GABAA receptors (GABAA-R) are pentameric membrane proteins that belong to the superfamily of cys-loop ligand-gated ion channels (LGIC), which operate as GABA-gated Cl--selective channels. GABAA-R mediate most of the fast inhibitory neurotransmission in the CNS [13]. Initially, two subunits of the GABAA-R named α and β were purified [4, 5] and subsequently their cDNAs were isolated [6]. Twenty related GABAA-R subunits have been so far identified in mammals (α1–6, β1–4, γ1–3, δ, ε, π, θ, and ρ1–3 [7, 8]), yielding a high degree of potential diversity. If all of these subunits could randomly co-assemble, more than one hundred thousand GABAA-R subtypes with distinct subunit composition and arrangement would be formed [9]. The composition of the most abundant GABAA-R type in the CNS is αβγ, and immunohistochemistry studies suggest that receptors containing α1, β2/3 and γ2 subunits are the most widespread GABAA-R subtype in adult mammalian brain and represent about 50% of the total receptor pool [2, 10].

Typical αβγ GABAA-Rs harbor two agonist (GABA) binding sites located at the two α/β subunit interfaces [2, 11]. The function of GABAA-Rs can be modulated by various compounds acting at different allosteric sites located on GABAA-Rs. The benzodiazepine (BZD) site, which is located at an α/γ interface [12, 13], is the most frequently targeted site for therapeutic agents, and ligands that enhance GABAA-R function through positive modulation at this site possess anxiolytic, sedative, myorelaxant, anesthetic and amnestic properties [2, 3, 10, 14]. Based on pharmacological studies in transgenic mice, it has been proposed that GABAA-Rs can be classified into the following pharmacological classes according to the effects of BZ site ligands: α1-containing receptors (GABAA1) that mediate sedative effects; α2-containing receptors (GABAA2) that mediate anxiolytic effects; α3-containing receptors (GABAA3) that mediate myorelaxation; and α5-containing receptors (GABAA5) that are involved in learning and memory processes [7, 15, 16]. This classification is consistent with the sedative/hypnotic profile of GABAA1-preferring compounds such as zolpidem and zaleplon [17], but pharmacological studies in wild-type animals and in man have raised questions regarding the attribution of anxiolytic effects to GABAA2 receptors. In particular, a number of compounds have been identified that exhibit an anxioselective profile in vivo despite lacking the expected GABAA2 selectivity. A series of compounds with mixed preference for α23-containing receptors has been reported to produce robust anxiolysis in animals without noticeable sedation, including one compound that exhibits selectivity for α3-containing receptors [1821]. Other compounds, such as ocinaplon [22] and DOV 51,892 [23], are anxiolytic in humans and animals without undesired side effects such as sedation and myorelaxation, but do not exhibit strong selectivity among GABAA-Rs sensitive to benzodiazepines (that is, those receptors containing α1–3 and/or α5-subunits)

One hypothesis that could explain the anxioselective profile of ocinaplon is the presence of one or more biotransformation products that exhibit selectivity at GABAA2 receptors. To test this hypothesis, we characterized the pharmacological properties of the major biotransformation product of ocinaplon in dogs, rats and man, DOV 315,090 (Fig. 1), using in vitro radioligand binding and two-electrode voltage-clamp electrophysiology. We now report that like its parent compound, DOV 315,090 acts as a positive modulator at GABA receptors, and like its parent, does not exhibit marked selectivity among α1–3 and α5 containing receptors. Thus, while DOV 315,090 may contribute to the pharmacological actions of ocinaplon, the anxioselective profile of ocinaplon cannot be explained on the basis of enhanced subunit selectivity on the part of DOV 315,090.

Figure 1
figure 1

Structures of diazepam, ocinaplon and DOV 315,090.


