Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Benzodiazepines (BZs) that positively modulate GABA at GABA_A receptors are widely used for the treatment of anxiety and insomnia (for review, see Woods et al., 1992). Despite safety and clinical efficacy, dependence on BZs can develop as evidenced by withdrawal signs (e.g., anxiety and insomnia) that emerge upon discontinuation of long-term BZ treatment (Woods et al., 1992). BZ withdrawal has been studied in various nonhuman primate species, including rhesus monkeys, squirrel monkeys, and baboons. Suspension of BZ treatment or administration of an antagonist (e.g., neutral GABA_A modulator flumazenil) produce a withdrawal syndrome (e.g., tremor, vomiting, and convulsions) as well as disruptions in schedule-controlled behavior (Yanagita and Takahashi, 1973; Lukas and Griffiths, 1984; Spealman, 1986; Gerak and France, 1997). Severity of BZ withdrawal is dependent on BZ dose (Lukas and Griffiths, 1984) and does not appear to differ among ligands varying in selectivity at BZ receptor subtypes (Griffiths et al., 1992; Ator, 2000).

BZs positively modulate GABA-mediated chloride flux by binding to sites on the GABA_A receptor complex. This complex includes a chloride ionophore comprising five protein subunits (, , , , , , or ) with functional GABA_A receptors requiring coexpression of -, -, and -subunits (for review, see Mehta and Ticku, 1999). BZ receptors are localized at the interface of - and -subunits with different BZ receptor subtypes comprising different -subunits. BZ₁ receptors comprise ₁-subunits and are suggested to mediate sedative-hypnotic effects, while BZ₂ receptors comprise ₂-, ₃-, or ₅-subunits and are suggested to mediate anxiolytic effects (McKernan et al., 2000; for review, see Sieghart, 1995). GABA_A receptors expressing ₄- or ₆-subunits do not bind conventional BZs and are referred to as diazepam-insensitive (Luddens et al., 1990). BZ₁ receptor-selective ligands such as zolpidem and zaleplon purportedly elicit hypnosis with greater potency than other BZ effects (Depoortere et al., 1986; Beer et al., 1997). Thus, drugs that bind selectively to BZ receptor subtypes might be used clinically to elicit some but not other BZ effects.

Despite increasing knowledge of how different GABA_A receptor subunits mediate acute effects of BZ site positive modulators, relatively less is known about possible subtype specificity underlying BZ dependence and withdrawal. Drug discrimination has been recently applied to the study of BZ dependence and withdrawal by training diazepam (5.6 mg/kg/day p.o.)-treated rhesus monkeys to discriminate flumazenil (0.32 mg/kg s.c.) from vehicle (Gerak and France, 1999). Under these conditions, temporary suspension of diazepam treatment produced signs of withdrawal that were accompanied by responding on the flumazenil-appropriate lever (Gerak and France, 1999). The flumazenil discriminative stimulus in diazepam-treated monkeys might therefore be a valid measure for investigating GABA_A receptor pharmacology underlying BZ dependence and withdrawal.

The present study evaluated positive modulators differing in their site of action on the GABA_A receptor complex for their ability to attenuate a flumazenil discriminative stimulus in diazepam-treated monkeys, i.e., to prevent diazepam withdrawal. In another group of untreated monkeys, positive modulators were tested for their ability to substitute for the discriminative stimulus effects of the nonselective BZ midazolam. It was hypothesized that the potency of positive GABA_A modulators to substitute for midazolam would predict their potency to attenuate flumazenil-precipitated withdrawal from diazepam. The BZ₁-selective ligands zaleplon and zolpidem were compared with nonselective BZs to determine whether subtype-specific ligands differentially modify flumazenil in diazepam-treated monkeys. Drugs acting at neuroactive steroid and barbiturate sites on the GABA_A receptor complex were also studied to examine how positive modulation through non-BZ sites modifies the flumazenil discriminative stimulus.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Subjects. Six adult female (midazolam discrimination) and four adult male (flumazenil discrimination) rhesus monkeys (Macaca mulatta) were housed individually on a 14-h light/10-h dark schedule, maintained at 95% free-feeding weight (range 3.8-11.5 kg) with a diet comprising primate chow (High Protein Monkey Diet; Harlan Teklad, Madison, WI), fresh fruit, and peanuts, and provided water in the home cage. Monkeys were trained previously to discriminate midazolam (Lelas et al., 1999) or flumazenil (Gerak and France, 1999). The animals used in these studies were maintained in accordance with the Institutional Animal Care and Use Committee, The University of Texas Health Science Center at San Antonio, and with the 1996 Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources on Life Sciences, National Research Council, National Academy of Sciences).

