Contribution of Resting Conductance, GABAA-Receptor Mediated Miniature Synaptic Currents and Neurosteroid to Chloride Homeostasis in Central Neurons

Abstract Maintenance of a low intraneuronal Cl– concentration, [Cl–]i, is critical for inhibition in the CNS. Here, the contribution of passive, conductive Cl– flux to recovery of [Cl–]i after a high load was analyzed in mature central neurons from rat. A novel method for quantifying the resting Cl– conductance, important for [Cl–]i recovery, was developed and the possible contribution of GABAA and glycine receptors and of ClC-2 channels to this conductance was analyzed. The hypothesis that spontaneous, action potential-independent release of GABA is important for [Cl–]i recovery was tested. [Cl–]i was examined by gramicidin-perforated patch recordings in medial preoptic neurons. Cells were loaded with Cl– by combining GABA or glycine application with a depolarized voltage, and the time course of [Cl–]i was followed by measurements of the Cl– equilibrium potential, as obtained from the current recorded during voltage ramps combined with GABA or glycine application. The results show that passive Cl– flux contributes significantly, in the same order of magnitude as does K+-Cl– cotransporter 2 (KCC2), to [Cl–]i recovery and that Cl– conductance accounts for ∼ 6% of the total resting conductance. A major fraction of this resting Cl– conductance is picrotoxin (PTX)-sensitive and likely due to open GABAA receptors, but ClC-2 channels do not contribute. The results also show that when the decay of GABAA receptor-mediated miniature postsynaptic currents (minis) is slowed by the neurosteroid allopregnanolone, such minis may significantly quicken [Cl–]i recovery, suggesting a possible steroid-regulated role for minis in the control of Cl– homeostasis.


Introduction
The intraneuronal Clconcentration, [Cl -] i , is critical for synaptic communication. The inhibitory actions of GABA and glycine depend on a relatively low [Cl -] i . [Cl -] i may, however, rise as a consequence of intense synaptic activity. When excitatory inputs depolarize the membrane, the driving force for inward Clflux through channels opened by GABA or glycine increases and may impose a strong Clload on the neuron. Clflows into the cell also during shunting inhibition (Doyon et al., 2016b). In some cases, the raised [Cl -] i changes the effect of GABA to excitatory (Khalilov et al., 2003;Astorga et al., 2015). Restoring [Cl -] i after large perturbations may thus be expected crucial for normal inhibitory GABAergic synaptic function. The K ϩ -Clcotransporter 2 (KCC2) is critical for the control of [Cl -] i and likely contributes significantly to [Cl -] i restoration (Doyon et al., 2016b). However, as a consequence of the negative resting potential maintained by neurons, passive Clflux through Clconducting membrane pathways may also contribute to the recovery of a low [Cl -] i . In the absence of Cltransporters, the Clequilibrium potential (E Cl ) will settle at the membrane resting potential. The relative importance of KCC2 and passive, "conductive" flux is at present unknown.
Conductive [Cl -] i recovery depends on the resting membrane Clconductance (g Cl ). There is, however, no generally accepted view on the resting Clpermeability or g Cl in central neurons. A low resting permeability has been suggested by Thompson et al. (1988), but a significant conductance has been described in other studies (Forsythe and Redman, 1988;Sacchi et al., 1999;Müller, 2000).
The molecular basis for resting g Cl is also not well understood. It has been proposed that ClC-2 channels are general contributors to the resting g Cl in neurons (Rinke, 2010; but see Stölting et al., 2014). However, the expression of ClC-2 is highly restricted to specific neuronal types (Smith et al., 1995) or neuronal compartments (Földy et al., 2010). Another contribution to resting g Cl may be provided by GABA A receptors, which underlie a tonic conductance in some neurons (Semyanov et al., 2004;. It seems possible that also the spontaneous GABA release that is seen as miniature postsynaptic potentials/ currents (minis) contributes significantly to resting g Cl . The functional role of such minis is not clear. Although they imply relatively small currents, we hypothesized that they may be important for conductive recovery of [Cl -] i after perturbations, at least when prolonged by endogenous neurosteroids that target GABA A receptors and in some conditions also enhance GABA release (Haage and Johansson, 1999).
In the present work, we estimated the contribution of passive Clflux to [Cl -] i recovery, in preoptic neurons from rat. We also developed a novel method to estimate the relative g Cl at rest and analyzed whether ClC-2 channels and GABA A receptors contribute to resting g Cl . Finally, we tested the hypothesis that spontaneous GABA release is important for [Cl -] i recovery. To improve the electrical recording conditions and to avoid possible [Cl -] i redistribution between neuronal compartments, we used acutely isolated neurons from which neurites were largely removed, but which retain attached presynaptic nerve terminals that enable analysis of spontaneous GABA release (Haage et al., 1998). Cells were loaded with Clby combining GABA or glycine application with a depolarized voltage and [Cl -] i was obtained from measurements of E Cl by gramicidin-perforated patch recordings with GABA or glycine application in combination with voltage ramps (Karlsson et al., 2011;Yelhekar et al., 2016).
Here, we show that passive Clflux contributes significantly to [Cl -] i recovery after a high load, and, using our novel method, we also show that Clconductance accounts for ϳ6% of the total resting conductance. A major fraction of this resting Clconductance is picrotoxin (PTX)-sensitive and thus likely due to GABA A receptors. Further, we show that when the decay of GABA A receptor-mediated minis is slowed by the neurosteroid allopregnanolone, such minis contribute significantly to [Cl -] i recovery, thus suggesting a possible steroidregulated role for minis in the control of Clhomeostasis. synaptic terminals remained attached to the postsynaptic cell bodies (Haage et al., 1998).

