Fast synaptic inhibition in the forebrain is mediated by transient activation of synaptic GABA_A receptors by high concentrations of GABA released from the presynaptic terminals. GABA in the extracellular space mediates a slow inhibition of neurons by persistent activation of extrasynaptic receptors, which is commonly termed tonic inhibition. Tonic inhibition has been well characterized in cerebellar granule cells (Brickley et al., 1996; Hamann et al., 2002) and more recently in hippocampal dentate granule cells (DGCs) (Nusser and Mody, 2002; Stell et al., 2003). For GABA_A receptors to be activated by the low concentrations of GABA present in the extracellular space, these receptors must have a high affinity for GABA. Because persistent activation of many types of GABA_A receptors leads to their desensitization, specific subsets of receptors with slow rates of desensitization would be required for maintaining tonic inhibition. Recombinant GABA_A receptors containing the δ subunit in combination with the α₄ or α₆ subunit form receptors that have a high affinity for GABA as well as a slow rate and limited degree of desensitization (Brown et al., 2002; Wohlfarth et al., 2002). The δ subunits are expressed perisomatically in hippocampal DGCs (Wei et al., 2003) and extrasynaptically in cerebellar granule cells, suggesting that these receptors are capable of mediating tonic inhibition.

In keeping with this hypothesis, previous studies on cerebellar and hippocampal granule cells demonstrated that tonic inhibition has pharmacological properties similar to those of recombinant receptors containing the δ subunit. For example, zolpidem does not affect holding (mean) current, whereas it enhances synaptic currents in DGCs in vivo (Nusser and Mody, 2002) and in cultured hippocampal neurons (Yeung et al., 2003). In cerebellar granule cells, receptors mediating tonic inhibition are sensitive to the neurosteroid THDOC (Stell et al., 2003) and furosemide (Hamann et al., 2002; Wall, 2002) but insensitive to diazepam (Hamann et al., 2002). These pharmacological properties are shared with recombinant receptors containing the α₆ and δ or α₄ and δ subunits. Postembedding electron microscopic studies further demonstrated that δ subunit-containing GABA_A receptors are present on extrasynaptic membranes in cerebellar granule cells (Nusser et al., 1998). Based on these studies it has been proposed that δ subunit-containing receptors mediate tonic inhibition. However, earlier studies do not provide direct evidence of the presence of high-affinity, slowly desensitizing GABA_A receptors in neurons that are necessary to mediate tonic inhibition.

We directly demonstrate that DGCs express persistently open GABA_A receptors with high GABA affinity. Furthermore, these cells express GABA_A receptors with a slow rate and limited extent of desensitization, which remain open when synaptic receptors have desensitized. We also characterized pharmacological properties of tonic currents, which were similar to those of recombinant receptors containing α₄, β₁, and δ subunits.

Materials and Methods

Adult male Sprague-Dawley rats (150–200 g) were anesthetized with halothane before decapitation according to the University of Virginia Animal Care and Use Committee guidelines. Brains were dissected free and immersed in ice-cold (2–4°C) artificial cerebrospinal fluid (ACSF) saturated with 95% O₂, 5% CO₂. The ACSF consisted of 127 mM NaCl, 2 mM KCl, 1.5 mM CaCl₂, 1.5 mM MgSO₄, 25.7 mM NaHCO₃, 1.1 mM KH₂PO₄, and 10 mM glucose (osmolarity, 300–305 mOsM). After cooling, the brains were blocked and mounted on a vibratome stage (Camden Instruments Ltd., Leicester, UK), and 300-μm-thick horizontal sections containing the ventral hippocampus were cut. Slices were maintained in continuously oxygenated ASCF, at 32°C in a holding chamber for 30 to 45 min, and then at room temperature in a recording chamber mounted on the stage of an Olympus BX51 microscope equipped with a 40× water-immersion objective, infrared-differential interference contrast optics, and video. DGCs were identified in dentate granule layer as small- and medium-sized neurons with typical oval-shaped soma and single process. Patch electrodes (final resistances, 6–8 MΩ) were pulled from borosilicate glass (Sutter Instrument Company, Novato, CA) on a horizontal Flaming-Brown microelectrode puller (model P-97; Sutter Instrument Company), using a two-stage pull protocol. Electrode tips were filled with a filtered internal recording solution consisting of 153.3 mM CsCl, 1.0 mM MgCl₂, 10.0 mM HEPES, and 5.0 mM EGTA (pH adjusted to 7.2 with CsOH; osmolarity, 285–295 mOsM). The electrode shank contained 3 mM ATP Mg²⁺ salt, 0.1 mM GTP Na⁺ salt. Neurons were voltage-clamped to –65 mV with an Axopatch 1D or Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA). Whole cell capacitance and series resistance were compensated by 70 to 75% at 10-μs lag. Recording was performed when series resistance after compensation was 20 MΩ or less. Access resistance was monitored with a 10-ms, –5-mV test pulse once every 2 min, and if the series resistance increased by 25% at any time during the experiment, the recording was terminated. Currents were filtered at 5 kHz and then digitized using a Digidata 1322 digitizer and acquired using Axoscope 8.0 software (Molecular Devices) on an IBM PC-compatible computer hard drive.

