The molecular mechanisms of ethanol tolerance and dependence appear to involve changes in GABA_AR function (Buck and Harris, 1990; Morrow et al., 1990; Kang et al., 1996). GABA_AR function, including responsiveness to many clinically important drugs, is highly dependent on the subunit composition of native receptors (Olsen et al., 1990; Seeburg et al., 1990; Whiting et al., 2000). Region-specific alterations in GABA_AR subunit composition have been observed in several animal models of chronic ethanol consumption (Mhatre et al., 1988, 1993; Buck et al., 1991; Devaud et al., 1995; Mahmoudi et al., 1997; Matthews et al., 1998).

The chronic intermittent ethanol (CIE) treatment of rats, which includes multiple (≥60) withdrawal episodes, has recently been validated as a model of human alcohol withdrawal and dependence (Cagetti et al., 2003). CIE treatment leads to a kindling-like state with a persistent decrease in pentylenetetrazol seizure threshold (Kokka et al., 1993), persistently reduced hippocampal GABA_A receptor (GABA_AR)-mediated synaptic function measured 2 to 40 days after the ethanol treatment is ended (Kang et al., 1996, 1998), and decreased sedative/hypnotic responses to positive allosteric modulators of GABA_ARs such as flurazepam, alphaxalone, and, to a lesser extent, pentobarbital (Cagetti et al., 2003). Immunoblotting revealed a decrease in α1 and δ subunit expression and an increase in γ2 and α4 subunits in hippocampus of CIE rats, confirmed by an increase in diazepam-insensitive binding for Ro15-4513, a marker for α4γ2 receptors. Analogous changes in α1 and α4 subunit peptide levels were previously reported in the hippocampus of rats after 40 days of continuous ethanol administration (Matthews et al., 1998). Recordings in hippocampal slices from CIE rats revealed that the decay time of GABA_AR-mediated miniature inhibitory postsynaptic currents (mIPSCs) in CA1 pyramidal cells was decreased, and potentiation of mIPSCs by positive modulators of GABA_ARs was also reduced, compared with controls. However, mIPSC potentiation by the α4-preferring benzodiazepine ligands bretazenil and Ro15-4513 was maintained, and increased, respectively (Cagetti et al., 2003).

Spillover of GABA released in the synaptic cleft and the presence of ambient GABA (Lerma et al., 1986; Tossman et al., 1986; Attwell et al., 1993) activate extrasynaptic GABA_ARs. Persistent activation of extrasynaptic GABA_ARs results in a tonic inhibitory influence on neurons. This small, but significant, GABAergic current has been observed in various brain regions (Otis et al., 1991; Brickley et al., 1996; Salin and Prince, 1996; Bai et al., 2001; Nusser and Mody, 2002). Immunocytochemical studies in these brain regions have provided evidence that the relative densities and subunit composition of extrasynaptic GABA_ARs are quite different from those of synaptic GABA_ARs (Soltesz et al., 1990; Nusser et al., 1995, 1998; Devor et al., 2001; Scotti and Reuter, 2001; Brunig et al., 2002). Functional studies determined that extrasynaptic GABA_ARs activate at lower GABA concentrations and desensitize more slowly than do the synaptic GABA_ARs (Brickley et al., 1999; Robello et al., 1999; Banks and Pearce, 2000; Hutcheon et al., 2000; Nusser and Mody, 2002). The differences in subunit composition also confer considerable pharmacological differences between the synaptic and extrasynaptic GABA_ARs (Robello et al., 1999; Hutcheon et al., 2000; Bai et al., 2001; Nusser and Mody, 2002).

