Recent evidence suggests that GABAA receptors containing an α1 subunit mediate the sedative effect of diazepam, whereas receptors with an α2 subunit mediate this benzodiazepine's anxiolytic effect. Thus, compounds selective for GABAA-α2 receptors may offer advantages, i.e., lack of sedation, over current benzodiazepines. Whether such compounds would offer additional advantages over benzodiazepines is unclear. Here, we address the issue of physical dependence by comparing the GABAA-α1 affinity-selective drug zolpidem, the novel compounds 7-(1,1-dimethylethyl)-6-(2-methyl-2H-1,2,4-triazol-3-ylmethoxy)-3-(2,5-difluorophenyl)-1,2,4-triazolo[4,3-b]pyridazine (L-838,417) and 6-fluoro-9-methyl-2-phenyl-4-(pyrrolidin-1-yl-carbonyl)-2,9-dihydro-1H-pyridol[3,4-b]indol-1-one (SL651498) with functional selectivity for certain non-α1 GABAA receptors, nonselective partial agonists [bretazenil, 1-[1-[3-(3-pyridyl)phenyl]benzimidazol-5-yl]ethanone O-ethyloxime (NS2710), and 5-furan-3-yl-1-(3-imidazol-1-phenyl)-1H-benzoimidazole (NS2664)], and nonselective full efficacy benzodiazepines, in a rapid precipitated withdrawal assay using the inverse agonist N-methyl-β-carboline-3-carboxamide (FG-7142). For all compounds, we determined in vitro IC50 values to displace [3H]flunitrazepam from rat cortex and in vivo ED50 values for displacement of [3H]flunitrazepam from mouse forebrain (including length of in vivo occupancy). In the precipitated withdrawal model, compounds were administered at a dose giving ∼80% receptor occupancy, obviating major differences in central nervous system bioavailability. Mice were administered compounds twice daily for 4 days and on day 5, 20 h after the final dose, given a dose of FG-7142 (40 mg/kg i.p.) that did not induce seizures in control animals. In mice treated with the three subtype-selective compounds, FG-7142 did not induce seizures. Moreover, there was a low propensity for FG-7142 to induce seizures in animals treated with the partial agonists, whereas seizures were clearly seen in animals treated with most benzodiazepines. Nonetheless, differences among the benzodiazepines themselves, similarities between the partial agonists and subtype-selective compounds, the in vitro/in vivo potency, and in vivo receptor exposure time data suggest a complex interaction among selectivity, efficacy, potency, and receptor exposure in determining physical dependence liability of benzodiazepine site modulators in mice.
Recent studies show that GABAA receptors containing a α1 subunit mediate diazepam's locomotor depressant action (Rudolph et al., 1999; McKernan et al., 2000), whereas GABAA receptors containing an α2 subunit mediate diazepam-induced anxiolysis in the elevated plus maze and lightdark box (Low et al., 2000). These studies raise the possibility that novel GABAA subtype-selective compounds may retain anxiolytic efficacy with sedative side effects obviated. However, it is unclear what additional advantages such compounds may offer over current benzodiazepines.
This point is relevant because benzodiazepines interact with ethanol, impair cognition, induce physical dependence, and abuse liability. Because of these safety issues, antidepressants are first line drugs for treating anxiety disorders despite a slower onset of efficacy (Nutt, 2005). However, nonselective partial agonists that bind to the benzodiazepine site such as bretazenil do not fully potentiate the effect of GABA at GABAA receptors (Haefely et al., 1990) and have reduced abuse, physical dependence, and memory impairment liabilities in animals and humans (Busto et al., 1994; Martin et al., 1995). Likewise, zolpidem, a drug that selectively binds to GABAA-α1-containing receptors, has a reduced propensity to induce physical dependence in rodents (von Voigtlander and Lewis, 1991; Perrault et al., 1992), baboons (Weerts et al., 1998), and humans (Shaw et al., 1992; Hajak et al., 2003).
