A Furosemide-Sensitive K⁺–Cl⁻Cotransporter Counteracts Intracellular Cl⁻ Accumulation and Depletion in Cultured Rat Midbrain Neurons

Cl⁻ gradient between soma and dendrites

When tight-seal whole-cell recordings were performed under conditions of 15 mm permeant anions in the patch pipette ([A⁻]_pip), 2 mm[K⁺]_o, and pharmacological blockade of ionotropic glutamate receptors, sIPSCs did not reverse in sign at a defined holding potential (V_H). Instead, at V_H in a range near the reversal potential (E_Cl = −61 mV as determined from the Nernst equation), inward and outward sIPSCs concurred in 46 of 51 cells (Fig. 1A). All sIPSCs were blocked by bicuculline (n = 5), suggesting that they were mediated by GABA_A receptors (Fig.1B). Inward and outward sIPSCs appeared to be separated (Fig. 1A1) or as sequences of inward currents curtailed by outward currents (Fig. 1A2). As shown in Figure 1A3, another pattern consisted of clustered inward sIPSCs that were accompanied by a continuous barrage of outward sIPSCs or vice versa. We analyzed the time courses of apparently isolated inward and outward sIPSCs in a single cell at aV_H at which amplitudes of inward and outward IPSCs were in a similar range (Fig. 1C,D). There was a clear difference in the time courses. Outward sIPSCs consistently had a slower rise time and time constant of decay than inward sIPSCs (rise time, 1.9 ± 0.1 msec vs 5.3 ± 0.4 msec; decay time constant, 16.2 ± 1.3 msec vs 38.6 ± 4.8 msec for inward and outward sIPSCs, respectively; n = 5 cells). On the basis of these observations, we assumed that outward sIPSCs were generated in dendritic compartments but that inward sIPSCs were generated near soma. Presynaptic GABAergic neurons of the network would generate inward and outward IPSCs if they formed synapses with all parts of the somatodendritic surface. Presynaptic neurons contacting restricted areas either near soma or in apical dendrites would generate a continuous barrage or clusters of inward or outward IPSCs consistent with their firing mode, continuous or in bursts. Assuming further that under the given recording conditions in the culture (see Discussion) soma and dendrites did not deviate considerably from isopotentiality, we had to postulate [Cl⁻]_i to be lower in dendritic than in somatic compartments.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

GABAergic network neurons generated inward and outward IPSCs in a single target neuron through GABA_Areceptors. A1, Spontaneous IPSCs were inward and outward current transients in a cell recorded at V_H−63 mV. Inward and outward sIPSCs occurred in isolation (1), or inward currents were curtailed by outward sIPSCs (2). Note the difference in calibration between 1 and 2. A3, In another cell inward sIPSCs occurred in clusters interrupted by “silent” periods, whereas outward sIPSCs occurred rather continuously (V_H −71 mV). B, The GABA_A antagonist bicuculline (bic) reversibly blocked all inward and outward sIPSCs. a, b, Expanded segments of the top trace where indicated. C, D, Inward and outward sIPSCs differed in their time courses. Inward (D1) and outward (C1) sIPSCs that occurred in separation could be fitted by a single exponential function (solid line; V_H −63 mV). For the sIPSCs shown, the time constant of decay was 26.2 msec (amplitude coefficient 39.4) and 6.0 msec (amplitude coefficient −33.1). For analysis, only sIPSCs that had no inflections in the rise or decay phase were accepted. Rise time (10–90%) histogram of all inward (D2) and outward (C2) sIPSCs showed that rise time was slower for outward sIPSCs. Similarly, histograms of decay time constant (C3, D3) revealed a much slower decay for outward sIPSCs (C3). For the histograms, 31 outward and 45 inward IPSCs (sampled in 1 min) were analyzed. All recordings in this Figure were obtained in the presence of ionotropic glutamate receptor antagonists and of 2 mm[K⁺]_o([A⁻]_pip 15 mm).