In the present study, we sought to determine the effect of CIE treatment on different pools of receptors, the extrasynaptic (tonic) and synaptic GABA_AR-mediated inhibition in CA1 neurons, and to obtain additional evidence supporting the interpretation that there is a subunit composition switch in this model. Our electrophysiological data demonstrate a loss of the potentiation of tonic GABA current by positive allosteric modulators (diazepam and zolpidem). Furthermore, reduced tonic current potentiation by Ro15-4513 and the directly acting agonist 4,5,6,7-tetrahydroisoxazolo[5,4-c]pyridin-3-ol (THIP; also known as gaboxadol) suggests reductions in extrasynaptic α1 and α4 subunit-containing GABA_ARs. Interestingly, THIP-induced decreases in mIPSC charge transfer in saline-treated rats are “switched” to potentiation in CIE-treated rats, consistent with the evident increases in α4 subunits in synaptic GABA_ARs. In addition, we demonstrate that the soporific effects of THIP are only modestly reduced and its anxiolytic effects are maintained in CIE rats. The novel enhancement by THIP of synaptic GABA receptors tempered by the reduced sensitivity of extrasynaptic receptors after CIE treatment suggests potential usefulness of THIP in ameliorating the symptoms of human alcohol withdrawal. Some of these results have been reported previously in abstract form (Liang et al., 2003a,b).

Materials and Methods

Production of CIE Rats. The Institutional Animal Care and Use Committee approved all animal experiments. Male Sprague-Dawley rats (170-190 g) were housed in the vivarium under a 12-h light/dark cycle and had free access to food and water. Intoxicating doses of ethanol (Pharmco Products, Brookfield, CT) were administered by oral intubation on a chronic regimen: for the first five doses, rats received 5 g/kg of body weight as a 25% (w/v) solution in saline once every other day, and for the following 55 doses, 6 g/kg ethanol 30% (w/v) once every day. The control group received saline (20 ml/kg of body weight). This ethanol regimen led rats to experience multiple cycles of intoxication and withdrawal phases. The blood ethanol concentrations peaked (300-400 mg/dl) at 1 h after gavage and declined rapidly such that they were undetectable at 8 h after treatment (Kang et al., 1998). Although saline-treated rats initially tended to gain weight faster than ethanol-treated rats, the weights of the two groups after 60 doses were not significantly different. After the treatment and 2 days of withdrawal, rats were subjected to behavioral experiments. Alternatively, rats were withdrawn for 2 to 5 days and euthanized, and tissues were prepared for experiments. Since it was reported that certain effects of chronic alcohol administration revert to control levels within 24 h of alcohol withdrawal, whereas others persist for more than a week (Buck and Harris, 1990; Buck et al., 1991; Devaud et al., 1996), we compared the properties of GABA_AR activation at 2 and at 5 days after the last CIE dose. There were no significant differences in the kinetics of mIPSCs or the tonic GABA_AR-mediated currents at 2 days (n = 10 cells, 4 rats) or at 5 days (n = 5 cells, 2 rats) after alcohol withdrawal. Furthermore, the effects of THIP (0.3 and 3 μM) application on synaptic and extrasynaptic currents did not differ between the 2-day and 5-day withdrawal groups. Therefore, results combine the electrophysiological data from CIE rats at 2 to 5 days of withdrawal.

Electrophysiology. Transverse slices (400 μm thick) of dorsal hippocampus were obtained using standard techniques (Spigelman et al., 1992; Kang et al., 1996). Whole-cell patch clamp recordings were obtained from cells located in the CA1 pyramidal layer at 34 ± 0.5°C during perfusion with artificial cerebrospinal fluid (ACSF) composed of 125 mM NaCl, 2.5 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂, 26 mM NaHCO₃, and 10 mM d-glucose. The ACSF was continuously bubbled with a 95/5% mixture of O₂/CO₂ to ensure adequate oxygenation of slices and a pH of 7.4. Patch pipettes contained 135 mM cesium gluconate, 2 mM MgCl₂, 1 mM CaCl₂, 11 mM ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid, 10 mM HEPES, 2 mM K₂ATP, 0.2 mM Na₂GTP; pH adjusted to 7.25 with CsOH. GABA_AR-mediated mIPSCs were pharmacologically isolated by adding tetrodotoxin (0.5 μM), d-(-)-2-amino-5-phosphonopentanoate (40 μM), 6-cyano-7-nitroquinoxaline-2,3-dione (10 μM), and CGP 54626 (1 μM) to the ACSF from stock solutions. Stock solutions of 6-cyano-7-nitroquinoxaline-2,3-dione, bretazenil, diazepam, zolpidem, and Ro15-4513 were made with pure dimethyl sulfoxide. The final concentration of dimethyl sulfoxide did not exceed 42 μM in the recording chamber. Signals were recorded in voltage-clamp mode with an amplifier (Axoclamp 2B, Axon Instruments Inc., Union City, CA). Whole-cell access resistances were in the range of 2.5 to 15 MΩ before electrical compensation by about 90%. During voltage-clamp recordings, access resistance was monitored by measuring the size of the capacitative transient in response to a 5-mV step command, and experiments were abandoned if changes >20% were encountered. At least 10 min was allowed for equilibration of the pipette solution with the intracellular milieu before commencing mIPSC recordings. Data were acquired with pClamp 8 software (Axon Instruments Inc.), digitized at 20 kHz (Digidata 1200B; Axon Instruments Inc.), and analyzed using the Clampfit software (Axon Instruments Inc.) and the Mini Analysis Program (versions 5.2.2 and 5.4.8; Synaptosoft, Decatur, GA).