For recording of GABA_A receptor-mediated spontaneous inhibitory postsynaptic currents (sIPSCs) and tonic GABAergic current, 50 μM dl-2-amino-5-phosphonopentanoic acid (dl-AP5) and 20 μM 6-cyano-7-nitroquinoxaline-2,3-dione (Tocris-Cookson Inc., Ellisville, MO) were included in the ACSF to block N-methyl-d-aspartate and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid/kainate receptor-mediated currents, respectively. All other reagents were obtained from Sigma-Aldrich (St. Louis, MO).

Acquisition and Analysis. The digitized current traces were analyzed with Mini Analysis software (Synaptosoft, Leonia, NJ). The software was used to detect and analyze sIPSCs. To detect sIPSCs, a detection threshold five times the root-mean-square (RMS) amplitude was used. RMS noise (I_rms) was defined as a square root of the average of the squares of the deviation from the I_avg (amplitude) over the chosen time interval using the equation where I_avg is an average amplitude, I_n is amplitude in an individual point, and n is the number of measurements in an epoch.

To study the tonic inhibition, transient events were manually removed from the current trace, so that it consisted only of holding current, the current required to voltage-clamp the cell. Two features of the holding current were studied: the mean current and the noise. Mean current was measured in 100-ms epochs with 1-s interval between epochs in 30 epochs. The measurements were taken 30 s before and 5 min after application of a drug. The segments of a current trace that contained synaptic event were omitted. Mean current (I_avg) was defined as an arithmetic mean of peak-to-peak amplitudes of individual points during that epoch. To assess the effect of a drug on I_avg in an individual neuron, the distribution of I_avg before application of the drug during the baseline period was compared with that after drug application by means of a Kolmogorov-Smirnov (KS) test. To compare the data obtained from a group of neurons, mean of I_avg from all neurons before and after drug application was compared.

A second measure of holding current, I_rms, was studied. The time interval (epoch) for each measurement was 50 ms and contained 250 amplitude measurements (5-kHz digitization rate). The time interval between two epochs was 500 ms. Sixty epochs were analyzed for each experimental condition (60 control and 60 after a drug application in each cell). To assess the effect of a drug on I_rms in an individual neuron, the distribution of I_rms in epochs before the application of a drug (during the baseline period) was compared with that after drug application by means of KS test. To compare data obtained from a group of neurons, I_rms values in individual epochs before and after drug application were averaged. Mean I_rms, mean I_avg, and mean sIPSC frequency and amplitude values were compared using t tests unless specified otherwise.

Results

Tonic Inhibition Revealed by Competitive Antagonists. Several recent studies have used competitive antagonists to measure tonic inhibition. To measure tonic inhibition, sIPSCs interspersed with the mean current were recorded from DGCs in the hippocampal slices by blocking excitatory neurotransmission with 50 μM dl-AP5 and 20 μM 6-cyano-7-nitroquinoxaline-2,3-dione. The neurons were voltage-clamped to –65 mV, under nearly symmetrical Cl^– conditions, at a room temperature (20–22°C; see Materials and Methods for details). After the input resistance became stable, sIPSCs and mean current were recorded for 5 min, and then 100 μM bicuculline was bath-applied. A typical recording with 100 μM bicuculline application is shown in Fig. 1A. In response to bicuculline, a slow outward current occurred, accompanied by loss of sIPSCs and reduction in baseline noise.

For analysis of tonic currents, I_avg and I_rms were measured. I_avg was analyzed in 30 successive 100-ms epochs (2000 points), and at every second interval they were analyzed after manually excluding rapid transient, synaptic currents from each epoch. A typical experimental change in I_avg after application of bicuculline was 67 pA. The experiment was replicated on four DGCs and the mean reduction of I_avg was 72 ± 19 pA. I_rms, the deviation from the mean of 250 individual digitized current measurements in each epoch lasting 50 ms (RMS noise; Fig. 1C), was measured in 60 epochs during the baseline period and 60 epochs during application of bicuculline after manually removing rapid transient synaptic currents. Each epoch was separated from the subsequent epoch by 500-ms intervals. The I_rms during application of bicuculline was less than that during the baseline period in each of four cells (KS test; p < 0.001). Pooled values from all four cells were compared using a paired t test, and mean I_rms during the baseline period was 7.082 ± 0.56 pA, and during 100 μM bicuculline application it was 5.997 ± 0.20 pA (p = 0.001).

Application of another competitive GABA_A receptor antagonist, SR 95531 (100 μM), to 5 DGCs produced similar results on synaptic currents in each cell (Fig. 1B). Unlike bicuculline, SR 95531 did not have an effect on I_avg in three DGCs (82.8 ± 6.9 versus 80.1 ± 3.6 pA; p = 0.57; paired t test). The I_rms was significantly decreased (Fig. 1C). Data were pooled from all five cells, and during baseline period the mean I_rms was 6.671 ± 0.11 pA and during SR 95531 application it was 5.496 ± 0.07 pA (KS test; D = 0.24; p < 0.001).