To examine compartmental differences in [Cl⁻]_i we applied GABA (1 mm) focally to somatic and dendritic regions (Fig.2). To provide comparable conditions of origin for sIPSCs and GABA-evoked currents (I_GABA), application was adjusted so that the amplitude of I_GABA was in the range of amplitudes of sIPSCs. When tested in a voltage range around the reversal potential (E_GABA ± 25 mV), slope conductances of dendritic and somatic I_GABA were linear (Fig. 2). However, slope conductance was smaller for dendritic than for somatic applications, and often pressure or application time for the focal application had to be increased to obtain a measurable dendriticI_GABA. The difference was probably caused by different density of receptors and/or surface size facing the application pipette if GABA was applied near soma or near dendrites. As expected from the observation of inward and outward sIPSCs, dendriticI_GABA (application >100 μm away from soma) was outward at a V_H at which somaticI_GABA of the same cell was inward (Fig. 2). In a singe cell, E_GABA differed by up to 25 mV (mean 12.9 ± 2.3 mV; n = 9) between soma and dendrites (Table 2).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Dendritic and somatic GABA applications generated Cl⁻ currents with different reversal potentials. In the same cell, currents induced by focal application of GABA (1 mm) to a dendrite (▪) or the soma (●) had different reversal potentials. Application sites as shown for a cell in theinset (s, somatic; d, dendritic). Recording conditions were 2 mm[K⁺]_o and 15 mm[A⁻]_pip. Letters mark the points on the current–voltage relationship at which the sample currents shown above had been recorded. Symbolsrepresent the mean of three consecutive applications. Note the voltage range at which GABA evoked an outward current in the dendrite and an inward current at the soma.

Dependence of the somatodendritic Cl⁻ gradient on a furosemide-sensitive transport

Calculation of inorganic [A⁻]_ifrom E_GABA values reported above revealed that dendritic [A⁻]_i was lower than somatic [A⁻]_i. If dendritic [Cl⁻]_i was lowered by a furosemide-sensitive transport, it should be possible to reduce the somatodendritic [Cl⁻] gradient by furosemide application. Therefore, we tested the effects of furosemide on sIPSCs and I_GABA. Furosemide altered the driving force for sIPSCs. At V_H near the expectedE_IPSC, amplitudes of inward sIPSCs increased and those of outward sIPSCs decreased or they reversed in sign (Fig. 3A). A new apparent steady state was achieved after maximally 1 min, and the effect of furosemide was fully reversible in a similar time. Effects of furosemide were observed at concentrations as low as 10 μm (Fig. 3B; n = 4 of 5 cells) and increased in a concentration-dependent manner (n = 9). The maximal concentration we used in all further experiments was 0.1 mm to avoid unspecific effects of furosemide (Cabantchik and Greger, 1992). To quantify the effect of furosemide onE_GABA, current–voltage relationships forI_GABA were determined in the presence and absence of furosemide (0.1 mm). In one series of experiments (n = 15), the position of the application pipette was guided by the current direction of sIPSCs. First, the range of V_H was determined at which inward and outward sIPSCs were measurable. Thereafter, a V_H at the positive side of this range was chosen to record predominantly outward sIPSCs that were likely generated in dendritic compartments (Fig.4A1). In this situation the position of the application pipette was chosen so thatI_GABA was inward. In most instances the compartment reached from the soma up to 100 μm away from the somatic patch pipette for smoothly tapering dendrites (Fig. 2,inset). At a distance from soma of ∼200 μm, current direction of I_GABA and sIPSCs concurred (Fig.4A2). Furosemide only slightly increased the amplitude of somatic I_GABA (Fig.4A1), whereas dendritic I_GABAreversed in sign (Fig. 4A2). As shown in Figure4B, a new steady-state for dendriticI_GABA was achieved within 30–45 sec. The effect of furosemide was fully reversible. The data indicate a pronounced shift of E_GABA under furosemide in dendrites but not at the soma.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Furosemide changed the reversal potential of sIPSCs. A, Spontaneous IPSCs were inward and outward atV_H of −61 mV. Furosemide (furo; 0.1 mm) increased the frequency and amplitude of inward sIPSCs (recorded in 2 mm[K⁺]_o, 15 mm[A⁻]_pip), whereas outward sIPSCs disappeared. The onset of the furosemide effect is already seen after 10 sec, and the effect is reversible. Letters mark time points that are shown at higher sweep speed in the lower traces.B, In another cell under otherwise identical recording conditions as in A, V_H was set to record outward sIPSCs. Application of increasing concentrations of furosemide first reduced the amplitude of outward sIPSCs and then reversed sign of sIPSCs.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Furosemide affected the reversal potential of dendritic I_GABA more strongly than of somatic I_GABA. A, Pressure ejection of GABA (▴) (distance to the soma is given inparentheses) induced an inward current when applied close to the soma (A1) but an outward current in dendrites (A2). Under control recording conditions (15 mm [A⁻]_pip; 2 mm [K⁺]_o; V_H −66 mV), sIPSCs were recorded at aV_H at which outward sIPSCs predominated. In the presence of tetrodotoxin (TTX) to eliminate sIPSCs, furosemide reversed the current direction of dendriticI_GABA (A2), whereas the amplitude of somatic I_GABA was increased only slightly (A1). Both effects were completely reversible. To obtain a clear signal in presence of sIPSCs, slightly larger ejection pressure was used than was needed in TTX.B, Chart recording of the time course of the furosemide-induced shift of E_GABA. In the same cell as in A, dendriticI_GABA measured in TTX reversed current direction on application of furosemide. This effect readily reversed on washout.

