α4β3δ GABAAReceptors Characterized by Fluorescence Resonance Energy Transfer-derived Measurements of Membrane Potential*

Selective modulators of γ-aminobutyric acid, type A (GABAA) receptors containing α4subunits may provide new treatments for epilepsy and premenstrual syndrome. Using mouse L(−tk) cells, we stably expressed the native GABAA receptor subunit combinations α3β3γ2,α4β3γ2, and, for the first time, α4β3δ and characterized their properties using a novel fluorescence resonance energy transfer assay of GABA-evoked depolarizations. GABA evoked concentration-dependent decreases in fluorescence resonance energy transfer that were blocked by GABAA receptor antagonists and, for α3β3γ2and α4β3γ2 receptors, modulated by benzodiazepines with the expected subtype specificity. When combined with α4 and β3, δ subunits, compared with γ2, conferred greater sensitivity to the agonists GABA, 4,5,6,7-tetrahydroisoxazolo-[5,4-c]pyridin-3-ol (THIP), and muscimol and greater maximal efficacy to THIP. α4β3δ responses were markedly modulated by steroids and anesthetics. Alphaxalone, pentobarbital, and pregnanolone were all 3–7-fold more efficacious at α4β3δ compared with α4β3γ2. The fluorescence technique used in this study has proven valuable for extensive characterization of a novel GABAA receptor. For GABAA receptors containing α4 subunits, our experiments reveal that inclusion of δ instead of γ2subunits can increase the affinity and in some cases the efficacy of agonists and can increase the efficacy of allosteric modulators. Pregnanolone was a particularly efficacious modulator of α4β3δ receptors, consistent with a central role for this subunit combination in premenstrual syndrome.

A variety of animal models of epilepsy lead to changes in the level of expression of ␣ 4 and ␦ subunit protein and mRNA in hippocampal dentate gyrus (17,(22)(23)(24)(25)(26) and thalamic relay nuclei (27), and acute pentylenetetrazol-induced seizures, to which mice lacking ␦ subunits are more susceptible (56), lead to an increase in ␦ subunit expression in neocortex (28). Elevated levels of ␣ 4 subunits are also implicated in an animal model of alcohol dependence (29) and in steroid-withdrawal models of premenstrual syndrome and postpartum or postmenopausal dysphoria, particularly the increased anxiety and incidence of seizures (30 -34). The association of these pathologies with changes in ␣ 4 and ␦ subunit expression and the observation that ligands with high affinity for ␣ 4 ␤␥ 2 GABA A receptors are amethystic (35,36) suggest that novel selective modulators of these GABA A receptors may, as well as leading to a better understanding of the properties and physiological roles of these subunits in the brain, have great therapeutic benefit. The development of such modulators has been held back on two counts. First, ␣ 4 ␤ 3 ␦ receptors cannot easily be expressed in transient recombinant systems, and so their properties remain unclear. Second, GABA A receptor drug-development programs have depended until now on difficult and time-consuming electrophysiological techniques or less sensitive radio-ion flux and pH methods for determining the effects of compounds on GABA A receptor function (6,(37)(38)(39). We have overcome these problems by creating a stable L(Ϫtk) mouse cell line in which expression of ␣ 4 ␤ 3 ␦ receptors is under the control of a dexa-methasone-induced promoter, and by developing an experimental system using fast ratiometric voltage-sensitive FRET (40) to measure GABA-evoked changes in membrane potential. Fluorescence measurements of GABA A receptor function offer significant advantages because they are safe, are sufficiently sensitive to detect small potentiations and inhibitions, and can be miniaturized for future ultrahigh throughput applications. Furthermore, unlike high throughput radioligand binding assays, which have also been used for the development of GABA A receptor modulators, they can identify modulators regardless of their site of action. Here we describe the use of this novel fluorescence technique to characterize the pharmacological activation and modulation of GABA A receptors with the subunit combinations ␣ 3 ␤ 3 ␥ 2 , ␣ 4 ␤ 3 ␥ 2 , and ␣ 4 ␤ 3 ␦.

EXPERIMENTAL PROCEDURES
cells were stably transfected, using a pMSGneo vector, with combinations of human GABA A receptor subunits. Expression of ␣, ␤, and ␥ subunits was controlled by a dexamethasone-inducible promoter as described previously (14,41), whereas expression of ␦ subunits was constitutive. Enzyme-linked immunosorbent assays using Myc-tagged subunits confirmed that ␦ subunits were only present at the cell surface if both ␣ 4 and ␤ 3 subunits were also present. Cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% serum (Fetalclone II) at 37°C in an atmosphere of 5% CO 2 and 95% air. Cells were passaged weekly and for experiments were transferred to 96-well black-sided microtiter plates at a density that gave confluent monolayers on the days of experiments. Receptor expression was induced 24 h before experiments by replacing 50% of the medium with medium containing dexamethasone (1 M final concentration).