Solutions used for preparation and for recording
A solution of the following composition was used to cut and incubate brain slices and for mechanical dissociation of individual neurons: 150 mM NaCl, 5.0 mM KCl, 2.0 mM CaCl 2 , 1.2 mM MgCl 2 , 10 mM HEPES, 10 mM glucose, and 4.93 mM Tris-base, pH 7.4 (95% O 2 , 5% CO 2 ). The extracellular solution used for recording contained: 137 mM NaCl, 5.0 mM KCl, 1.0 mM CaCl 2 , 1.2 mM MgCl 2 , 10 mM HEPES, and 10 mM glucose, pH 7.4 (NaOH). The standard pipette-filling solution contained: 140 mM K-gluconate, 3.0 mM NaCl, 1.2 mM MgCl 2 , 10 mM HEPES, and 1.0 mM EGTA, pH 7.2 (KOH). Gramicidin (Sigma-Aldrich) was prepared from a stock solution (120 mg/1.0 ml DMSO) to a final concentration of 600 g/ml of pipette-filling solution. Alternatively, amphotericin B was prepared from a stock solution (6 mg/100 l DMSO) and added to a final concentration of 120 g/ml of pipettefilling solution. Pipette tips were filled with a similar solution without gramicidin or amphotericin B. Cells were continuously (between test applications of agonists) perfused with gravity-fed extracellular solution provided by a custom-made pipette positioned 100-200 m from the studied cell. Computer-controlled exchange to solutions containing agonist via solenoid valves occurred with a time constant of ϳ50 ms, as measured by the change in offset potential on changing between high and low K ϩ concentrations in the perfusate (Karlsson et al., 2011).

Electrophysiological recording
Electrophysiological experiments to study [Cl -] i recovery were made by using the gramicidin-perforated patch technique, which enables recording of whole-cell currents without interference with the intracellular Clconcentration (Abe et al., 1994;Kyrozis and Reichling, 1995) and without current rundown (Wang et al., 2003). Experiments to estimate resting Clconductance, which depend on estimates of [Cl -] i but are compatible with slow Clequilibration with the patch pipette, were made using the quicker amphotericin B-perforated patch technique (Rae et al., 1991). Both perforated-patch techniques are compatible with rapid Clloading (Karlsson et al., 2011). The majority of experiments were made under voltage-clamp conditions, but a few control experiments were made under current-clamp conditions at zero current. Patch pipettes had a resistance of 3-4 M⍀ when filled with standard pipette-filling solution and immersed in the standard extracellular solution (see above). Series resistance was evaluated repeatedly during experiments, by the "membrane test" provided in the Clampex software (versions 9 and 10; Molecular Devices), and was typically 15-40 M⍀. Online series-resistance compensation, which can never be complete and is associated with extra noise, was not used. For more complete compensation, all voltages were corrected for series resistance offline with voltage error computed from raw current and series resistance. Liquid᎑junction potentials were calculated using the Clampex software and have been subtracted in all potentials given. All experiments were performed at room temperature (21-23°C).

Loading cells with Cland estimation of [Cl -] i recovery time course
Cells were loaded with Cl -, to concentrations up to ϳ70 mM, by using combined depolarization and application of GABA (1.0 mM) or glycine (1.0 mM; Karlsson et al., 2011). Estimates of [Cl -] i during loading as well as during recovery after loading were calculated from E Cl as obtained from I-V relations constructed from rapid voltage ramps (rate of Ϯ 1.6 V s Ϫ1 ) during brief applications of GABA or glycine (100 M to 1.0 mM) and after correction for series resistance and subtraction of leak currents. The studied cells generated stable (for Ͼ10 min) responses to GABA as well as to glycine during perforated-patch recordings . Since the method used to estimate [Cl -] i depends on the current reversal potential, it is not sensitive to changes in the absolute GABA-or glycine-evoked conductance or peak current. Under the conditions used, without HCO 3 in the solutions, the reversal potentials for currents evoked by GABA or glycine (E GABA and E glycine ) are equivalent to E Cl (Chavas and Marty, 2003). The probe applications of GABA/glycine were given at a voltage close to the estimated E Cl to minimize influence on [Cl -] i . The charge transfer associated with a probe application of GABA/glycine was estimated to affect [Cl -] i by Ͻ1.0 mM, as calculated from the "equivalent volume" of cytosol. The equivalent volume, v equ , was obtained from the Clloading procedure and represented the volume in which the recorded charge transfer, Q, would cause the observed [Cl -] i changes, ⌬[Cl -] i , according to the relation: v equ ϭ QF Ϫ1 (⌬[Cl -] i ) Ϫ1 , where F is the Faraday constant. Thus, v equ corresponds to the expected cytosolic volume (Karlsson et al., 2011). Leak currents, including small voltage-dependent components, were obtained from voltage-ramps given in the absence of GABA and glycine. The obtained leak I-V relation (after correction for series resistance) was fitted by a linear function or by the exponential function I ϭ A ϩ B e V/C (A, B, and C being constants). The latter function was used to calculate the leak current component when GABA or glycine was present. This procedure takes the differences in voltage, arising as a consequence of the series resistance and different current amplitudes, during ramps with and without GABA/glycine into account, and, therefore, provides a better estimate of leak currents than the commonly used technique to directly subtract the current recorded in the absence of ligand from that recorded in the presence of ligand. The similar results obtained with GABA and with glycine to load cells with Cl - (Karlsson et al., 2011) show that the method used for loading is not critically influencing the results. Thus, GABA and glycine were here used interchangeably and each agonist was used for estimation of [Cl -] i while blocking receptors for the other to clarify their role in [Cl -] i regulation.

