3-(aminosulfonyl)-5 (butylamino)-4-phenoxybenzoic acid
intracellular carbonic anhydrase
(2S)-3-[[(1S)-1-(3,4-dichlorophenyl)ethyl]amino-2-hydroxypropyl] (phenylmethyl) phosphinic acid
- E GABA
reversal potential of GABAA channel-mediated response
type A GABA receptor
GABAAR-mediated inhibitory postsynaptic potential
K+–Cl− cotransporter isoform 2
Na+-dependent K+–Cl− cotransporter
GABAergic excitatory [K+]o transients can be readily evoked in the mature rat hippocampus by intense activation of GABAA receptors (GABAARs). Here we show that these [K+]o responses induced by high-frequency stimulation or GABAA agonist application are generated by the neuronal K+–Cl− cotransporter KCC2 and that the transporter-mediated KCl extrusion is critically dependent on the bicarbonate-driven accumulation of Cl− in pyramidal neurons. The mechanism underlying GABAergic [K+]o transients was studied in CA1 stratum pyramidale using intracellular sharp microelectrodes and extracellular ion-sensitive microelectrodes. The evoked [K+]o transients, as well as the associated afterdischarges, were strongly suppressed by 0.5–1 mm furosemide, a KCl cotransport inhibitor. Importantly, the GABAAR-mediated intrapyramidal accumulation of Cl−, as measured by monitoring the reversal potential of fused IPSPs, was unaffected by the drug. It was further confirmed that the reduction in the [K+]o transients was not due to effects of furosemide on the Na+-dependent K+-Cl− cotransporter NKCC1 or on intraneuronal carbonic anhydrase activity. Blocking potassium channels by Ba2+ enhanced [K+]o transients whereas pyramidal cell depolarizations were attenuated in further agreement with a lack of contribution by channel-mediated K+ efflux. The key role of the GABAAR channel-mediated anion fluxes in the generation of the [K+]o transients was examined in experiments where bicarbonate was replaced with formate. This anion substitution had no significant effect on the rate of Cl− accumulation, [K+]o response or afterdischarges. Our findings reveal a novel excitatory mode of action of KCC2 that can have substantial implications for the role of GABAergic transmission during ictal epileptiform activity.
The K+–Cl− cotransporter isoform 2 (KCC2) is exclusively expressed in central neurons where it is responsible for maintaining an inwardly directed driving force for chloride ions, which is a prerequisite for hyperpolarizing inhibition (Payne et al. 1996; Rivera et al. 1999; Hubner et al. 2001; for review, see Blaesse et al. 2009). At early stages of development before KCC2 expression has reached a significant level, activation of ionotropic type A GABA receptors (GABAARs) generates responses that are depolarizing and even excitatory (Ben-Ari, 2002). In rat hippocampal pyramidal neurons, a second qualitative developmental change in GABAAR-mediated signalling takes place upon expression of intrapyramidal carbonic anhydrase (CA) isoform VII at around postnatal day (P) 12 (Ruusuvuori et al. 2004). While GABAAR-mediated inhibitory postsynaptic potentials (IPSPs) are hyperpolarizing in CA1 pyramidal neurons after P12, a more intense activation of GABAARs gives rise to biphasic responses where the initial hyperpolarization is followed by a depolarization that can excite pyramidal neurons and promote synchronous spiking (for review, see Voipio & Kaila 2000; Rivera et al. 2005). These responses resemble the dose-dependent multi-phasic GABAAR-mediated responses that have been observed in slices from mature hippocampi upon microinjections of GABAAR agonists (Alger & Nicoll, 1982).
The depolarizing and excitatory phase in stimulation-induced bicarbonate-dependent GABAergic responses in rat CA1 hippocampal pyramidal neurons has been shown to be due to anion redistribution followed by a transient increase in the extracellular concentration of K+ ([K+]o; Kaila et al. 1997; Smirnov et al. 1999; see also Wong & Watkins, 1982). Pronounced GABAergic [K+]o transients are generated only in the presence of CO2/HCO3− and functional intrapyramidal CA (Ruusuvuori et al. 2004), suggesting that pyramidal neurons are crucially involved in the generation of GABAergic [K+]o transients. Several observations have shown that prolonged activation of GABAARs leads to a large intracellular chloride load that is generated by a GABAAR-mediated Cl− influx driven by bicarbonate-dependent depolarization (Kaila & Voipio, 1987; Kaila et al. 1989; Thompson & Gähwiler, 1989a; Staley & Proctor, 1999). We have speculated before that the GABAergic [K+]o response could be accounted for by a recovery from this intraneuronal Cl− load by either a parallel one-to-one K+ and Cl− efflux via potassium and chloride channels or by a KCl efflux via KCC2 (Voipio & Kaila, 2000; Blaesse et al. 2009). However, the precise mechanism and cellular source underlying GABAergic excitatory [K+]o transients have remained unidentified.
GABAergic excitation resulting in synchronous firing of CA1 pyramidal neurons in mature hippocampal slices is readily inhibited by membrane-permeant carbonic anhydrase inhibitors such as acetazolamide (Grover et al. 1993; Staley et al. 1995), which has been used in the treatment of human epilepsies. This is interesting in the light of activity-related accumulation of extracellular potassium and its possible causal role in seizure initiation (Yaari et al. 1986; Fröhlich et al. 2008). Therefore, identification of the ionic mechanism that can lead from intensive activation of GABAARs to tonic excitation of neuronal networks by extracellular K+ is important for understanding both the physiology and pathophysiological states of the hippocampus.