In another set of experiments, furosemide (0.1 mm)-induced shifts in E_GABA were quantified for dendritic (application ∼200 μm apart from soma) and somaticI_GABA. As shown in Figure5A, furosemide shifted dendritic and somatic E_GABA in positive direction but did not reduce slope conductance. This suggests that furosemide changed [A⁻]_i but had no blocking action on GABA_A receptors. Furosemide (0.1 mm) shifted somatic E_GABA less than dendritic E_GABA (4.2 ± 1.0 mV vs 11.8 ± 1.4 mV; n = 8) (Fig. 5B), with the result that furosemide reduced the somatodendritic inorganic [A⁻] gradient. These findings showed that at 15 mm [A⁻]_pip and 2 mm [K⁺]_o the somatic inorganic [A⁻]_i was mainly governed by A⁻ diffusion from the patch pipette into the cytoplasm of the cell body. Dendritic inorganic [A⁻] was significantly lowered with respect to somatic [A⁻]_i by the operation of a furosemide-sensitive Cl⁻ outward transport.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Furosemide reduced the difference of reversal potentials of I_GABA between dendrites and soma. A, Current–voltage relationship of somatic (A1) and dendritic (A2)I_GABA could be fitted by a linear regression line (15 mm [A⁻]_pip and 2 mm [K⁺]_o). Furosemide shifted the current–voltage relationship to more positive values, whereas the slope conductance was unchanged. The effect was reversible on washout. E_GABA in the absence of furosemide was more negative for dendriticI_GABA than for somaticI_GABA. In furosemide, the difference between dendritic and somatic E_GABA was small, indicating that the change of E_GABA was more pronounced at the dendrites than at the soma. B,Bar charts summarize meanE_GABA (and SEM) for the different recording conditions. The number of cells is nine for somaticI_GABA and six for dendriticI_GABA.