Derivation of relative Clconductance at rest
Previous methods for estimating the relative Clconductance of the neuronal membrane at rest have not been New Research straight-forward and dependent on many assumptions (Forsythe and Redman, 1988). We here aimed at a method with which we could estimate the relative Clconductance from voltage ramps applied under conditions with different Clconcentrations on the in-or outside of the membrane. The starting assumption was that the membrane current reversal potential could be described by the "conductance" version of the Goldman-Hodgkin-Katz equation (Goldman, 1943;Hodgkin and Katz, 1949;see, e.g., Koester and Siegelbaum, 2013): where E rev is the reversal potential for the total current, i.e., the resting potential, g is conductance with subscripts T, K, Na, and Cl for total, K ϩ , Na ϩ , and Cl -, respectively, and E K , E Na , and E Cl are the equilibrium potentials for respective ions. From Equation 1, we obtain (2) If we manipulate E Cl , e.g., by loading cells with Clas described above, without changing ion conductances and without changing equilibrium potentials for Na ϩ and K ϩ , then for the two conditions, before (subscript 1) and after (subscript 2) Clloading, (3) and if g T does not change (as was experimentally verified, see below), then (4) From Equation 4, we obtain the relative Clconductance: The relative Clconductance may thus be obtained from the shift in reversal potential of total current relative to the shift in Clequilibrium potential.
We here used continuous voltage ramp sequences, in a regular zigzag pattern, with application of glycine during part of the ramp sequence ( Fig. 1). By this procedure, estimates of conductance and reversal potential were obtained before as well as throughout the glycine application, with a time resolution determined by the ramp cycle frequency, here 20 Hz. The control ramps (i.e., before or in the absence of glycine application) were used to estimate total conductance (g T ) and reversal potential (E rev ), but also for subtraction of leak currents from the currents evoked by glycine to estimate (E Cl ). A first set of g T , E rev and E Cl estimates were obtained from voltage ramps symmetrically around a holding potential of -74 mV (i.e., near the resting potential). Subsequently the holding voltage was raised to -14 mV (with ramps symmetrically around this voltage, as indicated in Figure 1) and [Cl -] i was raised by glycine (1.0 mM) application for 20 s. E Cl was measured at the end of the 20-s glycine application (E Cl 2 a in Figure 1). Next, holding voltage was returned to -74 mV and a second estimate of g T (g T 2) and E rev (E rev 2) was obtained. A final estimate of E Cl (E Cl 2 b ) was made with ramps around a holding voltage of -34 mV. A linear interpolation between E Cl 2 a and E Cl 2 b was used to estimate E Cl at the time of g T 2 (and E rev 2), to account for [Cl -] i changes during this time interval. g T 2 was compared with g T 1 to verify that glycine-activated conductance had decayed and to satisfy the conditions for Equation 4. Variations Ͻ30% between the two measurements in the same cell were accepted (to account for fluctuations in resting conductance), and absence of any significant systematic difference was confirmed (see Figure 5 below). These measurements were made with 100 M Cd 2ϩ , to minimize transmitter release from adhering presynaptic terminals, and 30 mM TEA (replacing an equimolar amount of Na ϩ in the standard extracellular solution), to minimize voltagegated K ϩ currents, tetrodotoxin (TTX; 2.0 M), to block voltage-gated Na ϩ channels, and with the KCC2 blocker VU0255011-1 (10 M) to minimize outward Cltransport between the times for measuring E Cl 2 a and E Cl 2 b . To relate the derived resting g Cl to the total membrane conductance in the absence of blockers, we also considered the effect of the used blockers on the total resting conductance g T . Control experiments showed that the total resting conductance in the presence of all four blockers (TEA, Cd 2ϩ , TTX, and VU0255011-1; n ϭ 14) was, on average, only 54% (p ϭ 0.0038, Mann-Whitney test) of that without blockers (n ϭ 10). This effect could be ascribed to TEA and/or Cd 2ϩ since the total conductance in a combination of TTX and VU0255011-1 (n ϭ 14) was not significantly different from that without blockers (n ϭ 10).

Computation of [Cl -] i recovery time course
The time course of theoretically expected [Cl -] i recovery was computed for a simplified model of a spherical onecompartment cell (Karlsson et al., 2011). The model included Cltransport mediated by KCC2 as well as Clcurrent through a leak conductance. [Cl -] i was computed numerically according to where and t is time, ⌬t is the integration time step (here, 50 ms), g KCC2 is a transport proportionality factor ("apparent conductance", dependent on the number and transport capacity of KCC2 molecules) for KCC2, ⌬ K,Cl is the driving force for KCC2: and F is the Faraday constant and v equ is the equivalent volume, i.e., the cytosolic volume where Cl Ϫ equilibrates, taken to be 50% of the total cell volume (here, assumed 5.0 10 Ϫ13 l, i.e., 50% of the volume of a sphere of radius 6.2 m; cf Karlsson et al., 2011). The proportionality factor g KCC2 was adjusted to match KCC2-mediated [Cl -] i recovery to experimental results (experiments with minimized influence of g Cl , compare Fig. 2F), whereas g Cl was as experimentally recorded (compare Figure 5 below). The computations were made using Turbo Basic software (Borland).

