Mechanisms of anabolic androgenic steroid inhibition of mammalian ɛ-subunit-containing GABA_A receptors

Abstract

GABAergic transmission regulates the activity of gonadotrophin-releasing hormone (GnRH) neurons in the preoptic area/hypothalamus that control the onset of puberty and the expression of reproductive behaviours. One of the hallmarks of illicit use of anabolic androgenic steroids (AAS) is disruption of behaviours under neuroendocrine control. GnRH neurons are among a limited population of cells that express high levels of the ɛ-subunit of the GABA_A receptor. To better understand the actions of AAS on neuroendocrine mechanisms, we have characterized modulation of GABA_A receptor-mediated currents in mouse native GnRH neurons and in heterologous cells expressing recombinant α₂β₃ɛ-receptors. GnRH neurons exhibited robust currents in response to millimolar concentrations of GABA and a picrotoxin (PTX)-sensitive, bicuculline-insensitive current that probably arises from spontaneous openings of GABA_A receptors. The AAS 17α-methyltestosterone (17α-MeT) inhibited spontaneous and GABA-evoked currents in GnRH neurons. For recombinant α₂β₃ɛ-receptors, 17α-MeT inhibited phasic and tonic GABA-elicited responses, accelerated desensitization and slowed paired pulse response recovery. Single channel analysis indicated that GABA-evoked events could be described by three open dwell components and that 17α-MeT enhanced residence in the intermediate dwell state. This AAS also inhibited a PTX-sensitive, spontaneous current (open probability, ∼0.15–0.2) in a concentration-dependent fashion (IC₅₀ ≈ 9 μm). Kinetic modelling indicated that the inhibition induced by 17α-MeT occurs by an allosteric block in which the AAS interacts preferentially with a closed state and promotes accumulation in that state. Finally, studies with a G302S mutant ɛ-subunit suggest that this residue within the transmembrane domain TM2 plays a role in mediating AAS binding and modulation. In sum, our results indicate that inclusion of the ɛ-subunit significantly alters the profile of AAS modulation and that this allosteric inhibition of native GnRH neurons should be considered with regard to AAS disruption of neuroendocrine control.

Anabolic androgenic steroids (AAS) are synthetic derivatives of testosterone that are valuable therapeutic agents, but whose clinical use has been overshadowed by illicit self-administration of supratherapeutic doses. In contrast to physiological levels of endogenous steroids, which are in the nanomolar range (for review, see Shahidi, 2001), serum levels of AAS have been estimated to reach micromolar concentrations in human subjects who abuse them (Masonis & McCarthy, 1995; Wu, 1997; Daly et al. 2001). One of the hallmarks of AAS use is disruption of puberty and of the expression of normal reproductive behaviours (for review, see Clark & Henderson, 2003). These behaviours are regulated by gonadotrophin-releasing hormone (GnRH) neurons located in the preoptic area/hypothalamus which are under a tonic inhibitory tone exerted by GABA_A receptors (for review, see Herbison et al. 1991; Moenter et al. 2003). A critical and unaddressed question is whether AAS alter GABAergic control of GnRH neurons and thus neuroendocrine regulation.

The native GABA_A receptor is a pentameric ionotropic transmembrane protein for which 16 different receptor subunit genes (α_1–6, β_1–3, γ_1–3, δ, ɛ, π and θ) and numerous alternatively spliced mRNAs have been identified in mammals (for review, see Whiting et al. 1999). The ɛ-subunit shows a highly restricted pattern of expression (Davies et al. 1997; Whiting et al. 1997; Moragues et al. 2000; Sinkkonen et al. 2000) and in rodents is enriched in the ventromedial nucleus of the hypothalamus, the medial preoptic area (mPOA), the septum and the amygdala (Moragues et al. 2000, 2002, 2003; Sinkkonen et al. 2000; McIntyre et al. 2002); all regions intimately involved in the generation of reproductive behaviours (for review, see Clark & Henderson, 2003). Of particular relevance, virtually all GnRH neurons in the preoptic region express marked levels of this subunit (Moragues et al. 2003).

Recombinant receptors composed of α₁β₃ɛ- or α₂β₁ɛ-subunits are both spontaneously active and gated by GABA (Neelands et al. 1999; Davies et al. 2001; Maksay et al. 2003; Wagner et al. 2005). All reports to date indicate that ɛ-containing receptors are not sensitive to nanomolar concentrations of high-affinity benzodiazepine (BZ) binding site agonists (Davies et al. 1997, 2001; Whiting et al. 1997; Thompson et al. 1998; Neelands et al. 1999; Maksay et al. 2003), but micromolar concentrations of BZs have been reported to inhibit these ɛ-containing receptors (Maksay et al. 2003; cf. Neelands et al. 1999), potentially at a separate low-affinity site. The ability of anaesthetics and anaesthetic neurosteroids to potentiate ɛ-containing channels has been somewhat controversial (Davies et al. 1997, 2001; Whiting et al. 1997; Thompson et al. 1998), but the percentage potentiation elicited by these modulators has been shown to be inversely proportional to the level of ɛ-subunit expression in recombinant systems (Thompson et al. 2002). Similarly, expression of the ɛ-subunit in native neurons has been correlated with diminished sensitivity to BZs (Kasparov et al. 2001), neurosteroids (Jorge et al. 2002) and anaesthetics (Irnaten et al. 2002; Sergeeva et al. 2005).

Anabolic androgenic steroids are both structurally and functionally distinct from the neurosteroids (for review, see Lambert et al. 1995, 2003; Clark et al. 2004). The AAS 17α-MeT acts as a positive modulator of currents elicited by application of millimolar GABA to recombinant α₂β₃γ₂-receptors, but is without effect on such currents elicited at α₁β₃γ₂- or α₂β₃δ-receptors. Conversely, 17α-MeT potentiates tonic currents elicited by micromolar concentrations of GABA at α₁β₃γ₂-receptors, but is without effect on tonic currents produced by stationary levels of micromolar GABA at α₂β₃γ₂-receptors. Thus the profile of allosteric modulation elicited by this AAS is dependent upon both subunit composition and the concentration/duration of GABA exposure (Yang et al. 2002, 2005; Clark et al. 2004). GnRH neurons are highly heterogeneous in GABA_A receptor subunit expression (Sim et al. 2000; Todman et al. 2005). While the complement of surface receptors in GnRH neurons is not known, the preferential expression of α₂- and β₃-subunits in the preoptic area/hypothalamus (Wisden et al. 1992; Herbison & Fénelon, 1995; Pirker et al. 2000; McIntyre et al. 2002; Penatti et al. 2005), coupled with the abundant expression of the ɛ-subunit in GnRH neurons (Moragues et al. 2003), suggests that α₂β₃ɛ-receptors are likely to constitute an important receptor class in these cells.