Dependence of the direction of the Cl⁻transport on [K⁺]_o and [Cl⁻]_i

The concentrations of furosemide used in our study inhibit the K⁺–Cl⁻ cotransporter (KCC2) in heterologous expression systems. At concentrations of [Cl⁻]_i (4–10 mm) and [K⁺]_o (2–10 mm) that can be expected to occur in the mammalian brain, KCC2 operates near its thermodynamic equilibrium. The direction of Cl⁻transport depends on [Cl⁻]_i and [K⁺]_o (Payne, 1997). We tested whether these characteristics also apply to the regulation of Cl⁻ homeostasis in neurons.

I_GABA at 15 mm[A⁻]_pip

Increasing [K⁺]_o from 2 to 10 mm shifted dendritic E_GABA to positive values (Table 2). An apparent, new steady state was attained only after the inward current elicited by [K⁺]_o had peaked (Fig.6A1). When measured at 2 and 10 mm [K⁺]_o in the same cells, the shift of dendritic E_GABA mounted to 16.5 ± 3.3 mV (n = 3). In 10 mm[K⁺]_o, furosemide (0.1 mm) shifted dendritic E_GABA to even more positive values (3.3 ± 0.7 mV; n = 4), but the shift was consistently smaller than it had been at 2 mm[K⁺]_o (11.8 ± 1.4 mV;n = 8) (Fig. 6A2–A4). This result corresponds to the expected diminution of the driving force for the K⁺–Cl⁻ cotransporter at high [K⁺]_o. Assuming that somatic inorganic [A⁻] was dominated by [A⁻]_pip, elevation of [K⁺]_o should not result in large changes in somatic E_GABA. As expected, increasing [K⁺]_o from 2 to 10 mm hardly changed somatic E_GABA(1.0 ± 0.5 mV; n = 3) when tested in the same cells, and furosemide had no effect on somaticE_GABA at 10 mm[K⁺]_o (Table 2). Consequently, in 10 mm [K⁺]_o and furosemide, there was little if any somatodendritic gradient for inorganic [A⁻].

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

The furosemide-sensitive K⁺–Cl⁻ cotransporter extruded or accumulated Cl⁻ depending on [K⁺]_o and [Cl⁻]_i. A, The shift of dendritic E_GABA induced by furosemide was larger in 2 mm [K⁺]_o than in 10 mm [K⁺]_o when [A⁻]_pip was high (15 mm).A1, Elevating [K⁺]_ofrom 2 to 10 mm shifted E_GABA of dendritic I_GABA (▴ marks application of GABA) to positive values, which resulted in a reversal of current direction (V_H is indicated below chart records). Washout of furosemide had only a small effect onI_GABA in 10 mm[K⁺]_o (A2) but a strong effect in 2 mm [K⁺]_o.A4, Current–voltage relationship for the cell shown inA1–A3. For clarity, control and recovery data were pooled. Arrows indicate direction of the shift in dendritic E_GABA. B, In another cell recorded with low [A⁻]_pip (4.5 mm), furosemide induced a small positive shift at 2 mm[K⁺]_o but a small negative shift of dendritic E_GABA at 10 mm[K⁺]_o.

I_GABA at 4.5 mm[A⁻]_pip

In the experiments presented above, the patch pipette was used to load the somatic compartment with moderate inorganic [A⁻]. To evaluate the role of [Cl⁻]_i, we lowered [A⁻]_pip (4.5 mm). With 4.5 mm [A⁻]_pip, dendritic I_GABA was more negative than with 15 mm [A⁻]_pip (Table 2). According to the smaller driving force for the K⁺–Cl⁻ cotransporter at 4.5 mm [A⁻]_pip and 2 mm [K⁺]_o, the furosemide (0.1 mm)-induced shift of dendriticE_GABA was small (7.0 ± 0.9 mV;n = 6) (Fig. 6B). Increasing [K⁺]_o from 2 to 10 mmstrongly shifted dendritic E_GABA to positive values (27.8 ± 2.8 mV; n = 6). At 10 mm [K⁺]_o, furosemide induced small shifts of dendriticE_GABA; however, they were in the direction opposite to the shifts in 2 mm[K⁺]_o (−5.0 ± 0.7 mV;n = 7) (Fig. 6B). These data indicate that at low [A⁻]_pip the direction of the Cl⁻ transport in the dendrites had been reversed by elevating [K⁺]_o. At 2 mm [K⁺]_o, [A⁻]_pip and the furosemide-sensitive Cl⁻ outward transport determined dendritic [Cl⁻]_i, whereas at 10 mm [K⁺]_o, dendritic [Cl⁻]_i was governed by the Gibbs–Donnan equilibrium and the furosemide-sensitiveinward transport.