Detection and Analysis of mIPSCs and Tonic Currents. The recordings were low-pass filtered off-line (Clampfit software) at 2 kHz. The mIPSCs were detected (Mini Analysis Program) with threshold criteria of 5 pA amplitude and 20 pA · ms area. Frequency of mIPSCs was determined from all automatically detected events in a given 100-s recording period. For kinetic analysis, only single-event mIPSCs were chosen during visual inspection of the recording trace. This included mIPSCs with a stable baseline, sharp rising phase, and exponential decay. Double- and multiple-peak mIPSCs were excluded. The mIPSC kinetics were obtained from analysis of the averaged chosen single events (>120 events/100-s recording period) aligned with half-rise time in each cell. Decay time constants were obtained by fitting a double exponential to the falling phase of the averaged mIPSC in each neuron. The tonic current magnitudes were obtained from the mean baseline current during the 100-s recording periods. The investigator performing the recordings and mIPSC analysis was blind to the treatment (saline or CIE) that the rats received. All comparisons of group differences in mIPSC kinetics and drug effects were made with ANOVA (Sigmastat; SPSS Inc., Chicago, IL).

Sleep Time Assay. The hypnotic effect of THIP (30 mg/kg i.p.; H. Lundbeck A/S, Copenhagen, Denmark) was tested on CIE- and saline-treated rats (n = 11). THIP was dissolved in 0.9% saline. Injection volume was 2 ml/kg. The latency time (the time required after drug injection to lose the righting reflex) and sleep time were measured. Sleep time was determined as follows. After drug injection and loss of righting reflex, rats were placed on their backs in a V-shaped trough and a timer was started. The sleep time period ended when the animal was able to flip over three times in 30 s. Latency and sleep time are reported as mean (minutes) ± S.E.M. Statistical significance was calculated by ANOVA.

Pentylenetetrazol (PTZ) Seizure Threshold and Anticonvulsant Effect of THIP. After 2 days of treatment withdrawal, a group of CIE- and saline-treated rats were tested for the anticonvulsant effect of THIP against PTZ-induced seizures. First, the seizure threshold to PTZ was determined in saline-treated (n = 6) and CIE rats (n = 6). Rats were gently restrained and PTZ (25 mg/ml in 0.9% saline) was injected slowly, at a constant rate, into the tail vein via a 24 gauge shielded i.v. catheter (BD Medical, Franklin Lakes, NJ). The endpoint was taken as the time to the first myoclonic twitch of the head. Seizure thresholds were calculated from the volume injected for body weight and are presented as milligrams per kilogram of PTZ. Once the seizure threshold for each rat was established, 2 days later the rats were tested for the anticonvulsant effect of THIP (they had no ethanol during the extra 2 days). Rats were injected with 8 mg/kg THIP (diluted in 0.9% saline, 2 ml/kg i.p.) 30 min before infusion of PTZ, and seizure threshold was measured again. Data were analyzed by ANOVA.

Anxiolytic Effect of THIP on the Elevated Plus-Maze. Rats were tested for the anxiolytic effect of THIP (1.25 and 2.5 mg/kg) on the elevated plus-maze. Rats were brought to the procedure room 2 h before testing. The plus-maze was constructed as described previously (Pellow et al., 1985). Rats were randomly divided into the following groups: saline controls treated with vehicle (saline) or THIP, and CIE rats treated with vehicle or THIP. Rats were injected 1 h before testing (volume 2 ml/kg i.p.). Rats were placed on the central area of the maze, tested for 5 min, and videotaped in a room absent of any humans. The following measures were scored: number of entries into open arms, closed arms, or center platform, and time spent in each of these areas. An arm entry was defined as the entry of all four feet into one arm. Data are reported as mean ± S.E.M. percentage of entries into the open arms and percentage of time spent in open arms. Statistical differences were determined using ANOVA.

