the transport process
In order to model the Cl− transport, it was necessary to know the affinity of the transport process for [Cl−]i. In the next experiments, we loaded neurons with Cl− via the recording electrode. If Cl− is actively transported out of the neuronal cytoplasm by a saturable KCl cotransporter (Misgeld, 1986; Kaila, 1994), then loading neurons with Cl− via the recording electrode should result in a steady state between the rate of Cl− influx from the electrode and the rate of outward Cl− cotransport. Although the rate of Cl− influx from the electrode cannot be quantified, [Cl−]i can be estimated from the reversal potential of the GABAA response (Fig. 7A and B), so that the activity of the cotransporter can be characterized in terms of [Cl−]i. Figure 7C illustrates this approach: for any given electrode Cl− concentration, the calculated [Cl−]i is lower than the electrode Cl− concentration, and the difference between the electrode Cl− concentration and [Cl−]i is proportional to [Cl−]i. In these experiments, nominally HCO3−-free intra- and extracellular solutions were used to minimize HCO3− flux, and the holding potential was set equal or positive to the predicted EGABA to minimize conductive Cl− efflux through ClC-2 (Staley, 1994; Staley et al. 1996). The increase in error in calculated [Cl−]i at higher electrode Cl− concentrations in Fig. 7C may reflect the combined effects of an increase in the somatic-to-dendritic Cl− gradient and the experimental variation in the subcellular distribution of activated GABAA receptors. When the transport velocity was expressed as the difference between the electrode Cl− and the cytoplasmic Cl−, and the concentration of the transported species (Cl−) was expressed as the cytoplasmic Cl− concentration, then the kinetics of Cl− transport could be expressed using Michaelis-Menten kinetics, with a vmax of 38 mM DEC−1 and a KD of 15 mM (Fig. 7D). The vmax is not a physiologically relevant value, since it reflects the difference between electrode and cytoplasmic Cl− concentration when Cl− cotransport was maximal, and as calculated here (see Methods) included the term DEC, the electrode-to-cytoplasm diffusion coefficient whose value is unknown. The vmax of 5-7 mmol l−1 s−1 obtained from the experiments in Figs 3-5 is more useful. However, this experiment does provide a useful KD of 15 mM, which represents the cytoplasmic Cl− concentration at which somatic Cl− cotransport is half-maximal.
One problem with characterizing Cl− cotransport using enzyme kinetic models is that the concentration of the substrate, [Cl−]i, affects not only the probability of binding to the transporter, but also the amount of energy required to transport Cl− across the neuronal membrane. The Michaelis-Menten model assumes that increasing Cl− affects only the probability of binding to the transporter, but changing [Cl−]i also changes ECl. The free energy gained by Cl− during transmembrane transport is proportional to the difference between the membrane potential and ECl, thus changing [Cl−]i also changes the amount of work done by the transporter. The source of this additional energy is the transmembrane K+ gradient, and is therefore proportional to the difference between the membrane potential and EK (Thompson & Gähwiler, 1989b; Alvarez-Leefmans, 1990). Thus for any given membrane potential, the difference between EK and ECl represents the free energy available for KCl cotransport. The data used in Fig. 7D can also be expressed as in Fig. 7E to show that the potential difference generated by the cation-Cl− cotransporter is proportional to the free energy available for transport. This indicates that at physiological concentrations of K+ and Cl−, either the affinity of the transporter for Cl− or the free energy gradient for K+vs. Cl− (Thompson et al. 1988; Thompson & Gähwiler, 1989b) may have limited Cl− cotransport measured in these experiments.