We next tested the effects of [K⁺]_ochanges on somatic E_GABA at low [A⁻]_pip. At 2 mm[K⁺]_o, somaticE_GABA was more positive than dendriticE_GABA (Table 2). However, elevation of [K⁺]_o and application of furosemide (0.1 mm) had effects similar to those on dendriticE_GABA (Table 2). At 2 mm[K⁺]_o, furosemide shifted somatic E_GABA to more positive values (4.2 ± 0.5 mV; n = 8). Elevating [K⁺]_o from 2 to 10 mmshifted somatic E_GABA in the same direction (7.0 ± 0.1 mV; n = 4) and reversed the shift of somatic E_GABA under furosemide (−3.4 ± 0.6 mV; n = 6). In 10 mm[K⁺]_o, therefore, no somatodendritic gradient for inorganic [A⁻]_i existed in either the presence or absence of furosemide.

sIPSCs

Finally, we tested whether inward and outward sIPSCs were altered by [Cl⁻]_i, [K⁺]_o, and furosemide as expected from the data reported on I_GABA. At 2 mm [K⁺]_o and low [A⁻]_pip, the mean amplitude of sIPSCs measured at different V_H could not be fitted by a linear regression because sIPSCs appeared as inward and outward currents near E_IPSC (n = 4) (Fig. 7A). Increasing [K⁺]_o from 2 to 5 mm while recording from the same cell allowed the current–voltage relation to be fitted by a first order regression because the sIPSCs reversed in sign at a defined V_H (Fig. 7B). At 5 mm [K⁺]_o, the ratio of [Cl⁻]_o/[Cl⁻]_iapproximately equaled [K⁺]_i/[K⁺]_o; hence no driving force for an electroneutral K⁺–Cl⁻ cotransport existed, leading to a collapse of the somatodendritic [A⁻] gradient. Furthermore, the time course of inward IPSCs measured near the reversal potential became slower in 5 mm[K⁺]_o than apparent inward IPSCs had been in 2 mm [K⁺]_o in the same cell (rise time, 1.48 ± 0.04 msec vs 1.00 ± 0.04 msec; time constant of decay, 16.5 ± 0.1 msec vs 7.1 ± 0.3 msec for 59 inward sIPSCs in 5 mm[K⁺]_o and 37 inward sIPSCs in 2 mm [K⁺]_o atV_H −77 and −87 mV, respectively). As expected from the thermodynamic properties of the K⁺–Cl⁻ cotransporter, increasing [K⁺]_o to 8 mmat 4.5 mm [A⁻]_pipreversed the transport direction of the Cl⁻transport. Under these conditions, furosemide decreased the amplitude of inward sIPSCs and increased the amplitude of outward sIPSCs (n = 3) (Fig. 8). The changes were the opposite of the changes induced by furosemide at 15 mm [A⁻]_pip and low [K⁺]_o (Fig. 3A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 7.

Near the thermodynamic equilibrium of the K⁺–Cl⁻ cotransporter, GABAergic network neurons generated sIPSCs with the same reversal potential at the soma and the dendrites of the target neuron. A, In a cell (4.5 mm[A⁻]_pip), inward and outward sIPSCs occurred at a V_H −80 to −90 mV. Current–voltage relationship for the sIPSCs was determined by sampling peaks of inward or outward currents, respectively (A1).A2, Traces show representative examples for inward and outward sIPSCs at two V_H values that were 5 mV apart. B1, In the same cell as in A, the current–voltage relationship for peaks of all sIPSCs could be well fitted by a linear regression line (r = 0.96) at 5 mm [K⁺]_o. In the plots, mean values and SE bars are shown unless SE bars remained within symbol sizes. Minimum to maximum of sIPSC numbers sampled at eachV_H ranged from 48 to 132.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 8.

