Hochman, Daryl W., Raimondo D'Ambrosio, Damir Janigro, and Philip A. Schwartzkroin. Extracellular chloride and the maintenance of spontaneous epileptiform activity in rat hippocampal slices. J. Neurophysiol. 81: 49–59, 1999. Previous studies showed that furosemide blocks spontaneous epileptiform activity without diminishing synaptic transmission or reducing hyperexcited field responses to electrical stimuli. We now test the hypothesis that the antiepileptic effects of furosemide are mediated through its blockade of the Na+,K+,2Cl− cotransporter and thus should be mimicked by a reduction of extracellular chloride ([Cl−]o). In the first set of experiments, field recordings from the CA1 cell body layer of hippocampal slices showed that spontaneous bursting developed within 10–20 min in slices perfused with low-[Cl−]o (7 mM) medium but that this spontaneous epileptiform activity ceased after a further 10–20 min. Intracellular recordings from CA1 pyramidal cells showed that normal action potential discharge could be elicited by membrane depolarization, even after the tissue was perfused with low-[Cl−]o medium for >2 h. In a second set of experiments, spontaneous bursting activity was induced in slices by perfusion with high-[K+]o (10 mM), bicuculline (100 μM), or 4-aminopyridine (100 μM). In each case, recordings from the CA1 region showed that reduction of [Cl−]o to 21 mM reversibly blocked the bursting within 1 h. Similar to previous observations with furosemide treatment, low-[Cl−]o medium blocked spontaneous hypersynchronous discharges without reducing synaptic hyperexcitability (i.e., hyperexcitable field responses evoked by electrical stimulation). In a third set of experiments, prolonged exposure (>1 h after spontaneous bursting ceased) of slices to systematically varied [Cl−]o and [K+]o resulted in one of three types of events: 1) spontaneous, long-lasting, and repetitive negative field potential shifts (7 mM [Cl−]o; 3 mM [K+]o); 2) oscillations consisting of 5- to 10-mV negative shifts in the field potential, with a period of ∼1 cycle/40 s (16 mM [Cl−]o; 12 mM [K+]o); and 3) shorter, infrequently occurring negative field shifts lasting 20–40 s (21 mM [Cl−]o; 3 mM [K+]o). Our observations indicate that the effects of low [Cl−]o on neuronal synchronization and spontaneous discharge are time dependent. Similar effects were seen with furosemide and low [Cl−]o, consistent with the hypothesis that the antiepileptic effect of furosemide is mediated by the drug's effect on chloride transporters. Finally, the results of altering extracellular potassium along with chloride suggest that blockade of the Na+, K+,2Cl− cotransporter, which normally transports chloride from the extracellular space into glial cells, is key to these antiepileptic effects.
Epileptiform activity is identified with spontaneously occurring synchronized discharges of neuronal populations (Schwartzkroin 1993). Such hypersynchronized activity is typically induced in experimental models of epilepsy by either increasing excitatory or decreasing inhibitory synaptic currents. It is therefore often assumed that hyperexcitability per se is the defining feature involved in the generation and maintenance of epileptiform activity; consequently, focus on synaptic hyperexcitability has been a guiding principle in basic research on the mechanisms of epileptogenesis (McNamara 1994) and in the design and discovery of new antiepileptic drugs (Macdonald and Greenfield 1997; Upton 1994). It has recently been shown, however, that it is possible to dissociate hypersynchronous epileptiform activity from hyperexcitability (Hochman et al. 1995). In that study, a variety of treatments, involving a spectrum of different physiological mechanisms, was used to elicit spontaneous discharges in hippocampal slices. Furosemide, a cation-chloride cotransport inhibitor, reversibly blocked the synchronized bursts in each model without reducing the hyperexcited synaptic responses of CA1 pyramidal cells to electrical stimulation of the Schaffer collaterals. Those observations suggested that furosemide affected mechanisms, common to all of the models, that were necessary for the maintenance of spontaneous epileptiform activity but were independent of synaptic hyperexcitability.
Because the primary pharmacological action of furosemide is inhibition of the Na+,K+,2Cl− and K+,Cl− cotransporters (Alvarez-Leefmans 1990; Geck and Heinz 1986; Geck and Pfeiffer 1985), it is likely that the antiepileptic effects of furosemide are mediated through the blockade of these cotransporters. The Na+,K+,2Cl− and K+,Cl−-cotransport systems are gradient driven, with transport-rate dependencies phenomenologically described as functions of the “ion products,” [Na] × [K+] × [Cl−]2 or [K+] × [Cl−], respectively (Cala 1990; Geck and Pfieffer 1985). According to this formulation, furosemide-sensitive cotransport is sensitive to [Cl−]o. Indeed, both furosemide and low [Cl−]o have been shown to similarly block activity-induced volume changes in the extracellular space (ECS) in hippocampal and cortical slices (Holtoff and Witte 1996; MacVicar and Hochman 1991) and K+-driven cell volume changes in astrocytes (Walz 1992; Walz and Hertz 1984).