Apparatus. During experimental sessions, monkeys were seated in chairs (model R001; Primate Products, Miami, FL) that provided restraint at the neck, and placed in ventilated, sound-attenuating chambers equipped with two response levers, stimulus lights, and a food cup to which pellets could be delivered from a dispenser. For monkeys discriminating midazolam under a schedule of stimulus-shock termination (SST), feet were placed in shoes containing brass electrodes through which brief electric shock (3 mA, 250 ms) could be delivered from an a.c. shock generator. An interface (Med Associates, St. Albans, VT) connected the chambers to a computer that controlled and recorded experimental events.

Midazolam Discrimination Procedure. Experimental sessions consisted of multiple 15-min cycles each comprising a 10-min time-out period, during which responses had no programmed consequence, followed by a 5-min response period, during which red stimulus lights were illuminated and the FR5 schedule of SST was in effect. Illumination of stimulus lights located above each lever signaled the beginning of the response period in which shock was to be delivered every 15 s. Five consecutive responses on the lever designated correct by the injection administered during the first minute of the cycle extinguished the stimulus lights and postponed the shock schedule for 30 s. The selection of saline- and midazolam-appropriate levers varied among monkeys and remained the same for an individual throughout the study. Responding on the incorrect lever reset the response requirement on the correct lever. Response periods ended after 5 min or after the delivery of four shocks, whichever occurred first.

Flumazenil Discrimination Procedure. Diazepam was given 3 h prior to experimental sessions. Multiple-cycle procedures were similar to those described above except that the 5-min response period comprised an FR5 schedule of food presentation during which a maximum of 10 food pellets was available. When the maximum number of food pellets was obtained in less than 5 min, the remainder of the response period was a time-out. Test sessions were conducted when animals satisfied the criteria specified above for monkeys discriminating midazolam. Cumulative flumazenil dose-effect tests were conducted by injecting the flumazenil vehicle solution during the first minute of the first cycle followed by increasing doses of flumazenil during the first minute of subsequent cycles with the cumulative dose increasing by 0.25 or 0.5 log unit per cycle. Test sessions ended when 80% of the total responses occurred on the flumazenil-appropriate lever or when response rate decreased to below 20% of control response rate. The following drugs were injected s.c. 45 min prior to the first cycle of cumulative flumazenil dose-effect tests: diazepam (3.2 or 10 mg/kg), amobarbital (3.2 or 10 mg/kg), and pentobarbital (3.2 or 10 mg/kg). The following drugs were injected at the beginning of the first cycle prior to cumulative doses of flumazenil: midazolam (1 or 3.2 mg/kg), triazolam (0.32 or 1 mg/kg), zolpidem (3.2-32 mg/kg), zaleplon (5.6 or 10 mg/kg), and pregnanolone (1-5.6 mg/kg). Control flumazenil dose-effect functions were determined every two or three tests throughout the course of these experiments.

