Classic benzodiazepine agonists induce their clinical effects by binding to a site on gamma-aminobutyric acid type A (GABAA) receptors and enhancing receptor activity. There are conflicting data regarding whether the benzodiazepine site is allosterically coupled to gamma-aminobutyric acid binding versus the channel open-close (gating) equilibrium. The authors tested the hypothesis that benzodiazepine site ligands modulate alpha1beta2gamma2L GABAA receptor gating both in the absence of orthosteric agonists and when the orthosteric sites are occupied.
GABAA receptors were recombinantly expressed in Xenopus oocytes and studied using two-microelectrode voltage clamp electrophysiology. To test gating effects in the absence of orthosteric agonist, the authors used spontaneously active GABAA receptors containing a leucine-to-threonine mutation at residue 264 on the alpha1 subunit. To examine effects on gating when orthosteric sites were fully occupied, they activated wild-type receptors with high concentrations of a partial agonist, piperidine-4-sulfonic acid.
In the absence of orthosteric agonists, the channel activity of alpha1L264Tbeta2gamma2L receptors was increased by diazepam and midazolam and reduced by the inverse benzodiazepine agonist FG7142. Flumazenil displayed very weak agonism and blocked midazolam from further activating mutant channels. In wild-type receptors activated with saturating concentrations of piperidine-4-sulfonic acid, midazolam increased maximal efficacy.
Independent of orthosteric site occupancy, classic benzodiazepines modulate the gating equilibrium in alpha1beta2gamma2L GABAA receptors and are therefore allosteric coagonists. A Monod-Wyman-Changeux coagonist gating model quantitatively predicts these effects, suggesting that benzodiazepines minimally alter orthosteric ligand binding.
CLASSIC benzodiazepine agonists such as diazepam and midazolam are drugs that induce muscle relaxation, sedation, suppression of seizures, anxiolysis, and amnesia by enhancing the function of specific types of γ-aminobutyric acid type A (GABAA) receptors.1GABAAreceptors are inhibitory ligand-gated ion channels formed from five homologous subunits arranged around a gated chloride-selective pore. Multiple GABAAreceptor subunit classes (α, β, γ, δ, ε, π, and ρ) and isoforms have been identified, each sharing the structural motif of the cys-loop ion channel superfamily: a large N-terminal extracellular domain, four transmembrane domains (M1 to M4), and a large cytoplasmic domain between M3 and M4.2Most synaptic GABAAreceptors in mammalian brain are formed from α, β, and γ subunits with stoichiometry 2α:2β:1γ.3At the interfaces between α and β extracellular domains, each receptor-channel complex forms binding sites for the neurotransmitter γ-aminobutyric acid (GABA). When GABA binds these two orthosteric agonist sites, gating of the chloride channel is triggered.4Sedation and anxiolysis effects of benzodiazepine agonists are associated with another binding site on GABAAreceptors, characterized by nanomolar affinity for its ligands. The amino acids that determine benzodiazepine binding are located at the interface between extracellular domains of the γ subunit and one of the two α subunits. The benzodiazepine site is a structural homolog of the GABA sites.5,6
Benzodiazepine agonist binding to GABAAreceptors allosterically increases apparent GABA binding affinity and enhances electrophysiologic responses to low concentrations of GABA. Benzodiazepine site competitive antagonists block the enhancing effects of benzodiazepine agonists, whereas inverse agonists allosterically reduce GABA binding affinity and reduce electrophysiologic responses to GABA.7,8In principle, modulation of low GABA responses by benzodiazepine site ligands could be due either to altered GABA binding at the orthosteric sites or to altered gating efficacy (the transition from closed to open states) of agonist-bound receptors, or both.9In comparison to studies of general anesthetics that clearly affect the gating equilibrium of GABAAreceptors,10–12experiments investigating gating modulation by classic benzodiazepines have provided conflicting results.11,13–17The absence of consistent evidence for gating modulation has led some to infer that benzodiazepines act by allosterically altering the microscopic binding of GABA.11,16
We have examined gating modulation of recombinantly expressed α1β2γ2LGABAAreceptors by several high-affinity benzodiazepine site ligands: diazepam, midazolam, the competitive antagonist flumazenil, and the inverse agonist FG7142. Receptor-mediated chloride currents were monitored electrophysiologically in Xenopus oocytes using the two-microelectrode voltage clamp technique. To test for gating effects in the absence of GABA site agonists, we introduced a pore domain mutation, α1L264T, that induces constitutive channel activation.18,19These spontaneously active mutant channels provide enhanced sensitivity for gating modulation; if benzodiazepines alter channel gating, the fraction of active channels, and hence chloride current, should change. We also tested midazolam for gating modulation in agonist-bound wild-type receptors using an apparently saturating concentration of the partial agonist piperidine-4-sulfonic acid (P4S). We compared the sensitivity of α1L264Tβ2γ2Lversus α1β2γ2Lreceptors to midazolam by measuring leftward shifts in GABA concentration responses.