Radioligand Binding Assays

HEK293 cells (CRL 1573, American type Culture Collection, Rockville, MD, USA) were cultured in Dulbecco's modified Eagle's medium (D-MEM, Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (Invitrogen, Carlsbad, CA, USA) and 1% MEM Non-Essential Amino Acids Solution (Invitrogen, Carlsbad, CA, USA). cDNAs encoding rat GABAA-R subunits were in the following vectors: α1 and α5 in pRc/CMV, α2, α3, γ2S and γ3 in pcDNA3 and β2 in pcDNA1. The cells were transiently transfected (5 μg of each cDNA per 100 mm dish) using FuGene™ (Roche Diagnostics Corporation) at a 3:1 FuGene:DNA ratio. Transfection efficiency was 50–80% as measured by co-transfection with green fluorescent protein cDNA (2.5 μg/100 mm dish). Forty-eight hours after transfection, cells were washed with ice-cold PBS, harvested and homogenized. Cell homogenates were centrifuged (100,000 g, 25 min) and washed three times by homogenization in ice-cold PBS buffer followed by centrifugation at 100,000 g for 25 min. The final pellets were stored at -20°C until needed.

For competition binding, 100 μg of membrane protein was incubated in 500 μl of PBS buffer with 0.5 nM [3H]Ro15–1788 (78.6 Ci/mmol, PerkinElmer Life Sciences) in the presence of diazepam (1 nM – 10 μM, Sigma-Aldrich), ocinaplon (0.1 – 250 μM, DOV Pharmaceuticals) or DOV 315,090 (0.1 – 50 μM, DOV Pharmaceuticals) for 1 h at 0°C. The samples were then diluted with 5 ml of ice-cold buffer and filtered under vacuum through glass-fiber filters (GF/B Whatman). Filters were washed 3 times with 5 ml of buffer and the radioactivity was quantitated by liquid scintillation counting in 5 ml of Ecolite scintillation fluid (ICN). Non-specific binding determined in the presence of 100 μM Ro 15–1788 (Sigma-Aldrich) was subtracted from total binding to calculate specific binding. Data were analyzed by non-linear regression (Prism, Graph-Pad software).

Recording of GABA-Gated Currents from GABAA Receptors Expressed in Xenopus Oocytes

cRNAs encoding GABAA-R α1, α2, α3, α5, β2 and γ2S subunits were injected into oocytes from Xenopus laevis. Forty-eight hours later, measurements of the effects of diazepam, ocinaplon and DOV 315,090 on GABA-gated Cl- currents from oocytes expressing GABAA-Rs were performed using a Warner TEVC amplifier (Warner Instruments, Inc., Foster City, CA) (Park-Chung et al., 1999). GABA (Sigma) was prepared as a 1 M stock solution in ND96. Microelectrodes of 1–3 MΩ when filled with 3 M KCl were used to record from oocytes in a recording chamber continuously perfused with ND-96 buffer solution. During data acquisition, oocytes were clamped at a holding potential of -70 mV. Drugs were applied by perfusion at a rate of approximately 50 μl sec-1 for 20 s followed by a 120 s wash. At the end of each experiment 3 μM of diazepam was applied as a potentiation control. All experiments were performed at room temperature (22–24°C).

GABA concentration-response data was obtained for each subunit combination, and the GABA EC10 was determined by nonlinear regression using the logistic equation. This concentration of GABA was used for modulation studies. Peak current measurements were normalized and expressed as a fraction of the peak control current measurements. Control responses to an EC10 concentration of GABA were re-determined after every 2 – 4 applications of modulator + GABA. Percent potentiation is defined as [I(GABA + Drug)/IGABA)-1] × 100, where I(GABA + Drug) is the current response in the presence of diazepam, and IGABA is the control GABA current. Potentiation data from each oocyte was fitted to the equation Potentiation = Emax × [Drug]/([Drug + EC50) by non-linear regression (Prism, Graph-Pad software). Due to a decline in the response at high diazepam concentrations, concentrations of diazepam above 3 μM were excluded from the fit. Some oocytes expressing α1β1γ2 receptors appeared to exhibit a biphasic modulatory response to diazepam, suggesting the possible presence of an additional component of modulation with a sub-nM EC50. For 6 of 8 oocytes, the fit was significantly improved by adding a second, higher-potency component of modulation, but the affinity of this second component was not well resolved in fitting due to its small amplitude. Given the lack of consistency of this possible high affinity effect, we have omitted it in fitting our concentration-effect curves. The choice of fitting to a monophasic or biphasic equation had only a small effect on the EC50 for the major component of modulation. For diazepam, the mean EC50 of the major component was increased from 35 nM to 42 nM when a two-component fit was used for those oocytes in which it produced a significant improvement in the sum of squares.