Statistical analysis
Data are presented as the mean Ϯ SEM, with n representing the number of cells studied. Differences between groups were analyzed with the paired Wilcoxon signedrank test for paired data and with the Mann-Whitney test for unpaired data; p Ͻ 0.05 was considered statistically significant. Statistical analyses were performed using the software OriginPro 2016 (OriginLab).

Transporter-dependent recovery of [Cl -] i
To obtain an overall view of the [Cl -] i recovery capacity in dissociated medial preoptic nucleus (MPN) neurons, we first addressed transporter-dependent recovery after a Clload to concentrations significantly above baseline [Cl -] i , previously estimated to ϳ9 mM in this preparation (Karlsson et al., 2011). To estimate the capacity, we loaded the cells with Clby combined depolarization and GABA or glycine application. This resulted in an exponential (time constant 3.8 Ϯ 0.4 s, n ϭ 19) increase in [Cl -] i to 60 Ϯ 3 mM (n ϭ 19; Fig. 2A-C). After loading, [Cl -] i recovery was studied by voltage ramps (to estimate E Cl ) given from a holding potential that was kept as close as possible to the gradually changing E Cl , to minimize influence of passive conductive recovery and also to minimize influence of the GABA/glycine applications during recovery (Fig. 2D,E). This required test runs to obtain a rough estimate of the E Cl recovery time course. Under these conditions, favoring mainly nonconductive, transporter-dependent changes, [Cl -] i recovery varied somewhat between cells, following a roughly exponential, a doubly exponential or a nearly linear time course (Fig. 2F). For simplicity, we used the best fitted single-exponential for comparison between cells and recording conditions. In control conditions, the time constant was 499 Ϯ 49 s (n ϭ 19). Since the recovery time course was considerably slower than estimated for some other cell types, and with other techniques (Staley and Proctor, 1999;Wagner et al., 2001;Földy et al., 2010;Rinke et al., 2010;Deisz et al., 2011), we made a few recordings from neurons dissociated from the CA1 region in brain slices containing the hippocampus. However, in the three hippocampal cells tested, [Cl -] i recovery occurred with a time constant of 1877 Ϯ 60 s, thus even slower than for the MPN neurons.
To verify that the [Cl -] i recovery described was mediated by transporter activity, we applied several types of blockers. The time course of recovery was not significantly affected by a blocker (bumetanide, 10 M) of the Na ϩ -K ϩ -Clcotransporter 1 (NKCC1; Fig. 3A,D) but was dramatically slowed by furosemide (0.5 mM; Fig. 3B,D), which affects also KCC2, as well as by the selective KCC2 blocker VU0255011-1 (10 M; Fig. 3C,D). These effects are consistent with the idea that a major fraction of the observed recovery is mediated by KCC2. The kinetic details of transporter-dependent [Cl -] i recovery were not investigated further here.

Passive, conductive recovery of [Cl -] i
If a substantial conductive component contributes to [Cl -] i recovery, this should be seen as a voltagedependent component. KCC2-mediated Cltransport does not depend on voltage (see, e.g., Alvarez-Leefmans and Delpire, 2009). We therefore studied [Cl -] i recovery as above, but with membrane voltage kept at either -14 or -74 mV during the recovery phase. Test pulses of GABA/ glycine were still applied at a holding voltage close to E Cl , to minimize GABA/glycine-induced Clflux. The time at these voltages was, however, very small compared with the total recovery time. The recovery time constant observed at -14 mV was 843 Ϯ 208 s (n ϭ 8), significantly (p ϭ 0.0078; paired Wilcoxon signed-rank test) longer than that at -74 mV (375 Ϯ 95 s; n ϭ 8) in the same cells. The difference obtained by changing the holding voltage was reversible (Fig. 4A,B). In addition, [Cl -] i recovered to a lower level (4.3 Ϯ 1.2 mM, n ϭ 8) at -74 mV than at -14 mV (10.9 Ϯ 3.0 mM; n ϭ 8; p ϭ 0.039; evaluated from the asymptote of the fitted exponential function). Thus, the data are compatible with a substantial contribution of conductive Clflux to [Cl -] i recovery.
To quantify the conductive component of [Cl -] i recovery in isolation, we repeated the experiments with different holding voltages in the presence of the KCC2 blocker VU0255011-1 (10 M): There was only little recovery (7.9 Ϯ 2.7% of the loaded [Cl -] i recovered after 20 min; n ϭ 5) at -14 mV, whereas at -74 mV, [Cl -] i recovered with a time constant of 651 Ϯ 171 s (n ϭ 5) and 60 Ϯ 5% of the [Cl -] i increase during loading was eliminated after 20 min of recovery (Fig. 4C,D). Thus, it is clear that without KCC2 there is very small, if any, Clextrusion capacity to counteract the effect of conductive Cltransport on [Cl -] i at -14 mV and that at -74 mV, conductive Cltransport accounts for a recovery time course in the same order of magnitude as does KCC2 in the absence of blockers (compare Fig. 3, black filled symbols, with Fig. 4D, brown filled symbols).