High-Affinity GABA_A Receptors. Studies with bicuculline and SR 95531 confirmed that tonic inhibition was present in DGCs; however, these drugs blocked synaptic inhibition along with tonic inhibition. It has been suggested that the GABA_A receptors specifically mediating tonic inhibition possess the high affinity for GABA that makes them less sensitive to competitive GABA_A receptor antagonists than those receptors mediating synaptic inhibition. However, this high affinity should render these receptors mediating tonic inhibition more sensitive to agonists than synaptic receptors. We determined whether a low concentration of GABA (1 μM), similar to that found in the extracellular space, could selectively activate GABA_A receptors mediating tonic inhibition without altering synaptic currents.

After making a baseline recording for 5 min, a low concentration of GABA (1 μM) was bath-applied, and 5 min was allowed to elapse to allow the drug to equilibrate within the slice. The holding current and sIPSCs in the subsequent 5 min were compared with those during the 5-min baseline period before GABA application. A typical experiment is shown in Fig. 1. Note that there was no apparent change in I_avg on visual examination of the trace (Fig. 2A). A detailed quantitative examination of the I_avg confirmed the visual impression. The distribution of I_avg in 30 epochs before GABA application was compared with that in 30 epochs after application KS test (Fig. 2B), and the two distributions were not different (71.3 ± 11.2 versus 69.3 ± 6.7 pA; D = 0.04; p = 0.94). The experiment was repeated in four DGCs, and 1 μM GABA did not change I_avg. in any of the cells.

Visual inspection of the trace also suggested that the baseline trace was thicker after application of 1 μM GABA, and when expanded further it seemed to possess greater noise (Fig. 2C). The I_rms during baseline period was 5.30 pA (Fig. 2D, top trace), and 5 min after application of 1 μM GABA it was 6.0 pA. The distribution of I_rms measurements from each of 60 epochs before and after GABA application was compared using the KS test, indicating that they were significantly different (D = 0.68; p = 0.001; Fig. 2E). This finding was confirmed in four DGCs, and in each one the KS test revealed that I_rms was significantly increased. To further understand the impact of 1 μM GABA on I_rms, an I_rms amplitude frequency distribution histogram was constructed (Fig. 2F), which demonstrated that 1 μM GABA increased the frequency of large I_rms epochs. In four DGCs tested, the mean I_rms increased from 6.29 ± 0.02 to 6.66 ± 0.03 pA (p < 0.0001) in response to 1 μM GABA.

Using a different approach, we studied the slope of the current-voltage relationship (slope conductance) to confirm that 1 μM GABA indeed increased GABA_A receptor conductance. Recordings were obtained in DGCs at baseline and after application of 1 μM GABA voltage-clamped to –65, –35, and –15 mV. I_rms in 60 epochs was obtained for each holding potential, and these values were plotted against the membrane holding potential (Fig. 2G). During the baseline period, the I_rms decreased proportionally to the driving force (r² = 1) and ordinate intercept was 2.9 pA. The regression line crossed the abscissa at close to 0 mV, the chloride reversal potential, suggesting that passage of chloride ions through the membrane contributed to I_rms. The slope of the RMS noise (current)-voltage relationship in the control condition was 18 pS. GABA (1 μM) increased I_rms at each holding potential, and the ordinate intercept was also close to 0 (2.9 mV). The slope of the RMS noise (current)-voltage relationship in the presence of 1 μM GABA in this neuron was 26.2 pS; therefore, 1 μM GABA increased the slope conductance. These findings were confirmed in four DGCs, and in each cell 1 μM GABA increased the slope conductance of I_rms.

We tested whether increased I_rms by 1 μM GABA was accompanied by inhibition of sIPSCs, because low concentrations of GABA are known to desensitize synaptic GABA_A receptors and diminish synaptic currents in DGCs (Overstreet and Westbrook, 2001). The mean amplitude of sIPCSs remained unchanged after application of 1 μM GABA, 31.2 ± 5.3 pA in baseline and 33.7 ± 8.6 pA (approx. 500 events; paired t test; p = 0.82; n = 4). The mean decay time constant was also unchanged (9.9 ± 0.9 versus 10.12 ± 1.1 ms; p = 0.88). The frequency of sIPSCs was 1.38 ± 0.7 Hz at baseline and 1.49 ± 0.7 Hz after application of 1 μM GABA (p = 0.92). Therefore, 1 μM GABA selectively enhanced baseline noise but did not have an effect on synaptic currents.

GABA Concentration-ΔI_rms Relationship. Theoretical studies on receptor-gated channel noise and subsequent recordings predict that the relationship between agonist concentration and I_rms is parabolic, with minimum noise occurring at low and high concentrations of the agonist and maximum noise occurring at intermediate concentrations (Traynelis and Jaramillo, 1998). As noted above, 1 μM GABA increased mean I_rms by 0.37 pA. In three DGCs, 10 μM GABA increased mean I_rms by 1.69 pA at the maximum change of I_avg. However, when the concentration of GABA was increased further to 30 μM, the increase in mean I_rms was smaller, 0.41 pA. These results indicated that the GABA concentration versus I_rms noise change relationship was parabolic, thus further confirming that GABA_A receptor conductance contributes to I_rms.