Use of anionic transport data to model the GABAA depolarizing response
The data characterizing dendritic and somatic anionic homeostasis are sufficient to test the hypothesis that large GABAA currents cause a rundown in the transmembrane Cl− gradient and thereby shift EGABA towards EHCO3 (Fig. 8A). We used a standard model of membrane potential (Fig. 8B; Finkelstein & Mauro, 1977; see Methods) incorporating a GABAA conductance (Fig. 8B and C), a leak conductance responsible for the resting membrane potential and input conductance, a Cl− transport system with a vmax of 5 mmol l−1 s−1 and a KD of 15 mM, a resting Cl− leak sufficient to keep the resting Cl− at 2 mM in the face of outward transport, a stable HCO3− gradient, and a variable volume into which the GABAA current flowed. This simple model accurately predicts the duration and waveform of the hyperpolarizing and depolarizing GABAA responses in structures with volumes similar to dendrites and soma (Fig. 8D; compare with Fig. 1A). The kinetics of the hyperpolarizing and depolarizing response vary substantially depending on the volume into which the Cl− current flows; Cl− accumulation results in much smaller changes in [Cl−]i in large structures due to dilution, so ECl is more stable in larger structures such as the soma (Alger & Nicoll, 1982; Staley et al. 1995). The time courses of Cl− accumulation and extrusion (Fig. 8E) are consistent with observations regarding the direction and amplitude of a second GABAA current evoked shortly after a conditioning GABAA current (Figs 2-4 and 6). The accuracy of this simple model suggests that the proposed mechanism of anionic gradient collapse is a sufficient explanation of the depolarizing GABAA response.
Recent findings suggest that activity-dependent increases in [K+]o may contribute to GABAA receptor-mediated, HCO3−-dependent, post-tetanic membrane depolarization (Kaila et al. 1997). Although the increase in [K+]o may directly depolarize the cell membrane, increased [K+]o will also decrease KCl cotransport (Thompson & Gähwiler, 1989b). We tested the effects of increased [K+]o in the Cli− accumulation model by altering the maximum velocity of the KCl cotransporter (by decreasing EK; Thompson & Gähwiler, 1989b). GABAA receptor-mediated responses were elicited by single electrical stimuli in s. moleculare of CA1. The stimulus intensity was increased until the late membrane potential response was depolarizing, and then the stimulus was decreased until no depolarizing component was present. At this stimulus intensity, the late depolarizing component reappeared when the extracellular [K+] was increased from 2.5 to 8.5 mM (n= 5; Fig. 9). In each case the control and high-[K+]o waveforms could be well approximated using the model shown in Fig. 8. The only model parameters altered between the control and high-[K+]o responses were gGABA (McBain, 1994) and vmax. Thus the Cli− accumulation model is sufficient to predict the effects of physiologically relevant alterations in GABAA receptor-mediated Cl− flux or Cl− transport on the kinetics and direction of the GABAA response.
These experiments demonstrate that neuronal transmembrane Cl− transport is rate limited, that there is significant alteration of dendritic [Cl−]i and ECl as a result of intense GABAA receptor activation, and that the transmembrane gradients for Cl− at the soma and HCO3− in the dendrites are significantly less affected by large GABAA currents. Incorporation of these observations into a standard model of membrane potential accurately reproduces both the hyperpolarizing and depolarizing GABAA responses.
It would be conceptually simpler to perform the experiments demonstrating the stability of ECl at the soma (Fig. 2) using gramicidin recordings in order to eliminate the possibility that the somatic Cl− concentration was buffered by the electrode solution ‘reservoir’. However, gramicidin recordings were not sufficiently stable during the steps to and from -30 mV to be useful. Conductive channels in the membrane are formed from gramicidin dimers, and large voltage steps are likely to cause alterations in gramicidin dimerization in the membrane, as a consequence of either voltage-dependent electrostatic protein interactions or changes in intracellular Ca2+ during voltage steps (Lundbaek et al. 1997). Fortunately such recordings are not necessary to demonstrate stability of somatic ECl. KCl cotransport decreased the somatic Cl− concentration to a steady-state value that was less than that of the electrode solution: 7 mM in cytoplasm based on ECl (Fig. 2D) vs. 10 mM in the pipette solution. Thus the direction of Cl− diffusion is from the electrode to the cytoplasm, so that the stability of the somatic ECl cannot be ascribed to Cl− flux from the soma to the electrode.