Our results indicate that high-affinity benzodiazepine ligands allosterically alter the gating equilibrium of both agonist-bound and unbound α1β2γ2LGABAAreceptors. These results are explained by a Monod-Wyman-Changeux (MWC) allosteric coagonist mechanism.
Materials and Methods
Xenopus laevis maintenance and oocyte harvest procedures were approved by the Subcommittee on Research and Animal Care of the Massachusetts General Hospital, Boston, Massachusetts.
Drugs and Solutions
γ-Aminobutyric acid, dimethyl sulfoxide (99%), diazepam, midazolam (maleate salt), P4S, picrotoxin, and all buffers and salts were purchased from Sigma (Saint Louis, MO). FG7142 and flumazenil (Ro15-1788) were obtained from Tocris Cookson (Ellisville, MO), and propylene glycol was from Fisher Scientific (Fair Lawn, NJ).
Diazepam was dissolved in 100% propylene glycol to make a 1-mm stock solution, which was diluted in recording solution. The highest concentration of propylene glycol (1%) in electrophysiology experiments had no effect on resting leak currents or on GABA-activated currents. Midazolam was dissolved in double-distilled water (1 mm) and diluted in recording solution. FG7142 and flumazenil stock solutions (20 mm) were prepared in dimethyl sulfoxide and diluted into recording solution. The maximal concentration of dimethyl sulfoxide (0.05%) had no effect on resting leak currents or on GABA-activated currents. Picrotoxin (2 mm) was dissolved directly in recording solution by stirring for 30 min in the dark. GABA stock (1 m) in double-distilled water was stored at −80°C, and fresh aliquots were thawed each day and diluted into recording solution. P4S was dissolved directly in recording solution.
Plasmids containing complementary DNAs (cDNAs) for bovine α1and human β2and γ2LGABAAreceptor subunits were kindly provided by Dr. Paul Whiting (Merck Sharp & Dohme Research Labs, Essex, United Kingdom). The L264T mutation was introduced into the α1cDNA as previously described.19The presence of the mutation and the absence of stray mutations were confirmed by dideoxynucleotide sequencing. Messenger RNAs (mRNAs) were transcribed in vitro from linearized cDNA templates (Ambion Inc., Austin, TX), isolated using spin columns (Ambion Inc.), and stored at −80°C.
Oocytes were harvested from female Xenopus laevis (Xenopus One, Ann Arbor, MI) as described previously.10Defolliculated stage V and VI oocytes were injected with 25–50 nl subunit mRNA mixtures with a weight/weight ratio of 1α1:1β2:4γ2to ensure efficient incorporation of γ2subunits and consistent sensitivity to classic benzodiazepines.20,21Oocytes were incubated for 3–6 days at 18°C in ND-96 solution (96 mm NaCl, 2 mm KCl, 10 mm HEPES, 1.8 mm CaCl2, 1.0 mm MgCl2; pH 7.5) supplemented with 0.5 U/ml penicillin and 5 μg/ml streptomycin.
Electrophysiologic recordings using the two-microelectrode oocyte voltage clamp technique were performed at room temperature (22°C). Oocytes were placed in a 0.02-ml flow chamber, impaled with borosilicate pipettes filled with 3 m KCl (0.5–2 MΩ), and voltage-clamped at −50 mV (model OC-725C; Warner Instrument Corp., Hamden, CT). Cells were superfused at a rate of 4–5 ml/min with ND-96 recording solution (without antibiotics) from glass syringe reservoirs via a polytetrafluoroethylene valve and tubing system. Delivery of superfusion solutions from reservoirs to the flow chamber was controlled by computer-activated solenoid valves. Currents were digitized at 100 Hz (Digidata 1200; Axon Instruments, Foster City, CA) and recorded on a personal computer running commercial software (Clampex8; Axon Instruments).