Assessments of GABAergic control of GnRH neurons have been facilitated by the generation of lines of transgenic mice in which green fluorescent protein (GFP) is expressed from the GnRH promoter (Spergel et al. 1999; Suter et al. 2000; Han et al. 2004). Here, we have taken advantage of one of these transgenic lines (Suter et al. 2000) to assess how the commonly abused AAS 17α-MeT modulates GABA_A receptor-mediated phasic and spontaneous currents in GnRH neurons. To further define the mechanism of AAS action at ɛ-containing receptors, we have assessed modulation by 17α-MeT of both GABA-gated and spontaneous currents through recombinant α₂β₃ɛ-receptors to characterize how these abused steroids may influence GABAergic activity in GnRH neurons.

Methods

GFP-GnRH transgenic mice and primary neuron preparation

GFP-GnRH mice were generously provided by Dr Suzanne Moenter (Department of Medicine and Cell Biology, University of Virginia, Charlottesville, VA, USA). In these mice, a portion of the mouse GnRH promoter was used to drive expression of enhanced GFP (Suter et al. 2000). Briefly, a portion of the mouse GnRH promoter (−3446 to +23) was used to drive expression of a transgene consisting of the B intron of rabbit β-globin as a splice donor/acceptor, the coding sequence for enhanced GFP and the polyadenylation signal from human growth hormone. The line was generated in a CBB6 strain, and 99.5% of neurons expressing GFP were found to be GnRH positive (Suter et al. 2000). For experiments performed here, mice were killed with a rising concentration of CO₂, decapitated, and their brains removed directly into an artificial cerebral spinal fluid (ACSF; mm): 124 NaCl, 2.4 CaCl₂, 10 d-glucose, 5 KCl, 1.3 KH₂PO₄, 1.3 MgSO₄ and 24 NaHCO₃, superfused with 95% O₂−5% CO₂. The brain was transected coronally at the level of the optic chiasm, and 400 μm slices were made in this oxygenated ACSF from the rostral portion of the brain using an Electron Microscopy Sciences OTS-4000® vibroslicer (Hatfield, PA, USA). Slices were transferred to a sterile 100 mm Petri dish containing 5 ml of Hibernate-A (BrainBits, Springfield, IL, USA), a Mops-buffered medium designed for use in ambient O₂, which was supplemented with 1 × B27 (Invitrogen Corp., Gaithersburg, MD, USA). Regions containing the mPOA were dissected and incubated in a solution of 2 mg ml⁻¹ papain (Worthington Biochemical Corp., Lakewood, NJ, USA) in Hibernate-A at 30°C for 30 min. Tissue was then treated with 1.2 U μl⁻¹ Dnase I type II (Sigma-Aldrich Co., St Louis, MO, USA) and triturated, and the resulting cell suspension was centrifuged for 2 min at 450 ×g. The supernatant was decanted, the pellet was resuspended in Neurobasal-A medium (Invitrogen Corp.) supplemented with B27, and the cells were plated onto plastic dishes and maintained at 37°C and 5% CO₂ for at least 1 h prior to recording.

All animal care procedures were approved by Institutional Animal Care and Use Committee at Dartmouth. Procedures were performed to minimize the use of animals and any pain or discomfort to them, and are in agreement with the guidelines and recommendations of the National Institutes of Health and the American Veterinary Medical Association.

Recombinant GABA_A receptor cDNAs and expression in heterologous cells

Transfections and cell maintenance

Human Embryonic Kidney (HEK) 293 cells (American Type Culture Collection, Manassas, VA, USA) were grown on tissue culture dishes (BD Falcon, Franklin Lakes, NJ, USA) and maintained in Dulbecco's Modified Eagle's Medium (Invitrogen Corp.) supplemented with 10% fetal bovine serum (Invitrogen Corp.), 2 mm l-glutamine, 50 IU ml⁻¹ penicillin and 50 μg ml⁻¹ streptomycin (all from Mediatech Inc., Herndon, VA, USA) at 37°C and in 5% CO₂−95% O₂. Constructs encoding the rat β₃- and γ_2L- and the human α₂-subunits were kindly provided by Dr Stefano Vicini (Georgetown University Medical School, Washington, DC, USA). The human ɛ-subunit cDNA was provided by Dr Paul Whiting (Merck, Sharp & Dohme, Harlow, UK). HEK293 cells were transiently transfected according to Yang et al. (2002, 2005) at equal ratios of 0.8 μg of each plasmid using the Lipofectamine PLUS® protocol (Invitrogen Corp.) for HEK293 cells. The pGreenLantern plasmid (Invitrogen Corp.) was cotransfected (0.8 μg) to permit selection of transfected cells expressing GFP with fluorescence optics. After transfection, cells were maintained in culture medium containing 20 μm bicuculline methiodide (Sigma-Aldrich Co.), which promoted cell survival, and recordings were made approximately 24–48 h after transfection.

Mutagenesis of GABA_A receptor cDNAs

The mutant G302S GABA_A receptor construct was engineered from the cDNA encoding the human ɛ-subunit subcloned into the pCDM8 expression vector (Invitrogen Corp.) using the QuickChange System (Stratgene, La Jolla, CA, USA) and according to the manufacturer's instructions. Numbering of the amino acid 302 refers to position relative to the start of the mature peptide. Mutant DNA was amplified by transformation of XL1-Blue supercompetent cells, transformed colonies were identified, and plasmid DNA was isolated using the Qiagen MiniPrep system (Valencia, CA, USA). Successful mutations were identified by sequencing in both forward and reverse directions using SP6 and T7 primers, respectively. Primers were purchased from Integrated DNA Technologies (Coralville, IA, USA).

The ɛ(G302S) primers were: forward, 5′-CTG ACC ATG ACC ACG TTG AGC ACC TTT TCT CG-3′; and reverse, 5′-CGA GAA AAG GTG CTC AAC GTG GTC ATGGTC AG-3′.