The previous considerations suggest that a reduction of [Cl−]o should have antiepileptic effects similar to those of furosemide. However, previous studies of the electrophysiological effects of low [Cl−]o indicated that a reduction of [Cl−]o leads to the development of epileptiform activity, presumably through the blockade of inhibitory synaptic mechanisms (Avoli et al. 1990; Yamamoto 1972; Yamamoto and Kawai 1969). Hyperpolarizing inhibitory postsynaptic potentials (IPSPs) in cortical and hippocampal neurons are generated by the influx of chloride ions down a gradient that is maintained by a furosemide-sensitive outwardly directed chloride transport mechanism (Misgeld et al. 1986; Thompson et al. 1988; Thompson and Gähwiler 1989). Because γ-aminobutyric acid (GABA) receptor-mediated inhibitory current is determined by the difference between the equilibrium potential for Cl− and the neuronal membrane potential (Thompson 1994), reduction of [Cl−]o would be expected to cause a depolarizing shift in the GABA reversal potential (E GABA), leading to hyperexcitability and interictal-like spiking similar to that caused by blockers of GABAA receptor function (Schwartzkroin and Prince 1980).
A similar analysis of the effects of furosemide on intracellular chloride might also lead one to expect that furosemide should induce, rather than block, spontaneous bursting. Recordings obtained from cortical and hippocampal neurons have shown that the mechanism of Cl− extrusion responsible for keeping an adequate driving force for the IPSP is an outwardly directed K+,Cl− cotransport that is blocked by furosemide (Misgeld et al. 1986; Thompson and Gähwiler 1989; Thompson et al. 1988). After furosemide treatment of hippocampal or cortical slices, the chloride equilibrium potential (ECl−) in pyramidal cells is determined by a passive Donnan distribution, and hyperpolarizing IPSPs are abolished. Neither furosemide nor low-[Cl−]o alters intrinsic electrophysiological properties such as resting membrane potential, input resistance, or action potential generation; furthermore, excitatory synaptic transmission is not reduced by either of these treatments (Avoli et al. 1990; Holthoff and Witte 1996; MacVicar and Hochman 1991). These data suggest that the effects of both low [Cl−]o and furosemide on neuronal physiology are primarily due to the induction of a loss of the transmembrane chloride gradient.
Given the apparent inconsistency of these previous studies with the recent demonstration of an antiepileptic furosemide effect, we carried out additional experiments to better understand the mechanism of furosemide action. We hypothesized that the effects of low-chloride and furosemide might vary in a time-dependent fashion with at least three distinct phases of action: 1) Initial reduction of [Cl−]o (or application of furosemide) would lead to a diminished neuronal transmembrane chloride gradient, resulting in hyperexcitibility and generation of epileptiform activity; 2) longer exposure to low [Cl−]o (and furosemide-containing) medium would block spontaneous bursting activity, but general hyperexcitability would be maintained; and 3) prolonged tissue exposure to low-[Cl−]o medium would lead to a series of membrane oscillations, dependent on the direction of Cl− and K+ transport by the chloride cotransporters. Results of some of these studies were reported in preliminary form (Hochman and Schwartzkroin 1997).
Sprague-Dawley adult rats were prepared as previously described (Hochman et al. 1995).
Transverse hippocampal slices, 400 μm thick, were cut with a vibrating cutter. Slices typically contained the entire hippocampus and subiculum. After cutting, slices were stored in an oxygenated holding chamber for ≥1 h before recording. All recordings were acquired in an interface type chamber with oxygenated (95%O2-5%CO2) artificial cerebrospinal fluid (ACSF) at 34–35°C. Normal ACSF contained (in mmol/l) 124 NaCl, 3 KCl, 1.25 NaH2PO4, 1.2 MgSO4, 26 NaHCO3, 2 CaCl2, and 10 dextrose.
In some experiments, normal or low-chloride medium was used containing bicuculline (20 μM), 4-aminopyridine (4-AP) (100 μM), or high-K+ (12 mM). Low-chloride solutions (7, 16, and 21 mM [Cl−]o) were prepared by equimolar replacement of NaCl with Na+-gluconate (Sigma). All solutions were prepared so that they had a pH of ∼7.4 and an osmolarity of 290–300 mosmol at 35°C and at equilibrium from carboxygenation with 95% O2-5% CO2.
Sharp-electrodes filled with 4 M potassium acetate were used for intracellular recordings from CA1 pyramidal cells. Field recordings from the CA1 or CA3 cell body layers were acquired with low-resistance glass electrodes filled with NaCl (2 M). For stimulation of the Schaffer collateral pathway, a small monopolar electrode was placed on the surface of the slice midway between areas CA1 and CA3. Spontaneous and stimulation-evoked activities from field and intracellular recordings were digitized (Neurocorder, Neurodata Instruments, New York, NY) and stored on videotape. AxoScope software (Axon Instruments) on a personal computer was used for off-line analyses of data.