Because α1L264Tβ2γ2Lreceptors are very sensitive to GABA, particular care was taken to eliminate GABA and other orthosteric agonists from the apparatus during investigation of the direct effects of benzodiazepines. The potent GABAAreceptor inhibitor picrotoxin (2 mm) was used to measure the spontaneously active currents. After complete washout of picrotoxin (10 min), benzodiazepine solutions were applied to oocytes, and benzodiazepine-elicited currents (IBZ) were normalized to the spontaneous picrotoxin-sensitive current (IPTX). Benzodiazepine applications varied in duration from 15 to 120 s, depending on the rate of current activation for the specific study solution, followed by a washout of at least 5 min in ND-96. Separate experiments on a different set of oocytes were performed to determine the ratio of IPTXto maximal GABA-activated current (ImaxGABA).
Midazolam effects on P4S efficacy in wild-type GABAAreceptors were studied as follows: In one set of oocytes, P4S concentration responses were measured to determine the concentration range associated with maximal receptor activation (3–10 mm). Every other sweep was activated with 1 mm P4S for normalization. These oocytes were then exposed to 1 μm midazolam for 10 s before activation with 1 μm midazolam plus 10 mm P4S for 30 s. A second set of oocytes was used to compare the maximal efficacy of 10 mm P4S (ImaxP4S) to that of 1 mm GABA (ImaxGABA).
γ-Aminobutyric acid concentration–response studies in oocytes expressing wild-type or mutant receptors were performed in the absence and presence of midazolam. The GABA concentration ranges used for wild-type (up to 1 mm) and mutant (up to 100 μm) receptors were based on previous studies.10,19Currents were activated with GABA for 15–30 s followed by a 3- to 5-min washout. For normalization of peak currents, every other sweep was activated with maximal GABA. After GABA responses were recorded in the absence of midazolam, a second set of experiments in the same oocyte was performed in the presence of midazolam (added to both ND-96 wash and GABA solutions).
Data Analysis and Statistics
Analysis of current traces was conducted off-line.
After leak correction, currents were normalized to the average of pretest and posttest control responses in the same oocyte (elicited with 2 mm picrotoxin, 1 mm P4S, or maximal GABA). These values were renormalized relative to maximal GABA responses based on the independent determinations of IPTX/ImaxGABAand ImaxP4S/ImaxGABAratios.
Leak-corrected and normalized concentration responses in individual oocytes were fitted by nonlinear least squares with Hill (logistic) equations of the general form
where X is the varied activating ligand (GABA or a benzodiazepine agonist), ImaxGABAis the maximally evoked current, EC50is the concentration of X eliciting half of its maximal effect, and n is the Hill coefficient of activation. For GABA concentration–response studies, we defined EC500as the half-effect GABA concentration in the absence of other ligands and EC50MDZin the presence of midazolam. Because GABA EC50s varied more than twofold from oocyte to oocyte, GABA EC50left-shift ratios (EC50MDZ/EC500) were calculated for individual oocytes.
Normalized current data were also used to estimate the overall open probability (Popen) of both wild-type and mutant receptors. To estimate Popen, we combined the ligand-stimulated receptor currents (IX, where X represents either GABA or benzodiazepine agonist) and the basal spontaneous (picrotoxin-sensitive) current (IPTX). The total current was normalized to the sum of the basal current and the maximal stimulated GABA current (ImaxGABA+ IPTX). Finally, estimated Popenvalues were renormalized assuming that Popenin the presence of maximal GABA is 0.85 for wild-type α1β2γ2Land 1.0 for α1L264Tβ2γ2L(based on previous macrocurrent and single-channel studies).10,18For wild-type currents, where IPTXis too small to measure, this simply required multiplying IX/ImaxGABA× 0.85. In a generalized format, estimated Popen(Popenest) was calculated as follows:
where all currents are used as positive amplitudes. Logistic fits were also performed on the data displayed as Pestopenusing equations of the form
where X represents the variable stimulating ligand (GABA or a benzodiazepine) and Ibasalrepresents either the spontaneous current (IPTX) or when midazolam was present during GABA concentration responses, IPTX+ IMDZ.
Analysis Using the Monod-Wyman-Changeux Model.
Open probability in the MWC coagonist model, which assumes two equivalent GABA sites and a single benzodiazepine site, is defined by the following equation9,22:
where L0is the basal gating equilibrium constant ([closed]/[open] for unliganded receptors), KGis the equilibrium dissociation constant for GABA binding to closed receptors, c is the GABA efficacy for each of the two orthosteric sites (the ratio of dissociation constants in open vs. closed receptors), KBZis the equilibrium dissociation constant for benzodiazepine binding to closed receptors, and d is the benzodiazepine efficacy.