Penicillin Modulates Tonic and Synaptic Inhibition. Bicuculline and SR 95531 blocked synaptic inhibition along with tonic inhibition. When the GABA_A receptor is going through open and desensitized states, the agonist remains bound to it (agonist trapping) and competitive antagonists cannot bind to the receptor or close it (Bianchi and Macdonald, 2001). Therefore, bicuculline cannot selectively inhibit tonic currents, which are mediated by bound open receptors. We reasoned that a drug that preferentially binds to GABA_A receptors in the open state is likely to discriminate between receptors mediating tonic and synaptic inhibition. Penicillin G is a noncompetitive antagonist of the GABA_A receptor (Macdonald and Barker, 1977) that inhibits GABA_A receptors by blocking the chloride channel in all open states (Twyman and Macdonald, 1992). Penicillin at a concentration as low as 250 μM reduces total average current in a single channel by 38% (Twyman et al., 1992).

The effect 300 μM penicillin on tonic and synaptic currents was studied in seven DGCs. A typical experiment is shown on Fig. 3, where penicillin did not change the I_avg but it did diminish I_rms (Fig. 3, A and B). The decrease was significant for each cell, as confirmed by KS test with p < 0.001. The mean I_rms decreased from 4.364 ± 0.30 to 3.609 ± 0.21 pA (p = 0.002; n = 7). At the same time, penicillin effects on synaptic currents were subtle. It did not change peak amplitude and frequency of sIPSCs but significantly decreased decay time constant (Fig. 3C). The mean frequency of sIPSCs during the baseline period in seven neurons was 1.190 ± 0.19 Hz, and in the presence of 300 μM penicillin it was 1.25 ± 0.25 Hz (p = 0.84). The mean amplitude of sIPSCs was unchanged, 45.1 ± 13 pA during the baseline period and 43.7 ± 14 pA (p = 0.95) in the presence of penicillin. However, penicillin is an open channel blocker, and it is believed to inhibit channels after activation. We tested whether the decay of sIPSCs was accelerated by penicillin. The average decay time constant of sIPSCs decreased from 9.310 ± 0.9 to 5.240 ± 0.9 ms (p = 0.002; n = 4). The decrease of decay resulted in decreased charge transfer. It decreased from 512.1 ± 56.9 to 256.0 ± 33.4 pA/ms (p = 0.006; n = 4).

Desensitizing and Nondesensitizing GABA_A Receptor Currents Elicited by 10 and 100 μM GABA. Several studies have shown that δ subunit-containing receptors desensitize less extensively and more slowly than γ2 subunit-containing receptors (Saxena and Macdonald, 1994; Haas and Macdonald, 1999; Bianchi et al., 2002). We tested whether higher concentrations of GABA (10 and 100 μM) could differentially desensitize lower affinity synaptic GABA_A receptors while persistently activating high-affinity, slowly desensitizing receptors that are believed to mediate tonic inhibition.

After 5 min of baseline recording, 10 μM GABA was applied for 10 min. This resulted in an inward shift in I_avg, which reached a maximum (trough) and then decayed to persistent current that did not return to baseline I_avg for the duration of application of GABA (Fig. 4, A and B). In five DGCs tested, the peak I_avg evoked by 10 μM GABA was 134.7 ± 0.81 pA. The I_avg decayed to a persistent current (90.25 ± 0.57 pA), which was measured 5 min after application of 10 μM GABA was started, and the residual nondesensitizing component I_avg was 44.5 ± 1 pA (n = 5). Therefore, the extent of desensitization of I_avg was 33%. The channel noise can decrease because of opening or closing of channels; therefore, I_rms was not used to study desensitization of receptors.

We determined the extent of desensitization of synaptic currents by 10 μM GABA. GABA (10 μM) was applied after recording of sIPSCs at baseline from a DGC, and it reduced the amplitude and number of sIPSCs. The reduction of sIPSC amplitude was probably caused by desensitization of synaptic receptors; however, reduction of number of sIPSCs could be because of diminished presynaptic release of GABA (decrease of frequency) or could be because of disappearance of smaller amplitude events into increased baseline noise (Stell and Mody, 2002). To determine whether reduction of number of sIPSC was merely increase of failures to detect events, the Cl^– driving force was diminished. sIPCSs were recorded from three DGCs voltage-clamped to –65 mV for 5 min and then to –20 mV, thus reducing the driving force by 60%. The frequency of sIPSCs decreased from 2.05 to 1.23 Hz (by 40%), and the mean amplitudes decreased from 4.6 ± 18.6 to 40.6 ± 12.9 pA. Therefore, decrease of number of sIPSCs was proportional to decrease of driving force, which suggests that decrease of frequency with decrease of driving force reflects increase of failures to detect, rather than changes in presynaptic release of GABA.