Ion-selective microelectrodes were fabricated according to standard methods previously described (Janigro et al. 1997; Lux 1974; Nicholson and Rice 1988). Double-barreled pipettes were pulled and broken to a tip diameter of ∼3.0 μm. The reference barrel was filled with ACSF, and the other barrel was sylanized and the tip backfilled with a resin selective for K+ (Corning 477317). The remainder of the sylanized barrel was filled with KCl (140 mM). Each barrel was led, via Ag/AgCl wires, to a high-impedance dual-differential amplifier (WPI FD223). Each ion-selective microelectrode was calibrated by the use of solutions of known ionic composition and was considered suitable if it was characterized by a near-Nernstian slope response and if it remained stable throughout the duration of the experiment.
After placement in the interface chamber, slices were superfused at ∼1 ml/min. At this flow-rate, it took ∼8–10 min for changes in perfusion media to be completed. All of the times reported here have taken this time delay into account and have an error of approximately ±2 min. This accuracy is sufficient for the purposes of this study.
In all of our experiments, [Cl−]o was reduced by equimolar replacement of NaCl with Na+-gluconate. Our decision to use gluconate rather than other anion replacements was made for several reasons. First, patch-clamp studies demonstrated that gluconate appears to be virtually impermeant to Cl− channels, whereas other anions (including methylsulfate, sulfate, isethionate, and acetate) are permeable to varying degrees (Bevan et al. 1985; Gray and Ritchie 1986). Second, transport of extracellular potassium through glial Na+,K+,2Cl− cotransport is blocked when extracellular chloride is replaced by gluconate but is not completely blocked when replaced by isethionate (Walz 1995). Because this furosemide-sensitive cotransporter plays a significant role in cell swelling and volume changes of the ECS, we wished to use the appropriate anion replacement so that the effects of our treatment would be comparable with previous furosemide experiments (Hochman et al. 1995). Third, formate, acetate, and proprionate generate weak acids when employed as Cl− substitutes and lead to a prompt fall in intracellular pH; gluconate remains extracellular and was not reported to induce intracellular pH shifts (Kaila and Voipio 1990; Roos and Boron 1981). Fourth, for purposes of comparison we wished to use the same anion replacement that was used in previous studies examining the effects of low [Cl−]o on activity-evoked changes of the ECS (Holthoff and Witte 1996; MacVicar and Hochman 1991).
There is some suggestion that certain anion replacements might chelate calcium (Christoffersen and Skibsted 1975). Although subsequent work failed to demonstrate any significant ability of anion substitutes to chelate calcium (Pollard et al. 1977), there is still some concern in the literature regarding this issue (Kaila 1994). Calcium chelation did not appear to be a problem in our studies because resting membrane potentials remained normal, and synaptic responses (indeed, hyperexcitable synaptic responses) could be elicited even after several hours of exposure to medium in which [Cl−]o was reduced by gluconate substitution. Further, we confirmed that calcium concentration in our low-[Cl−]o medium was identical to that in our control medium by measurements made with Ca2+-selective microelectrodes (these ion-sensitive microelectrode measurements were performed in the laboratory of Dr. Bruce Ransom, University of Washington, Seattle, WA).
Effects of low [Cl−]o on CA1 field recordings
The focus of these experiments was to observe the times of onset and possible cessation of low-[Cl−]o–induced hyperexcitability and hypersynchronization. Data regarding the temporal evolution of the effects of low-[Cl−]o exposure, which are necessary to test our hypothesis, were not previously reported.
Slices (n = 6) were initially perfused with normal medium until stable intracellular and field recordings were established in a CA1 pyramidal cell and the CA1 cell body layer, respectively. In two experiments, the same cell was held throughout the entire length of the experiment (>2 h) (see Fig. 1). In the remaining experiments (n = 4), the initial intracellular recording was lost during the sequence of solution changes, and additional recordings were acquired from different cells. Patterns of neuronal activity in these experiments were identical to those seen when a single cell was observed. The field and intracellular electrodes were always placed in close vicinity to one another (<200 μm). In each case, after ∼15–20 min exposure to the low-[Cl−]o medium (7 mM), spontaneous bursting developed, first at the cellular level and then in the field. This spontaneous field activity, representing synchronized burst discharge in a large population of neurons, lasted from 5 to 10 min, after which time the field recording became silent. Figure 1 A shows the continuous traces of a field and intracellular recording just before and during the cessation of the spontaneous field events. When the field first became silent the cell continued to discharge spontaneously. This result suggests that population activity was “desynchronized,” whereas the ability of individual cells to discharge was not impaired. After ∼30 min exposure to low-[Cl−]o medium, intracellular recording showed that cells continued to discharge spontaneously, although the field remained silent (Fig. 1 B). The insets in Fig. 1, A and B, show the response of this cell to intracellular current injection at these two time points, demonstrating that the cell's ability to generate action potentials was not impaired by low-[Cl−]o exposure. Further, electrical stimulation in CA1 stratum radiatum elicited burst discharges, indicating that a hyperexcitable state was maintained in the tissue.