In this study, we put forward a novel mechanism that accounts for the GABAAR-dependent increase in [K+]o by a release of K+ via KCC2, and provide experimental evidence obtained in rat hippocampal slices to support this hypothesis. In brief, GABAARs mediate a net uptake of Cl−, which leads to increased extrusion of Cl− and K+ by K+–Cl− cotransport. Cl− uptake is driven by the depolarizing action of HCO3− efflux via GABAARs (i.e. an inward current), which does not fade because of rapid replenishment of HCO3− by intracellular CA activity. Hence, the key molecules in this mechanism, in addition to GABAA receptors, are the intrapyramidal carbonic anhydrase (Ruusuvuori et al. 2004) and the neuronal K+–Cl− cotransporter KCC2 (Rivera et al. 1999). The present data fully support the view that pyramidal neurons constitute the main source of K+ release.
The results presented in this study reveal a functional link between CA activity and KCC2 in CA1 pyramidal neurons that may provide a significant contribution to the [K+]o transients that are characteristic during epileptiform activity (Yaari et al. 1986; Fröhlich et al. 2008). Hence, it is possible that inhibition of the mechanism studied in this work is involved in the anti-epileptic actions of furosemide (Hochman et al. 1995; Gutschmidt et al. 1999; Holtkamp et al. 2003) and of carbonic anhydrase inhibitors (Thiry et al. 2007).
All experiments were approved by the Ethics Committee for Animal Research at the University of Helsinki, and all our experiments comply with the ethical policies and regulations of The Journal of Physiology (Drummond 2009).
Slices and solutions
All experiments were performed on rat hippocampal slices at P15–P35. Rats were anaesthetized using an intraperitoneal injection of pentobarbital (40–60 mg kg−1). After decapitation the brain was quickly removed and placed into ice-cold standard solution containing (in mm): 124 NaCl, 3 KCl, 2 CaCl2, 25 NaHCO3, 1.1 NaH2PO4, 2 MgSO4 and 10 d-glucose, and equilibrated with 95% O2–5% CO2. Transverse hippocampal slices (350–400 μm; 250 μm for microspectrofluorometry) were cut using a Leica VT1000S vibrating-blade microtome (Nussloch, Germany). Slices were allowed to recover for 30 min at 32°C in the standard solution that was continuously bubbled with 95% O2–5% CO2 (pH 7.4). Thereafter, the slices were stored for at least 30 min at room temperature before the experiments were started.
In the CO2/HCO3−-free Hepes-buffered solution, NaHCO3 was replaced with 20 mm Hepes (pH 7.4 with NaOH), and the solution was gassed with 100% O2. The formate-containing solution was made by substituting 20 mm sodium formate for 20 mm NaCl in the CO2/HCO3−-free Hepes-buffered solution (100% O2, pH 7.4). When monitoring transient changes in interstitial volume, 1.5 mm tetramethylammonium (TeMA+) chloride was added to the superfusing saline. Barium rather than cesium was used as a blocker of K+ channels, since Cs+ is known to have an inhibitory action on KCC2 (for references, see Blaesse et al. 2009). When BaCl2 (2 mm) was added to the saline, MgSO4 was replaced with 2 mm MgCl2 to prevent precipitation.
dl-2-amino-5-phosphonopentanoic acid (AP-5, 40 μm), 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxa-line-7-sulfonamide disodium salt (NBQX, 10 μm), bicuculline methiodide (BMI, 10 μm), picrotoxin (PiTX, 100 μm), (2S)-3-[[(1S)-1-(3,4-dichlorophenyl)ethyl]amino-2-hydroxypropyl] (phenylmethyl) phosphinic acid (CGP-55845, 1 μm), tetrodotoxin (TTX, 0.5–1 μm), isoguvacine hydrochloride (applied by microinjections or at 200 μm in bath), and 5-aminomethyl-3-hydroxyisoxazole (muscimol; microinjections) were from Tocris Cookson (Bristol, UK). Furosemide (0.5 or 1 mm), ethoxyzolamide (EZA, 100 μm), and 3-(aminosulfonyl)-5 (butylamino)-4-phenoxybenzoic acid (bumetanide, 10 μm) were from Sigma (St Louis, MO, USA). The membrane-impermeant carbonic anhydrase inhibitor benzolamide (BA) was a gift from Professor E. Swenson (Department of Medicine and Laboratory Medicine, University of Washington Medical Center, Seattle, WA, USA).
Electrophysiological recordings were carried out in a submerged-type recording chamber (800 μl, 33–34°C), where slices were continuously superfused on both sides at a flow rate of 2.0 ml min−1 (4.7 ml min−1 in experiments with bath-applied agonists). Measurements were done within stratum pyramidale of the CA1 region. Potassium-sensitive microelectrodes were made using a valinomycin-based membrane solution (Cocktail B, Fluka) and backfilled with 150 mm NaCl plus 3 mm KCl (Voipio et al. 1994). In TeMA+-sensitive electrodes the membrane solution consisted of 100 mg of potassium tetrakis (4-chlorophenyl) borate in 2 ml of 1,2-dimethyl-3-nitrobentzene and the backfilling solution was 150 mm TeMA-Cl. In simultaneous recordings with valinomycin-based K+ microelectrodes and TeMA+-sensitive microelectrodes, their tips were placed within 50 μm from each other in CA1 stratum pyramidale. Extracellular field potentials were recorded using glass capillary microelectrodes filled with a solution containing 150 mm NaCl and 3 mm KCl, and having a resistance of 2–5 MΩ. Sharp intracellular microelectrodes were filled with 1 m potassium acetate plus 5 mm KCl (pH ∼6.8) and they had a resistance of 90–160 MΩ.