Recently, two ligands, L-838,417 (McKernan et al., 2000) and SL651498 (Griebel et al., 2001), have been described that show selectivity for non-α1-containing GABAA receptors, i.e., a profile that is in many ways opposite to that of the GABAA-α1-selective drug zolpidem (see below). Therefore, it is relevant to ask whether this selectivity profile imparts a different side effect profile for these compounds compared with benzodiazepines, in addition to their reduced sedative properties already described (McKernan et al., 2000; Griebel et al., 2001). Zolpidem's affinity selectivity for GABAA-α1 receptors means it selectively potentiates the effect of GABA at GABAA-α1 receptors, although this is concentration-dependent because it only shows 10- to 20-fold selectivity for GABAA-α1 receptors over GABAA-α2 or α3 receptors. However, zolpidem does show >1000-fold selectivity for GABAA-α1 over GABAA-α5 receptors (Faure-Halley et al., 1993; Sieghart, 1995). As described above, the selectivity of zolpidem for GABAA-α1 receptors may be important in the reduced physical dependence properties of this drug. However, neither L-838,417 nor SL651498 have selective affinity for non-α1 GABAA receptors (Atack, 2003). Rather, both compounds have been described as functionally (or efficacy) selective modulators of non-α1 GABAA receptors. That is, both compounds show no affinity difference in binding to GABAA receptors containing α1, α2, α3, or α5 subunits but differentially modulate these receptors when bound, leading to functional selectivity. Thus, L-838,417 does not potentiate effects of GABA at GABAA-α1 receptors but does so at GABAA-α2, -α3, and -α5 receptors; L-838,417 is a partial agonist at these latter subtypes. SL651498 fully potentiates effects of GABA at GABAA-α2 receptors and has marginally lower efficacy at GABAA-α3 receptors, with least efficacy (partial agonist profile) at GABAA-α1 and -α5 receptors. SL651498 has reduced physical dependence liability compared with benzodiazepines in a mouse-precipitated withdrawal model (Griebel et al., 2001). This compounds partial agonism at certain GABAA receptor subtypes, and/or its subtype selectivity may account for this profile. As described above, a nonselective partial agonist like bretazenil that does not fully potentiate the effect of GABA at GABAA receptors also has a reduced propensity to induce physical dependence in various species. Moreover, in the current study, we introduce two novel nonselective partial agonists, NS2710 and NS2664, which also have a reduced physical dependence liability in mice (for structures of nonselective partial agonists and selective modulators described, see Table 1).
In addition to intrinsic efficacy and receptor selectivity, the liability of benzodiazepine site modulators to induce physical dependence may be dependent upon other factors including potency, half-life, length of treatment, and differences between continuous/intermittent treatment (Woods et al., 1995). For example, short-intermediate half-life benzodiazepines like midazolam, triazolam, and lorazepam result in more severe rebound night-time insomnia/daytime anxiety in humans than longer acting agents like diazepam and clonazepam. However, in rhesus monkeys Yanagita (1983) demonstrated a more severe withdrawal syndrome with diazepam rather than zopiclone, despite diazepam's longer half-life. There is also evidence that high-potency compounds such as triazolam are particularly noted for inducing withdrawal symptoms (Vgontzas et al., 1995; Chouinard, 2004).
We chose a precipitated withdrawal model of physical dependence and compared the propensity of the β-carboline FG-7142, a partial inverse agonist at the benzodiazepine receptor, to precipitate seizures in mice (von Voigtlander and Lewis, 1991; Martin et al., 1995) treated with nonselective full efficacy benzodiazepines that differed with respect to half-lives and potency. Thereafter, we tested nonselective partial agonists: bretazenil and two benzimidazoles, NS2710 (Mirza et al., 2003) and NS2664 (Mathiesen et al., 2003); the imidazopyridine zolpidem, a GABAA-α1-selective sedative-hypnotic; and SL651498, a pyridoindole, and the triazolopyridazine L-838,417, described above.