Drugs. The vehicle for oral administration of diazepam was fruit punch combined with suspending agent K (Bio-Serv, Frenchtown, NJ) in a concentration of 1 g of suspending agent per liter of fruit punch. Tablets containing 10 mg of diazepam (Zenith Laboratories, Inc., Northvale, NJ) were dissolved in vehicle, mixed in a blender, and administered using a 12-gauge drinking needle attached to a 60-ml syringe. To obtain a dose of 5.6 mg/kg diazepam, a standard concentration of diazepam was given in a volume adjusted to individual body weights. The diazepam mixture was prepared immediately before administration.

Data Analyses. Drug discrimination data are expressed as the percentage of total responses occurring on the drug-appropriate lever averaged among monkeys (±S.E.M.) and plotted as a function of dose. Substitution for the training drug was defined as 80% responding on the drug-appropriate lever. When a test with a given compound was conducted more than once, the determinations were averaged for an individual subject for further analyses. Doses of a compound required to produce 50% drug-appropriate responding (ED₅₀) and the 95% confidence limits (95% CL) were estimated using linear regression by using more than two appropriate data points, otherwise by interpolation. These values were determined first for individual monkeys and then averaged among all monkeys. ED₅₀ values for flumazenil-appropriate responding following administration of a positive GABA_A modulator were compared with the average of control flumazenil ED₅₀ values determined just before and after each combination test. The magnitude of rightward shift elicited by a given positive modulator was determined first for individual monkeys and then averaged among all monkeys. ED₅₀ values were considered to be significantly different when corresponding 95% CLs did not overlap. To graphically compare potencies of the same compounds in the two procedures, the magnitude of shift in the flumazenil dose-effect function (±S.E.M.) was plotted as a function of positive modulator dose expressed as a multiple of its potency (ED₅₀) to substitute for midazolam.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Substitution of Positive GABA_A Modulators for Midazolam Discriminative Stimulus. Cumulative doses of triazolam, midazolam, zaleplon, zolpidem, diazepam, amobarbital, and pentobarbital increased midazolam-lever responding in a dose-related manner and substituted (80% midazolam-lever responding) for the midazolam discriminative stimulus (Fig. 1, top). Single doses of pregnanolone (3.2, 5.6, or 10 mg/kg) were administered in separate tests at the beginning of six to eight cycles with the larger two doses increasing midazolam-lever responding across cycles in a time-related manner. Maximum levels of midazolam-lever responding were observed within 30 to 45 min after injection of pregnanolone (data not shown). The pregnanolone dose-effect function (Fig. 1, top) was constructed from the percentage of midazolam-lever responding observed 45 min after injection. The order of potency for midazolam substitution was as follows: triazolam = midazolam > zaleplon > zolpidem = diazepam > pregnanolone > amobarbital = pentobarbital. ED₅₀ values and 95% CLs for midazolam substitution are presented in Table 1. The vehicle solutions for each positive GABA_A modulator occasioned predominantly saline-appropriate responding (data not shown).

View larger version (27K): [in this window] [in a new window]   Fig. 1.   Discriminative stimulus and rate effects of positive GABAA modulators in rhesus monkeys discriminating 0.56 mg/kg midazolam from saline. Abscissae: dose of test compound in milligrams per kilogram of body weight. Ordinates: mean (±S.E.M.) percentage of responding on the drug-appropriate lever (%DR = drug responding, top) and mean (±S.E.M.) response rate expressed as percentage of control rate [rate (% of control), bottom]. Data represent average values from at least three monkeys. , triazolam; , midazolam; , zaleplon; , zolpidem; , diazepam; , pregnanolone; , amobarbital; , pentobarbital.

Attenuation of Flumazenil Discriminative Stimulus in Diazepam-Treated Monkeys. In monkeys treated daily with diazepam, administration of cumulative doses of flumazenil dose dependently increased responding on the flumazenil-appropriate lever with cumulative doses of 0.1 or 0.32 mg/kg producing 80% flumazenil-lever responding (Figs. 2-9, top). Administration of the flumazenil vehicle solution during the first cycle of these tests occasioned predominantly vehicle-appropriate responding (Figs. 2-9, top). The range of control flumazenil ED₅₀ values was 0.043 to 0.067 mg/kg with the overall average ED₅₀ (95% CL) being 0.056 mg/kg (0.038-0.075).