Biotransformation of ocinaplon into DOV 315,090 in vivo

As shown in Figure 2, DOV 315,090 appears rapidly in plasma following i.v. or oral administration of a behaviorally active dose of ocinaplon (5 mg/kg) to rats. At 1 h, corresponding to the time at which the anticonflict effect of ocinaplon was evaluated [22], the plasma concentration of DOV 315,090 is ~38% of the concentration of parent compound.

Figure 2
figure 2

Pharmacokinetics of ocinaplon and DOV 315,090. Blood levels of ocinaplon (●,○) and DOV 315090 (▲,△) were determined at various times after i.v. (●,▲) or oral (○,△) administration of 5 mg/kg ocinaplon to rats. Plotted results do not include one animal that exhibited a low blood level (0.47 μg/ml) of ocinaplon at the initial 10 min time point after oral administration and proportionally lower levels of both compounds throughout the duration of the experiment. This animal may have regurgitated a portion of the dose (of the suspension).

Comparison of the binding properties of diazepam, ocinaplon and DOV 315,090

Figure 3 and Table 1 document the binding properties of diazepam, ocinaplon and DOV 315,090 in HEK293 cells expressing different GABAA-R subunit combinations. Examination of binding constants shows that ocinaplon and DOV 315,090 have lower affinity than diazepam at all of the receptor subunit combinations tested. The binding profile of DOV 315,090 is similar to that of ocinaplon, with little selectivity among the subunit combinations tested. In contrast to diazepam, which exhibits markedly lower affinity for α1β2γ3 and α2β2γ3 receptors than for α1β2γ2 s and α2β2γ2 s receptors, replacement of γ2S with γ3 had little effect on the affinity of either ocinaplon or DOV 315,090 for any subunit combination (Table 1). Also, whereas diazepam has similar affinity for α1-containing and α2-containing receptors, both ocinaplon and DOV 315,090 have 3–4 fold lower affinity for α2-containing receptors. Specific [3H]Ro15–1788 or [3H]flunitrazepam binding to membrane preparations from cells transfected with α3, β2 and γ3 subunits was not detected, suggesting that these subunits failed to assemble in the HEK293 cells.

Figure 3
figure 3

Displacement curves of [3H]Ro 15–1788 binding by diazepam (DZ), ocinaplon and DOV 315,090 in homogenates of HEK293 cells transfected with different subunit combinations. Smooth curves are calculated from the mean parameter values in Table 1.

Table 1 Binding affinity of diazepam, ocinaplon and DOV 315,090 for GABAA-Rs with different subunit composition.

Modulation of GABAA-R function by diazepam, ocinaplon and DOV 315,090

Consistent with previous studies [22, 23], the potency and efficacy of ocinaplon were lower than diazepam at the four receptor subtypes analyzed. The highest efficacy was observed at receptors containing α3 subunits (Table 2). DOV 315,090 also exhibited the highest maximal potentiation at α3-containing receptors; however, its Emax values were similar to those of diazepam at receptors containing α1 or α3 subunits (Table 2).

Table 2 Properties of diazepam, ocinaplon and DOV315090 determined by two-electrode voltage clamp electrophysiology using Xenopus oocytes injected with cRNA.

DOV 315,090 and ocinaplon exhibited similar efficacies (150% vs. 139% potentiation, respectively) and EC50s (12.5 μM vs. 9.12 μM, respectively, n = 4) at α2β2γ2S receptors (Figure 4, Table 2). In contrast, whereas ocinaplon and DOV 315,090 were approximately equipotent at α3β2γ2S receptors (EC50 = 8.01 μM and 10.21 μM, respectively), the efficacy of DOV 315,090 was almost 1.87 fold greater than that of ocinaplon (340% vs 181% potentiation) (Figure 4, Table 2). Finally, DOV 315,090 was less efficacious and potent than ocinaplon at α5β2γ2S receptors (Figure 4, Table 2). The rank order of potency (EC50) of the pyrazolopyrimidines at enhancing GABA-gated chloride currents in receptors containing different α subunits was: α2≈α3≈α5 < α1 for DOV 315,090, compared to α2≈α3 < α5≈α1 for ocinaplon. Furthermore, DOV 315,090 and ocinaplon had different efficacy (Emax) profiles; the rank order of absolute efficacy was α5 < α2 < α1 < α3 for DOV 315,090, as compared with α5 < α1 < α2 < α3 for ocinaplon.