Estimate of resting Clconductance
From above, it is clear that although KCC2 is required for a rapid and complete recovery of [Cl -] i to normal resting levels, a significant recovery takes place also by passive, conductive pathways in the absence of functional KCC2. The rate of such recovery will depend on the conductance to Cl -, g Cl . Resting g Cl will also be an important determinant of resting [Cl -] i . We therefore derived a method for estimating the resting g Cl , as described in Materials and Methods and Figure 1. Since the method depends on a total membrane conductance, g T , that does not change with Clloading, we first verified that g T before (g T 1) and shortly after (g T 2) Clloading was not significantly different (Fig. 5A). Subsequently, we computed the relative resting g Cl , from Equation 4 and the conductance and reversal potential measures obtained as described in Figure 1. A final computation accounted for the effect of used background blockers (TEA, Cd 2ϩ , TTX, and VU0255011-1) on the total membrane conductance (see Materials and Methods). The obtained relative resting g Cl was 6.0 Ϯ 1.4% (n ϭ 12; relative to total resting conductance in the absence of   blockers), corresponding to an absolute resting g Cl of 54 Ϯ 13 pS (n ϭ 12).
We next addressed the molecular basis for the resting g Cl . We first tested the hypothesis that open GABA A receptors accounted for part of g Cl . Relative resting g Cl obtained from experiments with PTX (100 M) added to the external solution was only 2.2 Ϯ 1.4% (n ϭ 12; Fig.  5B), or ϳ37% of that without PTX. Since 200 M PTX does not significantly affect the near-steady glycineevoked current in the studied cell type (Karlsson et al., 2011), this suggests that GABA A receptors account for a major fraction of resting g Cl .

ClC-2 channels do not contribute significantly to resting conductance in MPN neurons
Because Cl --permeable ClC-2 channels have been reported to contribute substantially to the resting conductance in CA1 pyramidal cells , we used a voltage protocol suitable for detection of Clcurrents through ClC-2 channels (Fig. 6A, bottom) to establish whether the MPN neurons express such channels. However, the current responses in eight cells tested were only composed of linear capacitive and small leak currents in the relevant voltage range (Fig. 6A, top, B), making it unlikely that ClC-2 contributes significantly to the resting conductance in MPN neurons.

Contribution of GABA A receptors to conductive [Cl -] i recovery
The above results show that there is a significant resting Clconductance in MPN neurons and suggest that a major fraction of this conductance is due to GABA A receptors. We therefore expected that GABA A receptors may contribute to [Cl -] i recovery after a high Clload. In dissociated MPN neurons, GABA A receptors may be activated as a consequence of spontaneous GABA release from adhering presynaptic nerve terminals (Haage et al., 1998). We speculated that possibly such spontaneous GABA release could contribute to g Cl and [Cl -] i recovery. To clarify this, we studied the time course of conductive [Cl -] i recovery (i.e., in the presence of 10 M VU0255011-1 to block KCC2) under voltage-clamp conditions at -74 mV. A few control recordings showed that in the absence of additional blockers, a roughly similar time course of [Cl -] i recovery is seen under current-clamp conditions at zero current (Fig. 7A). We first compared [Cl -] i recovery at -74 mV without and with TTX (2.0 M) to block action potential-evoked release of GABA, however, without noting any significant difference (Fig. 7A). We subsequently added 3.0 M gabazine (SR-95531 hydrobromide), previously shown to block spontaneous IPSCs in the studied cell type (Karlsson et al., 2011), but also this was without significant effect on conductive [Cl -] i recovery, as was also the glycine-receptor blocker strychnine (10 M; Fig. 7A). The lack of effect of gabazine was somewhat surprising in light of the above suggested contribution of GABA A receptors to resting g Cl . However, while PTX, which was used above for resting g Cl , is noncompetitive and usually considered as a pore blocker (Takeuchi and Takeuchi, 1969;Etter et al., 1999), gabazine is a competitive GABA A receptor antagonist (Heaulme et al., 1986;Hamann et al., 1988) and may fail to block receptor-channels that are spontaneously open in the absence of GABA (Birnir et al., 2000;Wlodarczyk et al., 2013). Indeed, consistent with the effect of PTX on resting g Cl , PTX significantly reduced conductive [Cl -] i recovery in the studied 10-min interval (Fig. 7B), suggesting that spontaneously open GABA A receptors in the absence of GABA do contribute to recovery.

Neurosteroid regulates conductive [Cl -] i recovery via miniature IPSCs (mIPSCs)
Although the above results suggested that spontaneous, action potential-independent GABA release does not significantly affect [Cl -] i recovery, we hypothesized that such release and the generated mIPSCs may still play a physiologic role for the Clhomeostasis. Hypothetically, mIPSC potentiation by an endogenous modulator may enable such a functional role for the spontaneously released transmitter. The endogenous neurosteroid 3␣hydroxy-5␣-pregnane-20-one (allopregnanolone), which may be produced de novo in the brain, e.g., during stress  A). Note the small amplitude and the linearity in the range below -65 mV. (Purdy et al., 1991), increases the frequency as well as prolongs the decay time of mIPSCs in MPN neurons (Haage and Johansson, 1999). We therefore analyzed the effect of allopregnanolone on [Cl -] i recovery at -74 mV in the presence of 10 M VU0255011-1 to block KCC2. It was clear that, in the presence of TTX (2.0 M) to block action potential-evoked GABA release, 20 nM allopregnanolone, a concentration that mainly affects the time course of mIPSCs (Haage et al., 2002) and may be reached in vivo (Purdy et al., 1991), significantly potentiated [Cl -] i recovery as quantified 5 min after loading with Cl - (Fig. 7C). This effect of allopregnanolone was abolished by gabazine (3.0 M; Fig. 7C), suggesting that the steroid effect was mediated by GABA A receptors that were activated by GABA, as expected for mIPSCs. Allopregnanolone at 20 nM concentration did not affect the baseline current in 11 cells tested (p ϭ 0.55; paired sample Wilcoxon signed-rank test), ruling out significant ef- fects on tonic GABA A receptor-mediated currents at this steroid concentration.
Higher, micromolar concentrations of allopregnanolone may directly activate GABA A receptors in MPN neurons (Haage and Johansson, 1999). Here, the effect of 1.0 M allopregnanolone on [Cl -] i recovery was dramatic and highly significant (Fig. 7D, compare top and lower curves). The effect was abolished in the presence of 100 M PTX, confirming an action on GABA A receptors (Fig. 7D).
It seems possible that some of the effect of 1.0 M allopregnanolone may be mediated by ␦-subunitcontaining GABA A receptors, which are extra-or perisynaptically located and highly sensitive to potentiation by neuroactive steroids (Stell et al., 2003). They may also display a degree of spontaneous activity (Jensen et al., 2013;Wlodarczyk et al., 2013) that should contribute to the background Clconductance. We took advantage of the positive allosteric modulator ␦-selective compound 2 (DS2; Jensen et al., 2013), which, although not perfectively selective, may preferentially potentiate ␦-subunit-containing GABA A receptors at concentrations Ͻ2 M (Ahring et al., 2016). Here, DS2 (1.0 M) significantly enhanced the [Cl -] i recovery as measured 5 min after loading with Cl - (Fig. 7E), suggesting that [Cl -] i recovery may be pharmacologically enhanced via positive modulation of ␦-subunit-containing GABA A receptors. This effect was additive to the effect of 20 nM allopregnanolone (Fig. 7E).