View larger version (20K): [in this window] [in a new window]   Fig. 2.   Discriminative stimulus and rate effects of flumazenil in diazepam-treated monkeys that received diazepam. Diazepam (3.2 or 10 mg/kg) was administered 45 min prior to vehicle (V) and 60 min prior to the administration of the first dose of flumazenil. See Fig. 1 for other details.

View larger version (20K): [in this window] [in a new window]   Fig. 3.   Discriminative stimulus and rate effects of flumazenil in diazepam-treated monkeys that received midazolam. Midazolam (1 or 3.2 mg/kg) was administered at the start of the session 15 min prior to the administration of the first dose of flumazenil. See Fig. 1 for other details.

View larger version (21K): [in this window] [in a new window]   Fig. 4.   Discriminative stimulus and rate effects of flumazenil in diazepam-treated monkeys that received triazolam. Triazolam (0.32 or 1.0 mg/kg) was administered at the start of the session 15 min prior to the administration of the first dose of flumazenil. See Fig. 1 for other details.

View larger version (20K): [in this window] [in a new window]   Fig. 5.   Discriminative stimulus and rate effects of flumazenil in diazepam-treated monkeys that received zaleplon. Zaleplon (5.6 or 10.0 mg/kg) was administered at the start of the session 15 min prior to the administration of the first dose of flumazenil. See Fig. 1 for other details.

View larger version (21K): [in this window] [in a new window]   Fig. 6.   Discriminative stimulus and rate effects of flumazenil in diazepam-treated monkeys that received zolpidem. Zolpidem (3.2, 10.0, or 32.0 mg/kg) was administered at the start of the session 15 min prior to the administration of the first dose of flumazenil. See Fig. 1 for other details.

View larger version (22K): [in this window] [in a new window]   Fig. 7.   Discriminative stimulus and rate effects of flumazenil in diazepam-treated monkeys that received pregnanolone. Pregnanolone (1, 3.2, or 5.6 mg/kg) was administered at the start of the session 15 min prior to the administration of the first dose of flumazenil. Data for 5.6 mg/kg pregnanolone represent values from one monkey. See Fig. 1 for other details.

View larger version (20K): [in this window] [in a new window]   Fig. 8.   Discriminative stimulus and rate effects of flumazenil in diazepam-treated monkeys that received amobarbital. Amobarbital (3.2 or 10 mg/kg) was administered 45 min prior to vehicle (V) and 60 min prior to the administration of the first dose of flumazenil. See Fig. 1 for other details.

View larger version (21K): [in this window] [in a new window]   Fig. 9.   Discriminative stimulus and rate effects of flumazenil in diazepam-treated monkeys that received pentobarbital. Pentobarbital (3.2 or 10 mg/kg) was administered 45 min prior to vehicle (V) and 60 min prior to the administration of the first dose of flumazenil. See Fig. 1 for other details.

View larger version (27K): [in this window] [in a new window]   Fig. 10.   Rightward shift in the flumazenil dose-effect function elicited by positive GABAA modulators expressed as a multiple of their midazolam substitution ED50. Abscissa: multiple of the ED50 of the appropriate positive GABAA modulator in substituting for midazolam. Ordinate: mean (±S.E.M.) rightward shift in the flumazenil dose-effect function expressed as flumazenil ED50 following pretreatment with the appropriate positive GABAA modulator divided by the corresponding control flumazenil ED50. Vertical dashed line represents the ED50 for midazolam substitution.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Chronic treatment of anxiety, insomnia, or convulsions with positive GABA_A modulators can produce tolerance as well as physical dependence, as evidenced by withdrawal signs that emerge upon discontinuation of treatment (Woods et al., 1992). Despite increasing knowledge of how different GABA_A receptor subunits mediate the acute behavioral effects of BZs (Lelas et al., 2000b), little is known about subtype specificity of BZ dependence and withdrawal. In addition, it is not clear how drugs differing in their site of action on the GABA_A receptor complex modify BZ dependence and withdrawal. To begin studying these questions, various positive GABA_A modulators were tested for their ability to modify a flumazenil discriminative stimulus in diazepam-treated monkeys.