Data acquisition and analysis

Pipette fabrication and general recording conditions

Patch pipettes used for all recordings were fabricated from borosilicate glass (Sutter Instrument Co., Novato, CA, USA). Typically, electrodes used for whole-cell recordings had open tip resistances between 3 and 5 MΩ. Data were acquired and analysed using an EPC-9 patch clamp amplifier and HEKA Pulse software (Instrutech Corp., Port Washington, NY, USA). Whole-cell and nucleated patch recordings were made at room temperature (22–24°C) with a pipette solution containing (mm): 140 CsCl, 1 MgCl₂, 10 Hepes, 5 EGTA and 4 Mg-ATP (pH 7.2, with CsOH). A bath solution composed of the following (mm): 130 NaCl, 6 KCl, 2 MgCl₂, 1 CaCl₂, 10 Hepes, 10 D-glucose and 10 sucrose (pH 7.4, with NaOH) was used for all recordings.

Ultrafast perfusion

GABA and other modulators were applied to the patch using the LSS-3100 High Speed Positioning System (Burleigh Instruments Inc., Fishers, NY, USA) and a double-barrelled length of theta-glass (Sutter Instruments Co.; tip diameter 100–120 μm) as previously described (Yang et al. 2002, 2005). Open tip currents showed that the 10–90% of the peak on- and off-response was achieved in less than 1 ms. Fast perfusion of transfected HEK293 cells was made using nucleated outside-out patches. Fast perfusion of primary neurons was made using the whole-cell configuration because these acutely isolated cells did not tightly adhere to the dish surface, which made patch excision difficult. All primary neurons were of small diameter, were bereft of major processes following dissociation and had comparable surface area to nucleated outside-out patches, with an average capacitance of 4.6 ± 0.2 pF for the GnRH neurons and 6.9 ± 0.4 pF for the nucleated outside-out patches from HEK293 cells. For experiments examining modulation of GABA-evoked responses, modulators (17α-MeT or zolpidem) were present in both streams from the theta-glass in order to pre-equilibrate the patches. When switching between control, drug and wash conditions, or between different concentrations of modulators, a minimum of 2 min was allowed to elapse to ensure full solution exchange. The ultrafast perfusion set-up was plumbed with Teflon® tubing (Small Parts, Inc. Miami Lakes, FL, USA).

Dish perfusion

For experiments examining the effects of AAS on currents induced by tonic exposure to GABA, HEK293 cells were transfected and replated at low density, and recordings were made in the whole-cell configuration 5–8 h after replating. Modulators were applied to the bath by gravity flow from a series of manually controlled reservoirs connected with polyethylene tubing to a low volume perfusion manifold (Automate Scientific, San Francisco, CA, USA). Dish fluid volume was kept low (< 500 μl) to minimize total solution exchange time (∼20 s).

Single channel recordings

Single channel activity was acquired based on the methods of Steinbach & Akk (2001). Briefly, recordings were made in the on-cell configuration with a pipette solution containing (mm): 125 KCl, 20 TEA-Cl, 10 MgCl₂, 0.1 CaCl₂, 10 d-glucose and 10 Hepes (pH 7.4, with KOH) at a pipette potential (V_pip) of +80 mV. In separate experiments, resting potentials measured in HEK293 cells immediately after establishing a whole-cell configuration were found to be −20 to −40 mV. Therefore we assume that the total membrane potential for experiments in the cell-attached configuration was −100 to −120 mV. When GABA or allosteric modulators were used, they were dissolved in the pipette solution. Patch pipettes for single channel recording had resistances of ∼8–10 MΩ. Recordings were filtered at 10 kHz and stored to VHS tapes at 100 kHz using a VR-10B pulse code module (Instrutech Corp.). For subsequent analysis, recordings were digitized at 50 kHz using Acquire software (Bruxton Corp., Seattle, WA, USA). Single channel events were detected after digital filtering to 3 kHz using the half-amplitude detection algorithm in QuB (http://www.qub.buffalo.edu), and event files were exported for analysis using TacFit software (Bruxton Corp.). The system dead and rise times were filter-limited (60 and 111 μs, respectively), and the distributions of dwell times were corrected for spuriously short events according to Colquhoun & Sigworth (1995). Analyses were restricted to portions of the recording with no evidence of multiple concurrent channel openings within 500 ms. Subconductance openings (which may have gone undetected using the 50% threshold) accounted for only a small percentage (< 2%) of events detected by the experimenter. Events collected from several patches under the same conditions were pooled before dwell time analysis. Dwell time distributions were fitted using a maximum-likelihood procedure employing correction for missed events (Colquhoun & Sakmann, 1985). Additional kinetic components were added until there was no significant change in the log-likelihood of the fit (Colquhoun & Sigworth, 1995). Assessment of closed duration dwell components were well fitted by four or five components where the longest closures (τ₃ through τ₅) may reflect closure of one receptor followed by sojourns to an open state of a different receptor, as well as long-lived desensitized states of a single receptor. Because these alternatives cannot be differentiated without knowledge of the numbers of channels in the patch (Colquhoun & Sakmann, 1985), long closures are likely to be underestimated. Bursts of openings that exclude these longest-lived closed states are, however, believed to reflect the activity of a single receptor (Sakmann et al. 1980; Colquhoun & Sakmann, 1985; Gibb &Colquhoun, 1991, 1992; Akk et al. 2001; Steinbach & Akk, 2001; Akk & Steinbach, 2003). Bursts were identified as groups of openings separated by a critical duration (T_crit= 17 ms) defined by the method of equal misclassifications (Colquhoun & Sakmann, 1985) such that short and intermediate gaps (closed times represented by τ₁ and τ₂) were classified as being within a burst (Colquhoun & Sakmann, 1985; Gibb & Colquhoun, 1991, 1992). The resulting burst distributions were fitted as described for open dwell time distributions.

Spectral analysis

For spectral analysis, recordings were digitized at 100 kHz through a 6 kHz Butterworth filter, and 250 sweeps of 4096 points each were captured. The fast Fourier transform (FFT) of each sweep was calculated after applying a tapered cosine window, and the resulting 250 individual power spectra were averaged. Both control and AAS spectra were readily fitted with two Lorentzian (1/f²) components (Colquhoun & Hawkes, 1977; Korn & Horn, 1988). Time constants (τ_i) were calculated from corner frequencies (f_c) by: τ_i= 1/(2πf_ci).

