Sedative-hypnotic drugs such as diazepam (a benzodiazepine) can impose graded effects on the function of the central nervous system (CNS), resulting in a spectrum of clinical actions. This range of therapeutic actions progresses from sedation at low doses, to induction of anesthesia at significantly higher doses. BZs act at nanomolar concentrations to modulate chloride ion flux through GABAA-receptor channels, thereby affecting synaptic transmission1,2,3. The crucial amino acids responsible for this potent action of BZs in part have been determined, and are located within the N-terminal domains of α and γ subunits1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16. The importance of two α subunit isoforms (α1 or α2) in mediating sedative or anxiolytic action of diazepam, respectively, has been demonstrated using transgenic animals17,18,19,20.

Certain BZs such as diazepam, midazolam and lorazepam are also known for their anesthetic properties21 and are routinely used in surgical procedures. However, in comparison to the extensive studies on the high-affinity action of BZs, little is known about the molecular determinants of the anesthetic action of BZs.

In this study, we have demonstrated that diazepam, in the presence of low GABA concentrations, produces two distinct components of potentiation within the nanomolar (nM) and micromolar (μM) concentration ranges for the α1β2γ2-receptor channel. In contrast to the nM component, the μM component of diazepam action is independent of the γ2 subunit and is insensitive to the selective BZ antagonist flumazenil. Mutations at equivalent residues within the TM2 of α, β and γ subunits, proven important for the action of other anesthetics, nullify the μM component without significantly affecting the nM component of response to diazepam. Furthermore, the μM action of diazepam can also be conferred upon the BZ-insensitive ρ1-receptor channel by converse mutation of the corresponding TM2 residue and a TM3 residue. The physiological significance of the μM action of diazepam is discussed in the light of the dynamics of benzodiazepine concentrations, the persistence of low concentrations of GABA and the tonic level of GABAA receptor activity within the CNS.


Biphasic potentiation of α1β2γ2 by diazepam

Complementary RNA (cRNA) from rat wild-type α1, β2 and γ22s)22 subunits were co-injected into Xenopus laevis oocytes. Three days after injection, GABA-elicited currents were recorded in the absence and presence of a broad range of diazepam concentrations (2 to 150,000 nM) using the two-electrode voltage-clamp technique. Figure 1a and b depicts current traces and the concentration–response relationship for diazepam at EC3, EC8 and EC16 (effective concentration giving rise to 3%, 8% and 16% of the maximal GABA current, respectively) GABA for the α1β2γ2-receptor channel. These data reveal two different patterns of response to diazepam, which are dependent on the GABA concentration. At all three GABA concentrations tested, diazepam produced a two- to three-fold potentiation of the GABA-evoked currents, with EC50 values between 32 and 54 nM, and a Hill coefficient approaching unity (nM component, Table 1). On the other hand, at EC3 and EC8 GABA, μM concentrations of diazepam (20 μM and above) evoked a second component of potentiation, further increasing GABA-elicited currents at EC3 from three-fold (the nM component) to approximately eight-fold. At EC3 GABA, the μM component of diazepam had an apparent EC50 of 63 ± 8.5 μM and a notably greater Hill coefficient (3.04 ± 0.37) than that of the nM component (Table 1). The extent of the μM diazepam-induced potentiation was markedly dependent upon the GABA concentration; potentiation was diminished significantly at EC8 GABA, and eventually abolished at EC16 GABA.

Figure 1: Diazepam modulation of wild type α1β2γ2- or α1β2-receptor channels.
figure 1

(a, b) The current traces and concentration–response relationship for diazepam at EC3 (3.2 μM), EC8 (4.6 μM) and EC16 (7.75 μM) GABA for the α1β2γ2-receptor channel. The presence of the diazepam μM component of potentiation is highly dependent on GABA concentration. (c, d) The diazepam concentration– response relationship (10 to 150,000 nM) and current traces at EC3 GABA (0.2 μM) for the α1β2-receptor channel. Diazepam at μM concentrations (but not at nM concentrations) elicited a significant potentiation for the α1β2-receptor channel.