Figure 4
figure 4

Potentiation of GABA-gated currents by diazepam, ocinaplon and DOV 315,090. Rat GABAA-Rs consisting of α1β2γ2S, α2β2γ2S, α3β2γ2S and α5β2γ2S subunits were expressed in Xenopus oocytes. Potentiation was determined using an EC10 concentration of GABA (~10 μM for α1β2γ2S, α2β2γ2S and α3β2γ2S; ~5 μM for the α5β2γ2S). Curves were calculated by normalizing values of relative currents obtained following administration of diazepam (○), ocinaplon (●) or DOV 315,090 (□) in the presence of GABA (from at least four oocytes harvested from at least two batches) to the value obtained following application of GABA. The dose-response curves of diazepam were fitted up to 3 μM. Higher concentrations (in parentheses) were excluded from the fit due to a decline in potentiation at higher concentrations. Smooth curves are calculated based on mean parameter values given in Table 2. Asterisks indicate fits for which the extrapolated Emax is more than 25% greater than the maximum potentiation observed at highest drug concentration.


In the CNS, classical 1,4-BZDs such as diazepam, as well as other ligands of the BZD binding site, act on GABAA-Rs that are composed of α, β, and γ subunits. The majority of GABAA receptors contain α1–6, β2/3 and γ2 subunits, whereas the β1 and γ1/3 subunits have very restricted patterns of expression [2]. It has been shown that BZD pharmacology is primarily dependent upon the α subunit subtype present (α1–3 or α5), whereas receptors containing α4 or α6 subunits are insensitive to "classical" 1,4-BZDs [7, 24, 25]. Studies of animals in which genes coding for specific α subunits have been deleted or mutated to eliminate BZD sensitivity (e.g. the α1H101R mutation, which disrupts the BZD binding site) led to the hypothesis that the sedative effects of the BZDs are mediated by α1-subunit containing receptors (designated GABAA1-R), whereas anxiolytic effects are mediated by α2-subunit containing receptors (GABAA2-R) [7, 17, 26, 27]. GABAA-Rs containing α5 subunits are thought to be responsible for the impairment of learning and memory that is induced by BZDs [28]. These finding raised the attractive prospect that BZD-like drugs that specifically target GABAA-Rs that contain a specific α-subunit will be able to produce the intended pharmacological effect (e.g sedation or anxiolysis) with reduced incidence of side effects. Because BZD-like drugs function as allosteric modulators and do not occupy the GABA binding site, specificity may be potentially achieved on the basis of either differences in potency or on differences in modulatory efficacy at specific receptor subtypes.

Compounds such as zolpidem and zaleplon, which exhibit higher affinity for α1-containing receptors relative to other subtypes, have been promoted as sedative agents, driven in part by the hypothesis that selectivity for GABAA1-Rs would translate into an improved side-effect profile, particularly with respect to tolerance, withdrawal, and abuse liability. Although these compounds are effective sedative agents, consistent with the identification of GABAA1-Rs as mediating sedation, the selectivity of these compounds for GABAA1-Rs vs. GABAA-Rs containing other α-subunits is generally an order of magnitude or less, and it is unclear to what extent the hypothesized benefits are achieved in clinical practice [17].

However, the situation is less clear for compounds possessing anxiolytic properties. Recently published articles describe the pharmacological properties of two novel anxioselective compounds – ocinaplon [22] and DOV 51892 [23]. These compounds do not exhibit a marked selectivity among GABAA-Rs containing different diazepam-sensitive subunits (e.g. α1–3 and α5), yet are reported to be anxioselective, lacking sedative and myorelaxant side effects at anxiolytic doses. In particular, DOV 51892 exhibits higher efficacy than diazepam at GABAA1-Rs.