For high-frequency stimulation (HFS), a bipolar electrode was inserted in stratum radiatum near stratum pyramidale border in the vicinity (≤500 μm) of the recording site, and trains of stimuli (40 pulses of 3–20 V and 50–60 μs at 100 Hz) were delivered at 10–15 min intervals. Pressure microinjections of isoguvacine hydrochloride (2–10 mm, dissolved in the CO2/HCO3−-free solution) were applied via a glass capillary microelectrode (tip diameter 2–4 μm) positioned close to the recording site using brief pulses of pressure (18 psi or 124 kPa, duration 0.25–2 s) every 10–15 min. Muscimol was applied iontophoretically from micropipettes with 2–4 μm tips and filled with 10 mm muscimol in 10 mm HCl (+50 to +100 nA pulses of 2–4 s with 10 min intervals on top of a continuous backing current of −1 to −5 nA to prevent leakage). For bath application of isoguvacine, the perfusion rate was increased to 4.7 ml min−1, and the chamber was superfused with saline containing 200 μm isoguvacine for 30 s periods with 15 min intervals.
Recorded signals were low-pass filtered (field and membrane potential (Vm) signals at 1.6–2.5 kHz, ion-sensitive electrode signals at 160 Hz), digitized at 5 kHz, and stored for off-line analysis with WinEDR and WinWCP software (courtesy of Dr John Dempster, University of Strathclyde, Glasgow, UK) and SigmaPlot 10.0 (Systat Sofwatre, Inc.) and Microsoft Office Excel 2003 (Microsoft Corp.).
Microspectrofluorometric measurement of intracellular pH
Intracellular pH (pHi) measurements were done on 250 μm thick hippocampal slices at P19–P32. After at least 60 min recovery, slices were moved to a submersion recording chamber (volume 800 μl, perfusion at 3–4 ml min−1, 32°C). The pH-sensitive indicator BCECF-AM (50–75 μm; Molecular Probes, Eugene, OR, USA) was pressure injected into the CA1 stratum pyramidale using a glass microelectrode with a tip diameter of 4–8 μm (Stenkamp et al. 2001). The slices were washed for at least 20 min before experiments. This method provided a local population of typically 10–20 fluorescent pyramidal neurons. Acutely isolated CA1 pyramidal neurons were prepared as described before (Ruusuvuori et al. 2004). The fluorescence imaging system (PTI, Lawrenceville, NJ, USA) consisted of a 75 W xenon-arc lamp, a monochromator-chopper unit, an inverted microscope (Nikon Fluor 20× objective, NA 0.75) and an intensified CCD camera (4–16 averaged frames at 0.5–1.5 Hz, excitation wavelengths 440 and 495 nm). The pHi signal was calibrated with the high-[K+]o–nigericin technique (Thomas et al. 1979).
The possible inhibitory effect of furosemide on intrapyramidal carbonic anhydrase activity was examined using a direct functional assay, in which catalytic activity is detected on the basis of the rate of intracellular alkalinisation induced by withdrawal of CO2/HCO3− (Ruusuvuori et al. 2004). In slice experiments, the initial rate of change of pHi was measured first in the standard saline and thereafter either after 30–40 min in the presence of furosemide or in the presence of 100 μm ethoxyzolamide (EZA), which is a membrane-permeant CA inhibitor that slows down the pHi response. Control experiments were performed in the presence of 10 μm benzolamide, a poorly permeant CA inhibitor, in order to rule out the possible contribution of extracellular CA activity. In experiments on acutely isolated pyramidal neurons, alkaline transients were first induced in the standard saline, then after exposure for 20 min to furosemide, and finally after subsequent exposure to EZA.
In experiments where standard solution was changed to the formate-containing (20 mm) CO2/HCO3−-free solution, steady-state pHi values were measured from individual pyramidal neurons (19–20 in each slice) as their mean values within a 10 min period immediately before and 50 min after the solution change.
Quantification of the rate of GABAA receptor-mediated intracellular Cl− accumulation
Evoking pharmacologically isolated IPSPs by tetanic stimulation in the CA1 area of rat hippocampal slices results in a biphasic Vm shift in which the early phase consists of fused IPSPs that gradually become less negative, and a subsequent depolarization mediated by a rise in [K+]o that develops after the 400 ms stimulation period (Kaila et al. 1997; Smirnov et al. 1999). We used sharp microelectrode current clamp recordings in the presence of inhibitors of ionotropic glutamate receptors and GABAB receptors to measure the HFS-induced gradual positive shift in the reversal potential of the GABAA receptor-mediated response (EGABA). As shown in Fig. 4A, HFS applied at different Vm levels brought about by 2 s constant current pulses in bridge-mode current clamp revealed an at least 10-fold increase in input conductance throughout the HFS period (n= 4). Furthermore, large negative deflections of Vm recorded in bridge mode upon brief pulses of injected current (−0.2 nA, 50 ms; not illustrated) were decreased to below noise level when HFS was applied, indicating an at least 15-fold increase in input conductance (n= 4). The HFS-induced large increase in input conductance was caused by activation of GABAARs, since it was blocked by 10 μm bicuculline (Fig. 4B). Therefore, the HFS-induced GABAA conductance is high enough to clamp Vm to EGABA, which allows estimation of the time-dependent shift in EGABA from a single response to HFS (see also Fig. 1B lower panel). Such a shift in the reversal potential of the GABAA receptor-mediated response has been shown to reflect a conductive net uptake of Cl− driven by the depolarizing action of HCO3− efflux via GABAARs (Kaila & Voipio, 1987; Kaila et al. 1989; Smirnov et al. 1999).
The instantaneous rate of change in intracellular [Cl−] (d[Cl−]i/dt) was obtained by first solving for [Cl−]i the Goldman–Hodgkin–Katz voltage equation written for Cl− and a permeant weak acid anion A− (see e.g. Kaila 1994):
and thereafter taking the time derivative of both sides while assuming that [A−]i is constant (see below):
In the above equations PA/Cl is the relative permeability of GABAA receptor channels to anion A− with respect to Cl−, the brackets with subscripts i and o denote intracellular and extracellular concentrations, respectively, and R, T and F have their usual meaning. Analysis was carried out using PA/Cl= 0.2 and PA/Cl= 0.5 where A denotes bicarbonate or formate, respectively (Bormann et al. 1987; Kaila 1994).