Fluorescence Measurements of Membrane Potential-All experiments were performed in a low Cl Ϫ buffer (160 mM sodium D-gluconate, 4.5 mM potassium D-gluconate, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM D-glucose, 10 mM HEPES, pH 7.4). Cells were washed twice, leaving a 25-l residual volume, and 55 l of dye solutions were added to give final concentrations of 4 M chlorocoumarin-2-dimyristoyl phosphatidylethanolamine (CC2-DMPE) and 1 M bis(1,3-diethyl-2-thiobarbiturate)trimethineoxonol (DisBac 2 )(3). After a 30-min incubation at room temperature in darkness, cells were washed again, and 65 l of dye solutions were added to give final concentrations of 1 M DisBac 2 (3) and 0.5 mM tartrazine. Microtiter plates were then placed in a voltage/ion probe reader (VIPR TM ; Aurora Biosciences Corp.), which performs automated additions of pharmacological stimuli and records fluorescence emission. Briefly, the VIPR TM consists of a Hamilton 2200 pipetter, an automated microplate positioning stage, and a fiber-optic illumination and detection system capable of measuring two emission wavelengths from eight wells simultaneously (40). A 400DF15 filter was used in the excitation pathway, and 460DF45 and 580DF60 filters were used in the respective emission pathways. In all experiments, basal fluorescence was read for 8 s before addition of modulators, and then GABA was added 22 s later. Fluorescence emission from wells was recorded at 1 Hz.
Data Analysis-For each time point and for each fluorescence emission wavelength, we subtracted background fluorescence recorded from wells without cells in the same microtiter plate and calculated the ratio of fluorescence at 460 nm to that at 580 nm. GABA-evoked depolarizations were then expressed as a fractional change in this ratio. Algorithms written as Excel 97 (Microsoft Corp.) macros were used for automated calculations of fluorescence ratio and GABA responses (39), and an iterative curve-fitting program (Prism, GraphPad Software Inc.) was used to fit concentration-effect relationships to a four-parameter logistic equation.

RESULTS
Previously, optical sensors of membrane potential operated through a slow redistribution of permeant ions or a rapid but insensitive perturbation of dyes attached to one face of the membrane (42)(43)(44). However, a recently developed membrane potential indicator, described in Fig. 1, uses FRET to provide a fluorescent readout of membrane potential that is both rapid and robust (45). Before using this technique to characterize cell lines expressing ␣ 4 subunit-containing GABA A receptors, we first established its pharmacological utility using cells expressing the well characterized subunit combination ␣ 3 ␤ 3 ␥ 2 . In low chloride medium, GABA-evoked depolarizations of cells expressing ␣ 3 ␤ 3 ␥ 2 GABA A receptors and loaded with CC2-DMPE and DisBac 2 (3) were rapidly transduced into decreased FRET, and, therefore, an increase in the ratio of fluorescence emission at 460 nm to that at 580 nm was seen (Fig. 2). The fluorescence emission ratio rose to a concentration-dependent plateau within 5 s that was sustained for Ͼ15 s. For the plateau phase of the response, measured as the mean normalized fluorescence emission ratio between 10 and 15 s after application of agonist, the half-maximal concentration (EC 50 ) of GABA was 2.1 Ϯ 0.2 M, and the Hill slope (n H ) was 1.5 Ϯ 0.1 (mean Ϯ S.E. of three experiments, Fig. 2b). We next examined receptor pharmacology by pretreating cells with compounds known to be active at ␣ 3 ␤ 3 ␥ 2 GABA A receptors and then applying a half-maximal concentration of GABA. ␣ 3 ␤ 3 ␥ 2 responses were blocked by the antagonists bicuculline (competitive) and picrotoxin (noncompetitive), potentiated by the benzodiazepine agonist zolpidem, and partially inhibited by the benzodiazepine inverse agonist dimethoxy-4-ethyl-␤-carboline-3-carboxylate (DMCM) (Fig.  2c). These findings are highly consistent with those from electrophysiological experiments (38).