The classic BZD diazepam has been shown to act with high efficacy and similar potency across a broad spectrum of GABAA-Rs [1, 10, 22] (Table 2). This lack of selectivity with respect to either potency or efficacy among the major GABAA-R types have been hypothesized to account for the side effects associated with the use of diazepam when used as an anxiolytic, which include sedation, myorelaxation, narcosis, and amnesia. However, as has been confirmed by in vivo behavioral studies, such side effects are not observed with ocinaplon (e.g. in motor activity test, inclined screen and rod walking) or for DOV 51892 (e.g. rotarod and grip strength tests), even at doses well in excess of those that enhanced punished responding in the thirsty rat test [22, 23]. Further, ocinaplon is an effective anxiolytic in humans at doses that do not produce BZD-like side effects [22]. The present study was designed to test whether the anxioselective profile of ocinaplon is due to metabolism into subtype-selective metabolites. Our pharmacokinetic data demonstrate that in rats, the major metabolite of ocinaplon is a 4'-N' oxide, DOV 315,090. Whereas DOV 315,090 is active as a GABAA-R modulator, and its in vitro binding affinities for recombinant α1β2γ2S, α2β2γ2S, and α3β2γ2S receptors differ only marginally from ocinaplon, its affinity for α5β2γ2S receptors is only slightly lower than that of ocinaplon (~2-fold).

Comparison of the pharmacological profile of ocinaplon and DOV 315,090 using two electrode voltage clamp electrophysiology (Table 2) shows that the greatest difference in efficacy occurred at α3β2γ2S receptors. Although a clear maximum was not attained due to solubility limits, the extrapolated maximum potentiation by DOV 315,090 was 1.87-fold greater, followed by a 1.45-fold difference at α1β2γ2S receptors compared to ocinaplon. In contrast, maximum potentiation by DOV 315,090 was lower than that of ocinaplon at the α5β2γ2S receptor subtype. The efficacies of DOV 315,090 and ocinaplon at α2β2γ2S receptors were similar.

These results do not support the hypothesis that the anxioselective profile of ocinaplon is attributable to enhanced selectivity of its metabolite DOV 315,090 for α2-containing receptors. Thus, compared to ocinaplon, DOV 315,090 does not exhibit enhanced affinity or potency for α2-containing receptors over α1-containing receptors, whereas the difference in efficacy favors α3-, α5-, or α1-containing receptors over α2-containing receptors. The present experiments examined GABAA-Rs in two different heterologous expression systems (Xenopus oocytes and HEK 293 cells), which may be lacking modulatory proteins or regulatory mechanisms that are only present in neurons. While we cannot exclude the possibility that such interactions somehow confer differences in modulator binding or efficacy, such a hypothesis would require that such interactions modify the structure of the benzodiazepine binding site, which is located in the extracellular domain of the GABAA-R, in such a way as to selectively alter its interactions with different ligands.

Recent studies suggest that GABAA3-Rs receptors are also important in mediating anxiolysis [18, 20, 3134]. DOV 315,090 has relatively high efficacy at α3β2γ2S, so it is likely that modulation of GABAA3-Rs by DOV 315,090 contributes to the anxioselective profile of ocinaplon; however, adipiplon (NG2-73), an α3-selective positive modulator, has been reported to have sedative/hypnotic activity [35], suggesting that α3 selectivity is not sufficient to confer anxioselectivity.

In summary, transgenic mice in which the BZD recognition site of the α2 subunit is disabled exhibit reduced diazepam sensitivity in behavioral tests considered to be predictive of anxiolytic activity, and a similar modification to the α1 subunit reduces sensitivity in tests held to be predictive of sedation [15, 26]. These observations have led to optimism that it will be possible to achieve the long-desired goal of developing a nonsedating anxiolytic [36]. And indeed, there has been substantial progress in identifying such compounds [1922, 31, 3740], yet ironically, they do not in general conform to the expected paradigm of favoring α2-containing over α1-containing receptors. This suggests that anxiolysis in humans may prove to be more complex than is suggested by a simple reading of the results from transgenic mice in behavioral models thought to be indicative of anxiety. It remains to be determined whether differences in the design of the behavioral assays [41, 42], interspecies differences [43, 44], or a combination of these factors account for these discrepancies. Translating such promising results into clinically useful compounds is likely to require an improved understanding of the ways in which BZD-like ligands act at different GABAA-R subtypes and the consequences of these effects upon neural system-mediated behavioral outputs.


1. DOV315090 is a major metabolite of the anxioselective GABAA-R modulator ocinaplon.

2. DOV 315,090 possesses modulatory activity at α1-, α2-, α3-, and α5-containing GABAA-Rs with a selectivity profile similar to that of ocinaplon.

3. The anxioselective properties of ocinaplon, demonstrated in both preclinical and clinical studies, are not a consequence of enhanced subtype selectivity by DOV315090.