The MWC allosteric model was fitted by nonlinear least squares to subsets of Popenestvalues derived from electrophysiologic data. The two independent L0parameters for the mutant and wild-type receptor models were constrained during fitting. L0for mutant receptors was calculated based on mutant basal activity results (L0mut= ImaxGABA/IPTX= 9.1). L0for wild-type receptors (L0wt= 40,000) was calculated based on the model’s predicted relation between GABA EC50and L0mutfor mutations such as α1L264T that selectively affect the gating equilibrium18,19:
Binding and efficacy parameters for both GABA and benzodiazepines were assumed to be global parameters (the same in both mutant and wild-type models). Using Popenestvalues calculated from direct activation results in mutant receptors, benzodiazepine binding and efficacy parameters were fitted to equation 4with L0mut= 9.1 and GABA = 0. KGand c were fitted to wild-type GABA concentration–response Popenestvalues using equation 4with L0wt= 40,000 and BZ = 0.
Predicted midazolam effects on GABA concentration responses for both wild-type and mutant models were calculated with equation 4using the appropriate L0values and midazolam and GABA values varied to simulate experimental conditions.
Nonlinear least-square fits and model calculations were performed using Origin 6.0 software (Microcal Inc., Northampton, MA). Fitted values are given as mean ± SD.
The Student t test was used to statistically establish minimally active concentrations of benzodiazepines and to compare responses elicited by P4S and those elicited by P4S plus midazolam. The Wilcoxon test was applied to compare GABA EC50s in the absence versus presence of midazolam. A P value of less than 0.05 was considered to be statistically significant.
Constitutive Activity in GABAAReceptors
Oocytes expressing α1L264Tβ2γ2LGABAAreceptors displayed high holding currents when voltage clamped at −50 mV in ND-96. Picrotoxin (2 mm) caused an apparently outward current in oocytes expressing α1L264Tβ2γ2LGABAAreceptors (fig. 1A), which represents inhibition of the current through constitutively active chloride channels.10,18Higher concentrations of picrotoxin did not further inhibit mutant channel activity. For all oocytes tested, the picrotoxin-sensitive leak current (IPTX) averaged 11 ± 3.9% (n = 15) of the maximal inward current evoked by 1 mm GABA (ImaxGABA).
Direct Activation of α1L264Tβ2γ2LGABAAReceptors by Classic Benzodiazepine Agonists
In the absence of GABA or other orthosteric agonists, diazepam reversibly induced inward currents in oocytes expressing α1L264Tβ2γ2LGABAAreceptors (fig. 1B). Diazepam-induced current magnitudes were concentration dependent. In some oocytes, 1 nm diazepam caused a detectable deviation from basal activity, but on average, this was not significantly different from holding current (P = 0.19), whereas at 10 nm (P = 0.001) and at all higher concentrations studied (P < 0.001), diazepam consistently evoked inward currents. Midazolam also directly activated α1L264Tβ2γ2LGABAAreceptors in a concentration-dependent manner (fig. 1C). At 10 nm (P = 0.003) and at higher concentrations (P < 0.001), midazolam consistently evoked inward currents.
Effects of FG7142 and Flumazenil on α1L264Tβ2γ2LGABAAReceptors
In the absence of agonist, the benzodiazepine inverse agonist FG7142 elicited apparently outward currents in oocytes expressing α1L264Tβ2γ2LGABAAreceptors (fig. 1D), indicating reduced channel activity (n = 8). Reduced mutant channel activation was consistently observed at 10 nm FG7142 (P = 0.03) and at higher concentrations (P < 0.001). Flumazenil weakly activated inward currents in oocytes expressing α1L264Tβ2γ2LGABAAreceptors (n = 8; fig. 1D). Flumazenil at 10 nm (P = 0.005) and higher concentrations (P < 0.005) consistently increased mutant channel activity. When oocytes expressing α1L264Tβ2γ2LGABAAreceptors were exposed to 10 μm flumazenil followed by 10 μm flumazenil plus 100 nm midazolam, flumazenil alone activated a small inward current, but no further activation was observed after the addition of 100 nm midazolam, which alone elicited more than three times the current stimulated by flumazenil (fig. 1E; n = 3). Therefore, flumazenil blocked midazolam activation of mutant channels.