Having established that fluorescence measurements of GABA A receptor function appear to reliably report receptor pharmacology, we then examined GABA-evoked changes in FRET using L(Ϫtk) cells expressing either ␣ 4 ␤ 3 ␥ 2 GABA A receptors or the previously uncharacterized subunit combination ␣ 4 ␤ 3 ␦. The kinetics of GABA-evoked depolarization were similar for these cells to those for cells expressing ␣ 3 ␤ 3 ␥ 2 GABA A receptors (Fig. 3a). GABA, muscimol, and THIP were between 3 and 6 times more potent at ␣ 4 ␤ 3 ␦ receptors compared with ␣ 4 ␤ 3 ␥ 2 (Figs. 2b and 3b). The first detectable response to mus-cimol occurred 1 s earlier than that for GABA or THIP, but thereafter the three agonists evoked changes in fluorescence ratio with similar kinetics (Fig. 3c). Although less potent than GABA, THIP was a fully efficacious agonist at ␣ 4 ␤ 3 ␥ 2 receptors and a superagonist at ␣ 4 ␤ 3 ␦ (Fig. 3b). Responses mediated by both ␣ 4 ␤ 3 ␦ and ␣ 4 ␤ 3 ␥ 2 receptors were inhibited by pretreatment with picrotoxin and bicuculline. Whereas picrotoxin (30 M) inhibited the responses to all concentrations of GABA, bicuculline (30 M) inhibited only submaximal responses, causing a 30-fold shift in the GABA concentration-response curve (Fig. 3d).

DISCUSSION
In this study we have developed a novel fluorescence technique that provides rapid and sensitive measurements of GABA A receptor function, and have used it to characterize a novel cell line expressing GABA A receptors with the composition ␣ 4 ␤ 3 ␦. Our initial experiments, using cell lines expressing the previously characterized GABA A receptor subunit combinations ␣ 3 ␤ 3 ␥ 2 and ␣ 4 ␤ 3 ␥ 2 , demonstrated that GABA A receptor-mediated chloride fluxes were rapidly and reliably transduced into decreased FRET. In contrast to traditional fluorescence assays of membrane potential utilizing oxonol redistribution, GABA-evoked depolarization of cells loaded with CC2-DMPE and DisBac 2 (3) and excited with 410 nm light leads to a change in fluorescence emission that occurs within seconds rather than minutes. As previously reported, substitution of ␣ 4 subunits for ␣ 3 did not affect GABA potency, which was similar to that previously reported for the same subunit combinations expressed in mammalian cells (6,19). GABA-evoked responses were blocked by picrotoxin and bicuculline, and at ␣ 3 ␤ 3 ␥ 2 and ␣ 4 ␤ 3 ␥ 2 receptors, the efficacies and potencies of the benzodiazepines tested were very similar to published values (6,19,20,38,47). Thus FRET-derived measurements of membrane potential proved to be a sensitive and reliable indicator of GABA A receptor pharmacology. They gave results that were essentially indistinguishable from those for electrophysiological experiments and are likely to prove useful for studies of multiple receptor classes.
While GABA A receptors composed of ␣, ␤, and ␥ subunits have been studied extensively, relatively little is known about the functional and pharmacological properties of receptor iso-forms containing ␦ subunits. Receptors containing ␦ subunits in combination with ␣ 4 , as occur in situ, have never been characterized. We therefore created a novel L(Ϫtk) cell line in which expression of ␣ 4 ␤ 3 ␦ GABA A receptors was under the control of a dexamethasone-inducible promoter, and used FRET-derived measurements of membrane potential to directly compare them to ␣ 4 ␤ 3 ␥ 2 receptors. We found that ␦ subunits, compared with ␥ 2 , conferred higher affinity for all the agonists tested. The rank order for agonist potency muscimol Ͼ GABA Ͼ THIP was unchanged, but THIP acted as a superagonist at ␣ 4 ␤ 3 ␦ receptors, evoking substantially larger changes in FRET than either GABA or muscimol. Partial agonists at other GABA A receptor subtypes have been described, but no agonist has shown greater efficacy than GABA. An equally valid interpretation of this data, therefore, is that ␦ subunits, when combined with ␣ 4 and ␤ 3 , confer partial agonism to GABA. The different potency, and perhaps efficacy, of GABA at ␣ 4 ␤ 3 ␥ 2 and ␣ 4 ␤ 3 ␦ receptors suggest quite different physiological roles for these receptor isoforms. Low affinity receptors containing ␥ 2 subunits may be suited to synapses where GABA is plentiful and a rapid dissociation rate is beneficial to high frequency signaling.