Kinetic modelling

Simulation of channel activity was used to determine kinetic transitions altered by AAS. Models were solved using Q-matrix techniques applied to non-stationary conditions (Colquhoun & Hawkes, 1977, 1995) using programs written (Brian L. Jones) in MatLab 6.1 (The Math Works; Natick, MA, USA; Yang et al. 2002, 2005). For spontaneous currents, the effects of 17α-MeT were simulated using a simple three-state kinetic scheme (in the absence of AAS) parameterized directly with microscopic rates estimated from macroscopic current data (see Fig. 7): a slowly equilibrating, non-conducting resting state (R) and rapidly equilibrating closed (C) and open (O) states. The rates connecting R to C (1/τ_Relax and 1/τ_Tail; 0.43 and 0.39 s⁻¹, respectively) were derived from the measurements of relaxation of peak inhibition and decay of the tail current. Closing rate (α) was estimated at 2564 s⁻¹ from spectral analysis. Opening rate (β) was adjusted to 900 s⁻¹ to provide an equilibrium open probability, P_o, of 0.15. The model provided a qualitatively accurate fit to the data for a wide range of estimates of P_o (0.15–0.5). To model the effects of 17α-MeT, a fourth AAS-bound state (B) was added. Blocking and unblocking rates (k_block= 2.1 × 10⁵m⁻¹ s⁻¹; k_unblock= 2.8 s⁻¹) were obtained from the macroscopic data.

Figure 7. Kinetic assessments of macroscopic currents suggest that 17α-MeT promotes inhibition of spontaneous openings by an allosteric block.

A–C, ultrafast perfusion was used to apply a maximally inhibitory concentration of 17α-MeT (100 μm). A, representative response illustrating the relaxation of inhibition during a 10 s application of 17α-MeT (τ_relax= 2.7 ± 0.3 s; n = 8) and the rebound current following cessation of 17α-MeT application (τ_tail= 2.5 ± 0.2 s; n = 8). B, currents elicited by progressively longer applications of 17α-MeT (0.5–32 s) were characterized by asymptotically increasing tail currents. C, time course of this increase in I_tail expressed as the percentage of I_holdversus the duration of exposure (τ_rebound= 2.8 ± 0.6 s, n = 6–7 per point). D, representative currents elicited by 1000 ms pulses of 17α-MeT (1, 3.16, 10 and 31.6 μm). The onset of inhibition could be described by a single exponential component (τ_block) that was dependent on [17α-MeT]. The decline in inhibition was fitted as a double exponential with one component (τ_unblock) that describes release from the blocked state (B; see F) and a second fixed component (τ_tail) that describes relaxation to the resting state (R; see F). E, linear regression indicated that the rate of inhibition (1/τ_block) was dependent on concentration (slope = 2.1(1) × 10⁵m⁻¹ s⁻¹; intercept = 2.8 ± 0.3 s⁻¹; r² > 0.98). The concentration dependence of unblocking (1/τ_unblock) was minimal, and the slope of the fit was constrained to zero (intercept = 2.7 ± 0.2 s⁻¹, n = 5–7 cells per point). F, macroscopic spontaneous current data could be modelled with a simple kinetic scheme parameterized directly with rates estimated from macroscopic current data that included a non-conducting and slowly equilibrating closed or resting state (R), a rapidly equilibrating closed state (C), an open state (O), and an AAS-bound and non-conducting blocked state (B). The rates connecting R to C (1/τ_relax and 1/τ_tail; 0.43 and 0.39 s⁻¹, respectively) were derived from the measurements of relaxation of peak inhibition and decay of the tail current. Closing rate (α) was estimated at 2564 s⁻¹ from spectral analysis (see Fig. 8). Opening rate (β) was adjusted to 1300 s⁻¹ to provide a P_o of 1.5–0.2. Blocking and unblocking rates were obtained from data as illustrated in D and E (k_block= 2.1 × 10⁵m⁻¹ s⁻¹; k_unblock= 2.8 s⁻¹). The three simulations shown are for AAS durations and concentrations of 10 s at 100 μm (G), 0.5–32 s at 100 μm (H) and 500 ms at 10, 31.6 and 100 μm (I) to simulate data shown in A, B and D, respectively.

Drugs

Stock solutions of zolpidem (N,N-6-trimethyl-2-(4-methylphenyl)-imidazo[1,2-a]pyridine-3-acetamide), 17α-MeT (17α-methyl-4-androsten-17β-ol-3-one) and picrotoxin (PTX; Sigma-Aldrich Co.) were made with cell culture grade dimethyl sulphoxide (DMSO) as the solvent and were diluted to achieve a final bath concentration ≤ 0.01% DMSO. Control experiments demonstrated that DMSO up to 0.1% had no effect on current properties (data not shown). Bicuculline methiodide (BMI; Sigma-Aldrich Co.) was prepared as an aqueous solution and kept frozen until time of use.

Statistical analysis

Values are reported as means ± s.e.m. For each cell, current parameters in the presence of a modulator (17α-MeT or zolpidem) were calculated as the percentage of the parameters measured under control conditions, and statistical significance was assessed using one-way two-tailed student's t tests. One-way ANOVA was used to establish the significance of differences measured under more than two experimental conditions; subsequent pairwise multiple comparisons were performed with student's t tests using the Bonferroni post hoc correction. For direct comparisons of dose–response and binding rate curves, two-way ANOVA with factors of receptor types and concentrations followed by a Bonferroni post hoc correction were used. This method allowed for model-free comparison of these data. This method of analysis is a useful adjunct to comparison of parameters derived from fitting of the empirically derived Hill equation, since the dose–response relationship described by the Hill equation is unlikely to apply to the gating mechanisms of real ion channels (Colquhoun, 1998). The uncertainties reported with parameters derived from curve fitting procedures are the s.e.m. values recovered from the fitting process. One- or two-sample two-tailed independent student's t tests were performed to determine significance between the responses of different combinations of modulators unless otherwise specified.

Results

Properties of GABA_A receptor-mediated currents in GnRH neurons

GABA-evoked responses in GnRH neurons

To determine the properties of GABA_A receptor-mediated currents in native GnRH neurons, whole-cell recordings were made from acutely dissociated fluorescent neurons derived from a transgenic mouse expressing GFP under the GnRH promoter (Suter et al. 2000). All animals used in these studies were males between the ages of postnatal days 20 and 30. To mimic the time course and concentration parameters of synaptic release, 1 mm GABA was applied for a brief duration (3 ms) using ultrafast perfusion. This application protocol elicited currents from GnRH neurons that were characterized by an average peak current (I_peak) of −2500 ± 500 pA, a maximum current density (σ_max) of −580 ± 90 pA pF⁻¹ and a total charge transfer (Q_tot) of −180 ± 20 pC (n = 15) (Fig. 1). Currents deactivated with a tri-exponential decay well described by time constants τ₁= 13 ± 1 ms (43 ± 3% contribution to I_peak), τ₂= 78 ± 6 ms (48 ± 3%) and τ₃= 357 ± 34 ms (11 ± 1%). Decay could also be summarized by a single weighted time constant, τ_w= 75 ± 7 ms (n = 15; Fig. 1).