During HFS, IPSPs fused and generated a linear shift in Vm starting from the time of the third IPSP. Therefore, the initial d[Cl−]i/dt was calculated at this time point. The obtained numerical values of d[Cl−]i/dt are slight underestimates, since [A−]i was assumed to stay constant during the shift in EGABA (see Kaila 1994). It is worth pointing out that taking into account the time-dependent decrease in [A−]i would make absolute values of d[Cl−]i/dt larger, but such an effect would cancel out when considering relative changes in d[Cl−]i/dt caused by e.g. furosemide.
The values of [A−]i in the standard saline ([HCO3−]i) and in the formate-containing saline ([HCOO−]i) were calculated as described previously (Roos & Boron 1981; Kaila 1994) on the basis of steady-state intrapyramidal pH measured with BCECF fluorescence, and assuming an interstitial pH (pHo) of 7.3 within the slice (Voipio & Kaila, 1993), from:
Unless otherwise stated, data are given as mean ±s.e.m. Statistical significance was tested using paired or unpaired t test, and P values < 0.05 were considered statistically significant.
An increase in [K+]o can be evoked in a hippocampal slice preparation by intense activation of GABAA receptors (Barolet & Morris, 1991; Kaila et al. 1997; Smirnov et al. 1999). Recent evidence points to a specific role of pyramidal neurons in the generation of the GABAergic potassium response (Ruusuvuori et al. 2004), but the molecular mechanisms underlying the rise in [K+]o have remained unknown.
Blocking potassium channels has opposite effects on the GABAergic increase in [K+]o and the depolarization of pyramidal neurons
Blocking potassium channels should not decrease HCO3−-driven Cl− accumulation via GABAARs, but it should have an inhibitory effect on the rise in [K+]o if K+ is released to the extracellular space by a channel-mediated mechanism. In order to see whether the K+ channel blocker Ba2+ inhibits GABAergic [K+]o responses seen with ion-sensitive microelectrodes, we used bath application or microinjections of the GABAA receptor-specific agonist isoguvacine under conditions where spiking activity and ionotropic glutamate and GABAB receptors were blocked by TTX, NBQX, AP-5 and CGP-55845.
Exposure of the slice to 200 μm isoguvacine for 30 s at 15 min intervals gave rise to an increase in [K+]o of 0.96 ± 0.10 mm (n= 7; Fig. 1A) in CA1 stratum pyramidale. The increase in [K+]o was enhanced to 1.50 ± 0.14 mm (n= 7; P= 0.0003) in the presence of 2 mm Ba2+. In these experiments we used bath application of isoguvacine to activate GABAA receptors in a spatially uniform manner throughout the slice in order to reduce the possible consequences of Ba2+-induced changes in glial spatial buffering of K+ (Jauch et al. 2002; Djukic et al. 2007). Under control conditions, the [K+]o responses were paralleled by a depolarization of 9 to 13 mV in pyramidal neurons (n= 3) that seemed to mask an initial hyperpolarizing shift in Vm. Application of 2 mm Ba2+ generated a reversible, slowly increasing positive shift in Vm that was variable in amplitude and resulted in rather stable Vm levels of up to −35 mV (n= 6). In simultaneous recordings of Vm and [K+]o under these conditions (n= 3), isoguvacine induced a Vm response that started with a prompt negative shift of 4 to 10 mV followed by a positive shift that did not cross the baseline level when the resting Vm had depolarized by more than 25 mV (Fig. 1A). Variability in Vm responses was seen also when using other isoguvacine concentrations or durations of application (data not shown). These observations can be accounted for by overlapping membrane currents that are mediated by GABAARs and by the simultaneous increase in [K+]o. Whole cell voltage clamp recordings provided direct support for this conclusion (see Supplemental data).
Pressure microinjections of 2–10 mm isoguvacine to CA1 stratum pyramidale evoked biphasic Vm responses in pyramidal neurons, with an initial hyperpolarization followed by a pronounced depolarization that had a substantially similar time course (time to peak 5.25 ± 0.41 s, n= 4) compared to [K+]o transients evoked by similar isoguvacine microinjections (10 to 90% rise time 5.40 ± 0.16 s, n= 7). The peak amplitudes of both the [K+]o and depolarizing responses upon isoguvacine microinjections were dependent on dose and the distance between the sites of injection and recording. The isoguvacine-induced depolarization measured with respect to the depolarized baseline level of Vm was selectively inhibited in the presence of 2 mm Ba2+ by ≥85% (n= 2; Fig. 1B). Ba2+ had no discernible effect on the isoguvacine-induced conductance (n= 2; superimposed traces in Fig. 1B), and the gradual positive shift in Vm during the pronounced GABAA conductance was not markedly affected by Ba2+. When 2 mm Ba2+ was applied, there was a transient fall in [K+]o which recovered back to the baseline level within 20 min, and the sustained depolarization of pyramidal cells in the presence of Ba2+ was not associated with an increase in the steady-state [K+]o.
The above results suggest that the GABAergic [K+]o increase may, indeed, be generated by increased extrusion of Cl− and K+ from pyramidal neurons by K+–Cl− cotransport (see Fig. 6 and Discussion). The experiments that will be described below were designed to test this hypothesis.