To ascertain if in vitro potency correlated with propensity for physical dependence, we determined IC50 values for all compounds to displace [3H]flunitrazepam from rat cortex. Moreover, we determined the ED50 for all compounds to displace [3H]flunitrazepam from mouse forebrain in vivo and thereby selected doses giving ∼80% receptor occupancy for the withdrawal studies. Thus, we could compare the relative tendency of compounds to induce physical dependence based on their selectivity and intrinsic efficacy rather than central nervous system bioavailability. We also determined the time course for in vivo displacement of [3H]flunitrazepam to ascertain whether length of receptor exposure was a factor in physical dependence liability. However, in vivo receptor occupancy is potentially a combination of the parent compound administered and any active metabolite(s). Moreover, because of active metabolites, it is important to consider that receptor exposure time is not necessarily equivalent to either the plasma or brain half-life of the parent compound (Table 1). Because some benzodiazepines and active metabolites have a long plasma half-life (e.g., diazepam and desmethyldiazepam) that may contribute to prolonged receptor exposure, it is feasible that a challenge dose of FG-7142 may simply induce a seizure in an animal by competitively displacing the agonist (parent and/or metabolite) from its binding site. To circumvent this problem, to the extent possible, FG-7142 was administered 20 h after the last administration of each compound.
Materials and Methods
Animals. Female NMRI mice (Harlan Scandinavia, Allerød, Denmark) weighing 20 to 25 g were housed and habituated for at least 7 days before experiments in Macrolon III cages (20 × 40 × 18 cm) with eight mice per cage. Food (Altromin) and tap water were available ad libitum, with lights on at 6:00 AM and off at 6:00 PM. All behavioral testing was conducted during the light phase. All experiments were performed according to the guidelines of the Danish Committee for Experiments on Animals.
Drugs and Solutions. Diazepam, chlordiazepoxide, clonazepam, lorazepam, and FG-7142 were all purchased from Sigma-Aldrich (Vallensbæk Strand, Denmark), whereas alprazolam and triazolam were sourced from Cambrex (Charles City, IA), and midazolam was a gift from Roche (Basel, Switzerland). Bretazenil, zolpidem, SL651498, L-838,417, NS2710, and NS2664 were all synthesized at the medicinal chemistry department, NeuroSearch A/S. All compounds were administered i.p. at a dose volume of 10 ml/kg, dissolved in 5% Cremophor. [3H]Flunitrazepam (88 Ci/mmol) was purchased from Amersham Biosciences UK, Ltd. (Little Chalfont, Buckinghamshire, UK). All other chemicals for binding studies described below were purchased from regular commercial sources and were of the purest grade available.
Precipitated Withdrawal Procedure. Mice were administered vehicle or test substances (n = 8–10) twice a day (8:00–9:00 AM and 2:00–3:00 PM) for 4 consecutive days (Monday–Thursday). Approximately 20 h after the last dose (Friday, 10:00–11:00 AM) mice were administered FG-7142 (40 mg/kg i.p.) and individually placed in separate observation cages for 30 min. Animals were continuously observed for any signs of seizure activity during this time. In the experiments reported here, the seizures observed were always clonic seizures involving the forelegs and facial muscles. Full-blown whole-body tonic-clonic seizures were never observed and were not anticipated with FG-7142 (Little et al., 1988). In preliminary studies a dose-response curve for FG-7142 (10–160 mg/kg) showed that this partial inverse agonist never engendered seizures in control animals (data not shown). FG-7142 was chosen for these experiments because partial inverse agonists seem to be more sensitive in detecting dependence liability of benzodiazepines site modulators compared with antagonists such as flumazenil (Martin et al., 1995).
The compounds tested in this study were administered at a dose that in most cases gave ≥80% benzodiazepine receptor occupancy in mouse forebrain based on in vivo binding experiments (see below). Consequently, the doses we have chosen for some compounds are generally lower than those used in previous studies assessing the liability of benzodiazepine site modulators to induce physical dependence (e.g., von Voigtlander and Lewis, 1991). In all experiments, triazolam (1.4 mg/kg) was used as a positive control (n = 6–10), consistently resulting in seizures in 80 to 100% of mice after challenge with FG-7142, whereas no seizures were seen in vehicle-treated animals (n = 8–10) (for doses of each compound used in the precipitated withdrawal model, see Table 2, column 4).