Table 1 Parameters obtained from fitting the Hill equation to GABA, diazepam and flurazepam data.

Monophasic potentiation of α1β2 by diazepam

Co-expression of cRNA for α1 and β2 subunits also yielded GABA-gated receptor channels. The α1β2-receptor channel can be distinguished from the α1β2γ2-receptor channel due to α1β2's unique pharmacological and kinetic characteristics. For example, the α1β2-receptor channel is more sensitive to GABA in comparison to the α1β2γ2-receptor channel (Table 1). Figure 1c and d shows the diazepam concentration–response relationship (10 to 150,000 nM) at EC3 GABA for the α1β2-receptor channel. Diazepam elicited only a single component of potentiation within the μM concentration range for the α1β2-receptor channel, increasing the control EC3 GABA currents by more than twelve-fold (Fig. 1c and d). Furthermore, the EC50 and the Hill coefficient values for the μM component of diazepam action were similar for both α1β2 and α1β2γ2-receptor channels (at EC3 GABA, Table 1). The lack of α1β2-receptor channel sensitivity to diazepam within the nM concentration range has been documented previously, as the presence of the γ subunit is essential in conferring diazepam's nM action11.

Zn2+ does not block the μM action of diazepam

Zinc inhibits the GABA-evoked currents from α1β2- but not from α1β2γ2-receptor channels23,24 (Fig. 2). Zinc exhibited properties of an open channel blocker for α1β2, inhibiting the GABA-evoked currents back to the baseline. In comparison, the GABA currents for α1β2γ2-receptor channels were reduced to 64 ± 4% of the control in the presence of 10 μM Zn2+, suggesting that a fraction of the total GABA current may have arisen from α1β2-type-receptor channels. To confirm that the μM potentiation of α1β2γ2 at EC3 GABA was not due to the presence of α1β2-receptor channels (see above), the α1β2γ2 diazepam concentration–response relationship at EC3 GABA was determined in the presence of 10 μM Zn2+ (Fig. 2b). In these experiments, the characteristics of α1β2γ2's biphasic potentiation by diazepam were unaltered in the presence of Zn2+ (Table 1). For the nM and the μM components, the EC50/Hill coefficient values obtained in the presence of Zn2+ were 0.048 ± 0.003/1.23 ± 0.05 and 60.5 ± 4.7/2.97 ± 0.25, respectively. Moreover, the absolute magnitudes of potentiation for the two components were 3.44 ± 0.26 for the nM and 4.34 ± 0.26 for the μM component of diazepam action. These values are similar to the values determined for the α1β2γ2-receptor channel in the absence of Zn2+ (Table 1). These data indicate that the significant potentiation observed at μM concentration of diazepam for α1β2γ2 (at EC3 GABA) does not arise from the concomitant presence of α1β2-receptor channels.

Figure 2: Diazepam biphasic potentiation of the α1β2γ2-receptor channel is unaltered in the presence of zinc.
figure 2

(a, b) Differential effect of Zn2+ (10 μM) on the α1β2- versus α1β2γ2-receptor channel. (a) The GABA-evoked currents (3.2 μM) for the α1β2-receptor channel are completely blocked by 10 μM Zn2+. (b) The current traces for the diazepam concentration– response relationship for the α1β2γ2-receptor channel in the presence of Zn2+ (10 μM). Although Zn2+ partially blocks the GABA-evoked currents for the α1β2γ2-receptor channel (see current traces in b), it does not significantly affect the biphasic potentiation of the α1β2γ2-receptor channel by diazepam at EC3 GABA.