Furosemide blocks GABAergic [K+]o increase in the CO2/HCO3−-buffered saline
First we used furosemide to inhibit the hypothesized KCl efflux via KCC2. In line with our previous results (Kaila et al. 1997; Ruusuvuori et al. 2004), local HFS in the presence of NBQX, AP-5 and CGP-55845 induced an increase in [K+]o of 6.09 ± 0.48 mm (n= 4) that peaked in 2.03 ± 0.16 s and was paralleled by a negative shift and population afterdischarges in the field potential signal in CA1 stratum pyramidale (Fig. 2). Furosemide at 1.0 mm inhibited the [K+]o increase by 71.2 ± 2.6% (P= 0.0029) and strongly inhibited or fully blocked the synchronous afterdischarges. Bicuculline (10 μm) induced a further reduction of only 4.7 ± 1.0% in the mean amplitude of the [K+]o increase (P= 0.016), indicating that the GABAergic component in the HFS-induced [K+]o response was nearly fully abolished by 1.0 mm furosemide in all the four experiments of this kind. Qualitatively similar results were obtained with 0.5 mm furosemide in five other slices, but the mean inhibitory effect on the [K+]o increase was weaker (40.0 ± 5.7%; P= 0.016) and, correspondingly, the subsequent inhibitory effect of bicuculline larger (16.9 ± 6.2%; P= 0.085; not illustrated). The K+ transient that remained in the presence of bicuculline was blocked by TTX suggesting that it was due to spiking-related release of K+.
The effects of furosemide were not due to its known inhibitory action on the Na+-dependent K+–Cl− cotransporter NKCC1, since 10 μm bumetanide (Blaesse et al. 2009) did not inhibit the field potential responses, and it caused a non-significant change of 10.7 ± 6.2% in the [K+]o responses when applied for 10–15 min (n= 4, P= 0.17) and a slight increase of 16.1 ± 4.0% when applied for 45 min or longer (n= 4, P= 0.023).
Stimulation-induced shrinkage of the interstitial space (Dietzel et al. 1980) plays a minor role in the generation of HFS-induced GABAergic [K+]o transients (Ruusuvuori et al. 2004). In simultaneous microelectrode recordings of [K+]o and [TeMA+]o in the presence of 1.5 mm TeMA+ and the aforementioned receptor blockers, the mean increase in [K+]o at the peak of the HFS-induced response was 93.6 ± 11.9% while the mean increase in [TeMA+]o was only 4.1 ± 1.2% (n= 9; not illustrated), confirming that the GABAergic K+ transients are almost solely due to a cellular net K+ release.
The inhibitory effect of furosemide on the GABAergic [K+]o transient is not mediated by inhibition of intraneuronal carbonic anhydrase
Furosemide is a rather non-specific drug that, in addition to its inhibitory actions on cation–chloride cotransporters (Blaesse et al. 2009), can inhibit intracellular carbonic anhydrases (CAi; Temperini et al. 2009) and could thereby diminish the GABAergic [K+]o transients that are dependent on intraneuronal CA activity. Therefore we used a direct functional assay of intrapyramidal CA activity based on BCECF fluorescence measurement of pHi (Ruusuvuori et al. 2004). In hippocampal slices, withdrawal of extracellular CO2/HCO3− caused a rapid alkaline shift in pHi of CA1 pyramidal cells. Notably, the rate of rise of pHi was not affected by 30–40 min application of 1 mm furosemide (control 0.063 ± 0.006 min−1, furosemide 0.063 ± 0.008 min−1 (mean ±s.d.); P= 0.76; n= 18 cells in four slices), whereas exposure to 100 μm EZA caused a significant reduction (control 0.067 ± 0.009 min−1, EZA 0.046 ± 0.007 min−1 (mean ±s.d.), P= 0.00049, n= 6 cells in one slice; not illustrated) that is identical to what we have reported before (Ruusuvuori et al. 2004). In two out of the four slice experiments with furosemide, 10 μm benzolamide, a poorly permeant CA inhibitor, was applied 2 min before and during each CO2/HCO3− withdrawal in order to rule out any possible effect of inhibition of extracellular carbonic anhydrase activity (Chesler, 2003). Since no difference was observed between experiments with and without benzolamide, data were pooled. The results obtained using slices were confirmed in experiments on acutely isolated pyramidal neurons where faster solution changes can be achieved than in slice preparations and where inhibition of intraneuronal CA activity by EZA slows down the induced pHi responses even more than in slices (Ruusuvuori et al. 2004). Superfusion for 20 min with 1 mm furosemide had no effect, whereas a subsequent exposure to EZA caused a clear reduction in the rate and amplitude of the alkaline shift upon CO2/HCO3− withdrawal (n= 4; Fig. 3). These results rule out the possibility that the effect of furosemide on the GABAergic K+ response could be mediated by inhibition of intraneuronal CA activity.
The inhibitory effect of furosemide on the GABAergic [K+]o transient is not caused by a suppression of the intraneuronal chloride load
The utility of furosemide in the present experiments could be compromised by a possible inhibitory action of this drug on the channel-mediated accumulation of Cl−. This effect could result from either an antagonistic effect on GABAA receptor channels (Korpi et al. 1995; Bosman et al. 2002), or from an increase in the resting level of [Cl−]i causing reduction in the driving force of conductive Cl− influx.