In Vitro [3H]Flunitrazepam Binding (Rat). Rat cerebral cortical membranes were prepared from male Wistar rats as described by Johansen et al. (1993). Cerebral cortices were removed rapidly after decapitation, homogenized for 5 to 10 s in 10 volumes of 30 mM Tris-HCl (pH 7.4), and centrifuged at 27,000g for 15 min. All procedures were performed at 0 to 4°C unless otherwise indicated. After washing the pellet three times (resuspension in 10 volumes of ice-cold buffer and centrifugation at 27,000g for 10 min), the pellet was homogenized in Tris-HCl buffer, incubated on a water bath (37°C) for 30 min, and then centrifuged at 27,000g for 10 min. The pellet was then homogenized in buffer and centrifuged for 10 min at 27,000g. After one more wash, the final pellet was resuspended in 10 volumes of buffer, and the preparation was stored at –20°C. On the day of the experiment, the membrane preparation was thawed and centrifuged at 2°C for 10 min at 27,000g. The pellet was washed twice with Tris-citrate (50 mM, pH 7.1) and centrifuged for 10 min at 27,000g. The final pellet was resuspended in Tris-citrate (500 ml of buffer/g of original tissue) and then used for binding assays. Aliquots of 500 μl of tissue were added to 25 μl of compound at the indicated concentration and 25 μl of [3H]flunitrazepam (1 nM, final concentration), mixed, and incubated for 40 min at 2°C. Nonspecific binding was determined in the presence of 1 μM clonazepam. Compounds were tested at 7 to 10 concentrations ranging from 0.03 nM to 30 μM. Binding was terminated by rapid filtration over Whatman GF/C glass fiber filters (Whatman, Clifton, NJ).
In Vivo [3H]Flunitrazepam Binding (Mouse). The in vivo binding studies were conducted as described by Jensen et al. (1983). Briefly, groups of three female NMRI mice (25–28 g) were injected i.p. with compound solutions prepared in 5% Tween 80. Ten minutes after compound administration, the mice were injected i.v. via the tail vein with 5.0 μCi [3H]flunitrazepam in 0.2 ml of saline. Twenty minutes after injection of [3H]flunitrazepam (i.e., 30 min after administration of compound), the mice were killed by decapitation, and the forebrains were rapidly excised and homogenized in 12 ml of ice-cold Tris-citrate (50 mM, pH 7.1) using an Ultra-Turrax homogenizer. Three aliquots of 1 ml were immediately filtered through Whatman GF/C glass fiber filters and washed with 2 × 5 ml of ice-cold buffer. Groups of vehicle-treated mice served as controls for estimation of total binding. To determine nonspecific binding, groups of mice were injected with clonazepam (2.5 mg/kg i.p.) 10 min before [3H]flunitrazepam injection. The 30-min pretreatment time was chosen to ensure an ED50 value for all compounds regardless of differences in half-life. Four to five doses ranging from 0.03 to 100 mg were tested for determination of ED50 values. In time course experiments, compounds were administered relative to the decapitation time to separate groups (n = 3 at each time point) of female NMRI mice, and the dose tested was the same as used in the behavioral studies.
Although it is possible that using a nonselective drug such as [3H]flunitrazepam may underestimate the receptor occupancy of ligands with affinity selectivity for subtypes of GABAA receptors, in the current study, we encountered no such problems with the GABAA-α1-selective modulator zolpidem (Faure-Halley et al., 1993; see Results). In all binding assays, the amount of radioactivity on the filters was determined by conventional liquid scintillation counting.
Data Analysis. Behavioral data are shown as the percentage of animals showing clonic seizures. In in vitro binding studies, IC50 values were estimated from the equation B = 100– [100 × Dn/(IC50n + Dn)] by nonlinear regression using GraphPad Prism version 2.01 (GraphPad Software Inc., San Diego, CA), where B is the specific binding in percentage of total specific binding, D is the drug concentration (nanomolar), and n is the Hill coefficient. All results are given as mean ± S.E.M. ED50 values for inhibition of in vivo [3H]flunitrazepam binding were estimated from the equation B = 100– [100 × Dn/(ED50n + Dn)], where D is the administered dose of drug (milligrams per kilogram i.p.), and n is the Hill coefficient. In vivo receptor occupancy was estimated from the equation RO = D × 100/(ED50 + D).