The μM action of diazepam is insensitive to flumazenil

Flumazenil is a selective antagonist for the high-affinity (nM) component of BZ action upon the GABAA-receptor channel25. For the α1β2γ2-receptor channel at EC3 GABA, flumazenil at concentrations of 20 μM (or 10 μM, data not shown) abolished the nM, but not the μM, component of diazepam action (Fig. 3). The overall potentiation at the μM component was reduced from approximately 8 to 3.31 ± 0.59. However, the EC50 and the Hill coefficient values derived for the diazepam μM component in the presence and absence of flumazenil were similar for the α1β2γ2-receptor channel at EC3 GABA (Table 1). Concentrations of flumazenil up to 100 μM were co-applied with 100 μM diazepam (at EC3 GABA) to test the effect of flumazenil at equivalent concentrations on μM potentiation by diazepam. In the presence of such high concentrations of flumazenil, 100 μM diazepam also produced a significant potentiation (data not shown). These experiments suggest that the nM and μM actions of diazepam are mediated through distinct mechanisms.

Figure 3: Flumazenil abolishes the nM but not the μM component of diazepam action upon the α1β2γ2-receptor channel.
figure 3

Top, current traces for diazepam modulation of GABA currents in the presence of 20 μM flumazenil. The nM but not the μM component of diazepam action is eliminated by the co-application of flumazenil (bottom).

TM2 mutations abolished the μM action of diazepam

The diazepam concentration–response relationships for αS267IβN265IγS280I at EC3, EC8 and EC16 GABA were next determined. At all three GABA concentrations tested, diazepam produced only a single component of potentiation within the nM concentration range (Fig. 4b). The EC50 and Hill coefficient values derived for αS267IβN265IγS280I (32 to 55 nM and 1.1 to 0.8 nM, respectively; Table 1) were similar to the values determined for the nM component of wild-type α1β2γ2 (Table 1). The αS267IβN265IγS280I-receptor channel experiments further demonstrate that the nM and the μM actions of diazepam are distinct and separable.

Figure 4: Diazepam modulation of the αS267IβN265IγS280I-receptor channel.
figure 4

(a) Amino-acid sequence comparison for the TM2 domains of α1, β2, γ2 and ρ1 subunits. The mutated residues are shown in boxes. (b) The current traces and the concentration–response relationship for diazepam, for the αS267IβN265IγS280I-receptor channel at EC3 (0.09 μM), EC8 (0.17 μM) and EC16 (0.35 μM) GABA. At all three GABA concentrations examined, diazepam produced a single component of potentiation occurring within the nM concentration range. The line is derived from the fit of the Hill equation to the data up to 5 μM diazepam.

In comparison to the α1β2γ2,-receptor channel, the αS267IβN265IγS280I-receptor channel exhibited an approximately 40-fold higher sensitivity to GABA (Table 1). This difference in sensitivity indicates the importance of these TM2 residues in the GABA-dependent modulation of the α1β2γ2-receptor channel.

Imparting diazepam sensitivity to ρ1-receptor channels

ρ1-receptor channels are insensitive to BZs and other intravenous anesthetics26 (excepting neuroactive steroids27). This property of the homo-oligomeric ρ1-receptor channel has been used to study the molecular basis of anesthetic action by conferring anesthetic sensitivity upon ρ1. For example, the specific mutation of residue 307 within TM2 or residue 328 within TM3 of the ρ1 subunit (Fig. 5a) can impart pentobarbital sensitivity upon the ρ1-receptor channel28,29. Within the ρ1 subunit, residues 307 and 328 were substituted, either singly or in combination, to their amino acid counterparts present within the α1 (Ser, TM2) and β2 (Met, TM3) subunits. Diazepam did not modulate the GABA-elicited currents at EC4 in wild-type ρ1-, ρI307S- and ρW328M- receptor channels (Fig. 5b). In contrast, doubly mutated ρ1 became highly sensitive to diazepam (Fig. 5c and d). The EC4 GABA currents for ρI307S/W328M were markedly potentiated in the presence of diazepam, resulting in an increase of GABA currents up to 9-fold (20 μM diazepam). Figure 5e and f show the potentiation of GABA currents by diazepam (10 μM) at EC1, EC4, EC50 and EC99 GABA for ρ1, ρI307S, ρW328M and ρI307S/W328M mutants. For ρI307S/W328M, diazepam potentiation was most prominent at GABA concentrations below the EC50 value. The fractional potentiation by diazepam (10 μM) was greatest at low GABA concentrations (EC1 5.40 ± 0.50 and EC4 3.20 ± 0.50), whereas at higher GABA concentrations (EC99), little potentiation was observed. In comparison, ρ1, ρW328M and ρI307S did not respond to diazepam at any of the GABA concentrations tested.