In order to see whether GABAergic intraneuronal chloride accumulation is affected by furosemide, we performed sharp microelectrode recordings and analysed the HFS-induced time-dependent shifts in EGABA to obtain estimates of the rate of rise in [Cl−]i (see Methods and Fig. 4). In the standard saline containing NBQX, AP-5 and CGP-55845, HFS evoked IPSPs that fused to form a linearly depolarizing shift in Vm starting from the response to the third pulse. Furosemide (1 mm) induced a shift in EGABA, measured from the response to the third pulse 15–35 min after furosemide application, from −77.2 ± 2.3 mV to −71.7 ± 3.6 mV (n= 4; P= 0.047), while the resting Vm was −66.1 ± 1.8 mV in the control solution and −67.6 ± 2.7 mV in the presence of furosemide. We have previously shown that even a robust downregulation of KCC2 activity does not abolish hyperpolarizing IPSPs unless the cell's Cl− regulation is challenged by a Cl− load from the electrode (Rivera et al. 2004; see also Jin et al. 2005). Our present results with 0.5 and 1 mm furosemide suggest that a full block of KCC2 is not achieved with 1 mm furosemide. Since we used sharp microelectrodes filled with a low-Cl− solution, this accounts for the observation that the measured EGABA values remained somewhat hyperpolarizing in the presence of furosemide. Under control conditions, the initial rate of rise in [Cl−]i was 13.2 ± 1.5 mm s−1 and it was not significantly influenced by 1 mm furosemide (11.3 ± 0.8 mm s−1, P= 0.21). Thus, conductive Cl− accumulation is not significantly suppressed in the presence of furosemide. This is an important finding in the present context since in the presence of furosemide, the HFS-induced increase in [Cl−]i starts from a higher initial level, and it is therefore more likely to result in reversal of the driving force of Cl− (see Discussion). Hence, in case the GABAergic rise in [K+]o would be due to a channel-meditated parallel efflux of K+ and Cl− rather than KCC2-mediated KCl efflux, furosemide should not inhibit but rather enhance it.
GABAergic [K+]o transients and intraneuronal Cl− accumulation driven by formate instead of bicarbonate
The ionic mechanisms underlying 4-aminopyridine-induced GABAergic excitatory events remain functional in mature hippocampal slices bathed in saline where formate (HCOO−) is used as a HCO3− substitute (Lamsa & Kaila, 1997). This is because, like bicarbonate, formate is permeant through GABAARs, and both weak acid anions have an outwardly directed driving force as a consequence of cellular pH regulation (Kaila, 1994). Therefore, using a formate-containing saline devoid of CO2/HCO3− enables us to examine the mechanism of GABAergic excitatory [K+]o transients and, especially, the key role of a GABAAR-permeable weak acid anion in an independent manner.
The increase in [K+]o evoked by 30 s bath application of 200 μm isoguvacine in the CO2/HCO3−-buffered standard saline was strongly attenuated in a Hepes-buffered saline (to 28.3 ± 3.6%, n= 5, P= 0.0004), but thereafter was nearly fully re-established upon superfusion with the Hepes-buffered solution containing 20 mm formate (to 83.9 ± 10.0%, P= 0.19; not illustrated). Similar results were obtained in experiments with HFS, in which both the field potential afterdischarges and the GABAergic [K+]o transients were always preserved in the formate-containing saline containing NBQX, AP-5 and CGP-55845 (Fig. 5A). The mean amplitude of the HFS-induced [K+]o transients in formate was nearly identical to that in control (119.0 ± 9.6% of control, n= 11, P= 0.26).
BCECF fluorescence measurement of intraneuronal pH from CA1 stratum pyramidale in submerged slices showed that the mean steady-state pHi was slightly but significantly more acidic in the formate-containing CO2/HCO3−-free saline than in the standard saline (7.02 ± 0.07 vs. 7.11 ± 0.06 (mean ±s.d.), respectively, P≪ 0.0001; n= 58 cells in 3 slices; not illustrated). Therefore, the calculated steady-state intraneuronal concentration of formate was only somewhat lower than that of bicarbonate under control conditions (10.5 mm vs. 16.1 mm). This and the much higher relative permeability of GABAARs to formate than bicarbonate (5:2) predict that IPSPs reverse at a less negative level in the formate saline. Indeed, HFS in the presence of NBQX, AP-5 and CGP-55845 evoked IPSPs that fused and underwent a roughly linear positive shift starting at the third IPSP from –55.2 ± 1.8 mV (n= 5) in formate compared to −74.6 ± 2.4 mV in control (P= 0.00017; see insets in Fig. 5A). Analysis of Cl− accumulation on the basis of the time-dependent shift in EGABA indicated a slight but not significant increase in the initial rate of rise in [Cl−]i from 15.2 ± 1.8 mm s−1 in control to 18.2 ± 1.7 mm s−1 in formate (P= 0.3). Thus, in the formate-containing saline both the amplitude of the [K+]o transients and the conductive Cl− accumulation were similar to those in the presence of CO2/HCO3−. Furthermore, the [K+]o transients were suppressed by furosemide (Fig. 5B) in a similar manner to what was seen in the presence of CO2/HCO3−, providing further support to the idea that a depolarization resulting from a net efflux of a permeable weak acid anion species via GABAARs does, indeed, cause a conductive Cl− accumulation that leads to a rise in [K+]o upon increased extrusion of Cl− and K+ by K+–Cl− cotransport (see Fig. 6).
Field potential recordings performed simultaneously with measurements of [K+]o always showed a significant reduction in HFS-induced afterdischarge activity in the presence of furosemide (Fig. 5B). While reduced spiking is likely to be a consequence rather than a cause of the smaller [K+]o transients (see Kaila et al. 1997; Ruusuvuori et al. 2004), the contribution of spiking-related K+ release cannot be ignored on the basis of these results. Therefore we confirmed the inhibitory action of furosemide on the GABAergic [K+]o transients using pressure microinjections of isoguvacine or iontophoretic microinjections of muscimol to CA1 stratum pyramidale in the presence of TTX and the synaptic antagonists. The GABAergic component in the isoguvacine-evoked [K+]o transients was strongly inhibited by 1 mm furosemide (by 75.5 ± 4.3%, n= 7, P= 0.0017) and to a lesser extent by 0.5 mm furosemide (by 46.9 ± 48.9%, n= 4, P= 0.031). Furosemide at 0.5–1.0 mm suppressed muscimol-induced [K+]o responses (by 39.3 ± 5.0%, n= 6, P= 0.0047; not illustrated).