Theoretically expected influence of conductive Clflux on [Cl -] i recovery
The experimental results described above show that MPN neurons display a resting Clconductance that contributes significantly to recovery of [Cl -] i after a high Clload. To clarify the theoretically expected contribution of the resting g Cl to [Cl -] i recovery, we used the model described by Karlsson et al. (2011) to compute the time course of [Cl -] i . In essence, the influence of relative g Cl and of the KCC2-mediated transport capacity (g KCC2 ) was analyzed in a single-compartment cell with dimensions, total resting conductance and ion (external K ϩ and Cl -, internal K ϩ ) concentrations matching experimental values (see Materials and Methods). The time course of computed [Cl -] i recovery mediated by KCC2 in the absence of resting g Cl was matched to the experimental results (Fig.  2F, control; experiment with minimized influence of g Cl ) by adjusting g KCC2 (Fig. 8A, red, middle). Note that the rate of recovery as well as the asymptotic steady-state [Cl -] i depends on V m . C, Recovery mediated by KCC2 and Clleak in combination at different relative g Cl , as indicated and at V m ϭ -74 mV (Ca) or V m ϭ -50 mV (Cb). Note that the speed of recovery increases with g Cl and that the asymptotic steady-state [Cl -] i also increases with g Cl . The latter effect is weak at -74 mV (Ca) but prominent at -50 mV (Cb). The transport capacity of KCC2 in (A-C) was adjusted by setting the transporter proportionality factor (apparent conductance; Johansson et al., 2016) g KCC2 to 6.7 10 Ϫ21 mol 2 V Ϫ1 C Ϫ1 s Ϫ1 , to match modeled KCC2-mediated [Cl -] i recovery time course to experimental results.
Assuming a relative resting g Cl of 6.0% (of a total conductance of 0.89 nS in blocker-free conditions, to match the experimentally obtained conductance of 0.48 nS in Figure 5A, left, and the effect of the used blockers on g T as described in Materials and Methods), but no KCC2mediated transport resulted in a computed [Cl -] i recovery (at -74 mV) that reasonably well matched the experimental recovery (Fig. 4D, lower curve) observed with KCC2 blocked (Fig. 8A, blue, top). The computed [Cl -] i recovery with g Cl as well as KCC2-mediated transport (Fig. 8A, black, bottom) was quicker, showing that even with a relative resting g Cl as low as 6.0%, a substantial effect on [Cl -] i is indeed expected. It is also clear that [Cl -] i settles at a slightly higher steady level due to the combined effects of g Cl and KCC2 as compared with KCC2-mediated transport only (Fig. 8A, black, bottom). It was recently shown that, in a cell where [Cl -] i is determined by g Cl and g KCC2 , at steady-state, E Cl will settle between the membrane potential and the equilibrium potential for K ϩ , depending on the relative influence of g Cl and of g KCC2 . As shown here (Fig. 8B), the membrane potential (V m ) influences the rate of [Cl -] i recovery as well as the steady-state [Cl -] i . As g Cl drives E Cl toward V m , there is a trade-off between the rate of [Cl -] i recovery (maximal at high g Cl ) and the steady-state [Cl -] i (minimal at low g Cl ). The effect of g Cl on steady-state [Cl -] i is small at strongly negative V m (-74 mV; Fig. 8Ca) but prominent at less negative V m (-50 mV; Fig. 8Cb). It is clear from Figure  8C that, at more negative potentials in particular, there is a strong potential to modulate the rate of [Cl -] i recovery by substances that affect g Cl , such as neuroactive steroids or GABA A receptor active compounds like DS2 and PTX. At more positive potentials, in particular, there is also a potential to modulate resting [Cl -] i by such substances.