Underestimation of distal dendritic currents could arise due to imperfect voltage control of the distal dendrites. Integrating the current over time to obtain transferred charge or using time-averaged currents limits this error to about 10 % (Staley & Mody, 1992). When Cl− transport is measured by the rate of change of ECl after a Cl− load (Figs 3, 4 and 6), the Cl− load will be underestimated, but the rate of change in ECl will not be underestimated as long as the ∼10 % error affects the amplitudes of each of the subsequent GABAA receptor-mediated currents equally. This could produce a 10 % underestimate of the maximum transport rate, since the actual Cl− load was larger than measured. When Cl− transport is measured from the decline in actual vs. expected current at positive test potentials (Figs 3C and 5B), space clamp errors will also lead to an underestimate of the maximum transport rate, because the current and the corresponding Cl− influx rate at which the response deviates from expected is larger than measured. If both distal and proximal dendritic GABAA receptors are activated at the same time, the space clamp error will be larger because of the decreased membrane resistance of the proximal dendrites: the amplitude of the depolarizing response in distal dendrites will be shunted by GABAA conductances in larger, more proximal structures (Staley & Mody, 1992). We addressed this problem by utilizing either focal application of GABA or in some cases by prior application of GABAA antagonists to the soma and proximal dendritic regions.
Cl− efflux via the ClC-2 Cl− channel may have added to KCl cotransport in some experiments, even though we had set the holding potentials to minimize activation of ClC-2 (Staley, 1994; Staley et al. 1996). This is likely to explain the correlation between ECl and the holding potential from which ECl was determined: in Fig. 2, the holding potential was -70 mV and ECl was -78 mV, while in Figs 5 and 6 the holding potential was -80 or -85 mV and ECl was ∼-98 mV; all these experiments were performed using an electrode Cl− concentration of 10 mM. The Cl− transport rate calculated from the experiments shown in Figs 3C and 5B were not affected by Cl− efflux via ClC-2, because the holding and test potentials were too far from the ClC-2 activation potential. The transport rate in these experiments is quite close to the transport rates calculated from Fig. 3D, indicating that Cl− efflux via ClC-2 was small during the time interval between the conditioning Cl− current and the test currents, reflecting the slow activation of ClC-2 (Staley, 1994).
Assumptions of the gradient collapse model
The basic assumption of the gradient collapse hypothesis is that Cl− influx through GABAA receptors can increase the local intracellular Cl− concentration so that EGABA is shifted in a positive direction. This seems contrary to the fact that only a tiny fraction of the ions present on either side of the membrane need to cross over in order to produce a charge imbalance sufficient to alter the membrane potential. For instance, the sequential activation of Na+ and K+ conductances during an action potential cause large changes in membrane potential but insignificant alterations in Na+ and K+ concentrations (Hille, 1992). This is because the ionic flux during the action potential is small due to the brevity of the conductances and the sequential activation of gNa and gK, which minimizes ‘short-circuiting’ of one current by the other. In contrast, bidirectional anionic flux through the GABAA receptor results in a severe short-circuit condition. Thus during prolonged activation of GABAA receptors, even though the membrane potential is at EGABA, a substantial fraction of the dendritic transmembrane Cl− gradient is being reduced by ongoing Cl− currents because the membrane potential cannot reach either EHCO3 or ECl.
A related assumption of the gradient collapse hypothesis is that the rate of Cl− transport can be exceeded by the rate of GABA-mediated Cl− influx (Wong & Watkins, 1982; Ballanyi & Grafe, 1985; Huguenard & Alger, 1986; Thompson & Gähwiler, 1989a). The experiments using high-intensity synaptic and exogenous activation of dendritic GABAA receptors (Figs 2-6) demonstrate that synaptic Cl− influx through the GABAA ionophore can exceed the maximum transport rate. The experiments illustrated in Fig. 2 support the assumption that at the soma, even if the rate of synaptic Cl− influx exceeds the transport rate, the volume of the soma precludes a significant change in [Cl−]i and the corresponding shift in ECl for at least several seconds (Fig. 8E and F; Ballanyi & Grafe, 1985), so that the GABAA current remains hyperpolarizing.
The final assumption of the gradient collapse hypothesis is that the dendritic HCO3− transmembrane gradient is more stable than the corresponding Cl− gradient. This is reasonable based on the pH alterations that would accompany a dissipation of the HCO3− gradient: at a constant pCO2, if EHCO3 was driven to RMP, the intracellular pH would be 6.3, which would denature proteins and damage the neuron. To avoid this, the intracellular HCO3− concentration is stabilized by diffusion and hydration of CO2 (Pasternak et al. 1993), while the pH is supported by intracellular buffers and H+ extrusion (Kaila & Voipio, 1987; Chen & Chesler, 1992; Staley, 1995). The data in Figs 2E and F and 5C-E confirm that the dendritic HCO3− concentration is maintained so that EHCO3 does not vary significantly during maximal dendritic GABAA receptor activation.