In each of eight DGCs studied, 10 μM GABA inhibited sIPSC amplitude and frequency (Fig. 4, C and D). The mean amplitude of sIPSC decreased from 64.3 ± 6.6 to 47.1 ± 6.1 pA (p < 0.05) (Fig. 4E), and the frequency decreased from 1.26 ± 0.3 to 0.36 ± 0.04 Hz (p = 0.01). The total charge transfer mediated by sIPSCs before the application of GABA was 120.24 pA/s, and in the presence of GABA it was 42.39 pA/s. Therefore, prolonged application of 10 μM GABA desensitized 65% of synaptic currents, substantially more than desensitization of mean current, which was 33%, suggesting synaptic currents were more susceptible to desensitization than mean current.

We confirmed these findings using a higher concentration of GABA. In five DGCs, higher concentration of GABA (100 μM) produced a 1642 ± 164 pA inward shift of I_avg at the maximum (trough), which decayed to a persistent 676.5 ± 148.5 pA current. The effect of 100 μM GABA on sIPSCs was more dramatic, because they disappeared in all five cells. The mean frequency of sIPSCs before GABA application was 0.92 ± 0.12 Hz, and no transient event could be detected after the mean current reached the peak.

Activation of presynaptic GABA_B receptors by GABA could have reduced applied to DGCs in the presence of CGP 55845 (10 μM) to block GABA_B receptors. There was no change in frequency of sIPSCs when recordings without CGP 55845 were compared with recordings in the presence of CGP 55845 (0.92 ± 0.17 Hz, n = 8 and 1.22 ± 0.05 Hz, n = 3; p = 0.36). The sIPSCs were completely inhibited by 100 μM GABA in the presence of 10 μM CGP 55845. Therefore, desensitization of postsynaptic GABA_A receptors alone without activation of GABA_B receptors could explain the loss of synaptic currents.

Saturating Concentration of GABA to Measure Total Residual Conductance. Studies with 10 and 100 μM GABA suggested that GABA could be used to differentially desensitize two kinds of receptors that mediate tonic and phasic inhibition in DGCs. GABA activated both kinds of receptors, but a fraction of receptors desensitized rapidly, whereas others were persistently active in the presence of GABA. We determined the residual GABA_A receptor conductance in the presence of a saturating concentration of GABA. GABA (1 mM) produced a large, whole cell inward I_avg with maximum amplitude of 1628 ± 194 pA (n = 8; see an example in Fig. 5A). Continuous application of 1 mM GABA (∼10 min) produced desensitization of GABA_A receptors. However, residual current remained regardless of duration of GABA application. This residual current had mean amplitude of 534 ± 126 pA. This residual current was probably mediated by a combination of slowly and rapidly desensitizing receptors.

Prolonged application of a saturating concentration of GABA could substantially affect the distribution of Cl^– ions, raising the possibility that a large persistent current in the presence of 1 mM GABA could result in the redistribution of Cl^– ions as opposed to increased conductance resulting from open GABA_A receptor channels. To address this question, the DGCs were voltage-clamped to progressively more depolarizing potentials such that no synaptic activity could be detected, typically these potentials were +2 to +4 mV and conductance pulses (2 mV; 10 ms) at 1-Hz frequency were applied. After measuring membrane resistance by applying conductance pulses during 5 min of the baseline recording, 1 mM GABA was applied for 5 min. In response to 1 mM GABA, the I_avg changed very little (Fig. 5A), but the amplitude of the conductance pulse first increased to a peak and then decayed to a persistent value (Fig. 5, B and C). In all six DGCs, GABA caused a small inward current (42 ± 6 pA) and reduced the membrane resistance from 211 ± 9.3 MΩ to a minimum of 57.5 ± 0.7 MΩ. The resistance then increased to a persistent level, 95.6 ± 0.8 MΩ. Therefore, the persistent change in resistance was 115.4 MΩ or an 8.66-nS change in conductance, similar to that when the DGCs were voltage-clamped to –65 mV (8.21 nS).

Pharmacological Properties GABA_A Receptors Mediating Tonic Currents. A metabolite of progesterone, 3a-OH-5a-pregnan-20-one or allopregnanolone, is a potent allosteric modulator of GABA_A receptors. Allopregnanolone, in concentrations of the physiological range (10–30 nM), enhances whole cell GABA_A receptor currents elicited from DGCs (Mtchedlishvili et al., 2001). Nanomolar concentrations of neurosteroid THDOC enhance GABA_A receptor currents elicited from recombinant receptors containing the δ subunit (Wohlfarth et al., 2002); thus, we tested whether low concentrations of allopregnanolone could modulate tonic inhibition in DGCs.