Figure 5: Imparting diazepam sensitivity to the α1-receptor channel by double mutation.
figure 5

(a) The position of the crucial amino acids within the putative second and third transmembrane domains. (b) GABA-evoked currents at EC4 from ρ1 (0.27 μM GABA), ρW328M (0.35 μM GABA) or ρI307S (0.11 μM GABA) are insensitive to diazepam concentrations up to 30 μM. (c, d) GABA currents (EC4 = 0.04 μM) from the ρI307S/W328M are markedly potentiated by μM concentrations of diazepam. (e, f) Poten-tiation of GABA currents by diazepam (10 μM) at EC1, EC4, EC50 and EC99 GABA for ρ1, ρI307S, ρW328M and ρI307S/W328M mutants. The potentiation of the ρI307S/W328M-receptor channel is dependent on the GABA concentrations. ρ1, ρW328M and ρI307S do not respond to diazepam across the range of GABA concentrations tested (relative to their EC50 values).

The co-application of 20 μM flumazenil and 20 μM diazepam at EC4 GABA did not significantly alter the diazepam-induced potentiation of the GABA currents elicited from ρI307S/W328M (data not shown). These data indicate that the μM action of diazepam on ρI307S/W328M and its nM action on the α1β2γ2-receptor channel are intrinsically different.

Diazepam at higher concentrations is an agonist for ρI307S/W328M (Fig. 6). The EC50 value for the direct action of diazepam was 93 ± 1.9 μM associated with a high Hill coefficient (3.3 ± 0.18, Table 1). Diazepam was highly efficacious relative to GABA, eliciting a relative maximum current of 0.8. In comparison, wild-type ρ1 and singly mutated ρ1 were not directly activated by diazepam at concentrations up to 300 μM.

Figure 6: The diazepam concentration– response relationship, and diazepam's relative efficacy with respect to GABA in its direct activation of ρI307S/W328M.
figure 6

Diazepam is an agonist for the ρI307S/W328M-receptor channel. (a) Current traces showing the direct activation of ρI307S/W328M by 30 to 300 μM diazepam. (b) Concentration– response relationship for diazepam. Diazepam activation of the ρI307S/W328M-receptor channel appears to be highly cooperative (high Hill coefficient, Table 1). (c) Diazepam was highly effective relative to GABA (current magnitude at 300 μM diazepam divided by current magnitude at 10 μM GABA; EC50 GABA for ρI307S/W328M, 0.095 ± 0.003 μM) for direct activation of the ρI307S/W328M-receptor channel. In comparison, diazepam alone, at concentrations up to 300 μM, did not evoke a significant response in either wild-type or singly mutated ρI.

Differential sensitivity of ρI307S/W328M to different BZs

The structure–function relationships of different BZs were assessed using the ρI307S/W328M model. Figure 7 shows the current traces and the relative efficacies of diazepam, midazolam, flunitrazepam and flurazepam in potentiating GABA responses (EC4) of the ρI307S/W328M-receptor channel. Diazepam was the most efficacious among the BZs tested. Nevertheless, midazolam and flunitrazepam also elicited a marked potentiation, increasing GABA currents by approximately 0.5-fold and 3-fold at 5 and 20 μM concentrations, respectively. In contrast, flurazepam at concentrations up to 20 μM did not significantly alter the GABA-evoked currents from the ρI307S/W328M-receptor channel.