␦ subunits were an important determinant of the effects of a variety of allosteric modulators, including benzodiazepines, steroids, and barbiturates. Substitution of ␥ 2 subunits with ␦ abolished sensitivity to modulators acting at the benzodiazepine binding site. Although ␤-carbolines, such as ␤-CCE and methyl-␤-carboline-3-carboxylate, inhibit GABA A receptors with high potency via the benzodiazepine binding site, they also potentiate GABA responses, with lower potency, via the loreclezole site present only on ␤ 2 -and ␤ 3 -containing receptors (10). Therefore the potentiation of both ␣ 4 ␤ 3 ␥ 2 and ␣ 4 ␤ 3 ␦ responses by ␤-CCE was almost certainly mediated by the binding site for loreclezole and does not indicate benzodiazepine sensitivity.
Barbiturates are thought to potentiate the response of GABA A receptors irrespective of their subunit composition (47). Both ␣ 4 ␤ 3 ␥ 2 and ␣ 4 ␤ 3 ␦ responses were potentiated by pentobarbital. However, ␦ subunits, compared with ␥ 2 , conferred 7 times higher efficacy to pentobarbital. At micromolar concentrations, barbiturates have a second effect, directly activating GABA A receptors (50,51). ␣ 4 ␤ 3 ␥ 2 receptors were activated by pentobarbital, but this effect was abolished when ␥ 2 subunits were substituted with ␦. We conclude that ␦ and ␥ 2 subunits affect both the modulation and activation of GABA A receptors by barbiturates. There may also be a role for ␤ subunits since ␣ 4 ␤ 3 ␥ 2 and ␣ 4 ␤ 2 ␥ 2 receptors are activated by pentobarbital (47), whereas the effect does not occur on ␣ 4 ␤ 1 ␥ 2 (20). ␣ 4 ␤ 3 ␥ 2 and ␣ 4 ␤ 3 ␦ receptors were differentially modulated by steroids. In contrast to the stimulatory effect at other GABA A receptors (52), receptors containing ␣ 4 subunits were inhibited by the naturally occurring neurosteroid pregnanolone. Furthermore, both pregnanolone and the synthetic anesthetic alphaxalone (52,53) were more efficacious at ␣ 4 ␤ 3 ␦ compared with ␣ 4 ␤ 3 ␥ 2 . These data demonstrate that ␦ subunits are a critical determinant of neurosteroid efficacy, possibly accounting for the reduced behavioral effects of alphaxalone and pregnanolone in mice lacking ␦ subunits (55). During the menstrual cycle and pregnancy in normal women, levels of pregnanolone correlate with those of progesterone from which it is synthesized (54). In addition to their effects on ␣ 4 subunit expression (30 -34), endogenous neuroactive steroids may therefore also modulate the function of GABA A receptors, particularly those containing ␦ subunits, and thereby contribute to the increased incidence of anxiety and seizures in premenstrual syndrome and postpartum and postmenopausal dysphoria. Our data imply that ␣ 4 ␤ 3 ␦ receptors may have a central role in these disorders and that new therapies might be developed by selective targeting of the steroid binding site of GABA A receptors containing ␦ subunits.
FRET-derived measurements of membrane potential provide the most robust and reliable high throughput assay of GABA A receptor function yet developed and will be an invaluable tool for characterizing novel subunit combinations and identifying new therapeutic modulators. When applied to cell lines expressing GABA A receptors with the subunit combinations ␣ 4 ␤ 3 ␥ 2 and ␣ 4 ␤ 3 ␦, this novel fluorescence technique revealed that ␦ subunits are an important determinant of the efficacy and potency of agonists and allosteric modulators. Of particular importance was the finding that ␣ 4 ␤ 3 ␦ receptors were markedly more sensitive to inhibition by pregnanolone, suggesting that this receptor subtype could be targeted for the treatment of premenstrual syndrome. FIG. 5. Modulation and activation of ␣ 4 ␤ 3 ␥ 2 and ␣ 4 ␤ 3 ␦ GABA A receptors by steroids and anesthetics. L(Ϫtk) cells expressing ␣ 4 ␤ 3 ␥ 2 (squares) or ␣ 4 ␤ 3 ␦ (circles) GABA A receptors were pretreated with different concentrations of alphaxalone (a, open), pregnenolone (a, filled), pregnanolone (b, filled), or pentobarbital (b, open) before application of GABA at half-maximal concentration. c, subtype-selective activation of GABA A receptors by alphaxalone and pentobarbital. L(Ϫtk) cells expressing ␣ 4 ␤ 3 ␥ 2 (squares) or ␣ 4 ␤ 3 ␦ (circles) GABA A receptors were stimulated with alphaxalone (10 M, filled) or pentobarbital (1 mM, open), addition of which is shown by the black bar. Responses were normalized with respect to maximum GABA as in Fig. 3. All data are the mean Ϯ S.E. of three experiments. At high concentrations, pregnenolone and pregnanolone interfered with fluorescence emission, and their effects could not be interpreted at concentrations above 30 M.