A physiological concentration of allopregnanolone (10 nM) was applied to a DGC after a 5-min recording of sIPSCs, which did not alter I_avg but increased mean I_rms (Fig. 6, A and B). The effect of allopregnanolone was studied in five DGCs, and it increased I_rms in each of the five DGCs tested (KS test). When the data from individual cells were pooled, 10 nM allopregnanolone increased I_rms from 9.165 ± 0.5 to 9.612 ± 0.7 pA (p < 0.001; n = 5). When data from five cells were combined, allopregnanolone (10 nm) increased mean I_avg from 46.8 ± 3.1 to 48.3 ± 0.6 pA (p = 0.4). A higher concentration of allopregnanolone, 30 nM, also increased mean I_rms from 7.887 ± 0.47 to 8.121 ± 0.4335 pA (t test; p = 0.003; n = 5) without any effect on mean I_avg (69.1 ± 8 and 71.6 ± 3.2 pA; t test; p = 0.38).

Previous studies suggested that neurosteroids modulate mean current, so we tested whether higher concentration of allopregnanolone (300 nM) would modulate mean current. The mean current was enhanced by 300 nM allopregnanolone as shown in Fig. 6D. In three DGCs, 300 nM allopregnanolone increased mean I_avg by 98.2 ± 17 pA from 62.6 ± 4.9 to 160.8 pA (p < 0.05). It also increased I_rms from 7.258 ± 0.06 to 8.151 ± 0.1 pA (p = 0.01). These results suggest that the low nanomolar concentrations of allopregnanolone, which is likely to be found in the hippocampus in vivo, enhanced GABA_A receptor-mediated tonic inhibition by increasing I_rms. The effect of 300 nM allopregnanolone on mean current could have been in part because of a direct effect of allopregnanolone on the chloride channel (Twyman and Macdonald, 1992).

Furosemide and Loreclezole. Furosemide is a noncompetitive antagonist, which inhibits GABA_A receptor currents elicited from α4 subunit-containing recombinant receptors (Wafford et al., 1996) and does not discriminate between γ and δ subunit-containing receptors (Korpi and Luddens, 1997). In six DGCs, furosemide (300 μM) decreased I_rms from 7.611 ± 0.54 pA during baseline to 6.498 ± 0.35 pA (p < 0.05; Fig. 7A), but it did not change I_avg, which was 143.7 ± 23 pA during baseline and 145.6 ± 35 pA (p = 0.38) during drug application. The presence of tonic GABA_A receptor-mediated current was confirmed by bath application of bicuculline (100 μM) in the presence of 300 μM furosemide. Bicuculline abolished all synaptic currents and decreased mean I_avg by 47 ± 8.31 pA (n = 3).

Loreclezole Decreased Tonic Current. Anticonvulsant loreclezole enhances β₂ or β₃, but it inhibits β₁ subunit-containing GABA_A receptors (Fisher et al., 2000). Loreclezole (30 μM; Fig. 8) decreased I_rms from 7.573 ± 0.75 to 7.355 ± 0.63 pA (p < 0.05) in four DGCs. Loreclezole at 30 μM had the tendency to reduce mean I_avg from 174.0 ± 1.1 pA during baseline to 168 ± 0.57 pA during drug application (p > 0.05; n = 4).

Diazepam and Zolpidem Did Not Modulate Tonic Current. Diazepam (100 nM) did not alter I_avg and I_rms (Fig. 9, A–C). To confirm the presence of tonic currents, bicuculline was bath-applied to the cell, which caused a reduction in mean current. This result was confirmed, and in four cells diazepam did not change mean I_avg, which was 46.8 ± 3.1 pA during baseline and 48.3 ± 0.6 pA (p = 0.4) during drug application. The mean I_rms was slightly but not significantly diminished from 3.62 ± 0.05 pA during baseline to 3.39 ± 0.12 pA (p = 0.15; Fig. 9). Thus, tonic inhibition was not modulated by diazepam.

The imidazopiridine zolpidem acts on the benzodiazepine binding site and exerts maximal potentiation of GABA_A receptor currents in the presence of α₁ subunits (Pritchett et al., 1989) and least enhancement in the presence of α₅ subunits. Zolpidem (100 nM) was bath-applied to DGCs. Mean current and RMS noise was measured 30 s before and 5 min after the flow of diazepam was started. There was no change in mean current (56.1 ± 2.1 and 59.3 ± 0.6 pA; t test; p = 0.48; n = 4). The RMS noise also remained unchanged (3.62 ± 0.05 versus 3.39 ± 0.12 pA; t test; p = 0.15; n = 4; Fig. 10). To confirm that tonic currents were present in cells insensitive to zolpidem, bicuculline was applied, and it caused reduction of mean current (Fig. 10A).

Discussion

This study for the first time demonstrated high-affinity, slowly desensitizing GABA_A receptors in hippocampal neurons. Furthermore, this study suggested that persistently open GABA_A receptors contribute to membrane current fluctuations (I_rms) recorded from DGCs in baseline conditions. GABA_A receptors contributing to tonic I_rms in DGCs were sensitive to allopregnanolone, furosemide, and loreclezole and insensitive to diazepam and zolpidem. These pharmacological properties were similar to those of recombinant receptors containing α₄, β₁, and the δ subunits.