Figure 7: The relative effectiveness of diazepam, midazolam, flunitrazepam and flurazepam in potentiating the GABA responses (EC4) of the ρI307S/W328M-receptor channel.
figure 7

Diazepam is the most effective among the benzodiazepines tested. In comparison, flurazepam up to 20 μM concentration did not potentiate the GABA-evoked currents for the ρI307S/W328M-receptor channel.

At EC3 GABA, the flurazepam concentration–response relationship was determined for the wild-type α1β2γ2-receptor channel to examine whether flurazepam could also produce a biphasic potentiation of GABA currents similar to that observed for diazepam (Fig. 8). In contrast to diazepam, flurazepam elicited only a single component of potentiation with an EC50 of 501 ± 96 nM and a Hill coefficient of 0.89 ± 0.06 (Table 1). Thus, at μM concentration, flurazepam neither produced a μM component of potentiation for α1β2γ2, nor augmented the GABA-evoked current from the ρI307S/W328M-receptor channel. Diazepam and midazolam, which possess a μM action upon ρI307S/W328M, are used routinely to induce and maintain anesthesia on their own, but flurazepam is not21.

Figure 8: Flurazepam elicits only a single component of potentiation within the nM concentration range.
figure 8

Note that the flurazepam induces inhibition at μM concentrations. The line is derived from the fit of the Hill equation to the data up to 20 μM flurazepam.

However, at higher concentrations (20 μM and above), flurazepam produced a significant inhibition of the current (Fig. 8) similar to the inhibitory effect observed at equivalent concentrations of diazepam for αS267IβN265IγS280I (Fig. 4) and wild-type α1β2γ2 (between 5 and 20 μM, Fig. 1). Previous studies have shown that impairing the high-affinity (nM) component of diazepam action within the α1β2γ2 by mutation of a single residue within the N-terminal domain of the α subunit causes the μM component to become more prominent5. On the other hand, eliminating the diazepam μM component (or its absence in the case of flurazepam) seems to reveal an inhibitory phase (which is otherwise masked by μM concentrations of diazepam; Figs. 4 and 8, respectively). The co-existence of an inhibitory phase associated with the nM component, as well as the apparent absence of an inhibitory phase following the abolition of the nM component5, suggests a close association between the diazepam inhibitory action and the nM component. BZs can enhance the desensitization of GABA currents30,31. It is thus possible that the inhibitory action observed here is a reflection of this induced desensitization upon the GABA current.


The nM and the μM components of diazepam action within the α1β2γ2-receptor channel have intrinsic differences. First, the nM component of diazepam action depends on the presence of the γ subunit. However, both α1β2- and α1β2γ2-receptor channels are sensitive to μM concentrations of diazepam. Second, for the α1β2γ2-receptor channel, flumazenil inhibits the nM but not the μM component of potentiation by diazepam. Third, there is a difference of three orders of magnitude in potency between the two components. Fourth, mutation of specific TM2 residues within the α1, β2 and γ2 subunits abolishes the μM but not the nM effect of diazepam. Fifth, the action of diazepam in eliciting the μM component of both α1β2γ2 and α1β2 receptors appears highly cooperative as inferred from the Hill coefficient. Diazepam's direct action upon the ρI307S/W328M-receptor channel and pentobarbital's direct action on ρW328M (ref. 28) also have a high Hill coefficient (Table 1), whereas the nM component persistently displays a low cooperativity (Hill coefficient 1). Sixth, there is a correlation between lipid solubility and the extent of the μM potentiation for the BZs tested. For example, diazepam, which has the highest octanol:buffer partition coefficient32 among the BZs tested, has the greatest efficacy for the ρI307S/W328M-receptor channel. In comparison, no such relationship has been shown for the nM component. These differences provide strong evidence that the nM and the μM component of diazepam action are mediated via distinct and separable mechanisms.