Discussion
In the present study, we have used voltage-ramp techniques to analyze how passive, conductive Clflux contributes to [Cl -] i recovery after a significant Clload introduced in central neurons. We have also compared with [Cl -] i recovery mediated by the transporter KCC2 and showed that conductive flux may account for a contribution in the same order of magnitude as KCC2. To estimate the relative resting Clconductance, for which little information is available concerning central mammalian neurons, we developed a novel technique based on [Cl -] i perturbation. We showed a relative resting Clconductance of 6%, with a major fraction mediated by open GABA A receptors. We also showed that conductive [Cl -] i recovery may be dramatically potentiated by the endogenous neurosteroid allopregnanolone, acting by enhancing the frequency and prolonging the time course of miniature postsynaptic currents. This suggests a novel role for neurosteroids in Clhomeostasis as well as a novel role for spontaneous transmitter release.

KCC2-mediated [Cl -] i recovery in isolated preoptic neurons
The time constant of [Cl -] i recovery in the studied MPN neurons was considerably longer than the few seconds reported in some earlier studies (Staley and Proctor, 1999;Wagner et al., 2001;Földy et al., 2010;Rinke et al., 2010;Deisz et al., 2011), including studies of hippocampal CA1 neurons, where whole-cell recording or sharp microelectrodes were used. The recordings we made from hippocampal neurons showed a still slower [Cl -] i recovery, suggesting that the difference from the earlier studies did not depend on the cell type, but likely on the recording techniques or type of preparational procedures. It seems likely that earlier studies with sharp microelectrodes or whole-cell recording may not have fully accounted for the introduced Clleak to/from the outside or equilibration with the patch pipette, the latter which may occur within a few seconds (Pusch and Neher, 1988). Further, the negative holding voltage used in some studies implies that conductive [Cl -] i recovery may have contributed significantly, whereas we reduced influence of conductive recovery by changing the holding voltage gradually while transporter-mediated recovery was estimated. Indeed, studies using the gramicidin-perforated patch technique report a slower recovery (Ͼ30 s up to several minutes; Chabwine et al., 2004;Lee et al., 2011;Pellegrino et al., 2011). Also measurements with optical Clsensors suggest a recovery time constant of ϳ30 s (Berglund et al., 2006) or several minutes (Friedel et al., 2015), although conductive recovery may have contributed also there. On the other hand, possibly the KCC2-mediated transport is dependent on the naturally surrounding milieu, which was largely eliminated by the dissociation process in the present work. Clearly, dendrites were removed in the present study. The [Cl -] i fluctuations in dendrites are expected to be much quicker than in the cell body (Doyon et al., 2011). Therefore, the presently reported time course of [Cl -] i recovery should be interpreted as reflecting the KCC2dependent transport capacity of isolated neuronal cell bodies.

Conductive [Cl -] i recovery
For any cell with a significant Clpermeability, a raised cellular [Cl -] i is expected to recover by conductive pathways to a level determined in part by the membrane potential. To our knowledge, such recovery has previously not been rigorously quantified. We here showed that, surprisingly, conductive [Cl -] i recovery at negative holding potentials was substantial, with a time course roughly similar to the time course of isolated KCC2-dependent recovery. This implies that the gain in speed of recovery contributed by KCC2 is not dramatic, if a sufficiently negative resting potential can be maintained. Nevertheless, KCC2 has been shown important for nervous function, with reduced KCC2 function being associated with several types of pathology, such as neuropathic pain (Coull et al., 2003), epilepsy (Huberfeld et al., 2007), autism (Tyzio et al., 2014Merner et al., 2015), schizophrenia (Merner et al., 2015), spasticity (Boulenguez et al., 2010), and ischemia (Jaenisch et al., 2010). There are several possibilities as to why KCC2 may be needed despite substantial conductive recovery. First, conductive recovery is dependent on the membrane potential, V m . Thus, when V m is raised, e.g., as a consequence of ongoing excitatory synaptic activity, conductive Clflux will drive [Cl -] i toward a higher level, where E Cl equals the momentary V m . In this respect, conductive Clflux may contribute to loading of Clat raised V m (as it does during the loading procedure used in the present work). Thus, KCC2 may be crucial in conditions of raised V m . Second, even at the resting potential in the absence of synaptic depolarization, conductive [Cl -] i recovery cannot lower E Cl beyond the membrane potential, and thus not sufficiently to produce hyperpolarizing GABA or glycine responses. The slight HCO 3 permeability of GABA A and glycine receptors implies that extra lowering of E Cl is needed for hyperpolarizing responses. Although GABA and glycine could still mediate shunting inhibition, it seems possible that hyperpolarizing responses may be of advantage in some conditions, due, e.g., to the different distribution in time and space of hyperpolarization and the underlying conductance change (Kaila et al., 2014). Third, the contribution by KCC2 to [Cl -] i recovery, even if small, still may be crucial to normal neural function. The finding that even mild KCC2 hypofunction may degrade neural coding (Doyon et al., 2016a) supports this possibility. Since conductive [Cl -] i recovery may be of a similar magnitude as KCC2dependent recovery, it seems likely that it will also be critical for neural coding and that reduced conductive recovery may be associated with pathology. Further, the potential for modulation of [Cl -] i recovery via altered g Cl , as illustrated in Figure 8Ca, suggests that the levels of neuroactive steroids or the presence of other GABA A receptor active substances, such as DS2, may be expected to affect the degree of disturbance in [Cl -] i -related pathologic conditions.
The expression of Cltransporters and [Cl -] i change during development, with fetal and early postnatal stages dominated by a low expression of KCC2, but a relatively high expression of NKCC1 and a relatively high [Cl -] i (see, e.g., Ben-Ari, 2002). If [Cl -] i is to be maintained lower than dictated by the thermodynamic equilibrium for NKCC1 (usually Ͼ60 mM), NKCC1 must be opposed by outward Clflux. In the absence of Clextruding transporters, the resting g Cl in combination with a negative resting potential may account for such outward Clflux. Thus, we may expect that conductive [Cl -] i recovery plays a relatively larger role early in development and in other situations when KCC2 expression is low, compared with when KCC2 expression is high.