Table 1 gives the structures of all compounds assessed in this study, along with available information on elimination half-lives of parent compound and any known active metabolites in human. Table 2 summarizes the in vitro and in vivo potency of all compounds and the dose chosen, based on in vivo receptor binding studies described below, in the precipitated withdrawal test. All compounds completely inhibited in vitro binding at the highest concentration tested, and in vivo, all benzodiazepines and partial agonists showed 90 to 100% inhibition at the highest dose tested. For the subtype-selective modulators the inhibition varied from ∼80 to 90% at the highest dose tested. The Hill coefficients were not significantly different from 1.
Benzodiazepines. All benzodiazepines potently inhibited in vitro [3H]flunitrazepam binding with IC50 values ranging from 0.62 to 19 nM, except for chlordiazepoxide, which had an IC50 of 930 nM (Table 2). Similarly, all benzodiazepines, with the exception of chlordiazepoxide, inhibited in vivo [3H]flunitrazepam binding with ED50 values in the range of 0.12 to 1.4 mg/kg, when measured 30 min after i.p. administration (Table 2). Chlordiazepoxide had an ED50 of 11 mg/kg. The calculated receptor occupancy at the doses tested in the behavioral studies described below for the benzodiazepines ranged from 74% (70–77%) for midazolam to 91% (87–95%) for alprazolam (Fig. 1A). Time course in vivo binding studies using the doses of each benzodiazepine selected for behavioral studies described below showed that the benzodiazepines generally inhibited in vivo [3H]flunitrazepam binding >70% for at least 6 h in the mouse brain (Fig. 2A). However, in vivo inhibition of [3H]flunitrazepam binding at 6 h postadministration for triazolam, midazolam, and clonazepam was 0, 0, and 53%, respectively, suggesting a shorter receptor occupancy half-life compared with the other benzodiazepines tested. Moreover, for midazolam, very little in vivo receptor occupancy (∼20%) was still observed at 2 h. This short-lived receptor occupancy profile of midazolam is similar to that observed with the majority of nonselective partial agonists and subtype-selective modulators described below.
Subchronic treatment with the various benzodiazepines engendered a differential propensity to seizure response in mice after a challenge dose of FG-7142 (Fig. 1A). Thus, although 80 to 90% of mice treated with triazolam, clonazepam, and diazepam had clonic seizures after challenge with FG-7142, no mice treated with the low-potency benzodiazepine chlordiazepoxide seized after administration of this β-carboline. Somewhat intermediate between these two extremes were the benzodiazepines lorazepam, midazolam, and alprazolam, where challenge with FG-7142 engendered seizures in 44 to 67% of mice.
Partial Agonists. The three nonselective partial agonists, NS2664, bretazenil, and NS2710, inhibited in vitro [3H]flunitrazepam binding with IC50 values ranging from 0.61 to 3.0 nM and potently inhibited in vivo binding with ED50 values between 0.15 and 1.7 mg/kg (Table 2). The calculated receptor occupancy at the doses tested in the behavioral study described below ranged from 83 to 89% (Fig. 1B). In vivo [3H]flunitrazepam time course studies using doses of each compound selected for behavioral studies described below showed that NS2664 and bretazenil were short lasting in mouse brain, with only 2 and 12% inhibition of [3H]flunitrazepam binding, respectively, 2 h postadministration (Fig. 2B). However, NS2710 and/or metabolites occupied the receptor for longer with 94 and 42% inhibition of [3H]flunitrazepam binding 2 and 6 h postadministration, respectively (Fig. 2B).
Mice treated with the nonselective partial agonist bretazenil showed a very low propensity to clonic seizures after the challenge dose of FG-7142, with only 20% of mice showing clonic seizures (Fig. 1B). This study was replicated to ensure the reproducibility of this data, giving essentially the same percentage of mice with a seizure response after FG-7142 challenge (Table 2). Likewise, only 30% of mice treated with NS2710, a nonselective partial agonist at GABAA receptors, had clonic seizures after challenge with FG-7142 (Fig. 1B). However, at the doses tested, no mice treated with the nonselective highly potent weak partial agonist NS2664 seized after FG-7142 challenge, following treatment twice daily for 4 consecutive days (Fig. 1B).