Figure 1. GABA_A receptor-mediated currents in GnRH neurons.

A, GFP-GnRH neurons in a 100 μm slice through the mPOA of a male mouse at postnatal day 20. B, representative current evoked by a 3 ms application of 1 mm GABA to a dissociated GnRH neuron (V_H=−60 mV). C, dissociated GFP-GnRH neuron in the absence of GABA displayed tonic currents that were blocked by PTX and 17α-MeT (500 ms applications), but were unaffected by bicuculline methiodide (BMI).

Spontaneous responses in GnRH neurons

In the absence of GABA, a baseline holding current (I_hold) of 70 ± 20 pA was evident in ∼70% of dissociated GnRH neurons (n = 7; holding potential, V_H=−60 mV). Baseline holding currents were reduced by 68 ± 19 pA (10 ± 4 pA pF⁻¹) by the non-competitive GABA antagonist, PTX (100 μm). The competitive antagonist of the GABA binding site, bicuculline methiodide (BMI; 20 μm), was without effect on I_hold (Fig. 1). The strong antagonism of this current by PTX, but not by bicuculline, is consistent with the pharmacology and biophysical properties of a conductance mediated by spontaneous openings of unliganded GABA_A receptors containing the ɛ-subunit (Maksay et al. 2003; Fig. 1).

Inhibition by AAS of GABA_A receptor-mediated currents from GnRH neurons

As noted above, other well-studied allosteric modulators of the GABA_A receptor have paradoxical effects on responses mediated by ɛ-subunit-containing receptors. To assess AAS modulation of currents that are likely to reflect activity of ɛ-subunit-containing GABA_A receptors in GnRH neurons, the effect of a high concentration of 17α-MeT (10 μm) on I_hold was compared to that of a maximally effective concentration of PTX (100 μm) in the same cell. Application of 17α-MeT diminished I_hold by 42 ± 20 pA (10 ± 4 pA pF⁻¹). When the effects of PTX and 17α-MeT were compared within individual cells, the ratio of antagonism of I_hold produced by 10 μm 17α-MeT to that produced by 100 μm PTX was 0.4 ± 0.2 (n = 3). These data indicate that 17α-MeT acts as a negative modulator of the spontaneous current in GnRH neurons. While this negative modulation may reflect AAS action at ɛ-subunit-containing receptors, the extensive heterogeneity in GABA_A receptor subunits expressed in GnRH neurons, especially in juvenile mice (Sim et al. 2000; Todman et al. 2005), introduces variability in assigning properties of AAS modulation in these primary neurons to any specific class of receptor. To better understand the mechanisms by which inclusion of the ɛ-subunit affects AAS modulation, experiments were subsequently performed on recombinant α₂β₃ɛ-receptors expressed in HEK293 cells.

Properties of macroscopic currents mediated by recombinant α₂β₃ɛ-receptors

Concentration–response parameters for currents elicited by GABA from α₂β₃ɛ-receptors

Application of GABA elicited concentration-dependent responses from recombinant α₂β₃ɛ-receptors with an EC₅₀ of 3.9 ± 0.9 μm and a Hill slope of 0.52 ± 0.05 (n = 4; V_H=−60 mV; Fig. 2 and Table 2). While there are no published data for α₂β₃ɛ-receptors, these values are in general agreement with those reported for α₁β₃ɛ-receptors (EC₅₀ of 0.8 μm; Hill slope of 0.8; Neelands et al. 1999) and for α₂β₁ɛ-receptors (EC₅₀ of 11.2 μm; Hill slope of 1.1; Davies et al. 1997). The apparent affinity of GABA for α₂β₃ɛ-receptors was higher than that observed for α₂β₃γ_2L-receptors (EC₅₀ of 17 ± 1 μm; Yang et al. 2005). The lower EC₅₀ of α₂β₃ɛ- versusα₂β₃γ_2L-receptors is in keeping with a role for the ɛ-subunit-containing receptors in mediating extrasynaptic and tonic conductances that would most probably be activated in response to low concentrations of GABA.

Figure 2. Phasic and tonic responses elicited by GABA from recombinant α₂β₃ɛ-receptors.

A, representative currents elicited by perfusion of a 62.5–1000 ms application of 0.1 μm to 10 mm GABA. B, concentration-response relationship for currents elicited by 125–1000 ms pulses of GABA from 0.1 μm to 10 mm. For each cell, the I_peak of each current was normalized to the maximal response for that cell. Values represent 4 cells per point.

Table 2.

Properties of current responses and modulation by 17α-MeT for wild-type and G302S receptors

	α2β3ɛ	α2β3ɛ(G302S)
Responses to GABA
EC50 (μm)	3.9 ± 0.9	1.0 ± 0.2
Hill slope	0.52 ± 0.05	0.9 ± 0.2
σmax (pA pF−1)	−120 ± 40	−180 ± 60
Modulation of spontaneous current by 17α-MeT
σ (pA pF−1)	−13 ± 4	−24 ± 3
IC50 (μm)	8.7 ± 3.0	15.8 ± 3.4
Hill slope	0.77 ± 0.14	0.68 ± 0.05
Binding parameters for 17α-MeT
kblock (m−1 s−1)	2.1(1) × 105	1.70(1) × 105
kunblock (s−1)	2.8 ± 0.3	4.4 ± 0.1
Kd (μm)	13	26

Values represent means of 5–7 cells for each concentration–response parameter, 18–20 cells for estimates of σ of spontaneous current, and 4–9 cells for 17α-MeT binding parameters. Numbers in parentheses are s.e.m. for the least significant digit in the rates.