Results

CA1 Neurons Exhibit Synaptic (Phasic) and Extrasynaptic (Tonic) GABA_AR-Mediated Inhibition. Initially, we sought to isolate the tonic component of inhibition in CA1 neurons from the synaptic component through the use of the GABA_AR antagonist gabazine (McCabe et al., 1988; Bai et al., 2001). We also wanted to ascertain that glycine receptor activation does not contribute to the Cl^- currents attributed to GABA_AR activation. Figure 1 illustrates that strychnine (1 μM) application had no effect on the kinetics of averaged GABA_AR-mediated mIPSCs or the tonic current (I_h). However, subsequent application of gabazine (1 μM) selectively abolished the mIPSCs without affecting the tonic current.

Loss of Diazepam and Decreased Ro15-4513 Effect on Tonic GABA_AR-Mediated Current in CIE-Treated Rats. Application of diazepam enhanced the isolated tonic current (without any appreciable desensitization) in CA1 neurons from saline-treated rats, whereas subsequent addition of the GABA_AR blocker (picrotoxin, 50 μM) blocked this response (Fig. 1; summarized in Fig. 3). By contrast, diazepam application in slices from CIE-treated rats had no effect on the tonic current. Picrotoxin was still able to reduce I_h, indicating that the GABA_AR-mediated tonic current was still present in CIE-treated CA1 neurons. These data suggested that the extrasynaptic GABA_ARs in CIE rats lose the diazepam-sensitive α1, -2, and -5 subunits and/or gain the diazepam-insensitive α4 subunits.

To test for this possibility, we next compared the effects of Ro15-4513 on the tonic current in saline- and CIE-treated rats. Ro15-4513 does not possess agonist activity at α1- and α2-containing GABA_ARs, but is a high-affinity agonist at α4-containing GABA_ARs (Wisden et al., 1991; Knoflach et al., 1996; Whiting et al., 2000). Application of Ro15-4513 in slices of saline-treated rats potentiated I_h from its baseline level, an effect that was blocked by subsequent picrotoxin application (Fig. 2; summarized in Fig. 3). This suggested that α4-containing GABA_ARs contribute to the tonic current in rat CA1 neurons. In slices from CIE-treated rats, Ro15-4513 still potentiated I_h, but this potentiation was smaller than that in slices from saline-treated rats. Quantification of I_h changes after application of various GABA_AR modulators confirmed these findings (Fig. 3).

Loss of Zolpidem Potentiation of Synaptic and Extrasynaptic GABA_ARs after CIE Treatment. The benzodiazepine, zolpidem, is an allosteric positive modulator of GABA_AR function with selectivity for the α1-containing GABA_ARs, with intermediate potency on α2/3 receptors, and no effect on α5-containing GABA_ARs (Whiting et al., 2000). Application of zolpidem potentiated both the mIPSCs and the tonic current in a concentration-dependent manner in CA1 neurons from saline-treated rats (Figs. 4 and 5). By contrast, zolpidem (0.1 μM) application in slices from CIE-treated rats had no effect either on mIPSCs or the tonic current, whereas 0.3 μM zolpidem had a very small potentiating effect on the synaptic and extrasynaptic GABA_AR-mediated currents. Examination of the mIPSC kinetics revealed that zolpidem increased the mIPSC area (total charge transfer) by prolonging the decay time constants without affecting mIPSC amplitude (Fig. 6). The mIPSC rise time was also increased by zolpidem, but only in saline-treated rats (Fig. 6). These data pointed to the loss of α1-containing extrasynaptic GABA_ARs, further extending our previous findings of an apparent decrease in synaptic α1-containing GABA_ARs (Cagetti et al., 2003), and also suggested that the zolpidem-insensitive α5-containing GABA_ARs do not solely contribute to the tonic current in CA1 neurons, since some zolpidem-enhanced receptors are present.