Under conditions of high [K⁺]_o and low [A⁻]_pip, furosemide shiftedE_IPSC to a more negative value.A, Furosemide reduced the amplitude of inward sIPSCs (recorded with 4.5 mm[A⁻]_pip and 8 mm[K⁺]_o). B, In the same cell recorded under identical conditions as inA but at a more positiveV_H, furosemide increased the amplitude of outward sIPSCs. Letters mark segments on the chart plots that were shown at higher sweep speed in the lower traces.

Reduction of a [Cl⁻]_i gradient between dendrites and soma by furosemide or [K⁺]_o

We observed a striking difference in driving forces for GABA_A currents between cell soma and dendrites. Both, bicuculline-sensitive sIPSCs and focal GABA applications revealed the gradient. At a V_H near the expected equilibrium potential for Cl⁻, GABA_A currents were inward near soma and outward in dendritic parts. Furosemide reduced the somatodendritic differences in the driving forces, inducing a shift inE_GABA that was more pronounced for dendritic currents than for somatic currents. A reduction of the somatodendritic gradient was also observed if [K⁺]_owas increased or [A⁻]_pip was reduced.

Dendritic [Cl⁻]_i resulting under our experimental conditions (Fig.9A) can be explained from Gibbs–Donnan equilibrium and the thermodynamic profile of an electroneutral K⁺–Cl⁻cotransport (Fig. 9B). K⁺ and Cl⁻ can be expected to be nearly in equilibrium across the neuronal membrane. [Cl⁻]_iequal to [K⁺]_o approximately satisfied the Donnan relationship as well as the principle of electroneutrality. [Cl⁻]_i was actually increased above [Cl⁻]_i predicted from the Donnan relationship or reduced below that concentration by a higher or lower [Cl⁻]_pip, respectively. Near soma, [Cl⁻]_pip determined [Cl⁻]_i more closely than in the dendrites. Depending on [K⁺]_o, dendritic [Cl⁻]_i was lower (Fig.9Aa–Ac) or higher (Fig. 9Ad) than somatic [Cl⁻]_i, indicating that Cl⁻ movement across conductivity pathways exceeded Cl⁻ diffusion from the patch pipette. The electroneutral K⁺–Cl⁻cotransporter contributed to the conductivity pathways according to its energy profile (Fig. 9B), as revealed by the shift in dendritic [Cl⁻]_i resulting from furosemide application. When [K⁺]_o was low and [Cl⁻]_pip was high (Fig.9Aa), the driving force for the K⁺–Cl⁻ cotransporter was sufficiently large to generate a pronounced somatodendritic gradient. The somatodendritic gradient was built up by the solution in the patch pipette, which served as a Cl⁻ source, and the K⁺–Cl⁻ cotransporter, which counteracted that Cl⁻ load. The development of a Cl⁻ gradient allowed us to study Cl⁻ regulation, but the experiments did not determine under which conditions such a gradient may exist in intact neurons.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 9.

Cl⁻ homeostasis in cultured neurons can be explained by the operation of an electroneutral K⁺–Cl⁻ cotransporter and the Gibbs–Donnan equilibrium. A, [Cl⁻]_i was calculated fromE_GABA listed in Table 2. Concentrations of the inorganic anions in the patch pipette were as indicated.Arrows indicate changes of [A⁻]_i calculated from changes inE_GABA of somatic or dendriticI_GABA induced by furosemide.Letters correspond to driving forces for the K⁺–Cl⁻ cotransporter derived from the energy profiles shown in B. B, Plot of the driving force for the K⁺–Cl⁻ cotransporter as a function of [K⁺]_o for different [Cl⁻]_i (taken fromE_GABA calculated for dendrites inA). Calculations are as described in Material and Methods. Positive values indicate outward transport.