The aforementioned transmembrane residues (αS267 and βM286 and the equivalent residues within ρ1) are centrally important for a number of anesthetic compounds28,29,33,34,35,36,37. Also, mutation of the equivalent residue (βN265) within human β2 or β3 subunits (to a Ser) of αβγ is sufficient to impair the action of loreclezole, a broad-spectrum anticonvulsant that acts at μM concentrations38. It is intriguing that so many structurally diverse and hydrophobic compounds mediate their action through these common residues.

The prevalent activation scheme for GABAA receptors postulates two open states39. In this scheme, one open state is derived from receptor channels occupied by a single GABA, whereas the other open state arises from doubly bound receptor channel. From the simple law of mass action, the proportion of the singlybound form of α1β2γ2-receptor channels may predominate at EC3 and EC8 GABA. Thus, the μM component of potentiation by diazepam could be the result of selective enhancement of currents from α1β2γ2-receptor channels, which are present in a mono-liganded state.

It has been suggested that the incremental CNS effects of benzodiazepines, from sedation to the induction of anesthesia, are all consequences of the relative occupancy of the benzodiazepine nM receptor. We postulate that the more marked actions of benzodiazepine upon the CNS, such as induction of anesthesia, could be the result of its μM action upon GABAA-receptor channels. Three observations suggest this view. First, the nM action of a benzodiazepine such as diazepam is dependent upon the isoform of the α subunit expressed as well as the presence of the γ subunit, and therefore, its nM activity is limited to only certain subtypes of GABAA receptors. On the other hand, diazepam's μM action seems less specific (since both α1β2- and α1β2γ2-receptor channels are sensitive to μM concentrations of diazepam) and thus may occur at all GABAA receptors. Second, diazepam produces the second component of potentiation only at low concentrations of GABA. Within the CNS, a low extracellular concentration of GABA has been documented40,41. In addition, a tonic level of functional GABAA channel activity in a number of CNS regions has been shown42,43. Third, concentrations of benzodiazepines can reach μM levels within the plasma44, and given their high lipid solubility, their concentrations can reach even higher levels within the CNS45. Finally, the additional chloride flux due to the μM effect upon GABAA-receptor channels can be predicted to dramatically increase diazepam's overall CNS depressant action. Therefore, under circumstances in which a large dose of diazepam is administered, the marked and ubiquitous action of diazepam at μM concentrations may be central in producing a state of general CNS depression, such as the induction of anesthesia.


Oligonucleotide-mediated site-directed mutagenesis was done according to the manufacturer's protocol (Altered Sites, Promega, Madison, Wisconsin). Successful mutagenesis was verified by DNA sequencing. The cDNAs were linearized with HpaI (for ρ1) or SspI (for α12, or γ2 subunits) leaving a 3′ tail several hundred base pairs long. These additional 3′ sequences may increase cRNA stability within the oocyte. The cRNA was transcribed from the linearized cDNAs by standard in vitro transcription procedures (Megascript, Ambion, Austin, Texas).

Xenopus laevis (Xenopus I, Ann Arbor, Michigan) were anesthetized by hypothermia, and oocytes were surgically removed and placed in oocyte Ringer (OR2) that consisted of 82.5 mM NaCl, 2.5 mM KCl, 10 mM HEPES, 1 mM CaCl2, 1 mM MgCl2, 1 mM Na2HPO4, 50 U/ml penicillin, 50 μg/ml streptomycin, pH 7.5. Oocytes were dispersed and then incubated in OR2 without Ca2+ plus 0.3% collagenase A (Boehringer Mannheim, Indianapolis, Indiana) for approximately 1 h. After isolation, the oocytes were thoroughly rinsed with OR2, and stage VI oocytes were separated and maintained at 18°C.