Tonic GABA_A Receptor Conductance Contributes to Membrane Noise. Membrane current fluctuations have long been used to study the properties of ion channels and receptors, but there are few studies demonstrating GABA_A receptor contribution to membrane noise. Several experiments in the present study suggested that persistently open GABA_A receptor channels contribute to the membrane noise (I_rms) in DGCs in hippocampal slices. The current fluctuations recorded from DGCs diminished in proportion to the chloride driving force, suggesting that passage of chloride ions through the membrane contributed to I_rms. Penicillin, which blocks open GABA_A receptors, reduced I_rms. GABA enhanced I_rms, suggesting that activation GABA_A receptors could contribute to I_rms. Pharmacological studies suggested that only a subset of GABA_A receptors contribute to baseline membrane noise. I_rms was enhanced by neurosteroid allopregnanolone, whereas benzodiazepine site agonists diazepam and zolpidem, which enhance synaptic GABA_A receptor currents in DGCs, did not enhance membrane noise.

We found that the membrane noise was more sensitive to GABA, penicillin, and allosteric modulators of GABA_A receptors than mean current, I_avg. Previous studies using competitive antagonists such as bicuculline found that these drugs reduce I_avg as well as membrane noise, I_rms (Stell and Mody, 2002). However, competitive antagonists block synaptic currents before inhibiting tonic currents. Penicillin altered I_rms without any change in I_avg. Penicillin had a minimal effect on synaptic currents; it did not decrease peak amplitude or frequency of IPSCs, but it hastened IPSCs decay, which resulted in a 2-fold decrease in total charge transfer. In contrast, bicuculline completely eliminated synaptic currents, and this could have contributed to changes in I_avg. Likewise, furosemide, another noncompetitive GABA_A receptor antagonist, reduced I_rms but not I_avg.

High-Affinity GABA_A Receptors. Experiments with low concentrations of GABA (1 μM and 300 nM) revealed a high-affinity GABA_A receptor on DGCs. Low concentrations of GABA increased membrane noise, but this could have resulted from closure of channels because of desensitization. Increased I_rms is not simply because of GABA_A receptor channel openings because the relationship between the noise variance (RMS noise) and mean current amplitude is parabolic, described by the binomial theorem [σ² = i² Np(1–p), where is σ² is I_rms, N is number of channels, i is single channel currents, and p is probability of single channel opening. That is, the current fluctuations at lowest and highest concentrations of a ligand are minimal because most channels are in a closed or open state, but the fluctuations are maximal at intermediate concentrations of a ligand (for review, see Traynelis and Jaramillo, 1998). Several lines of evidence suggested that the augmentation of I_rms by 1 μM GABA was because of an opening of receptor channels and not because of closure of open channels. Comparison of membrane slope conductance in the presence and absence of GABA demonstrated increased conductance, suggesting that GABA opened more channels than it had shut. Furthermore, 1 μM GABA did not alter the frequency or amplitude of sIPSCs, suggesting that this concentration did not desensitize synaptic GABA_A receptors. Finally, increasing the concentration of GABA to 10 μM resulted in a greater increased membrane noise than that observed with 1 μM GABA.

A number of studies report the extracellular GABA concentration to be in the range from tens of nanomoles to a few micromoles. Extracellular GABA concentration was reported to be 30 nM in neocortex, measured by high-performance liquid chromatography (Zhang et al., 2005). Based on microdialysis studies, it has been suggested that the extracellular concentration of GABA in the hippocampus varies in the 2.1 to 3.8 μM range (Shin et al., 2002). It seems that extracellular GABA concentration in vivo varies in the 0.3 to 3 μM range (Timmerman and Westerink, 1997). Studies in the past used blockade of GABA uptake to modulate tonic current in DGCs (Nusser and Mody, 2002), but the exact change in extracellular concentration of GABA in these studies is unknown. The current study demonstrated that a low concentration of exogenously applied GABA could activate chloride conductance.

This study also suggested that DGCs express GABA_A receptors with a high affinity to GABA. A likely explanation for high GABA affinity of receptors mediating tonic inhibition is that they contain the α₄ subunit. Recombinant GABA_A receptors containing the α₄ have a higher affinity for GABA than those containing the α₁ subunit (Wafford et al., 1996). The mRNA coding for the α₄ and polypeptide has been localized to DGCs in the past. Inhibition of tonic GABA_A receptor-mediated currents by furosemide suggests that α₄ subunit-containing GABA_A receptors participate in mediating tonic currents. Furosemide is known to inhibit α₄ subunit-containing receptors and was demonstrated to inhibit GABA_A receptor currents in dentate granule cells, which express the α₄ subunit (Wafford et al., 1996; Kapur and Macdonald, 1999). In addition to furosemide sensitivity, recombinant GABA_A receptors containing the α₄ subunit are insensitive to diazepam and zolpidem, which was also the case in our experiments with tonic currents recorded from DGCs.