To assess AAS effects on α₂β₃ɛ-receptors under conditions that mimic synaptic release, 3 ms pulses of 1 mm GABA were applied. This agonist protocol elicited currents that activated rapidly with an I_peak of −1200 ± 230 pA, a σ_max of −205 ± 26 pA pF⁻¹ and a Q_tot of −87 ± 15 pC. Currents deactivated with a tri-exponential decay with time constants τ₁= 12 ± 2 ms (33 ± 4%), τ₂= 64 ± 14 ms (40 ± 3%) and τ₃= 175 ± 9 ms (28 ± 4%). Overall current decay was described by a single time constant (τ_w) of 79 ± 6 ms (n = 10; Fig. 3A). These deactivation parameters are similar to those that described decay of currents elicited from primary GnRH neurons (see above: spontaneous responses in GnRH neurons). Zinc inhibition of α₁β₁-receptors is dramatically diminished by inclusion of the ɛ-subunit (IC₅₀= 0.24 for α₁β₁-receptors versus 41.9 μm for α₁β₁ɛ-receptors; Whiting et al. 1997). In the present experiments, we also found that inhibition by 10 μm zinc was marked for responses elicited from cells transfected with α₂ and β₃ cDNAs alone, that this inhibition was diminished in cells transfected at a ratio of 1:1:0.1 (α₂:β₃:ɛ cDNA) and that it became insignificant at a transfection ratio of 1:1:1 (response amplitudes not different from those in the absence of zinc; data not shown). These data suggest that α₂β₃-subunit-containing receptors are largely absent in cells transfected at a 1:1:1 ratio and that the fast component of decay in these cells reflects the kinetics of α₂β₃ɛ-receptors.

Figure 3. Modulation by AAS of GABA-dependent currents through α₂β₃ɛ-receptors.

A, representative currents showing the response elicited by a phasic (3 ms) application of 1 mm GABA and the inhibition of I_peak induced by application of 1 μm 17α-MeT. The inhibitory effect of the AAS was reversible (I_peak during wash > I_peak during AAS) and I_peak returned to control levels when scaled for the run-down of the response that occurs in response to prolonged (10–30 min) application of 1 mm GABA alone (not shown). B, representative responses to stationary application of 1 μm GABA and 1 μm 17α-MeT, demonstrating reversible inhibition of I_tonic by this AAS. C, graphic representation of the effects of 1 μm 17α-MeT on parameters of currents elicited by brief (3 ms) pulses of 1 mm GABA (phasic: I_peak, A_tot, τ_w, τ_1–3,%τ_1–3) and steady-state application of 1 μm GABA (I_tonic). *P < 0.05, significantly different from control values.

Modulation by AAS of phasic responses mediated by α₂β₃ɛ-receptors

The AAS 17α-MeT (1 μm) induced a modest, but nonetheless significant, and reversible inhibition of currents elicited by phasic (3 ms) application of 1 mm GABA. Specifically, 17α-MeT reduced I_peak (to 84 ± 4% of control values, P = 0.006) and Q_tot (to 76 ± 6% of control values, P = 0.003; Fig. 3A and C). With respect to decay kinetics, 17α-MeT significantly decreased τ₁ (to 78 ± 3% of control values, P = 0.01), but also increased percentage τ₂ (to 116 ± 2% of control values, P = 0.04; n = 8), which resulted in no significant effect on the overall time course of decay as indicated by τ_w (94 ± 3% of control values, P = 0.1; Fig. 3A and C).

Modulation by AAS of tonic responses mediated by α₂β₃ɛ-receptors

Other positive allosteric modulators of GABA_A receptors have been shown to enhance GABA-evoked tonic conductances mediated by δ-subunit-containing receptors selectively (Stell et al. 2003; Wei et al. 2004). For experiments performed here, tonic current was defined as the GABA-activated current elicited by sustained (bath) application of GABA that does not include any contribution from spontaneously opening α₂β₃ɛ-subunit-containing receptors. Bath application of a low concentration of GABA (1 μm) resulted in tonic currents (I_tonic) from recombinant α₂β₃ɛ-subunit-containing receptors of −670 ± 120 pA (n = 8; Fig. 3B). One micromolar 17α-MeT significantly inhibited these tonic currents (to 82 ± 3% of control values, P = 0.003, n = 8; Fig. 3B and C). In contrast, 1 μm 17α-MeT was without effect on tonic currents elicited by 1 μm GABA from α₂β₃γ_2L-receptors (103 ± 9% of control values, P = 0.79, n = 7), consistent with previous results (Yang et al. 2005). This concentration of 17α-MeT was also without effect on tonic currents elicited by 10 μm GABA, which represents a concentration of agonist that is predicted to produce a fractional response at α₂β₃γ_2L-receptors equivalent to 1 μm GABA at α₂β₃ɛ-receptors (data not shown). The high-affinity BZ site agonist zolpidem (1 μm) also induced a small but significant inhibition of I_tonic from α₂β₃ɛ-subunit-containing receptors (to 90 ± 3% of control values, P = 0.02, n = 6). This is in contrast to the potentiation of I_tonic from α₂β₃γ_2L-receptors by 1 μm zolpidem (to 453 ± 46% of control values, P < 0.001).

Modulation by AAS of desensitization, paired pulse recovery and high-frequency responses mediated by α₂β₃ɛ-receptors

To determine whether AAS alter desensitization of α₂β₃ɛ-subunit-containing receptors, the effect of 17α-MeT on currents produced by 500 ms applications of 1 mm GABA was studied (Fig. 4). GABA-evoked currents activated quickly with a 10–90% rise time of 2.1 ± 0.2 ms (n = 10) and were described by an average maximal current (I_GABA) of −1230 ± 240 pA, a σ_max of 210 ± 31 pA pF⁻¹ and Q_tot of − 39 ± 32 μC (n = 10). Current desensitization was biphasic in seven out of 10 cells exposed to GABA, with a fast component, τ_des-f= 29 ± 11 ms (23 ± 6%) and a slow component, τ_des-s= 521 ± 85 ms (76 ± 6%). For the three cells that showed monophasic desensitization, decay was well described by τ_des-s. Overall desensitization was described by a single weighted time constant of deactivation (τ_des) of 410 ± 80 ms (n = 10; Fig. 4A–C). The ratio of steady-state to peak current under these conditions was 34 ± 6% (n = 10).

Figure 4. Anabolic androgenic steroid alters the kinetics of GABA-induced desensitization and slows paired pulse recovery.

A, currents evoked by 500 ms application of 1 mm GABA show significant desensitization. Exposure to 1 μm 17α-MeT reduces I_peak and accelerates desensitization without altering deactivation. B, currents from A normalized to peak illustrate acceleration of desensitization. Inset: deactivation of representative currents elicited by 500 ms pulses of GABA in the presence and absence of 17α-MeT scaled to the current level immediately before the end of application show no differences in kinetics. C, bar graph summarizing AAS-induced changes in deactivation and desensitization parameters. Desensitization during the pulse could be described by a fast (τ_des-f) and a slow component (τ_des-f) or a single weighted time constant, τ_des. Current deactivation following the pulse could also be described by a fast (τ_off-f) and a slow (τ_off-s) component or a single weighted function (τ_off). D, AAS effects on recovery to paired responses. Insets show representative currents elicited by paired 3 ms pulses of 1 mm GABA in the presence and absence of 1 μm 17α-MeT. Comparison of the recovery time course of the second response in the paired pulse protocol shows that 1 μm 17α-MeT slows response recovery. Percentage recovery of the second pulse (% Recovery), as assessed by (I_peak2− onset₂)/(I_peak1− onset₁) × 100, was plotted as a function of interpulse interval and described by two exponential components with time constants, τ_r1 and τ_r2. Pulses with delays of 25–12 800 ms were used for analysis, but currents are only displayed to 3200 ms for clarity. E, exposure to 1 μm 17α-MeT inhibited the response to repetitive stimulation, significantly decreasing Q_tot. *P < 0.05, significantly different from control values.