[Cl⁻]_i calculated from somaticE_GABA was always somewhat higher than [Cl⁻]_pip. The small deviation was likely attributable to injected currents carried by anions because we set V_H negative to the membrane potential. Supporting this suggestion, the relative increase of [Cl⁻]_i above [Cl⁻]_pip was larger with [Cl⁻]_pip of 4.5 mm than with [Cl⁻]_pip of 15 mm. The low mobility of the main anion glucuronate promoted an increase of [Cl⁻]_i above [Cl⁻]_pip. Other possible factors such as production of endogenous CO₂ or the use of Br⁻ were less likely contributors. An equal pH inside and outside the cells probably prevented buildup of a HCO₃⁻ gradient, and removal of the gradient for Br⁻ had no measurable effect onE_GABA (see Materials and Methods).

The observed fast and slow sIPSCs could represent an artifact of an imperfect space clamp. A similar finding was reported for GABA_A responses in CA1 neurons of hippocampal slices (Pearce, 1993). In that study, arguments were put forward indicating that a space-clamp artifact did not account for the different time courses of synaptic currents generated at different locations of the neuron. In comparison, the analysis of I_GABA in cultured neurons that have high input resistances (≥1 GΩ) and short dendrites (≤300 μm) recorded in the presence of TTX, QX314, and Cs⁺ should be much less hindered by geometrical factors (Müller and Lux, 1993; Draguhn et al., 1997). Furthermore, the slowing of the time course of inward sIPSCs when [K⁺]_o was increased (Fig. 7) indicated that the time course of inward sIPSCs was shortened by superimposed small outward sIPSCs as long as the somatodendritic [Cl⁻]_i gradient existed.

The neuronal-specific K⁺–Cl⁻cotransporter is inhibited by furosemide (K_i 25 μm) (Payne, 1997). Furosemide has been used to establish the involvement of a K⁺–Cl⁻cotransporter in synaptic inhibition (Misgeld et al., 1986; Thompson, 1994). Furosemide, however, has various disadvantages. In non-neuronal preparations, furosemide at concentrations >0.1 mminhibits not only cation–anion cotransporters but also other transport proteins and Cl⁻ channels (Cabantchik and Greger, 1992). Here we could observe immediate effects of furosemide at concentrations below 0.1 mm. Furosemide has been reported to block GABA_A channels that are composed of subunits including α₄ and α₆ (Wafford et al., 1996). These subunits are uncommon in midbrain areas (Wisden et al., 1992). In neither this nor a previous study (Jarolimek et al., 1996) on midbrain culture did we obtain evidence that furosemide reduced GABA_A currents. Depending onV_H, the amplitudes of IPSCs increased when neurons were exposed to furosemide, and slopes of current–voltage relationships for GABA_A currents remained unaltered. As shown in the previous study (Jarolimek et al., 1996), shifts of reversal potentials of the same order of magnitude as those ofE_GABA can be observed for glycine currents.

Furosemide blocks Na⁺–K⁺–Cl⁻cotransporters more effectively than it blocks the K⁺–Cl⁻ cotransporter (Cabantchik and Greger, 1992). Thus, the furosemide effects described here do not allow discrimination between the two transporters. Our data, however, exclude a significant contribution of a Na⁺–K⁺–Cl⁻cotransporter in the net Cl⁻ transport of cultured midbrain neurons. Because of the large inwardly directed driving force for Na⁺, the driving force of the net Cl⁻ transport (Fig. 9Ac,d) would not reverse near [K⁺]_o and [Cl⁻]_i predicted from the driving force of a K⁺–Cl⁻ cotransporter (Fig. 9Bc, d).