Micropipettes for injecting cRNA were made on a Sutter P87 puller (Sutter Instrument Company, Novato, California), and the tips were cut with microscissors. The cRNA in diethylpyrocarbonate-treated water was drawn up into the micropipette with negative pressure and then injected into the oocytes by applying positive pressure using a Picospritzer II (General Valve Corporation, Fairfield, New Jersey). The oocytes were incubated in OR2 at 18°C for three days before the experiment. To ensure quality and to estimate the quantity of the cRNA, set dilutions of cRNA from mutants were electrophoresed on a 1% formaldehyde-containing agarose gel. The amount of cRNA was judged and matched by interpolation of lanes containing different dilutions of the corresponding cRNA. In addition, for nearly all mutants, two independent isolates were characterized and tested. The α1, β2 and γ2 subunits were co-injected into Xenopus laevis oocytes in the ratio of 1:1:1.8 to ensure the incorporation of γ2 subunits.

Three days after injection, oocytes were placed on a mesh machined in a small perfusing volume chamber (75 μl), with a t1/2 and clearance time of approximately 3 and 10 seconds, respectively. The chamber had an inlet in the top and an outlet in the bottom, which allowed continuous and rapid perfusion. Twenty separate reservoirs (100-ml glass containers) were connected to four six-way valves and the outlet of each of these six-way valves (the sixth position was continuous with the reservoir containing the control solution) was connected to one four-way valve. The outlet of the four-way valve led to the chamber. In this way, up to 20 different solutions could be bath-applied to an individual oocyte. Switching between the different solutions was controlled manually. The oocyte was continuously perfused with recording OR2 (without antibiotics and the 1 mM Na2HPO4, which was replaced with 1 mM NaCl) and switched to the test solution containing drug.

Recording microelectrodes were made with a Narishige PP-83 puller (Tokyo, Japan) and filled with 3 M KCl. Electrodes with resistances of 0.6–1.6 MΩ were used. A two-electrode voltage-clamp amplifier (Turbo TEC-05 npi, Adams and List, Westbury, New York) was used to record currents in response to the application of drugs. In all cases, membrane potential was clamped to −70 mV. Data were visualized on a Gould TA240 chart recorder (Valley View, Ohio) during the experiment and stored online using Pulse Fit or on a VCR using Instrutech 10b (Port Washington, New York) for subsequent off-line analysis.

The EC50 and Hill coefficients were estimated by fitting the concentration–response relationships to the Hill equation according to the following formula (Sigma plot, or software provided by D.S. Weiss).

Alternatively, the data containing two components of potentiation were fitted with a sum of two Hill equations.

Here, I is the peak current at a given concentration of agonist A, Imax is the maximum current, EC50 is the concentration of agonist yielding a half-maximal current, and n is the Hill coefficient.

The fractional potentiation was calculated using the following equation:

Here, IBZ is the measurement of GABA current in the presence of a given concentration of BZ, and the IGABA is the amplitude of the GABA current alone. In addition, a concentration of 70 nM BZ was bath-applied to the oocytes before collection of BZ data points.

The GABA concentration–response relationships were shifted to lower concentrations (two-fold) in oocytes that had high levels of expression. For this reason, oocytes used to collect the data were mostly limited to those that had less than 1 μA in maximal response to GABA. The EC values (for example, EC4) were first determined from the averaged concentration– response relationship. The obtained values were then empirically tested for each experiment by comparing the elicited GABA response at this concentration and the maximal response evoked by a GABA concentration equivalent to 20 × EC50. The values were then adjusted and retested accordingly until a true EC value was determined. The low EC values (for example, EC1) often had a standard error of 30% relative to expected values.

All BZs were purchased from Sigma, except for diazepam, which was purchased from Sigma or BIOMOL (Plymouth Meeting, Pennsylvania). Stock solutions of BZs at 100 mM were made in DMSO. Test solutions containing drugs were made by adding the BZ stock solutions to rapidly stirred recording OR2 or OR2 containing GABA. The presence of the vehicle solution DMSO at the maximum tested concentration (0.2%) did not alter the GABA-induced current from ρ1, α1β2γ2 or αS267IβN265IγS280I-receptor channels.

All data presented are the mean ± standard error (s.e.m.).