Slowly Desensitizing Receptors on DGCs. A large residual GABA_A receptor conductance was found in the equilibrium state after prolonged application of 100 μM and 1 mM GABA. This residual current was probably mediated by both synaptic and extrasynaptic receptors. Increasing the extracellular concentration of GABA inhibits synaptic GABA_A receptors (Overstreet and Westbrook, 2001) by slow desensitization (Overstreet et al., 2000). Furthermore, the disappearance of synaptic currents was temporally correlated with reduction of the whole cell conductance, suggesting these receptors were being desensitized. Finally, previous studies have suggested that α₁ and γ₂ subunit-containing receptors are expressed in GABAergic synapses on DGCs. Recombinant receptors containing these subunits desensitize rapidly and to a substantial extent (90%) (Bianchi et al., 2002). In contrast, the δ subunit-containing receptors desensitize very slowly (Saxena and Macdonald, 1996; Haas and Macdonald, 1999) and to a lesser extent (35%) (Bianchi and Macdonald, 2002). However, studies with recombinant receptors have shown desensitization of γ2 subunit-containing receptors is never complete. Because γ subunit-containing receptors produce larger currents compared with δ subunit-containing receptors, even after 90% decrement of the current, the residual current corresponding to slow and ultra-slow desensitization rates of γ subunit-containing receptors can be of similar amplitude as residual current mediated by δ subunit-containing receptors, which loose only 35% of current in response to 28-s application of 1 mM GABA (Bianchi and Macdonald, 2002). Therefore, it is likely that in our preparation, residual current after prolonged application of 1 mM GABA is generated by mixed population of γ and δ subunit-containing receptors. However, the current study does not permit to estimate the extent of participation of each subunit type-containing receptor because the ratio of expression of each subunit-containing receptor is unknown. Separate study will be needed to answer this question.

Pharmacological Properties of GABA_A Receptors Mediating Tonic Currents. Pharmacological properties of persistently open GABA_A receptors in DGCs were similar to those of recombinant receptors containing the δ subunits. Recombinant receptors containing the δ subunit are highly sensitive to neurosteroid enhancement. Enhancement of RMS noise by 10 nM allopregnanolone suggested that persistently open GABA_A receptors on DGCs are sensitive to a physiological concentration of allopregnanolone. This high sensitivity to allopregnanolone is likely to be caused by the presence of δ subunit-containing GABA_A receptors (Wohlfarth et al., 2002). Lack of diazepam and zolpidem sensitivity of receptors also suggested absence of γ₂ subunit and presence of δ subunit in these receptors.

We did not observe changes in mean current when low concentrations (10 and 30 nM) of allopregnanolone were applied; however, the RMS noise was increased. The mean current was increased by a 300 nM concentration of allopregnanolone (see Results). In contrast, low concentrations of the neurosteroid THDOC increased mean current in DGCs (Stell et al., 2003), which may suggest differential sensitivity of GABA_A receptors to THDOC and allopregnanolone. We have previously reported that 10 nM allopregnanolone strongly enhanced whole cell currents in DGCs (Mtchedlishvili et al., 2001), but it was not clear whether the enhancement was because of effect of allopregnanolone on synaptic currents, tonic currents, or both. The current study demonstrated that the enhancement of whole cell currents at least in part was because of enhancement of tonic currents.

The loreclezole inhibition of tonic currents suggested the expression of β₁ subunit in receptors mediating tonic currents. Loreclezole is an anticonvulsant that enhances peak GABA_A receptor currents elicited by subsaturating concentrations of GABA by acting at an allosteric regulatory site on β₂ and β₃ subunits (Wafford et al., 1994; Fisher and Macdonald, 1997). Peak currents elicited from receptors containing the β₁ subunit were not enhanced by loreclezole because the receptors lack the positive modulatory site present on the β₂ and β₃ subunits. In addition to potentiation of peak currents, loreclezole inhibited steady-state GABA_A receptor currents by acting at a site distinct from the positive modulatory site on β₂ and β₃ subunits and increased the apparent desensitization. This inhibitory action of loreclezole occurred regardless of the β subunit. Thus, GABA_A receptors containing the β₁ subunit may be inhibited by loreclezole or stay unaffected by it. Spontaneously opening GABA_A receptors are present in hippocampal neurons (Macdonald et al., 1989; Birnir et al., 2000). It is possible that spontaneously opening GABA_A receptors participate in the tonic current described in this article.

Conclusions. Ligand-gated ion channels can mediate at least three kinds of tonic currents: currents mediated by spontaneously open channels in the absence of the ligand, slowly desensitizing currents mediated by ligand-bound channels, and residual equilibrium currents recorded after desensitization of channels. These three forms of “tonic” currents may be mediated by overlapping pools of receptors as demonstrated in the current study. The current study demonstrated that on DGCs, some GABA_A receptors are persistently open. These channels have high affinity for GABA; therefore, they may be ligand-bound open channels. Properties of persistently open GABA_A receptors on DGCs were similar to recombinant receptors containing α₄, β₁, and δ subunits. A persistent current could be recorded after desensitization of GABA_A receptors on DGCs.