Partial Agonist, THIP, Potently Activates Tonic GABA_AR Currents in CA1 Neurons from Untreated Rats. To extend our observations on GABA_AR subunit changes, we decided to use a directly acting agonist at GABA_ARs, THIP. THIP has been demonstrated to act as a high-efficacy and reasonably high-affinity agonist at recombinant α4β3δ-containing GABA_ARs (Adkins et al., 2001; Brown et al., 2002) and is, therefore, a partial agonist at certain other GABA_ARs. Unlike GABA, THIP is not subject to rapid uptake and, thus, its application in brain slices would be somewhat analogous to increasing ambient [GABA] in the slice. Since the concentration-dependent actions of THIP on synaptic and extrasynaptic GABA_AR-mediated currents in CA1 neurons have not been previously characterized in detail, we first studied the effects of THIP application in slices from control (untreated) rats at 34.5°C. Bath application of THIP produced a concentration-dependent increase in I_h, without appreciable response desensitization at THIP concentrations up to 10 μM (Figs. 7, A and B, and Fig. 8). Some desensitization of the response was observed at THIP concentrations of 30 and 100 μM. The EC₅₀ for THIP was at approximately 1 μM (Fig. 7B).

THIP Inhibits mIPSCs in CA1 Neurons from Untreated Rats. Synaptic GABA_ARs in CA1 neurons were recently demonstrated to be biophysically and pharmacologically distinct from extrasynaptic receptors (Banks and Pearce, 2000; Bai et al., 2001; Yeung et al., 2003). Thus, synaptic GABA_ARs possess higher unitary conductance and lower affinity for GABA than do their extrasynaptic counterparts (Yeung et al., 2003). They also desensitize more rapidly than do extrasynaptic GABA_ARs and are selectively blocked by gabazine and penicillin (Banks and Pearce, 2000; Yeung et al., 2003). In the continued presence of a partial agonist such as THIP, these receptors would be expected to exhibit decreased responses to the full agonist (GABA) released in the synaptic cleft (Kurata et al., 1999; Bianchi and Macdonald, 2003). For example, one way in which a partial agonist can reduce GABA currents is to accelerate the desensitization of GABA-induced currents (Nakahiro et al., 1991; Kurata et al., 1999). Indeed, examination of averaged mIPSC current kinetics revealed a concentration-dependent decrease in mIPSC area by THIP (Fig. 7C). This decrease was primarily due to the speeding up of the mIPSC decay, since the rise time and amplitude of mIPSCs were little affected by THIP at concentrations ≤10 μM. In addition, there was a substantial concentration-dependent decrease in mIPSC frequency by THIP, which suggested THIP-mediated decreases in pre-synaptic GABA release. This decrease in frequency was not likely to be due to failure to detect mIPSCs of reduced amplitude since 1) THIP concentrations ≤10 μM had little affect on the amplitude of averaged mIPSCs, and 2) the current noise increases after THIP application were small. The root mean square of current noise increased from 4.5 ± 0.14 pA in the absence of THIP to a maximum of 6.3 ± 0.33 pA at 1 μM THIP, other concentrations producing smaller root mean square increases in current noise. At THIP concentrations of 30 and 100 μM, the average mIPSC amplitude also decreased (not shown), but the frequency of single analyzable events was too low to permit robust analysis of mIPSC kinetics.

Decreased Tonic Current Activation by THIP in CIE-Treated Rats. The concentration-response for THIP was very similar between slices from untreated and saline-treated rats (compare Fig. 7B and Fig. 9). However, application of THIP in slices from CIE rats was less effective in evoking the tonic GABA_AR-mediated current than in slices from saline-treated rats (Figs. 8 and 9). These data suggested an overall decrease in extrasynaptic GABA_ARs after CIE treatment. Also, given the high efficacy and reasonably high affinity of THIP at α4β3δ-containing GABA_ARs (Adkins et al., 2001; Brown et al., 2002), such GABA_ARs may be preferentially decreased at extrasynaptic locations after CIE treatment.