Effects of low [Cl−]o on high-[K+]o–induced epileptiform activity in CA1
The previous set of experiments showed that tissue exposure to low-[Cl−]o medium induced a brief period of spontaneous field potential bursting which ceased within 10 min. If a reduction of [Cl−]o is indeed eventually capable of blocking spontaneous epileptiform (i.e., synchronized) bursting, these results suggest that antiepileptic effects would likely be observable only after this initial period of bursting activity ceased. We therefore examined the temporal effects of low-[Cl−]o treatment on high-[K+]o–induced bursting activity (Fig. 2). Slices (n = 12) were exposed to medium in which [K+]o was increased to 12 mM, and field potentials were recorded with a field electrode in the CA1 cell body layer. Spontaneous field potential bursting (Fig. 2 B) was observed for ≥20 min, and the slices were exposed to medium in which [K+]o was maintained at 12 mM, but [Cl−]o was reduced to 21 mM. Within 15–20 min after the tissue was exposed to the low-[Cl−]o /high-[K+]o medium, the burst amplitude increased, and each field event had a longer duration (Fig. 2 C). After a brief period of this facilitated field activity (lasting 5–10 min), the bursting stopped (Fig. 2 D). To test whether this blockade was reversible, after ≥10 min of field potential silence, we switched back to high-[K+]o medium with normal [Cl−]o. The bursting returned within 20–40 min (Fig. 2 E). Throughout each experiment, the CA1 field response to Schaffer collateral stimulation was monitored (Fig. 2, right). The largest field responses were recorded just before the cessation of spontaneous bursting, during the period when the spontaneous bursts had the largest amplitude. Even after the blockade of spontaneous bursting, however, multiple population spikes were elicited by Schaffer collateral stimulation, indicating that synaptic transmission was intact and that the tissue remained hyperexcitable.
In four slices, intracellular recordings from CA1 pyramidal cells were acquired along with the CA1 field recording. During the period of high-[K+]o–induced spontaneous bursting, hyperpolarizing current was injected into the cell so that postsynaptic potentials (PSPs) could be better observed. After low-[Cl−]o blockade of spontaneous bursting, frequently occurring spontaneous action potentials and PSPs were still observed (data not shown). These observations further support the view that synaptic activity per se was not blocked by the low-[Cl−]o treatment.
Low-[Cl−]o blockade of epileptiform activity induced by 4-AP, high-[K+]o, and bicuculline in CA1 and CA3
We next wanted to test whether low-[Cl−]o treatment could block epileptiform activity elicited by different pharmacological treatments, as we had shown for furosemide treatment (Hochman et al. 1995). For this set of experiments, we chose to test the effects of low-[Cl−]o treatment on spontaneous bursting which was induced by high-[K+]o (12 mM) (n = 5), 4-AP (100 μM) (n = 4), and bicuculline (20 and 100 μM) (n = 5) (Fig. 3). In each set of experiments, field responses were recorded simultaneously from areas CA1 and CA3b; in each case, the spontaneous epileptiform activity (in both areas) was reversibly blocked ≤30 min after [Cl−]o in the perfusion medium was reduced to 21 mM. These data suggest that, like furosemide, low [Cl−]o reversibly blocks spontaneous bursting in several of the most commonly studied in vitro models of epileptiform activity.
Comparison between low [Cl−]o and furosemide on blockade of high-[K+]o–induced epileptiform activity
The data from the previous sets of experiments are consistent with the hypothesis that the antiepileptic effects of both low [Cl−]o and furosemide are mediated by their actions on the same physiological mechanisms. To further test this notion, we compared the temporal sequence of effects of low [Cl−]o (n = 12) and furosemide (2.5 and 5 mM) (n = 4) on high-[K+]o–induced bursting, as recorded with a field electrode in CA1 (Fig. 4). We found that both low [Cl−]o and furosemide treatment induced a similar temporal sequence of effects; an initial brief period of increased amplitude of field activity and then blockade (reversible) of spontaneous field activity. In both cases, electrical stimulation of the Schaffer collaterals elicited hyperexcited responses even after the spontaneous bursting was blocked.