Our results provide evidence for the idea that GABAergic excitatory [K+]o transients evoked by HFS and by exogenous GABAA agonists are generated by KCl efflux via KCC2. Such an involvement in an excitatory mechanism is a novel mode of action of this cation–chloride cotransporter and shows that KCC2 has a more versatile role in GABAergic transmission than has been demonstrated before.
We have previously shown that HFS-induced GABAergic [K+]o transients depolarize and excite CA1 pyramidal neurons by causing an inward K+ current, and that EGABA is more negative than Vm during the K+-mediated depolarization in spite of a positive shift in EGABA (Kaila et al. 1997; Smirnov et al. 1999). These observations provide a thermodynamic basis to conclude that the GABAergic [K+]o transients are not generated by parallel depolarization-driven net-electroneutral effluxes mediated by K+ and anion channels.
The finding that the HFS-induced GABAergic [K+]o increase is developmentally upregulated by the expression of cytosolic CA activity in CA1 pyramidal neurons (Ruusuvuori et al. 2004) suggested that pyramidal neurons are crucially involved in the mechanism(s) underlying K+ release upon activation of GABAARs. We propose here a novel mechanism to account for the generation of the GABAergic excitatory [K+]o transients (Fig. 6A and B). As schematically illustrated in Fig. 6B, the inward current mediated by the efflux of HCO3− through GABAA receptor channels depolarizes the membrane potential to a level that is significantly above the equilibrium potential of Cl−, thereby facilitating channel-mediated net uptake of Cl−. The efflux of HCO3− does not cause a directly proportional fall in the intraneuronal HCO3− concentration and thereupon a corresponding fall in the driving force of its efflux, because more HCO3− is produced by rapid net hydration of CO2 in a reaction catalysed by CAi. This accounts for the facilitating effect of CAi on conductive Cl− uptake. The only transmembrane current components shown in Fig. 6B are those via GABAARs, resulting in equal but opposite net fluxes of Cl− and HCO3− via GABAARs because of bulk electroneutrality. (This assumption holds even in the presence of other membrane currents in case the total membrane conductance is dominated by GABAARs and the membrane potential is at EGABA. It is worth emphasizing that significant net fluxes of HCO3− and Cl− flow via GABAARs even when Vm=EGABA and the net current via GABAARs is by definition zero.) As more HCO3− is produced within the cell, a net increase in the intracellular anion content occurs. This does not violate the requirement of bulk electroneutrality because H+ ions are produced at exactly the same rate as HCO3− anions in the hydration reaction of CO2. These protons bind to intrinsic buffers of the cytoplasm, which effectively oppose intraneuronal acidification and thereby assist in maintaining a driving force for the hydration reaction of CO2 when the intracellular concentration of HCO3− tends to fall. The increase in the intracellular Cl− concentration causes an increase in outward KCl transport by KCC2, and leads to accumulation of K+ into the interstitial space. Generation of a [K+]o transient high enough to have an excitatory action requires pronounced GABAAR activation in a large population of pyramidal neurons (Fig. 6A).
The mechanism depicted in Fig. 6B provides an immediate explanation of the inhibition of GABAergic [K+]o-mediated excitation by withdrawal of CO2/HCO3−, by block of CAi activity, and by furosemide-induced inhibition of KCC2. Depolarizing factors not shown in the figure such as glutamatergic inputs would increase the driving force of Cl− influx and increase KCl efflux via KCC2 (Buzsáki et al. 2007). The very rapid HFS-induced rise in [Cl−]i of 13–15 mm s−1 seen in the standard saline, together with the large volume fraction occupied by pyramidal cells in the stratum pyramidale (Miki et al. 2005; Hosseini-Sharifabad & Nyengaard, 2007), imply that the observed [K+]o transients can be accounted for by KCC2-mediated chloride regulation.
In this study we used furosemide that, in addition to its inhibitory action on both KCC2 and NKCC1 (Blaesse et al. 2009), might have an inhibitory action on pyramidal CA (Temperini et al. 2009) and is known to inhibit α6- and α4-subunit-containing GABAARs (Korpi et al. 1995; Tia et al. 1996; Wafford et al. 1996; Bosman et al. 2002; Wall, 2002). Therefore we performed control experiments to verify that the observed actions of furosemide were based on inhibition of KCC2. Specific inhibition of NKCC1 with a low concentration of bumetanide (Khirug et al. 2008; Blaesse et al. 2009) had no inhibitory effect on the responses. A direct functional assay based on BCECF fluorescence measurement of rapid changes in pHi was used to demonstrate that furosemide had no inhibitory action on intraneuronal CA activity under the present experimental conditions. The possibility that furosemide could act by inhibiting anion fluxes through GABAARs was excluded by quantification of the rate of Cl− accumulation evoked by HFS in the presence of inhibitors of ionotropic glutamatergic receptor channels and GABAB receptors. Finally, the mechanism of GABAergic [K+]o transients was demonstrated to be functional when bicarbonate was replaced in the saline by another weak acid anion that is permeant through GABAARs, namely formate (Bormann et al. 1987; Kaila, 1994; Lamsa & Kaila, 1997).