Of the 10 cells that were examined exposed to 500 ms pulses of 1 mm GABA, application of 1 μm 17α-MeT was successfully carried out in eight cells. Application of 1 μm 17α-MeT significantly shortened the 10–90% rise time in all eight cells (to 83 ± 4% of control values, P = 0.007), reduced I_peak in seven of eight cells (to 71 ± 9% of control values, P = 0.02) and accelerated desensitization in all eight cells (τ_des= 60 ± 8% of control values, P = 0.002) by accelerating the rate of slow desensitization τ_des-s (to 60 ± 9% of control values, P = 0.005; Fig. 4A–C). Current deactivation following 500 ms applications of GABA was biphasic, described by a fast and a slow time constant, τ_off-f and τ_off-s, respectively, or by a single weighted time constant, τ_off (135 ± 8 ms; Fig. 4C). Exposure to 1 μm 17α-MeT had no effect on deactivation following the pulse for currents produced in response to prolonged application of 1 mm GABA (P > 0.4 for all, n = 8; Fig. 4B and C).

Paired 3 ms applications of 1 mm GABA with varied interpulse intervals were used to assess the effect of 17α-MeT on the time course of recovery between successive applications of GABA. Ten paired applications with delays from 25 to 12 800 ms (doubling at each interval) were collected. The time course of recovery was bi-exponential with time constants of τ_r1= 95 ms (85%) and τ_r2= 2660 ms (15%; n = 10; Fig. 4D). Reponses recovered to 100% of the initial peak amplitude by 12 800 ms. Analysis of recovery curves by two-way ANOVA permitted a model-free comparison of AAS-induced changes in response recovery and indicated that 17α-MeT produced significant slowing of recovery (F_1,9= 4.43, P= 0.04, n= 7–10 per point) by increasing τ_r1 to 109 ms (155% of control values) and τ_r2 to 2450 ms (120% of control values; Fig. 4D).

Recordings made from native GnRH neurons indicate that these cells are subject to a high frequency of GABA_A receptor-mediated synaptic inputs (Sim et al. 2000; Nunemaker et al. 2003). To determine whether AAS exposure might alter overall charge transfer arising from high-frequency phasic responses mediated by ɛ-subunit-containing receptors, HEK293 cells expressing α₂β₃ɛ-receptors were exposed to trains (10 pulses, 3 ms in duration and 25 ms delay) of 1 mm GABA. This protocol produced an average net charge transfer during the train (Q_tot) of −4 ± 4 μC. Exposure to 1 μm 17α-MeT produced a significant decrease in Q_tot (to 64 ± 10% of control values, P = 0.01, n = 7) (Fig. 4E). This suggests that 1 μm 17α-MeT would have a net antagonistic effect on high-frequency phasic transmission mediated by ɛ-subunit-containing GABAergic synapses which may alter the episodic firing patterns observed in GnRH neurons (Sim et al. 2000; Nunemaker et al. 2003).

Properties of spontaneous current mediated by α₂β₃ɛ-recombinant receptors

Anabolic androgenic steroids administered in vivo, even at low doses, have been shown to diminish the release of luteinizing hormone (LH) dramatically (Bronson et al. 1996). The presence of a PTX-sensitive, bicuculline-insensitive current in GnRH neurons suggests that spontaneously active GABA_A receptors may play an important role in regulating the activity of these native neurons and their control of pituitary LH release, and thus may also be an important target with respect to AAS action. To determine how AAS interact with spontaneously active ɛ-subunit-containing GABA_A receptors, we first characterized the properties of spontaneous currents mediated by α₂β₃ɛ recombinant receptors and subsequently examined the effects of 17α-MeT on these currents.

Baseline I_hold(−240 ± 40 pA; −31 ± 7 pA pF⁻¹; V_H=−60 mV) was evident in all cells expressing α₂β₃ɛ-receptors (n = 10; Fig. 5). Picrotoxin (100 μm; 500 ms) reversibly antagonized I_hold (by 65 ± 7%). The portion of I_hold antagonized by PTX reversed at −0.3 ± 1 mV, consistent with the predicted reversal potential for Cl⁻ under our recording conditions, and showed subtle outward rectification with a ratio of 1.24 ± 0.08 for I_hold(+60mv):I_{hold(−60mV)} (Fig. 5D; n = 10). Taken together, these data indicate that I_hold arises from spontaneous openings of GABA_A receptors expressed in the HEK293 cells.

Figure 5. Recombinant α₂β₃ɛ-receptors in HEK293 cells open in the absence of GABA.

A, recombinant α₂β₃ɛ-receptors were activated by 1 mm GABA, but also displayed a notable level of spontaneous current in the absence of GABA that could be blocked by 100 μm of the non-competitive GABA_A antagonist, PTX. Blockade of spontaneous openings appears as an outward deflection in I_hold(V_H=−60 mV). B, voltage steps from −60 to +60 mV in control saline (bath) or in saline supplemented with 100 μm PTX (+PTX) show that application of PTX is associated with a decrease in I_hold at all potentials. C, membrane conductance at all potentials is decreased by 100 μm PTX. D, in solutions containing symmetrical chloride concentrations, I_hold reverses at ∼0 mV (V_r). Collectively, these observations indicate that I_hold arises from unliganded openings of GABA_A receptors.