THIP Potentiates GABA_AR-Mediated Synaptic Inhibition in CIE-Treated Rats. The concentration-dependent effects of THIP on mIPSC kinetics in saline-treated rats were almost indistinguishable from the effects in untreated (control) rats (compare Figs. 7C and 10). However, comparison of THIP effects on mIPSC kinetics between saline- and CIE-treated rats revealed a dramatic shift from a depressant action on mIPSCs in slices from saline-treated rats to a potentiating effect in CIE rats (Fig. 10). The predominant effect was a prolongation of both decay time constants, resulting in increased area of mIPSCs. Interestingly, the greatest potentiation was observed at the lowest concentration of THIP tested. Nevertheless, even at the highest concentration (10 μM), there was a significant increase in the mIPSC area. This behavior is consistent with a partial agonist model for THIP in the presence of the agonist GABA, with differential effects in CIE-versus saline-treated rats due to differential efficacy and kinetics of GABA_ARs with an altered subunit composition. There was also a concentration-dependent increase in mIPSC rise time in CIE- but not in saline-treated rats. The amplitude and frequency of mIPSCs were affected similarly by THIP in both saline- and CIE-treated rats.

Reduced Hypnotic and Maintained Anxiolytic Effects of THIP in CIE-Treated Rats. The dramatic shift from a depressant to a potentiating effect of THIP on mIPSCs coupled with a moderate reduction in its effect on the tonic current in CIE-treated rats was in marked contrast to the almost complete loss of potentiation by benzodiazepines such as diazepam and zolpidem of both synaptic and extrasynaptic GABA_ARs. Previously, we showed that the hypnotic effects of the benzodiazepine flurazepam were greatly attenuated after CIE treatment (Cagetti et al., 2003). Based on the current in vitro studies, we suspected that the in vivo actions of THIP may be affected differently from those of benzodiazepines after CIE treatment. To this end, we compared THIP for its anticonvulsant, soporific, and anxiolytic effects in saline- and CIE-treated rats. CIE rats showed partial tolerance to the hypnotic effect of THIP (Table 1). The mean sleep time of 92 ± 8 min for saline-treated rats was significantly reduced to 59 ± 6 min for CIE rats at 30 mg/kg (^*, p = 0.003). The latency time of 20 ± 2 min was significantly longer for CIE rats compared with 13 ± 1 min for controls (^*, p = 0.003).

The anticonvulsant effect of THIP (8 mg/kg) against PTZ seizure threshold was measured in CIE and control rats. As demonstrated previously (Kokka et al., 1993), the PTZ seizure threshold was significantly reduced in CIE rats compared with saline controls (Fig. 11): the seizure threshold was 28 ± 2 mg/kg PTZ in CIE (n = 6) and 48 ± 6 mg/kg PTZ in saline groups, respectively (p = 0.01). At the concentration tested, THIP failed to exhibit anticonvulsant activity in either group.

CIE- and saline-treated rats were tested on the elevated plus-maze with two doses of THIP (1.25 and 2.5 mg/kg). At both doses, THIP had no effect on the total number of entries in either saline or CIE rats. Total entries for saline rats treated with vehicle, 1.25 and 2.5 mg/kg THIP, were (mean ± S.E.M.): 13 ± 1, 10.5 ± 0.9, and 10.2 ± 0.9. For CIE rats, they were: 8.1 ± 0.7, 8.7 ± 0.9, and 8 ± 1.0. Total entries were significantly lower in CIE-compared with saline-treated rats (two-way ANOVA; p < 0.05). The lower dose of THIP did not have an anxiolytic effect in either group (Fig. 12, A and B). However, when tested at the concentration of 2.5 mg/kg, THIP had an anxiolytic effect in both groups. THIP increased significantly the number of entries into the open arms by saline controls (p = 0.02) and CIE rats (p = 0.04) and time spent in them (p < 0.05) (Fig. 12, A and B).

Discussion

The present study provides an electrophysiological and pharmacological analysis of synaptic and extrasynaptic GABA_ARs in a rat model of alcohol intoxication and withdrawal. Our data specifically point to subunit-selective alterations in CA1 hippocampal neuron GABA_ARs after CIE treatment. In addition, our data suggest that in contrast to classical benzodiazepines, certain GABAergic drugs that are active at α4-containing GABA_ARs may be more efficacious as sedative-hypnotic and anxiolytic agents in chronic alcoholism, as a result of up-regulation of such GABA_AR isoforms in this condition.