Bath application of furosemide resulted in a positive shift in EGABA indicating a higher initial level of [Cl−]i when applying HFS in the presence of this inhibitor. This evident consequence of KCC2 inhibition raises the question whether furosemide can be used as a tool to differentiate between transporter and channel-mediated effluxes of K+ and Cl−. However, Cl− accumulation starting from an elevated level of resting [Cl−]i is more prone to create an outward driving force for conductive Cl− efflux via e.g. ClC-2 chloride channels (Staley et al. 1996). Therefore, if the GABAergic [K+]o transients were a consequence of channel-mediated net-electroneutral fluxes of K+ and Cl− instead of KCl efflux via KCC2, they should be enhanced rather than inhibited by furosemide. In the standard CO2/HCO3−-containing saline, the effect of 1 mm furosemide on HFS-induced Cl− accumulation was non-significant but furosemide caused a near-complete block of the GABAergic component of the HFS-induced [K+]o response. Qualitatively similar results were obtained with 0.5 mm furosemide, and in the formate-containing saline. In addition, furosemide inhibited [K+]o responses evoked with microinjections of the GABAAR-specific agonists isoguvacine and muscimol under conditions where spiking activity was blocked.
Several studies have shown that blockage of neuronal K+–Cl− transport by furosemide has antiepileptic actions in vivo and in vitro, with minor effects on synaptic responses evoked by single electrical stimuli (Hochman et al. 1995; Gutschmidt et al. 1999; Hochman & Schwartzkroin, 2000; Margineanu & Klitgaard, 2006). The antiepileptic effects of furosemide have been explained e.g. by desynchronization (Hochman & Schwartzkroin, 2000), block of activity-evoked changes in extracellular volume (Hochman et al. 1995; Gutschmidt et al. 1999; Haglund & Hochman 2005), reduced excitability with consequent reduction of activity-induced increase in [K+]o (Gutschmidt et al. 1999), and enhancement of astrocytic inward potassium currents (Barbaro et al. 2004). Our present results point to an antiepileptic effect of furosemide by inhibition of excitatory GABAergic [K+]o transients, which is interesting as accumulation of extracellular potassium has long been thought to be capable of provoking epileptic activity (Green 1964; Fertziger & Ranck, 1970; Heinemann et al. 1986; Yaari et al. 1986; Somjen, 2004; Bazhenov et al. 2008; Fröhlich et al. 2008). Furthermore, the inhibitory action of furosemide on extracellular volume changes (cf. Dietzel et al. 1980; Walz, 1987) may be a consequence of the inhibition of GABAergic [K+]o increase.
We and others have suggested that an increase in [K+]o may inhibit or reverse the operation of KCC2 leading to intraneuronal accumulation of Cl− and to a positive shift in EGABA (Payne, 1997; Kaila et al. 1997; Jarolimek et al. 1999; DeFazio et al. 2000; Fröhlich et al. 2008; see also Thompson et al. 1988; Thompson & Gahwiler, 1989b). While this is evidently the case in a neuron that has a low [Cl−]i, our present results point to a mechanism of an opposite kind, namely to an activity-induced enhancement of KCl efflux via KCC2 as a mechanism that can cause an increase in [K+]o capable of depolarizing neurons in a non-synaptic manner. Moreover, there are several reports of agonists or positive modulators of GABAARs promoting epileptiform events (Reddy et al. 1993; D’Antuono et al. 2004; see also Avoli et al. 2002), which may well involve the mechanism described here.
Since either one of the steps mediated by KCC2 and intraneuronal CA activity can be rate limiting in the mechanism depicted in Fig. 6, an endogenous physiological or an exogenous pharmacological control of GABAergic [K+]o signals could be based on either KCC2 (and KCC3; Boettger et al. 2003) or CAVII (or any intraneuronal CA isozyme), or both. This explains why HCO3−-dependent GABAergic excitatory [K+]o transients can be observed only after expression of intrapyramidal CA activity (Ruusuvuori et al. 2004). In this context, it is interesting to note that the mechanism of GABAergic [K+]o transients implies that the efficacy of (initially) hyperpolarizing inhibitory transmission to inhibit target neurons depends on the temporal patterning of presynaptic activity in populations of inhibitory interneurons, whereas inhibition by shunting without hyperpolarization in the absence of functional KCC2 would lack this feature. Furthermore, our present results do not exclude the possibility that in a different experimental protocol a use-dependent positive shift in EGABA may result in direct synaptic excitation by GABAA receptors.
Experimentally induced seizures in vivo and pathophysiological activity in vitro cause downregulation of KCC2 and reduce Cl− extrusion capacity (Rivera et al. 2002, 2004; Fiumelli et al. 2005; see also Payne et al. 2003). In line with this, low levels of KCC2 mRNA or perturbed Cl− homeostasis have been detected in subicular pyramidal neurons from surgically resected epileptic human brain specimens (Palma et al. 2006; Huberfeld et al. 2007; see also Cohen et al. 2002, 2003). The present results predict that downregulation of KCC2 has an inhibitory action on excitatory GABAergic increase in [K+]o. This point is worth considering when making functional inferences from observations on compromised neuronal Cl− regulation on chronically epileptic tissue. In addition to promoting interictal events (e.g. Huberfeld et al. 2007), downregulation of KCC2 might have a suppressing effect on ictal events.
J.V. and K.K. conceived the project. All authors contributed to designing the experiments, to analysing and interpreting the data and to writing the paper. T.V. and E.R. conducted the electrophysiological and the microspectrofluorometric experiments, respectively. All experiments were done in the Laboratory of Neurobiology, Department of Biosciences, University of Helsinki, Finland.
This study was supported by the Academy of Finland, the University of Helsinki Research Funds, Sigrid Juselius Foundation, EPICURE/EFP6-037315 (K.K.). K.K. is a member of the Finnish Center of Excellence in Molecular and Integrative Neuroscience Research and of the Nordic Center of Excellence, WIRED.
Author's present address
T. Viitanen: Minerva Institute for Medical Research, Biomedicum 2U Helsinki, FI-00290 Helsinki, Finland.