Potency and Efficacy of Benzodiazepine Site Ligands in α1L264Tβ2γ2LGABAAReceptors
Figure 2Adisplays average leak-corrected normalized responses from direct benzodiazepine activation experiments in oocytes expressing α1L264Tβ2γ2LGABAAreceptors. The same raw data were used in equation 2(Materials and Methods) to derive estimated Popenvalues, shown in figure 2B. The major difference between figures 2A and Bis that the latter explicitly incorporates the spontaneous currents observed in experiments with mutant receptors. Data in both panels of figure 2were fitted with Hill equations (equation 1for fig. 2Aand equation 3for fig. 2B; Materials and Methods). The EC50s from Hill fits (n = 5 for each drug) were 40 ± 14 nm for diazepam and 17 ± 2.7 nm for midazolam. Maximal direct activation was seen at diazepam or midazolam concentrations of 1–10 μm. Maximum average activation by diazepam was 9% of ImaxGABA, whereas maximum activation by midazolam was approximately 16% of ImaxGABA(fig. 3).
Direct effects of both FG7142 and flumazenil on mutant receptor currents were concentration dependent, but because of their low efficacy, EC50s and Hill slopes for these drugs could not be accurately determined. Maximal effects of both FG7142 and flumazenil were seen at concentrations of 1–10 μm. For FG7142, the average maximum outward current was 1.8% of ImaxGABA, whereas flumazenil elicited maximal inward currents that were approximately 2% of ImaxGABA(fig. 3).
Leftward Shifts by Midazolam in Wild-type and Mutant Receptors
Midazolam modulation of α1β2γ2LGABAAreceptors was assessed by measuring leftward shifts in GABA concentration responses. Results from individual oocytes expressing wild-type receptors displayed variable GABA EC50s, averaging 66 ± 36.2 μm (± SD; n = 14). In the presence of 10 nm midazolam, a consistent and significant (P = 0.043) leftward shift of the GABA response curve was observed. An example for one oocyte is shown in figure 4(open symbols). The average wild-type EC50MDZ/EC500ratio at 10 nm midazolam was 0.77 ± 0.10 (n = 5). Left-shift ratios (EC50MDZ/EC500) at 100 nm (n = 3) and 1 μm (n = 6) midazolam averaged 0.50 ± 0.055 and 0.42 ± 0.10, respectively.
To compare midazolam modulation of α1β2γ2Land α1L264Tβ2γ2LGABAAreceptors, we also assessed leftward shifts in the presence of 10 nm midazolam in oocytes expressing α1L264Tβ2γ2LGABAAreceptors. In oocytes (n = 5) expressing mutant channels, GABA EC500was 1 μm ± 0.45 μm. In the presence of 10 nm midazolam, a significant (P = 0.04) leftward shift was consistently observed (fig. 4). The EC50MDZ/EC500ratio for mutant receptors averaged 0.67 ± 0.22. Therefore, in both wild-type and mutant GABAAreceptors, the leftward shifts caused by 10 nm midazolam were similar.
Midazolam Effects on P4S Efficacy in α1β2γ2LGABAAReceptors
The partial orthosteric agonist P4S elicited inward currents in oocytes expressing wild-type α1β2γ2LGABAAreceptors. Concentration-dependent activation by P4S showed an EC50of 290 ± 18 μm and a Hill slope of 0.94 ± 0.039 (n = 5; data not shown). There was less than a 5% increase in current response when P4S was increased from 3 to 10 mm. Therefore, 10 mm P4S (34 × EC50) was maximally activating in wild-type α1β2γ2Lreceptors and presumably occupies essentially all of the orthosteric agonist sites. Maximal P4S currents averaged 38 ± 3.4% (n = 3) of maximal currents elicited with 1 mm GABA. Currents elicited by 1 μm midazolam plus 10 mm P4S were significantly (P < 0.01) larger compared with peak currents elicited by 10 mm P4S alone (fig. 5). Midazolam increased maximal P4S efficacy by a factor of 2.5 ± 0.55 (n = 5).
We tested the hypothesis that ligand binding to the high-affinity benzodiazepine site formed at the interface of GABAAreceptor α and γ subunits is coupled to the channel open–closed gating equilibrium (gating allosterism). Electrophysiologic concentration–response data from peak currents in oocytes expressing α1β2γ2LGABAAreceptors were previously shown to mirror results obtained using smaller voltage clamped HEK293 cells, where currents were elicited with millisecond concentration jumps.10This is because α1β2γ2LGABAAreceptor currents desensitize slowly when the γ2subunit is overexpressed20and because we use relatively rapid solution exchange in our custom-built oocyte flow chamber.