Subtype-Selective Agents. L-838,417 had subnanomolar potency (IC50 = 0.83 nM) in inhibiting in vitro [3H]flunitrazepam binding, whereas SL651498 and zolpidem had IC50 values of 34 and 160 nM, respectively (Table 2). However, L-838,417 and SL651498 showed similar potency in inhibiting in vivo [3H]flunitrazepam binding (ED50 = 6.3 and 5.1 mg/kg, respectively; Table 2). Zolpidem was less potent with an ED50 of 19 mg/kg for inhibiting in vivo [3H]flunitrazepam to mouse forebrain. All three compounds were short-acting, and no inhibition of in vivo [3H]flunitrazepam binding in mouse forebrain was seen 2 h postadministration at the doses of L-838,417 (10 mg/kg), SL651498 (10 mg/kg), or zolpidem (30 mg/kg) tested for receptor occupancy over time (Fig. 2B).
The selective GABAA-α1 receptor drug zolpidem at doses of 30 and 100 mg/kg resulted in 62 and 85% occupancy of [3H]flunitrazepam binding sites in mouse forebrain, respectively (Fig. 1B shows the occupancy attained after the 100 mg/kg dose only). In mice treated with these doses of zolpidem, FG-7142 failed to engender any seizure response (Fig. 1B; Table 2). At a dose of 10 mg/kg, the two functionally selective compounds, L-838,417 and SL651498, were calculated to displace 71% (69–72%) and 75% (70–80%) of [3H]flunitrazepam binding in mouse forebrain, respectively, when measured 30 min after administration (Fig. 1B). However, from the time course study, it is clear that L-838,417 displaced 89% of [3H]flunitrazepam binding sites from mouse forebrain when binding was measured 15 min after administration (Fig. 2B), i.e., somewhat higher than seen when binding was measured 30 min after L-838,417 administration. In addition to testing both L-838,417 and SL651498 in the precipitated withdrawal study at a dose of 10 mg/kg giving ∼80% benzodiazepine receptor occupancy, both compounds were also tested at higher doses (15 and 30 mg/kg) in the behavioral assay. However, mice treated subchronically for 4 days with either agent at 10, 15, or 30 mg/kg did not show any seizures after a challenge dose of FG-7142 on day 5 (Fig. 1B; Table 2).
The current study shows that mice subchronically administered a range of benzodiazepines, with the exception of chlordiazepoxide, have an increased propensity to clonic seizures after a challenge dose with the β-carboline inverse agonist FG-7142, which at the dose tested (40 mg/kg) had no tendency to induce seizures in vehicle-treated animals; after treatment with the partial agonists bretazenil, NS2710, and NS2664, the percentage of mice with clonic seizures after a challenge dose of FG-7142 was generally lower than the percentage of mice seizing after treatment with benzodiazepines; and no mice treated with the subtype-selective agents zolpidem, L-838,417, and SL651498 and subsequently administered FG-7142 had seizures. One conclusion from this set of data may be that partial agonism and/or subtype selectivity lead to a reduced liability to physical dependence in mice. However, the in vitro and in vivo potency data and in vivo receptor occupancy time curves for all the compounds suggest that additional factors may be important in the propensity for a given benzodiazepine site modulator to engender physical dependence.
Thus, even among the benzodiazepines, there was a differential liability to induce physical dependence. For example, challenge with FG-7142 engendered no seizures in mice treated with chlordiazepoxide, whereas ≥80% of mice treated with triazolam, clonazepam, or diazepam seized after challenge with the β-carboline. Somewhat intermediate between these extremes were midazolam, lorazepam, and alprazolam, with 60, 44, and 67% of mice treated with these compounds seizing after FG-7142 challenge, respectively (Fig. 1A). Thus, the overall rank order of propensity for seizures after FG-7142 challenge was: triazolam = clonazepam = diazepam > alprazolam = midazolam = lorazepam ≫ chlordiazepoxide. Although it is reasonable to argue that the behavioral measure here may not be sensitive enough to truly rank compounds as such, the data nonetheless indicate obvious differences between chlordiazepoxide and other benzodiazepines. Because benzodiazepines are considered a homogenous group with respect to being full nonselective positive modulators at α1-, α2-, α3-, and α5-containing GABAA receptors (Sieghart, 1995), factors other than intrinsic efficacy and selectivity must account for the differences we see in mice. Comparing the rank order for seizures above with in vitro/in vivo potency data suggests potency may play some role because chlordiazepoxide has the lowest in vitro/in vivo potency and engendered no physical dependence at the dose tested. Although this may be the most parsimonious conclusion, future studies should address the issue of treatment length (see Introduction). Moreover, dose is certainly a factor with chlordiazepoxide because others (von Voigtlander and Lewis, 1991) have demonstrated precipitated withdrawal with flumazenil challenge in mice treated with high (150 mg/kg/day) but not low (1.5–15 mg/kg) doses of chlordiazepoxide in a 3-day procedure.