The differences in subunit composition confer considerable pharmacological differences between the synaptic and extrasynaptic GABA_ARs (Robello et al., 1999; Hutcheon et al., 2000; Bai et al., 2001; Nusser and Mody, 2002). Although precise physiological roles have not been identified for the tonic inhibitory currents mediated by extrasynaptic GABA_ARs, they are clearly subject to modulation by clinically important drugs. Our data confirm the presence of a tonic GABA_AR-mediated current in rat CA1 neurons (Bai et al., 2001; Yeung et al., 2003) and demonstrate that the potentiation of this current by several allosteric positive modulators is either reduced or eliminated after CIE treatment. Specifically, diazepam (0.3 μM) potentiation of the isolated tonic current is eliminated (Figs. 1 and 3), similar to the elimination of mIPSC potentiation we reported previously in CIE rats (Cagetti et al., 2003). Here we showed that zolpidem (0.1 and 0.3 μM) effects are similarly reduced, consistent with a reduction in α1-containing and/or increases in α4-containing GABA_ARs at both synaptic and extrasynaptic locations. However, whereas the potentiation of mIPSCs by Ro15-4513 is increased (Cagetti et al., 2003), the potentiation of the isolated tonic current by Ro15-4513 is reduced in CIE rats (Figs. 2 and 3). Since Ro15-4513 does not possess agonist activity at α1-containing GABA_ARs, but is a high-affinity agonist at α4-containing GABA_ARs (Wisden et al., 1991; Knoflach et al., 1996; Whiting et al., 2000), these data collectively suggest that a reduction in α1 subunits is compensated by the introduction of α4 subunits into synaptic GABA_ARs after CIE treatment. However, decreases in the α1 subunits at extrasynaptic GABA_ARs do not appear to be fully compensated by increases in α4 subunits because of the reduced effectiveness of Ro15-4513 (α4 subunit-selective) in potentiating the tonic current in CIE rats. Further evidence for decreases in extrasynaptic GABA_ARs was obtained by demonstrating that tonic current potentiation by THIP, a partial agonist with reported high efficacy at certain α4-containing GABA_ARs, was also reduced in CIE rats (Figs. 8 and 9). The large potentiation of synaptic currents by THIP in CIE rats is further evidence for the insertion of α4 subunits into synaptic GABA_ARs. Previously, increased expression of α4β3δ subunits in a rat model of progesterone withdrawal was shown to double the affinity and increase the efficacy of CA1 neuron GABA_ARs for THIP (Sundstrom-Poromaa et al., 2002). However, this subunit combination is unlikely to be inserted into CA1 synapses in CIE rats because 1) δ subunit has very low expression levels in the hippocampal CA1 region (Persohn et al., 1992), and 2) δ subunit protein levels are decreased by ∼50% in the hippocampus of CIE rats (Cagetti et al., 2003). Also, it should be noted that THIP was demonstrated to activate channels in rat CA1 neurons that are modulated by diazepam (Lindquist et al., 2003), and therefore, its actions are not limited to α4-containing GABA_ARs. Similarly, it should be noted that the allosteric actions of Ro15-4513 may not be entirely selective for α4-containing GABA_ARs, and therefore, the decreased potentiation of the tonic current after CIE treatment by Ro15-4513 could also be attributed to decreases in other extrasynaptic GABA_AR subunit combinations.

In saline-treated rats, THIP exhibits both anxiolytic and sedative-hypnotic properties (Fig. 12; Table 1). In hippocampal slices from these rats, THIP application activates the extrasynaptic GABA_AR-mediated tonic current but occludes mIPSCs generated by GABA release in the synaptic cleft (Figs. 7 and 8). This suggests that potentiation of extrasynaptic GABA_ARs alone may be sufficient to reduce anxiety or, at higher concentrations of this agonist, to induce sleep. We estimated the half-maximal concentration for THIP activation of the tonic current in control rats to be approximately 1 μM (Fig. 7B), which is substantially lower than the 80 μM value reported previously (Sundstrom-Poromaa et al., 2002). Interestingly, the plasma concentration of THIP that improves sleep quality in humans was also estimated to be about 1 μM (Schultz et al., 1981), and brain concentrations are unlikely to be much higher (Madsen et al., 1983; Faulhaber et al., 1997). THIP was demonstrated to promote deep, slow-wave sleep without affecting rapid eye movement sleep in both rodents and humans (Faulhaber et al., 1997; Lancel and Langebartels, 2000). Recent studies in normal human volunteers demonstrated a lack of tolerance to the soporific effects of THIP during long-term (3-week) daily treatment, this in contrast to zolpidem, to which patients exhibited marked tolerance (Vogel et al., 2003). Therefore, THIP holds promise as a sedative-hypnotic that lacks the major side effects of classical benzodiazepine drugs.