Our electrophysiologic experiments assessed the effects of benzodiazepine site ligands on gating in receptors that either had no orthosteric ligand bound or were agonist bound. We found that the classic benzodiazepine agonists diazepam and midazolam directly enhance the open probability of constitutively activated mutant GABAAreceptors in the absence of orthosteric agonists, whereas the benzodiazepine inverse agonist FG7142 reduced spontaneous channel openings (figs. 1 and 3). In addition, flumazenil, which is classified as a competitive benzodiazepine antagonist, seems to be a very weak partial agonist at the benzodiazepine site. These data unambiguously indicate that benzodiazepine site ligands allosterically alter gating in the mutant GABAAreceptors in the absence of orthosteric agonists.
To address the mechanism of benzodiazepine action in agonist-bound receptors, we used saturating concentrations of the orthosteric partial agonist P4S in wild-type GABAAreceptors. We found that midazolam increases the maximal efficacy of P4S (fig. 5). Because GABAAreceptor single-channel conductance is unchanged by classic benzodiazepines13,14and orthosteric site occupancy cannot increase at saturating P4S concentrations, our result indicates that gating of P4S-bound α1β2γ2LGABAAreceptors is allosterically enhanced in the presence of midazolam.
Walters et al. 23reported that, in addition to its actions at the high-affinity α/γ site, very high concentrations (> 10 μm) of diazepam can enhance the function of GABAAreceptors via one or more low-affinity sites, which are not affected by flumazenil. We found that flumazenil completely blocks further direct activation of α1L264Tβ2γ2LGABAAreceptors by the full benzodiazepine agonist midazolam (fig. 1E). Therefore, the low affinity sites described by Walters et al. do not mediate direct benzodiazepine activation under the conditions used in our experiments. Both flumazenil and other benzodiazepine actions we observed in mutant receptors are due to interactions at the high-affinity α/γ interfacial site.
The leftward shifts in GABA concentration responses caused by 10 nm midazolam were similar in both wild-type and mutant receptors (fig. 4). These results indicate that sensitivity to benzodiazepines is comparable in wild-type receptors and those containing the α1L264T mutation. Furthermore, the efficacy of midazolam was similar in both unbound and agonist-bound GABAAreceptors. In α1L264Tβ2γ2Lreceptors, channel activity was increased approximately 160% above constitutive by 1–10 μm midazolam, indicating approximately a 2.6-fold increase in the open probability. In wild-type receptor currents, 1 μm midazolam also increased the efficacy of P4S approximately 2.5-fold.
Allosteric coagonism, as depicted in figure 6A, is a mechanism that elegantly accounts for the gating effects by benzodiazepines in both unbound and agonist-bound GABAAreceptors. In this equilibrium scheme, based on classic MWC allosterism,22there are only two types of receptors, open and closed, and both GABA binding and benzodiazepine binding are coupled to the open–closed equilibrium via differential binding affinities. The scheme is highly constrained and defined by only five equilibrium parameters: L0(the basal equilibrium between closed [R] and open [O] states), KG(the closed state microscopic dissociation constant for GABA binding to its two equivalent orthosteric sites), c (GABA efficacy, defined as the ratio of dissociation constants in open vs. closed receptors), KBZ(the closed state microscopic dissociation constant for benzodiazepine binding), and d (benzodiazepine efficacy). In this and other MWC mechanisms, agonists are ligands that selectively bind to the open state and therefore have efficacy values less than 1.0, whereas inverse agonists bind more tightly to the closed state (efficacy > 1.0), and competitive antagonists have efficacy = 1.0.
We10and others18have previously shown that common GABA binding (KG) and efficacy (c) values but different L0values can be used to nearly quantitatively account for the GABA concentration responses in both wild-type and α1L264Tβ2γ2Lreceptors. Based on the equal left shifts seen at 10 nm midazolam (fig. 4), we also surmised that the underlying benzodiazepine binding (KBZ) and efficacy (d) parameters are common to both wild-type and mutant receptors. We estimated binding and efficacy parameters for diazepam and midazolam by fitting equation 4(with GABA = 0; Materials and Methods) to estimated Popenvalues derived from direct activation of α1L264Tβ2γ2Lreceptors (fig. 2B). In these fits, L0mutwas constrained to a value of 9.1, based on the average ratio of maximal GABA-activated current to picrotoxin-inhibited current in oocytes expressing α1L264Tβ2γ2LGABAAreceptors (9.1 ± 3.0). These fits are shown in figure 6B, and the fitted parameters are reported in the legend of figure 6. Our analysis with the MWC mechanism suggests that midazolam is an agonist that binds approximately threefold more tightly to open versus closed receptors at its single high-affinity site (d = 0.34), whereas diazepam shows only a twofold preference for open versus closed receptors (d = 0.5).