Consequences of prolonged exposure to low-[Cl−]o medium with varied [K+]o
It is probable that a K+,Cl− cotransporter plays a major role in the maintenance of the neuronal transmembrane chloride gradient by facilitating an efflux of KCl. This efflux is driven by the high intracellular and low extracellular concentrations of potassium (Misgeld et al. 1986; Thompson and Gähwiler 1989; Thompson et al. 1988). Similarly, the Na+,K+,2Cl− cotransporter is important in the movement of potassium from the extracellular spaces into glial cells by facilitating an influx of NaCl and KCl. This influx is driven, in part, by the transmembrane sodium and potassium concentration gradients (Walz 1992; Walz and Hinks 1986). The Na+,K+,2Cl− cotransporter is particularly sensitive to changes in [Cl−]o because its rate of transport is related to the square of the chloride concentration (Cala 1990; Geck and Pfieffer 1985). It might be expected then that changes in either [K+]o or [Cl−]o would affect the cotransported ion fluxes in both neurons and glial cells. Consequently, it would be predicted that changes in either [K+]o or [Cl−]o should affect spontaneous synchronized bursting activity.
In the course of the preceding experiments, we monitored field activity in some slices for >1 h after the spontaneous bursting was blocked by low-[Cl−]o exposure. After such prolonged low-[Cl−]o exposure, spontaneous, long-lasting, depolarizing shifts developed. The morphology and frequency of these late-occurring field events appeared to be related to the extracellular potassium and chloride concentrations. Motivated by these observations, we performed a set of experiments in which we systematically varied [Cl−]o and [K+]o and observed the effects of these ion changes on the late-occurring spontaneous field events.
In our first set of experiments, slices were exposed to medium containing low [Cl−]o (7 mM) and normal [K+]o (3 mM) (n = 6). After 50- to 70-min exposure to this medium, spontaneous events were recorded in area CA1 (Fig. 5); these events appeared as 5- to 10-mV negative shifts in the DC field, with the first episode lasting for 30–60 s (Fig. 5 A). Each subsequent episode was longer than the previous one (Fig. 5 B). This observation suggested that ion-homeostatic mechanisms were diminished over time as a result of the ion concentrations in the bathing medium. In some experiments (n = 2) in which these negative DC field shifts were induced, intracellular recordings from CA1 pyramidal cells were acquired simultaneously with the CA1 field recordings (Fig. 6 A). For these experiments, the intracellular and field recordings were acquired close to one another (<200 μm). Before each negative field shift (10–20 s), the neuron began to depolarize. Cellular depolarization was indicated by a decrease in resting membrane potential, an increase in spontaneous firing frequency, and a reduction of action potential amplitude. Coincident with the onset of the negative field shifts, the cells became sufficiently depolarized so that they were unable to fire spontaneous or current-elicited (not shown) action potentials. Because neuronal depolarization began 10–20 s before the field shift, it may be that a gradual increase in extracellular potassium resulted in the depolarization of a neuronal population, thus initiating these field events. Such an increase in [K+]o might be due to alterations of the chloride-dependent glial cotransport mechanisms that normally move potassium from extracellular to intracellular spaces. To test whether increases in [K+]o preceded these negative field shifts (and paralleled cellular depolarization), experiments (n = 2) were performed in which a K+-selective microelectrode was used to record changes in [K+]O (Fig. 6, B and C). In each experiment, the K+-selective microelectrode and a field electrode were placed in the CA1 pyramidal layer close to one another (<200 μm), and a stimulation pulse was delivered to the Schaffer collaterals every 20 s so that the magnitude of the population spike could be monitored. Multiple spontaneously occurring negative field shifts were evoke by perfusion with low-[Cl−]o (7 mM) medium. Each event was associated with a significant increase in [K+]o, with the [K+]o increase starting several seconds before the onset of negative field shift (Fig. 6, B and C, bottom traces). Figure 6, B and C, shows the first of a series of spontaneously occurring negative field shifts that occurred during one such experiment. A slow 1.5- to 2.0-mM increase in [K+]o occurred over a time interval of ∼1–2 min before the onset of each event. The stimulation-evoked field responses (arrows in Fig. 6 C) slowly increased in amplitude with the increasing [K+]o.
In a second set of experiments (n = 4), [K+]o was increased to 12 mM, and [Cl−]o was increased to 16 mM. After 50- to 90-min exposure to this medium, slow oscillations were recorded in area CA1 (Fig. 7). These oscillations were characterized by 5- to 10-mV negative DC shifts in the field potential and had a periodicity of ∼1 cycle/40 s. Initially, these oscillations occurred intermittently and had an irregular morphology (Fig. 7 A). Over time, these oscillations became continuous and developed a regular waveform (Fig. 7 B). On exposure to furosemide (2.5 mM), the amplitude of the oscillations was gradually decreased, and the frequency increased until the oscillations were completely blocked (Fig. 7 C). Such low-[Cl−]o–induced oscillations in tissue slices were not previously reported. However, the temporal characteristics of the oscillatory events bear a striking resemblance to the low-[Cl−]o–induced [K+]o oscillations that were previously described in a purely axonal preparation (Connors and Ransom 1984).
In a third set of experiments (n = 5) [Cl−]o was further increased to 21 mM, and [K+]o was reduced back to 3 mM. In these experiments, single, infrequently occurring negative shifts of the field potential developed within 40–70 min (data not shown). These events (5–10 mV) lasting 40–60 s, occurred at random intervals and maintained a relatively constant duration throughout the experiment. These events could sometimes be elicited by a single electrical stimulus delivered to the Schaffer collaterals.