To compare effects of these drugs in untreated monkeys, the positive modulators were tested for their ability to substitute for the nonselective BZ midazolam. Positive modulators substituted for midazolam with the following order of potency: triazolam = midazolam > zaleplon > zolpidem = diazepam > pregnanolone > amobarbital = pentobarbital. These data are consistent with the general finding that positive GABA_A modulators differing in their site of action substitute for midazolam and other BZs (Woudenberg and Slangen, 1989; Ator and Griffiths, 1997, for exception; Lelas et al., 1999; for review, see Lelas et al., 2000b). Moreover, these data support the view that the discriminative stimulus effects of different positive GABA_A modulators are mediated by a common mechanism, i.e., increased chloride flux.

In nonhuman primates treated chronically with a high-efficacy nonselective or subtype-selective BZ site ligand, administration of flumazenil elicits a withdrawal syndrome characterized by tremor, vomiting, and convulsions (Yanagita and Takahashi, 1973; Lukas and Griffiths, 1984; Griffiths et al., 1992; Ator, 2000). In addition to observational studies, drug discrimination has been used to study flumazenil-precipitated withdrawal by training diazepam (5.6 mg/kg/day p.o.)-treated monkeys to discriminate flumazenil (Gerak and France, 1999). Under these conditions, temporary suspension of diazepam treatment elicits signs of withdrawal that are accompanied by responding on the flumazenil lever (Gerak and France, 1999). Moreover, the discriminative stimulus effects of flumazenil are pharmacologically specific insofar as unrelated drugs (e.g., ketamine) do not substitute for flumazenil. The flumazenil discrimination is dose-dependent, and a supplemental dose of diazepam (10 mg/kg s.c. after the daily dose of 5.6 mg/kg) shifts the flumazenil dose-effect curve to the right (Gerak and France, 1999; this study). Thus, the flumazenil discriminative stimulus in diazepam-treated monkeys appears to be related to BZ withdrawal and probably results from its ability to antagonize the positive GABA_A modulatory effects of chronic diazepam.

Acute administration of a variety of BZ site positive GABA_A modulators, including the nonselective compounds diazepam, midazolam, and triazolam and the BZ₁-selective ligands zaleplon and zolpidem, attenuated the discriminative stimulus effects of flumazenil. Previous studies have confirmed simple, competitive interactions in vivo for flumazenil in combination with nonselective BZs (e.g., diazepam, midazolam, or triazolam; Lelas et al., 2000a) as well as the BZ₁ receptor-selective ligand zolpidem (Rowlett et al., 1999). Similarly, the discriminative stimulus effects of flumazenil in diazepam-treated monkeys were attenuated by subtype-selective as well as -nonselective BZ site positive modulators. Unlike BZ site positive modulators, the behavioral effects of barbiturates and neuroactive steroids are not antagonized by flumazenil (Herling and Shannon, 1982; L. R. McMahon and C. P. France, unpublished observations), consistent with these drugs acting at non-BZ sites on the GABA_A receptor complex. Nevertheless, these positive modulators also attenuated the discriminative stimulus effects of flumazenil. Thus, it appears as though the various different positive GABA_A modulators in this study attenuated the flumazenil discriminative stimulus in diazepam-treated monkeys, irrespective of their site of action on the GABA_A receptor complex or their selectivity for BZ receptor subtypes.