We also assumed in our calculations that changes in ECl are due to changes in [Cl−]i, because similar changes in the extracellular Cl− concentration produce insignificant changes in ECl (eqn (3)) and because activity-dependent changes in ECl have been demonstrated in cultured cells, where the supply of extracellular Cl− is essentially unlimited (Barker & Ransom, 1978; Dallwig et al. 1999). Although we have described the recovery of [Cl−]i as a function of active transport, the recovery could be due to both active transport and diffusion of Cl− away from the intracellular regions closest to the open GABAA ionophores. In the modelling studies of the GABA depolarizing response, the amount of Cl− entering the dendrite during large GABAA currents was sufficient to raise the [Cl−]i of the entire volume of the dendritic cytoplasm, so that there was no need to include a radial variation in [Cl−]i. Thus Cl− diffusion would only be significant at the ends of the active dendritic volume, and was neglected in the model.
Several explanations for the activity-dependent GABAA depolarizing response have been proposed (Alger & Nicoll, 1982; Staley et al. 1995; Perkins & Wong, 1996; Kaila et al. 1997). An intriguing recent proposal is that the GABA depolarizing response is due to the depolarizing effects of increased [K+]o which arises from activation of an interneuron network that is dependent on both GABA and HCO3−. However, increases in [K+]o are not necessary for the GABA depolarizing response, since responses with identical kinetics have been described in cultured cells, where [K+]o is stabilized by the large size of the extracellular space (Dallwig et al. 1999). The study by Dallwig et al. also supports the Cl− gradient collapse hypothesis by demonstrating the predicted increase in intracellular Cl− using Cl−-sensitive fluorescent dyes. A second difficulty with the K+o accumulation hypothesis is the lack of effect of TTX on the depolarizing response (Fig. 1A), which is difficult to reconcile with activation of an interneuron network. The increases in [K+]o of several millimolar occurred in response to prolonged tetanic stimulation (Kaila et al. 1997), whereas focal application of GABA agonists (Figs 1-4) produce much smaller changes in [K+]o (Müller et al. 1989). The increase in [K+]o generated by focal GABAA agonist application is similar in the dendrites and the soma (Müller et al. 1989); thus a direct membrane depolarization by [K+]o is not consistent with the lack of depolarizing response at the soma (Barker & Ransom, 1978; Andersen et al. 1980; Alger & Nicoll, 1982; Scharfman & Sarvey, 1987; Fig. 2). We suggest that GABA-dependent increases in [K+]o arise from KCl exchange (Müller et al. 1989; Payne, 1997) and that activity-dependent increases in [K+]o augment the depolarizing response primarily by decreasing the KCl transport rate, thereby enhancing Cl− accumulation (Fig. 9).
An argument against the Cl− gradient collapse hypothesis is that the stability of the HCO3− gradient appears to require both extra- and intracellular carbonic anhydrase activity (Fig. 8A) but benzolamide, a carbonic anhydrase inhibitor that is confined to the extracellular space, does not block the depolarizing GABA response (Kaila et al. 1997). However, if extracellular HCO3− is removed by other means such as diffusion or glial transport then inhibition of extracellular carbonic anhydrase will not limit HCO3− efflux. In addition, because the concentrations of HCO3− on both sides of the membrane are nearly symmetrical, a very large change in extracellular HCO3− would be required to significantly alter the HCO3− reversal potential (eqn (3)). For example, even if extracellular HCO3− increased by the same amount as intracellular Cl− (e.g. ∼7 mM calculated in Fig. 4), EHCO3 will only change by 7 mV, vs. 28 mV for ECl. For a RMP of -70 mV, this would represent a complete abolition of the driving force for Cl−, but the driving force for HCO3− would only decrease from 57 to 50 mV. Thus the small effect on the depolarizing GABA response produced by inhibition of extracellular carbonic anhydrase (Kaila et al. 1997), in contrast to the marked effect of inhibition of both intra- and extracellular carbonic anhydrase by acetazolamide (Staley et al. 1995), is entirely consistent with the proposed Cl− gradient collapse mechanism.