Finally, in a last set of experiments (n = 5), [Cl−]o was kept at 21 mM, and [K+]o was raised to 12 mM. In these experiments, late-occurring spontaneous field events were not observed during the course of the experiments (2–3 h).
This study was motivated by our interest to understand the apparently contradictory observations that furosemide blocks spontaneous burst activity (Hochman et al. 1995), whereas low [Cl−]o generates epileptiform activity (Avoli et al. 1990; Yamamoto 1972; Yamamoto and Kawai 1969). Our observations here suggest that there is a temporal dependence of the effects of low[Cl−]o on epileptiform activity. Initially, low [Cl−]o induced spontaneous activity and hyperexcitability. Longer exposure blocked spontaneous synchronous bursting but not hyperexcitability in a manner nearly identical to that of furosemide. Further, prolonged low-[Cl−]o treatment induced various patterns of negative-going field potential events, presumably reflecting synchronized neuronal depolarizations. By taking into account the temporal dependency of the effects of reduced [Cl−]o, it is possible to interpret all of the previously published data in a consistent manner. Indeed, Yamamoto and Kawai (1969), in their initial report of low-[Cl−]o–induced epileptiform activity in hippocampal slices, observed that, although afterdischarge activity could be evoked in slices at the beginning of their experiments (i.e., when low-[Cl−]o perfusion was initiated), a prolonged period of complete silence in the electrical activities often followed.
For proper comparison of our data to previously published work, it is important to note that there are significant differences between the reports as to the kind of electrophysiological phenomena that constitute “epileptiform” activity. In our studies, we define epileptiform activity as spontaneous synchronized discharges of a population of neurons, a significantly different phenomenon than the hyperexcited stimulus-evoked field responses or repetitive discharges of individual neurons. In contrast, much of the data reported in Avoli et al. (1990) was based on hyperexcited responses to electrical stimulation. Although some forms of spontaneous bursting were described in that study, no information was provided about the onset or cessation of spontaneous synchronized bursting with respect to the time of slice exposure to low-[Cl−]o medium.
The effects of low [Cl−]o on excitability and synchronization in our study closely parallel the effects of furosemide (see also Hochman et al. 1995). Because both furosemide and reduced [Cl−]o affect chloride cotransport, it is likely that the generation and maintenance of hypersynchronization associated with spontaneous epileptiform activity is dependent on either neuronal K+, Cl− or glial Na+,K+,2Cl− cotransport. In neurons, the chloride gradient necessary for the generation of hyperpolarizing IPSPs is maintained by the efflux of ions through a putative furosemide-sensitive K+,Cl− cotransporter (Misgeld et al. 1986; Thompson and Gähwiler 1989; Thompson et al. 1988). Under normal conditions, a high concentration of intracellular potassium, which is maintained by the 3Na+,2K+-ATPase electrogenic pump, serves as the driving force for the extrusion of Cl− against its concentration gradient (Fig. 8, box 1, left). Thus for neurons, decreases in either [K+]o or [Cl−]o should facilitate the efflux of these ion species from intracellular to extracellular spaces. In glial cells, the movement of ions via the furosemide-sensitive Na+,K+,2Cl− cotransporter, under both normal conditions and during increased neuronal activity, is from extracellular to intracellular spaces (Walz 1992; Walz and Hertz 1984). The ion gradients necessary for this cotransport are maintained, in part, by the “transmembrane sodium cycle”; sodium ions taken into glial cells via Na+,K+,2Cl− cotransport are continuously extruded by the 3Na+,2K+-ATPase pump so that a low intracellular sodium concentration is maintained (Walz and Hinks 1986) (Fig. 8, box 1, right). Thus for glial cells, decreases in either [K+]o or [Cl−]o will inhibit the influx of these ion species from extracellular to intracellular spaces. According to one formulation, the rate and direction of ion flux through the furosemide-dependent cotransporters are functionally proportional to their ion-product differences, written as ([K+]i × [Cl−]i − [K+]o × [Cl−]o) for neuronal K+,Cl− cotransport and as ([Na+]i × [K+]i × [Cl−]2 i − [Na+]o × [K+]o × [Cl−]2 o) for glial Na+,K+,2Cl− cotransport (Geck and Pfieffer 1985). We denote these quantities as QN and QG, for neuronal and glial ion-product differences, respectively. The sign of these ion-product differences shows the direction of ion transport, QN > 0 (neuronal ion-product difference is >0) and QG > 0 (glial ion-product difference is >0), implying the direction of ion cotransport in neurons and glia from intracellular to extracellular compartments. Depending on conditions, ion cotransport can take place in either direction. Also, the Na+,K+,2Cl− cotransporter is most sensitive to the chloride concentration because its rate of transport is dependent on the square of [Cl−] but is linearly dependent on [K+] and [Na+] (Cala 1990; Geck and Pfieffer 1985).