Regulation of the neuronal Cl⁻ homeostasis by the furosemide-sensitive K⁺–Cl⁻cotransporter

The electroneutral K⁺–Cl⁻cotransporter operates near equilibrium under conditions of normal internal and external ion concentrations (Jensen et al., 1993; Payne, 1997). If [K⁺]_o is kept at a high level (Fig. 9Ad), the K⁺–Cl⁻ cotransporter accumulates Cl⁻ in the cell. Under physiological conditions, Cl⁻ accumulation would be diminished by the concomitant removal of K⁺ from the extracellular space (Payne, 1997). During epileptiform activity, however, [K⁺]_o has been found to remain at high levels (10–12 mm) (Heinemann and Lux, 1977), with the result that the K⁺–Cl⁻transport actually accumulates Cl⁻ and hence intensifies epileptiform activity. Therefore, furosemide may have some antiepileptic potency under such an experimental condition (Hochman et al., 1995). In contrast, in neuronal culture we observed that furosemide induced burst activity (Jarolimek et al., 1996). In these experiments, [K⁺]_o was kept constant at 5 mm. In the presence of antagonists for ionotropic glutamate receptors, we observe frequent sIPSCs, suggesting that membrane potentials of the majority of cells are near to firing threshold, e.g. around −60 mV. The expected lower limit of [Cl⁻]_i is around 5.6 mmin 5 mm [K⁺]_o([Cl⁻]_o = 169 mm; [K⁺]_i assumed to be 150 mm). There is, however, an electromotor force (E_M − E_Cl) driving Cl⁻ into the cell. The hyperpolarizing gradient for synaptic inhibition will be maintained by the outwardly directed K⁺–Cl⁻ cotransport. The upper limit of [Cl⁻]_i is 20 mm if the K⁺–Cl⁻cotransporter is blocked by furosemide. Hence, in furosemide, the inhibitory potency of the GABA_A postsynaptic potentials is diminished or abolished.

CONCLUSIONS

In cultured neurons, the regulation of [Cl⁻]_i and the gradients for IPSCs is determined by the presence and activity of the K⁺–Cl⁻ cotransporter. The driving force for the net Cl⁻ transport is set by [K⁺]_o and [Cl⁻]_i. Values determined in this study indicate that [Cl⁻]_i is kept well below 10 mm as long as [K⁺]_o is 2 mm. Increases in [K⁺]_o will increase [Cl⁻]_i, but uptake of K⁺ into neurons on reversal of the transport will buffer [K⁺]_o changes (Payne, 1997). However, if [K⁺]_o remains high, the transport will accumulate Cl⁻ in the cells.In vivo additional factors determine gradients for IPSCs through GABA_A or glycine channels, complicating the analysis of their regulation. Despite this fact, a furosemide-sensitive K⁺–Cl⁻ cotransporter is likely to operate in central neurons because the KCC2 mRNA is expressed in hippocampal neurons (Payne et al., 1996). The permeability for HCO₃⁻ shiftsE_GABA by 10–15 mV (depending on pH_i) in a positive direction (Alvarez-Leefmans, 1990; Kaila, 1994). A regulation of GABAergic inhibition through [HCO₃⁻]_i (Kaila, 1994) is less likely given the low permeability ratio of GABA_Areceptors for HCO₃⁻/Cl⁻ (Bormann, 1988; Fatima-Shad and Barry, 1993). Furthermore, an inwardly rectifying chloride conductance (Smith et al., 1995), which does not appear to have a major influence in our cultured neurons (Jarolimek et al., 1996), may come into play. For an excitatory effect of GABA postnatally as opposed to an inhibitory effect in adult neurons (Cherubini et al., 1991), a transient dominance of a Na⁺–K⁺–Cl⁻cotransporter (LoTurco et al., 1995; Plotkin et al., 1997) probably coincides with a delayed expression of the K⁺–Cl⁻ cotransporter (Zhang et al., 1991).