Our data do not support the existence of a subset of GABAA receptors with high HCO3− or cationic permeabilities as an explanation for depolarizing GABAA responses: such receptors were not found when applying GABA directly to the dendrites (Figs 2-4). Additionally, in Fig. 4A the change in the direction of the initial current as a function of time since the last GABA application cannot be explained on the basis of diffusion of GABA to receptors with unique ionic permeabilities.
A distinction should be made between the phenomenon we have examined in this paper, namely the activity-dependent GABAA dendritic depolarization that occurs in neurons that are normally hyperpolarized by brief GABAA receptor activation, and the rapid activity-independent GABAA depolarizing response that occurs in many neurons during development (Cherubini et al. 1991; Owens et al. 1996), in some adult neurons (Staley, 1996) and following particular experimental manipulations (Van den Pol et al. 1996; Cerne & Spain, 1997). These shifts in EGABA occur under different conditions and on different time scales to the phenomena studied here, are HCO3− independent (Owens et al. 1996; Staley et al. 1996), and are thus likely to be related to long-term regulation of Cl− transport mechanisms (Rohrbough & Spitzer, 1996; Staley et al. 1996).
Electrical stimulation of GABAA terminals produced variable responses, presumably because variation in the length and distribution of the stimulated GABAergic axon terminals results in the activation of a mix of somatic and distal and proximal dendritic GABAA receptors that cannot be predicted from the location of the stimulating electrode. This variability applies to the effects of activation of individual interneurons: the interaction between the dendritic depolarizing response and somatic hyperpolarizing response will vary substantially as a function of the subcellular distribution of the GABAergic terminals of each interneuron (Freund & Buzsáki, 1996). Related to this problem, it has not been determined whether the discharge of a single interneuron or an interneuron network is required to elicit the depolarizing response.
At the soma, although Cl− transport can be saturated, ECl is stable because the volume is large enough to prevent significant alterations in [Cl−]i. The analogous question has not been answered for dendritic HCO3− homeostasis: is the rate of H+ transport so high that GABAA receptor-mediated HCO3− efflux does not alter the intracellular pH, or is the intradendritic H+ concentration stabilized by high buffer capacity? Further characterization of the molecular and functional characteristics of neuronal proton and anion transporters are necessary to resolve these issues.
Physiological significance of the GABAA depolarizing response
A large inefficiency in GABAA receptor function results from bi-directional anionic flux and the consequent dissipation of energy stored in the transmembrane ionic gradients (Fig. 8A). We suggest that this energetically inefficient bi-directional anionic flux is worthwhile because it provides a mechanism for frequency modulation of the GABAA response from inhibition to excitation. This activity-dependent change in GABA response is useful for regulating the voltage-dependent Mg2+ block of the NMDA receptor (Staley et al. 1995), and may contribute to the activity-dependent alterations in inhibition and NMDA receptor function that modulate synaptic plasticity (Bliss & Collingridge, 1993).
The dendritic location and activity dependence of the depolarizing GABAA response suggest that the GABA-releasing interneuron that elicits the response would need to fire in rapid bursts and have a significant axonal distribution in the distal dendrites of the pyramidal cells. Regulation of interneuron burst firing by neuromodulators (Freund & Buzsaki, 1996; Bergles et al. 1996) could therefore restrict the occurrence of the GABA depolarization to specific states. Alterations in synaptic inhibition are an important component of the pathophysiology of epilepsy (Prince & Connors, 1984). Thus dysregulation of GABA-mediated depolarization, for instance due to alterations in the anatomical distribution of the efferent and afferent connections of GABAergic interneurons (Babb et al. 1989), could contribute to epileptogenesis. The anticonvulsant action of the carbonic anhydrase inhibitor acetazolamide, an inhibitor of GABA-mediated depolarization (Staley et al. 1995), supports the idea that this phenomenon may contribute to some epileptic conditions, and suggests that manipulation of neuronal anionic homeostasis may be a promising anticonvulsant strategy.
This study was supported by NIH grants NS34360 and NS34700, the American Heart Association, and the Epilepsy Foundation of America.