Furosemide causes a depolarizing shift in E GABA by blocking chloride extrusion through K+,Cl− cotransport, thus allowing equilibration between intracellular and extracellular chloride concentrations. In contrast, reduction of [Cl−]o reduces the IPSP driving force by directly diminishing the transmembrane chloride gradient. In both cases, there is a reduction of synaptic inhibition. The resultant hyperexcitability was evidenced in our studies by two observations, 1) the reduction of [Cl−]o in slices that already displayed epileptiform bursts led to an increased magnitude and frequency of the spontaneous bursts and 2) after blockade of spontaneous bursting activity, hyperexcited field responses could be elicited by Schaffer collateral stimulation. Although both furosemide and low [Cl−]o result in this stimulus-evoked hyperexcitability, it is important to note that they have opposite effects on the outward K+,Cl− cotransporter; furosemide antagonizes, but low [Cl−]o facilitates, the cotransport of ions from intracellular to extracellular spaces (Fig. 8, top). However, both furosemide and the reduction of [Cl−]o have similar effects on glial Na+,K+,2Cl− cotransport because both treatments inhibit the movement of these ion species from extracellular to intracellular spaces. Given this background, we propose that furosomide and low [Cl−]o mediate their antiepileptic effects by antagonism of the Na+,K+,2Cl− cotransporter and not the K+,Cl− cotransporter because 1) blockade of GABAA-mediated inhibition itself (by either furosemide blockade of K+,Cl− cotransport or by the reduction of the transmembrane chloride gradient) appears to be neither necessary nor sufficient for the blockade of spontaneous epileptiform activity; indeed, specific blockade of GABAA-mediated inhibition through pharmacological means is known to be sufficient for inducing spontaneous bursting. Thus furosemide and low [Cl−]o must have mediated their antiepileptic effects in our experiments by affecting mechanisms independent of GABAA-dependent inhibition. 2) Furosemide and low chloride should have opposite effects on neuronal K+,Cl− cotransport but similar effects on glial Na+,K+,2Cl− cotransport. Because both furosemide and low [Cl−]o have nearly identical temporal effects on neuronal hypersynchronization, it is likely that glial rather than neuronal cotransport underlies the antiepileptic effects of these treatments.
The results of our experiments in which slices experienced prolonged exposure to low-[Cl−]o medium provide additional support for our hypothesis that the Na+,K+,2Cl− cotransporter modulates spontaneous, neuronal depolarization. Under conditions in which both [K+]o and [Cl−]o were minimal, long-lasting negative field events of increasing duration developed. When [K+]o and [Cl−]o were increased, shorter-lasting stable oscillations emerged. When [K+]o was reduced to control conditions but [Cl−]o was further increased, the field events were further diminished so that only sporadic and infrequently occurring negative shifts were observed. These results indicate a higher sensitivity of these late-occurring spontaneous events to [Cl−]o over [K+]o and implicate the Na+,K+,2Cl− cotransporter. The observation that furosemide blocked the oscillatory events induced by low-[Cl−]o/high-[K+]o supports the hypothesis that these late-occurring events are indeed mediated by ion fluxes through the chloride cotransporters.
We can propose a phenomenological model to explain the emergence of the late-occurring spontaneous field events that arise as a result of prolonged low-[Cl−]o exposure (Fig. 8). Initially, under control conditions, the ion-product differences for neurons are such that K+ and Cl− are cotransported from intracellular to extracellular spaces (QN > 0), and the ion-product differences for glia are such that Na+, K+, and Cl− are cotransported from the ECS to intracellular compartments (QG < 0) (Fig. 8, box 1). When [Cl−]o is reduced, the ion-product differences are altered so that neuronal efflux of KCl is increased, but the direction of glial ion cotransport is reversed (QG > 0) so that there is a net efflux of KCl and NaCl from intracellular to extracellular spaces (Fig. 8, box 2). These changes result in buildup of extracellular potassium over time. Eventually, [K+]o reaches a level sufficient to induce the depolarization of neuronal populations, resulting in an even larger accumulation of [K+]o (Fig. 8, box 3). This large accumulation of extracellular ions would then serve to reverse the ion-product differences so that KCl would now be moved from extracellular to intracellular spaces (QN < 0 and QG < 0). Further clearance of the extracellular potassium could occur through diffusion to the perfusion medium, eventually resulting in transmembrane ion gradients similar to the initial conditions. By cycling through this process, repetitive depolarizing field events and oscillations could be generated.