In contrast, receptor occupancy was approximately matched between the benzodiazepines, making central nervous system bioavailability of parent compound/metabolites an unlikely explanation for differences in physical dependence liability (Fig. 1A). Furthermore, receptor occupancy time curves (Fig. 2A) show that ∼70% in vivo benzodiazepine receptor occupancy [parent compound/metabolite(s)] was maintained for up to 6 h postadministration with most benzodiazepines, with the notable exceptions of midazolam and triazolam. Therefore, differences in receptor exposure time do not explain why chlordiazepoxide differs from other benzodiazepines. However, although absolute receptor occupancy and exposure cannot explain the differences in withdrawal liability between the benzodiazepines, the nature of potential metabolites that contribute to this receptor occupancy for the different benzodiazepines and the rate of receptor occupancy may be important determinants. For example, chlordiazepoxide and diazepam have long-acting active metabolites with pharmacological activity equivalent to the parent drug, whereas the metabolites of alprazolam, clonazepam, and lorazepam are short lasting and essentially pharmacologically inactive (Table 1). Furthermore, the rate of receptor occupancy is a variable we have not fully explored. In the field of dopamine reuptake inhibitors, for example, it has been demonstrated that both the rate and absolute occupancy of the dopamine transporter are key variables in liability of such compounds to be abused (Volkow et al., 2005).
Notwithstanding the differences between the benzodiazepines per se, the partial agonists had a considerably reduced liability to engender seizure response in mice challenged with FG-7142 compared with benzodiazepines (Fig. 1B). Thus, only ∼20 and 30% of mice treated with bretazenil and NS2710, respectively, had seizures after FG-7142 challenge. The data with bretazenil is commensurate with prior literature clearly showing that this compound has less liability to induce physical dependence in rodents and primates compared with benzodiazepines (Haefely et al., 1990; Busto et al., 1994; Martin et al., 1995). The benzimidazole structure NS2710 has an in vitro efficacy profile similar to bretazenil, i.e., a positive nonselective partial agonist at low GABA concentrations (Mirza et al., 2003). However, at high GABA concentrations, NS2710 has an inhibitory modulatory effect that is independent of the α subunit and that, unlike its positive modulatory effect, is not antagonized by flumazenil (Mirza et al., 2003). However, the importance of this dual mechanism effect in the propensity to physical dependence is unclear because NS2710 and bretazenil seem equivalent in our model. In contrast, NS2664, another benzimidazole structure, differed in that it engendered no seizures in mice challenged with FG-7142. NS2664, in patch-clamp electrophysiology studies on cloned human GABAA receptors expressed in HEK cells, is classified as a weak partial agonist/antagonist because its positive modulatory effects are often difficult to detect (Mathiesen et al., 2003). It is possible that the reduced intrinsic efficacy of NS2664 relative to NS2710 and bretazenil leads to a very low physical dependence liability. Certainly the profile of NS2664 relative to NS2710 and bretazenil cannot be explained by major differences in in vitro/in vivo potency, in vivo receptor occupancy, or receptor exposure time (Table 2; Figs. 1B and 2B).