Previous estimates of the basal gating equilibrium constant L0for wild-type α1β2γ2Lreceptors are in the range of 0.5–1 × 105.10,18,19Based on the predicted inverse square-root relation between L0and EC50for mutations such as α1L264T that selectively alter receptor gating (equation 5, Materials and Methods),9,19,24we calculated L0wtfrom the L0mut(9.1) and GABA EC50s for wild-type (66 μm) and mutant (1 μm) receptors. The resulting L0wtvalue of 0.4 × 105is close to the previous estimates and indicates that the basal open probability of unliganded wild-type receptors [P0= (L0+ 1)−1] is approximately 2.5 × 10−4.
To provide rough KGand c values, we fitted equation 4(with BZ = 0; Materials and Methods) to wild-type estimated Popenvalues from figure 4B. The fit is shown as the solid line in figure 6C. Altogether, we derived six parameters for our model from different subsets of our data: the distinct L0parameters for the mutant and wild-type models and the four global parameters KMDZ, d, KG, and c.
The MWC coagonist model was then used to predict the impact of midazolam on GABA concentration responses. First, we calculated the wild-type GABA concentration response in the presence of 10 nm midazolam (fig. 6C, wild type–midazolam, dashed line). In close agreement with experimental results, the model predicts an EC50MDZ/EC500ratio of 0.7 at 10 nm midazolam and an EC50ratio near 0.5 at both 100 nm and 1 μm midazolam (fig. 6D). The specific model shown in figure 6Cis derived from GABA concentration–response data in a single oocyte, whereas the leftward shift induced by midazolam in the model is entirely independent of the GABA binding and efficacy parameters. In the presence of high midazolam concentrations, the wild-type model Popen= (L0d + 1)−1≈ 7.1 × 10−5, which is nearly a tripling of the basal spontaneous P0but is too low to measure with macrocurrent electrophysiologic methods, which have a maximal sensitivity range (signal/noise) of approximately three orders of magnitude. Thus, the lack of evidence for direct activation in wild-type synaptic GABAAreceptors is a predictable consequence of the low efficacy of benzodiazepine gating effects.
When L0mutwas used in the coagonist model (fig. 6C), it correctly simulated both the low GABA EC50and the significant spontaneous activity of α1L264Tβ2γ2Lreceptors (fig. 6C, mutant–control, dashed line). The mutant model with 10 nm midazolam (fig. 6C, mutant–midazolam, dashed line) also correctly predicts both enhanced opening probability at 0 GABA (i.e. , direct midazolam activation) and a decrease in EC50to 0.7 × control. Therefore, despite the highly constrained nature of this simple MWC coagonist model, it is able to closely simulate our results in both the absence and the presence of GABA, for both wild-type and mutant channels, using four global parameters and a single additional basal gating parameter for each type of receptor.
The MWC model does not simulate the GABA concentration responses of mutant receptors as well as it does for wild-type receptors (fig. 6C). Specifically, the model predicts that the Hill slope of the GABA concentration responses for the mutant should be between 1.3 and 1.6, whereas our oocyte experiments consistently demonstrated much lower slopes near 0.8. Reduced Hill slopes could be due to rapid desensitization or inhibitory effects at high GABA, or negative GABA binding cooperativity. Currents elicited from rapidly superfused cell membrane patches expressing α1L264Tβ2γ2LGABAAreceptors desensitize very slowly19and do not show surge currents at high GABA, making the first two possibilities unlikely. Adding a variable parameter for GABA binding cooperativity would enable our model to fit the mutant Hill slope better, but this modification would not alter our conclusions regarding the mechanism of benzodiazepine actions. Similarly, the model only addresses the positive gating effects of benzodiazepines and therefore does not explain the surge currents observed with high diazepam (fig. 1B), which indicate the presence of a low-affinity inhibitory action.
We have recently described a similar MWC coagonist mechanism for R (+)-etomidate, a potent intravenous anesthetic that acts selectively at GABAAreceptors.10We found that etomidate modulation was consistent with the presence of two equivalent sites on α1β2γ2LGABAAreceptors, each with a dissociation constant of 36 μm and an efficacy factor of 0.0077 (a 130-fold open-state preference). As a result of its two sites and greater efficacy, etomidate enhances GABA-elicited responses much more than midazolam or diazepam, and high etomidate concentrations can directly activate wild-type GABAAreceptors. Comparing the distinct coagonist sites for etomidate and classic benzodiazepines on α1β2γ2LGABAAreceptors, the benzodiazepine site shows high affinity and low efficacy, whereas the etomidate sites display low affinity and high efficacy.