Potency of positive modulators to attenuate the flumazenil stimulus was examined by expressing the dose of positive modulator as a multiple of its potency in substituting for midazolam. Relative potency of positive modulators to attenuate flumazenil in diazepam-treated monkeys was not predicted by potency to substitute for midazolam in untreated monkeys. For instance, the flumazenil discriminative stimulus was attenuated by doses of BZ site positive modulators that were larger than the smallest doses that substituted for midazolam. Conversely, the flumazenil discriminative stimulus was attenuated by doses of barbiturate and neuroactive steroid site positive modulators that were smaller than doses that substituted for midazolam. When compared with midazolam substitution potency, amobarbital, pentobarbital, and pregnanolone were more potent than BZ site positive modulators in preventing the flumazenil discriminative stimulus. These results suggest that noncompetitive interactions at the GABA_A receptor complex (e.g., between flumazenil and pregnanolone) more potently attenuate flumazenil than competitive interactions at BZ receptors (e.g., between flumazenil and midazolam). Enhancement of GABA_A function through barbiturate and neuroactive steroid sites is apparently not amenable to antagonism by flumazenil, thereby underlying the more effective attenuation of the flumazenil discriminative stimulus with compounds acting at non-BZ sites.

These results might also be related to barbiturates and neuroactive steroids having greater affinity or positive modulatory efficacy than BZ site ligands at certain GABA_A receptor subtypes. For instance, flumazenil and diazepam have very low affinity for recombinant GABA_A receptors comprising ₄- and ₆-subunits (Huang et al., 2000), whereas pentobarbital potentiates GABA responses mediated through ₄₁₂ and ₆₁₂ subunit assemblies (Wafford et al., 1996). Thus, pentobarbital might more effectively attenuate flumazenil in diazepam-treated monkeys through actions at ₄₁₂ and ₆₁₂ subunit assemblies on neurons also containing subunit assemblies binding flumazenil (e.g., ₁₁₂, ₂₁₂, ₃₁₂, and ₅₁₂; for review, see Sieghart, 1995). Unlike flumazenil and diazepam, neuroactive steroids can positively modulate GABA through receptors not containing -subunits (Puia et al., 1990). In addition, potency of neuroactive steroids to positively modulate GABA is increased by substitution of a ₁- for a ₂-subunit, whereas BZ site ligands have low affinity for GABA_A receptors comprising ₁-subunits (Puia et al., 1993; Benke et al., 1996). Efficacy differences among positive modulators might account for the present results, with barbiturates and neuroactive steroids having greater efficacy than BZ site ligands at certain GABA_A subunit assemblies. The mechanism by which positive modulators increase chloride flux might also impact their potency to attenuate flumazenil. Potentiation of chloride flux with BZ site ligands is GABA-dependent, whereas potentiation of chloride flux with larger concentrations of barbiturates and neuroactive steroids is GABA-independent and occurs via direct activation of GABA_A receptors (Puia et al., 1990; Amin and Weiss, 1993).

The lower relative potency of BZ site positive modulators to prevent the flumazenil discriminative stimulus in diazepam-treated subjects could reflect a selective development of tolerance to BZ site ligands. Numerous studies have shown that tolerance develops to the anxiolytic, sedative-hypnotic, and anticonvulsant effects of BZs (for review, see File, 1985). Comparison of the effects of positive modulators on response rate between midazolam- and flumazenil-discriminating monkeys indicates that daily treatment with diazepam produced tolerance to the rate-decreasing effects of BZ site ligands. For instance, very large doses of BZ site positive modulators had little effect on response rate in diazepam-treated monkeys (Figs. 2-6, bottom), whereas much smaller doses decreased response rate in untreated monkeys (Fig. 1, bottom). Differences in the reinforcer (food versus SST) are not likely to account for these differences in sensitivity (L. R. McMahon and C. P. France, unpublished observations). A number of adaptations in GABA_A receptor function might account for BZ tolerance, including changes in GABA_A receptor mRNA expression or decreased efficacy of positive modulators acting at BZ sites, resulting from decreased allosteric coupling between BZ sites and GABA_A receptor-mediated chloride channels (Heninger et al., 1990; Hu and Ticku, 1994). In contrast to the well documented development of BZ tolerance, cross-tolerance does not reliably develop from BZs to barbiturates or neuroactive steroids (Cesare and McKearney, 1980; Rosenberg et al., 1983; Reddy and Rogawski, 2000). Similarly, there was no evidence for cross-tolerance to the rate-decreasing effects of non-BZ site positive modulators in diazepam-treated and -tolerant monkeys (compare Fig. 1, bottom, to Figs. 7-9). Thus, diazepam treatment might diminish the positive modulatory effects of drugs acting at BZ sites without affecting the positive modulatory effects of drugs acting at other sites. An apparent lack of cross-tolerance to amobarbital, pentobarbital, and pregnanolone might be partly responsible for the greater relative potency of the latter drugs in attenuating the flumazenil stimulus in diazepam-treated monkeys.