Resting Clconductance
The rate of conductive [Cl -] i change will necessarily depend on the membrane Clconductance, g Cl . Neuronal membranes are permeable to Cleven at rest (Russell and Brown, 1972;Strickholm and Clark, 1977), but for mammalian neurons, information on the quantity of the resting Clpermeability or g Cl is scarce. Such information is needed not only to understand recovery of a perturbed [Cl -] i . The background g Cl is also critically influencing neuronal firing properties, such as action potential frequency adaptation and "after discharge", and the control of cell volume (Berndt et al., 2011). Forsythe and Redman (1988) calculated the approximate resting g Cl in spinal motorneurons to 18% of the total (Na ϩ , K ϩ , and Cl -) conductances, but their estimate depended on several assumptions, including unknown changes in K ϩ conductance, and they emphasize the obstacles to estimation introduced by the complex neuronal morphology. In the present work, the latter type of obstacles was avoided by using isolated neuronal cell bodies. Further, the novel method we developed for quantifying g Cl employed a simple expression (Eq. 5), which depended mainly on the testable and verified assumption of constant total conductance in the experimental conditions used. The obtained resting g Cl was relatively low (mean relative g Cl 6% and absolute g Cl 54 pS), with a major (ϳ2/3) contribution by GABA A receptors, as suggested by the reducing effect of PTX. This corresponds to the conductance of only two open GABA A receptor channels (with most frequent single-channel conductance of 17 pS; Bormann et al., 1987). Since the presence of ␦-subunit-containing GABA A receptors, which may contribute to tonic currents (Jensen et al., 2013;Wlodarczyk et al., 2013), was suggested above (see Results), it seems likely that such receptors contribute to the GABA A receptor-mediated fraction of g Cl and [Cl -] i recovery. The novel method used here to estimate resting g Cl could easily be applied also to many other preparations. It may be noted that manipulating Clconcentrations in the external solution may in many cases provide a useful alternative to loading cells with Cl -, as done here.
A significant contribution of glycine receptors to the resting g Cl seems unlikely for several reasons. First, the competitive glycine-receptor blocker strychnine did not affect [Cl -] i recovery (Fig. 7A). Second, in contrast to some GABA A receptors, glycine receptors do not show any significant degree of spontaneous openings in the absence of ligand (Twyman and Macdonald, 1991;Beato et al., 2002;Bode et al., 2013).
Although ClC-2 channels seem to play a large role for the background conductance in CA1 pyramidal cells , we did not find evidence for such channels in MPN neurons. Thus, it is unlikely that they contribute to resting g Cl in these neurons. In cells where ClC-2 channels are present, their contribution to g Cl may be expected to enhance the impact of conductive [Cl -] i recovery even beyond that described in the present work. This may be of particular importance if the ClC-2containing neurons show a limited transporter-dependent recovery capacity, such as indicated by our recordings from hippocampal neurons in the present study.

Role of minis and neurosteroids in Clhomeostasis
Spontaneous neurotransmitter release may possibly play a special role in the control of conductive [Cl -] i changes. The small postsynaptic currents generated do most often not induce major voltage fluctuations (Miles and Wong, 1986;Sayer et al., 1989;Rotaru et al., 2015). Accordingly, our results showed that, in control conditions, spontaneously released GABA does not significantly affect the recovery of [Cl -] i after a high load. However, postsynaptic currents generated by GABA release onto MPN neurons may be prolonged by the action of the endogenous neurosteroid allopregnanolone (Haage and Johansson, 1999). We here showed that in the presence of a concentration (20 nM) that has been detected in vivo when animals are exposed to stress (Purdy et al., 1991), allopregnanolone significantly enhances the speed of [Cl -] i recovery. Thus, the neurosteroid has the capacity to control [Cl -] i dynamics in a situation when raised [Cl -] i seems particularly likely, as a result of stress-induced enhancement of neural activity. Although a stressinduced reduction of inhibition may be important for, e.g., corticotrophin-releasing hormone neurons (Hewitt et al., 2009), it seems likely that other neurons may need a maintained functional inhibition during stress and that allopregnanolone may facilitate this via enhanced conductive [Cl -] i recovery. Similarly, neurosteroid-induced enhancement of [Cl -] i recovery may likely partly compensate for low levels of KCC2. However, as shown in our computations (Fig. 8C), in such situations there will be a trade-off between the rate of recovery and the minimum steady-state [Cl -] i .
The observed effect of 20 nM allopregnanolone on [Cl -] i recovery can be ascribed to the postsynaptic currents (mIPSCs) generated by spontaneously released GABA, since baseline current was not affected. At micromolar concentrations, allopregnanolone directly activates GABA A receptors as well as potentiates mIPSC frequency and prolongs mIPSC decay in MPN neurons (Haage and Johansson, 1999). In other cells, direct activation of GABA A receptors may occur at 100 nM allopregnanolone (Shu et al., 2004), a concentration that may be reached during pregnancy (Nguyen et al., 2003), suggesting that steroid modulation of [Cl -] i homeostasis via resting g Cl through direct GABA A receptor activation may be expected under physiologic conditions. Neurosteroid-controlled regulation of Clhomeostasis via the channels opened by spontaneously released GABA may possibly provide a novel functional role for spontaneous neurotransmission.