The functionally selective positive modulators, L-838,417 and SL651498 (10–30 mg/kg), did not engender seizures after FG-7142 challenge, at doses attaining receptor occupancy equivalent to the benzodiazepines and partial agonists. The data were somewhat surprising with regard to SL651498, which has greater efficacy at α1, α2, α3, and α5 receptors compared with L-838,417 (Atack, 2003), although its efficacy values are generally intermediate between the full-efficacy benzodiazepines and the partial agonists described above (Sieghart, 1995; Atack, 2003). However, the data with SL651498 are in agreement with the lack of precipitated withdrawal after a challenge dose of flumazenil in mice administered 30 mg/kg SL651498 twice daily for 10 days (Griebel et al., 2001). The data seem to indicate that subtype-selective compounds with selectivity for non-α1 GABAA receptors have reduced liability to engender physical dependence. However, it is worth noting that FG-7142 has a modest 3- to 5-fold affinity selectivity for GABAA-α1 over GABAA-α2 and -α3 receptors and 20-fold selectivity over GABAA-α5-containing receptors. Because the partial agonists and the functionally selective compounds tested here have low or zero efficacy at GABAA-α1 receptors, this may explain why FG-7142 is more liable to engender convulsions in mice treated with nonselective modulators with full efficacy at GABAA-α1 receptors. However, zolpidem has affinity selectivity for GABAA-α1 receptors and is a full agonist at this receptor like the benzodiazepines but nonetheless has a considerably reduced seizure response after FG-7142 challenge as demonstrated in other species (von Voigtlander and Lewis, 1991; Perrault et al., 1992; Richards and Martin, 1998; Voderholzer et al., 2001; Hajak et al., 2003).
Thus, subtype selectivity per se, regardless of the specific receptor selectivity profile of a given compound, may be important in mitigating physical dependence. However, the partial agonist profiles of SL651498 and L-838,417 at select GABAA receptors suggest intrinsic efficacy is also a factor. Likewise, although zolpidem is a full agonist at GABAA-α1 receptors, it has a considerably reduced affinity at α5 receptors, suggesting it is unlikely to activate GABAA-α5 receptors in vivo (Sanger and Benavides, 1993; Richards and Martin, 1998). Therefore, at least in this mouse model, physical dependence is only induced by compounds with full efficacy at α1, α2, α3, and α5 receptors.
Although it might be concluded from the precipitated withdrawal study that only nonselective full efficacy modulators like the benzodiazepines will induce physical dependence, it is worth noting that for the majority of partial agonists and subtype-selective modulators, receptor exposure time was shorter compared with the benzodiazepines (compare Fig. 2, A with B). Therefore, this factor may be important in explaining their reduced liability to engender physical dependence. However, the nonselective full-efficacy benzodiazepine midazolam clearly engendered greater physical dependence in mice compared with the partial agonists and subtype-selective compounds, despite a similar receptor exposure time (Fig. 2, A and B). Conversely, the partial agonist NS2710 had a receptor exposure time (Fig. 2B) roughly intermediate between that of clonazepam and triazolam but, nonetheless, had a lower tendency to engender physical dependence compared with these benzodiazepines.
In conclusion, we have considered in vitro/in vivo potency, receptor occupancy/receptor exposure time, selectivity, and efficacy as potential factors in the predilection of positive benzodiazepine site modulators to engender physical dependence. Our data suggest that a combination of low intrinsic efficacy and selectivity are likely to lead to reduced physical dependence. However, future studies with novel, ideally, affinity-selective modulators may give insight into the relative importance of different GABAA receptors to dependence liability. Moreover, further consideration should be given to additional factors such as treatment length, metabolites, and the rate of receptor occupancy, when assessing current and novel benzodiazepine site positive modulators to better understand the underlying determinants of dependence.
We thank the technicians and academics in the Departments of Receptor Biochemistry and in Vivo Pharmacology for technical assistance.
- Received August 21, 2005.
- Accepted December 8, 2005.
ABBREVIATIONS: L-838,417, 7-(1,1-dimethylethyl)-6-(2-methyl-2H-1,2,4-triazol-3-ylmethoxy)-3-(2,5-difluorophenyl)-1,2,4-triazolo[4,3-b]pyridazine; SL651498, 6-fluoro-9-methyl-2-phenyl-4-(pyrrolidin-1-yl-carbonyl)-2,9-dihydro-1H-pyridol[3,4-b]indol-1-one; NS2710, 1-[1-[3-(3-pyridyl)phenyl]benzimidazol-5-yl]ethanone O-ethyloxime; FG-7142, N-methyl-β-carboline-3-carboxamide; NS2664, 5-furan-3-yl-1-(3-imidazol-1-yl-phenyl)-1H-benzoimidazole.
- The American Society for Pharmacology and Experimental Therapeutics