The BZ₁-selective ligands zaleplon and zolpidem were slightly less potent than diazepam and triazolam and approximately equipotent to midazolam in attenuating the flumazenil discriminative stimulus. These results suggest that BZ₁-selective ligands have actions similar to those of nonselective BZs despite their modestly higher affinity for ₁-subunits in vitro (Depoortere et al., 1986; Beer et al., 1997). This notion is consistent with zolpidem, zaleplon, and nonselective BZs maintaining comparable levels of self-administration, eliciting similar withdrawal syndromes in baboons, and having comparable subject-rated measures indicative of abuse potential (Griffiths et al., 1992; Rush et al., 1999; Ator, 2000). In contrast, other studies have suggested differences between BZ₁-selective ligands and nonselective BZs. For instance, unlike nonselective BZs, zolpidem or zaleplon do not always substitute for the discriminative stimulus effects of ethanol or other positive GABA_A modulators (Depoortere et al., 1986; Sanger et al., 1996; Sanger, 1997; but see Griffiths et al., 1992; Ator, 2000; this study). Moreover, some positive GABA_A modulators do not substitute for BZ₁-selective ligands (Sanger and Zivkovic, 1986; Vanover and Barrett, 1994; Rowlett et al., 1999), and BZ₁-selective ligands produce little tolerance in rodents (Depoortere et al., 1986; Sanger et al., 1996). These differences notwithstanding, the present results in nonhuman primates support the notion that some ligands with selectivity for BZ₁ receptors in vitro are not qualitatively different from nonselective BZs in vivo.

In summary, these results demonstrate that various positive GABA_A modulators substitute for a midazolam discriminative stimulus in otherwise untreated monkeys, with BZ site ligands being the most potent followed by the neuroactive steroid pregnanolone and the barbiturates amobarbital and pentobarbital. Each positive modulator attenuates the discriminative stimulus effects of flumazenil in diazepam-treated monkeys, i.e., prevents withdrawal. Unlike the low relative potency of pregnanolone, amobarbital, and pentobarbital in substituting for midazolam in untreated monkeys, these compounds are more potent than BZ site ligands in attenuating the flumazenil discriminative stimulus in diazepam-treated monkeys. The higher relative potency of neuroactive steroid and barbiturate site ligands might be due to actions at GABA_A receptor subtypes not modulated by BZ site ligands, to the development of tolerance to BZ site ligands without cross-tolerance to non-BZ site ligands, or to noncompetitive interactions at the GABA_A receptor complex. Relatively little is known about the nature of interactions among GABA_A modulators, and this paucity of data is especially true under conditions of chronic dosing (e.g., BZ dependence). These results predict that BZ withdrawal might be more effectively prevented by barbiturates and neuroactive steroids than BZs. The extent to which results of studies in diazepam-treated monkeys predict interactions among GABA_A modulators in monkeys treated with non-BZ site modulators (e.g., neuroactive steroids) is currently under investigation and should provide an additional rigorous test of the applicability of receptor theory to these behavioral measures.