These ion fluxes, resulting in dynamically changing osmotic gradients and subsequent water movement between intra- and extracellular compartments, would cause swelling and shrinking of the ECS. This model is consistent with a number of observations. 1) Furosemide blocked low-[Cl−]o–induced oscillations; antagonism of the chloride cotransporters would block the movement of ions between the ECS and intracellular compartments (Fig. 8, top). 2) The electrophysiological characteristics of the low-[Cl−]o–induced field shifts were more sensitive to changes in [Cl−]o than to changes in [K+]o, suggesting the involvement of the glial cotransporter, dependent on the chloride concentration squared. 3) Neuronal depolarization and increases in [K+]o were observed at times significantly before the negative field shifts (Fig. 6), suggesting that a gradual accumulation of extracellular potassium and consequent depolarization of a neuronal population generated the negative field shifts. 4) Previous observations showed that both furosemide and low [Cl−]o block activity-induced changes in the intrinsic optical signal and in the volume of the ECS (Holthoff and Witte 1996; MacVicar and Hochman,1991) while maintaining synaptic hyperexcitability (Hochman et al. 1995). These observations suggest that furosemide and low [Cl−]o block activity-dependent glial swelling and diminish the neuronal transmembrane chloride gradient (Fig. 8, box 2).
We assumed in our model that a sodium-independent K+,Cl− cotransporter is localized on neurons and that a Na+,K+,2Cl− cotransporter is localized on glia. These assumptions are based on functional data acquired from cortical and hippocampal neurons (Misgeld et al. 1986; Thompson and Gähwiler 1989; Thompson et al. 1988) and from glial cells in primary culture (Tas et al. 1987; Walz and Hertz 1984). However, it may be that there is an active Na+,K+,2Cl− cotransporter on neurons. Our own data cannot make this distinction (nor does our hypothesis depend on the localization of the Na+,K+,2Cl− cotransporter). Recently, a neuronal-specific isoform of a putative K+,Cl− cotransporter was identified in rat brain (Payne et al. 1996); in situ hybridization studies showed that the transcript of this isoform is expressed at high levels in neurons throughout the CNS, including hippocampal CA1 pyramidal cells, and is absent from glial cells. There is also a recent report of a neuronal-specific isoform of a Na+,K+,Cl− cotransporter (Plotkin et al. 1997); however, this study failed to find any significant expression of this cotransporter in hippocampal neurons. In another study, measurements from cultured hippocampal neurons stained with a Cl−-sensitive fluorescent probe suggested the possibility of a neuronal Na+,K+,2Cl− cotransporter localized on dendrites (Hara et al. 1992). None of the localization studies yet provide a molecular identification of a glial-associated Na+,K+,2Cl− cotransporter.
Our study provides evidence that chloride cotransport, presumed to be a glial Na+,K+,2Cl− cotransporter based on physiological studies (Tas et al. 1987; Walz and Hertz 1984), plays a critical role in the modulation of neuronal synchronization. There are a number of factors that might play a role in the modulation of synchronization and that could be affected by blockade of Na+,K+,2Cl− cotransport. These factors include 1) blockade of activity-induced changes in the ECS (Ballanyi and Grafe 1988; Holthoff and Witte 1996; MacVicar and Hochman 1995) leading to a reduction of ephaptic interactions and reduced transient extracellular ion concentrations (Jefferys 1995; McBain et al. 1990), 2) diminished capacity for ion homeostasis in the ECS and intracellularly in neurons and glia (Kimelberg 1990; Lux et al. 1986; Misgeld et al. 1986; Thompson and Gähwiler 1989; Thompson et al. 1988; Walz and Hertz 1984), and 3) possible changes in the pH dynamics of the intra- and extracellular spaces affecting putative pH-dependent signaling mechanisms (Chesler 1990; Gottfried and Chesler 1994; Pappas and Ransom 1994; Ransom 1992). Previous optical imaging studies of slices provided evidence that, concomitant with the blockade of spontaneous bursting, there is a blockade of activity-evoked changes of the extracellular space (Hochman et al. 1995). That observation suggested that mechanisms correlated to the regulation of activity-evoked changes in the ECS (Holthoff and Witte 1996) might be critical for the generation and maintenance of the hypersynchronized activity associated with epileptogenesis. However, other mechanisms were not explicitly addressed.
Whatever the specific combination of mechanisms through which the Na+,K+,2Cl− cotransporter exerts control over neuronal synchronization, it may be that pharmacological agents that preferentially target and antagonize this cotransporter could block seizure discharges without affecting synaptic transmission. Such an approach to seizure control would be distinctly different from presently prescribed antiepileptic medications that are thought to act by directly affecting synaptic interactions indiscriminately in both epileptogenic and normal areas in the brain (Macdonald and Greenfield 1997; Upton 1994).
The authors thank Drs. B. R. Ransom and A. Brown for assistance with the [Ca2+]o measurements in our low-[Cl−]o medium.
This work was supported by National Institute of Neurological Disorders and Stroke Grants NS-35548 to P. A. Schwartzkroin and NS-07144 to D. W. Hochman.
Address for reprint requests: P. A. Schwartzkroin, Dept. of Neurological Surgery, Box 356470, University of Washington, Seattle, WA 98195.