Our results agree with previous data supporting a gating effect for classic benzodiazepines. One previous study17has shown that maximal efficacy of a partial GABAAreceptor agonist, 5-(4-piperidyl)isoxazol-3-ol (4-PIOL), is enhanced by benzodiazepine agonists. Thompson et al. 25described another constitutively active GABAAreceptor mutant that is directly activated by benzodiazepine agonists via a site where flumazenil competes for occupancy. Benzodiazepine enhancement of extrasynaptic “tonic” GABAAreceptor currents in cultured neurons has also been reported.26,27It has been suggested that these tonic neuronal chloride currents are stimulated by GABA spillover from synaptic release,27but direct gating by benzodiazepines at spontaneously active receptors provides an alternative explanation for these observations. Several reports have previously concluded that flumazenil acts as a weak partial benzodiazepine agonist.28–30Boileau and Czajkowski31previously suggested that benzodiazepines act as coagonists, based on indirect structure–function evidence. Serfozo and Cash32suggested that benzodiazepine agonists may promote gating of GABAAreceptors bound by only one GABA molecule. The MWC coagonist mechanism implies that enhancement is independent of the number of GABA molecules bound.
Given the low impact on the gating equilibrium by benzodiazepines in comparison to general anesthetics such as etomidate or pentobarbital, it is not surprising that studies comparing their actions on GABAAreceptors have concluded that different coupling mechanisms were at work. That general anesthetics stabilize GABAAreceptor open states is supported by evidence that these drugs enhance the efficacy of partial orthosteric agonists,10–12lengthen the average single-channel open time of receptors,33,34and significantly slow deactivation of patch macrocurrents.35,36Single-channel studies of benzodiazepines on neuronal GABAAreceptors did not detect any increase in average channel lifetime.13,37Heterogeneous GABAAreceptor species in neurons makes interpreting these results difficult. Moreover, with GABA present at low concentrations that likely favor occupation of one GABA site (i.e. , the RG state in fig. 6A), benzodiazepines may primarily promote opening directly from this state (via the OGBz state), which might be characterized by brief openings. This idea is supported by other experiments32,38and could explain why single GABAAreceptor channels seem to open more frequently in the presence of benzodiazepine agonists but without increased mean open times.
Kinetic studies using rapid patch superfusion have given conflicting results on whether classic benzodiazepine agonists prolong deactivation. Krampfl et al. 15studied α1β2γ2Lreceptors in excised patches from HEK293 cells and observed slowed deactivation in the presence of diazepam. A similar study by Lavoie and Twyman16using α2β1γ2receptors reported no slowing of deactivation by diazepam. In another study of α2β1γ2receptors expressed in HEK293 cells, O’Shea et al. 11reported that midazolam produced a leftward shift of P4S responses without increasing maximal response, in contrast with our results and others17using α1β2γ2L. These discordant observations suggest that benzodiazepines may couple to both GABA binding and gating steps and that allosteric coupling to the benzodiazepine site may differ in α1β2γ2versus α2β1γ2receptors. The data for α1β2γ2receptors consistently favors a gating mechanism, and the ability of our MWC coagonist model to accurately predict benzodiazepine effects under a variety of experimental conditions suggests that binding allosteric interactions between the GABA and benzodiazepine sites are minimal. Further studies are needed to determine whether the mechanism of benzodiazepine modulation is altered for other GABAAreceptor subunit mixtures.
In summary, our results demonstrate that high-affinity classic benzodiazepines modulate α1β2γ2LGABAAreceptors via allosteric coupling to channel gating. An MWC allosteric coagonist mechanism quantitatively accounts for the modulation of GABA-activated currents in both wild-type receptors and spontaneously active mutant receptors. The same mechanism also accounts for the direct activation of spontaneously active GABAAreceptors by benzodiazepine agonists and the lack of similar observations in wild-type synaptic receptors. Classic benzodiazepines are high-affinity but low-efficacy GABAAreceptor coagonists, in contrast to many general anesthetic compounds that act on the same receptors with low affinity but high efficacy.
The authors thank Carol Gelb (Senior Laboratory Technician, Department of Anesthesia and Critical Care, Massachusetts General Hospital, Boston, Massachusetts) for technical support.