In CIE rats that exhibit large decreases in the hypnotic response to the benzodiazepine flurazepam and the steroid anesthetic alphaxalone (Cagetti et al., 2003), the anxiolytic effects of THIP are maintained (Fig. 12), and its sedative effects are only modestly reduced (Table 1), consistent with the decreased THIP actions at extrasynaptic receptors and a “switch” to potentiation of synaptic currents. By analogy with the rat CIE model, we expect that THIP would remain an active soporific agent in treating the insomnia symptoms of human alcohol withdrawal syndrome.

However, it should be underscored that these conclusions are based on studies of only one cell type in the central nervous system and, therefore, must be treated as tentative. For example, as is the case with THIP, the anxiolytic actions of diazepam (Cagetti et al., 2003) and alphaxalone (Cagetti et al., 2004) are maintained in CIE rats. Recent point mutation studies in mice have suggested that the anxiolytic properties of benzodiazepines such as diazepam are mediated through potentiation of α2-containing GABA_ARs (reviewed in Mohler et al., 2002). In hippocampal pyramidal cells, α2-containing GABA_ARs are concentrated in synapses at the axon initial segment (Nusser et al., 1996; Fritschy et al., 1998), and α2 mRNA does not appear to be reduced in CIE rats (Cagetti et al., 2003), whereas the potentiating effects of diazepam on both synaptic and extrasynaptic currents in CA1 neurons from CIE rats are lost. Thus, anxiolytic properties of diazepam in CIE rats are likely due to effects at GABA_ARs located elsewhere in the central nervous system.

Chronic ethanol treatment leads to increased GABA_AR internalization and changes in α1/α4 ratios, concurrently with changes in the way the protein kinase C (PKC) γ isoform couples to these subunits (Kumar et al., 2002, 2003). Protein phosphorylation plays an important role in regulating GABA_AR, including both GABA_AR channel activity and trafficking (Barnes, 1996, 2001; Kittler et al., 2002; Kittler and Moss, 2003). PKC also has important effects on allosteric modulation of GABA_AR by ethanol and neurosteroids (Weiner et al., 1997; Brussaard and Koksma, 2003; Harney et al., 2003; Koksma et al., 2003). For example, PKC activation in young rat dentate granule cells makes them sensitive to a previously ineffective concentration of an endogenous neurosteroid (Harney et al., 2003). Also, mutant mice lacking different PKC isoforms exhibit altered sensitivity to ethanol in vivo and altered GABA_AR responsiveness in vitro, such that PKCγ null mice have reduced ethanol modulation (Harris et al., 1995; Proctor et al., 2003) and PKCϵ null mice have increased ethanol modulation (Hodge et al., 1999; Proctor et al., 2003). To date, the possibility that changes in the GABA_AR phosphorylation state and trafficking may account for the altered responsiveness of GABA_AR to various allosteric modulators after CIE treatment has not been examined, and therefore, these changes represent alternative explanations to the subunit change hypothesis.

The current study extends our knowledge of GABA_AR pharmacology in the hippocampus of CIE rats and is consistent with plastic changes in subunit composition of GABA_ARs. In particular, the changes occur in both synaptically and extrasynaptically localized receptors which give rise to phasic and tonic inhibition in CA1 neurons, respectively. Differential sensitivity to benzodiazepine modulators and the partial agonist THIP suggests different subunit compositions before and after CIE treatment, possibly including decreases in α4 subunit-containing receptors at extrasynaptic locations and increases at synaptic locations, where they are coupled to γ2 subunits. These changes in CA1 GABA_ARs might explain some but not all of the behavioral plasticity observed in the CIE model of alcohol withdrawal and dependence. These changes are complex, and pathway-, cell-, and subcellular location-dependent.