To determine what percentage of the expressed GABA_A receptors contributes to I_hold, estimates of the proportion of α₂β₃ɛ-receptors open at equilibrium (P_o) in the absence of GABA were determined in two ways. First, assuming that the population of receptors that open spontaneously can also be activated by GABA, the ratio of maximal I_GABA plus I_hold (=I_max) to I_hold can provide an estimate of P_o for spontaneously opening GABA_A receptors. Prolonged (500 ms) applications of 1 mm GABA elicited currents from cells expressing α₂β₃ɛ-receptors that were characterized by an average amplitude (I_GABA) of −1200 ± 200 pA (V_H=− 60 mV; n= 10). Assuming that I_GABA elicited by 1 mm GABA reflects a P_o of ∼0.8, as determined from our kinetic modelling of macroscopic current data for responses to 1 mm GABA by α₂β₃ɛ-subunit-containing receptors (data not shown) and consistent with previous studies of activation of γ_2L-subunit-containing GABA_A receptors by 1 mm GABA (Jones & Westbrook, 1995; Akk et al. 2001), I_GABA at a maximal P_o= 1.0 would be predicted to be −1500 pA and I_max would represent this value plus the contribution from the spontaneous current (I_max= 1740 pA) and thus a ratio of I_hold:I_max of 0.14. No detectable spontaneous current was present in untransfected HEK293 cells or those transfected with α₂ and β₃, but not ɛ, subunit cDNAs, although the latter expressed appreciable GABA-induced currents, indicating that functional GABA_A receptors were present. Therefore, we believe that all of I_hold can be attributed to spontaneous openings of α₂β₃ɛ-receptors and that the ratio of I_hold:I_max therefore provides an estimate of P_o for spontaneous α₂β₃ɛ-receptors of 0.14. That 100 μm PTX does not provide a complete inhibition of I_hold for recombinant α₂β₃ɛ-receptors in HEK 293 cells is consistent with previous reports of recombinant α₁β₃ɛ-receptors expressed in oocytes (Maksay et al. 2003).

An approximation of P_o can also be made from spectral analysis of channel noise (Colquhoun & Hawkes, 1977). In the power spectra of channel noise from recombinant α₂β₃ɛ-receptors, the contribution of the 0.39 ms component, which we believe corresponds to the apparent mean channel open time, is 21 ± 3% (Fig. 8C). Data from these two approaches suggest that the P_o for unliganded α₂β₃ɛ-receptors is ∼0.15–0.2.

Figure 8. Anabolic androgenic steroid does not alter channel open dwell time.

A, representative spontaneously opening single channel currents (downward deflections) recorded in the on-cell configuration (V_pip=+80 mV). Most events are < 1 ms; attenuation of peak amplitude by filtering is evident. B, noise currents recorded in the presence or absence of a maximally inhibitory concentration of 17α-MeT (100 μm), which decreased mean current level and variance. C, power spectra for AAS and control conditions. Both control and AAS spectra were well described by two Lorentzian components (τ_f and τ_s). D, bar graph summarizing changes in spectral parameters. Anabolic androgenic steroid did not induce a change in the time constant of either the fast (τ_f) or the slow time constant (τ_s), but did decrease the absolute power of both components (S_f and S_s). *P < 0.05, significantly different from control values.

Modulation by AAS of spontaneous current mediated by recombinant α₂β₃ɛ-receptors

Bath application of 17α-MeT produced a concentration-dependent inhibition of I_hold with an IC₅₀ of 15 ± 7 μm and a Hill slope of 0.61 ± 0.07 (Fig. 6). Analysis of concentration–response data acquired using rapid agonist perfusion provided an IC₅₀ of 8.7 ± 3 μm and a Hill slope of 0.77 ± 0.14; differences that most probably reflect the better resolution of peak responses with the more rapid method of agonist application. The ratio of inhibition produced by a high concentration of 17α-MeT (10 μm) versus a high concentration of PTX (100 μm) was 0.58 ± 0.03 for recombinant α₂β₃ɛ-receptors; a value similar to that observed in native GnRH neurons (0.4 ± 0.2, see above: spontaneous responses in GnRH neurons).

Figure 6. Dose–response relationship for AAS inhibition of spontaneous GABA_A receptor-mediated current.

A, representative current trace showing dose-dependent inhibition of spontaneous current by 17α-MeT. To control for variability in this spontaneous holding current (I_hold), the peak of inhibition was normalized to that produced by a saturating concentration of PTX (100 μm). B, when fitted with the Hill equation, these data gave half-maximal inhibition (IC₅₀) of 15 ± 7 μm and a Hill slope of 0.61 ± 0.07. Values represent 5–7 cells per point.

Two notable kinetic features of modulation of I_hold by 17α-MeT were apparent in these records. First, even in the continued presence of this AAS, the peak of inhibition relaxed to a lower level. Second, removal of 17α-MeT produced a large, resurgent tail current (I_tail; Figs 6A and 7A and B). Kinetic assessments of the changes in peak inhibition during the application of 17α-MeT and in the decline of I_tail following the application of 17α-MeT provided estimates of the time constant for the relaxation of peak inhibition (τ_relax) of 2.7 ± 0.3 s (n = 8) and of the time constant describing the decline in I_tail (τ_tail) of 2.5 ± 0.2 s (n = 8; Fig. 7A). In experiments where 17α-MeT was applied for variable durations, the amplitude of I_tail increased asymptotically with the duration of application, and the relationship between pulse duration and I_tail amplitude could be described by a single exponential (τ_rebound) of 2.8 ± 0.6 s (Fig. 7B and C). That τ_relax and τ_rebound are effectively identical suggests that the ‘extra channels’ that open following removal of drug and give rise to the resurgent I_tail come from the same reservoir of closed channels that feeds the relaxation of peak inhibition (τ_relax). Furthermore, these data suggest that 17α-MeT may act by causing receptors residing in a stable, slowly equilibrating resting state (R) to move to a blocked state (B), from where transitions to the open state (O) are more likely to occur than are transitions to the open state from the resting state (Fig. 7F and Discussion).

To examine the dynamics of inhibition of I_hold by 17α-MeT, the rates of inhibition and relaxation of peak inhibition were assessed at 1, 3.16, 10 and 31.6 μm 17α-MeT (500 ms). The onset of inhibition was fitted as a single exponential (τ_block; Fig. 7D). For concentrations less than 1 μm, the signal-to-noise ratio of the recordings was inadequate for curve fitting. For other concentrations, the rate of relaxation of inhibition was assessed by fitting the decline in inhibition (I_tail) with a double exponential function with one component (τ_unblock) that described release from the blocked state and a second component (τ_tail) that described relaxation to the resting state (Fig. 7E). This procedure allowed the unblocking rate to be disentangled from the rate of relaxation of I_tail (τ_tail). The dependence of the rate of blocking (1/τ_block) on concentration was described by a linear relationship with a slope of 2.1(1) × 10⁵m⁻¹ s⁻¹ and an intercept of 2.8 ± 0.3 s⁻¹ (r² > 0.98; Fig. 7E). The concentration dependence of unblocking (1/τ_unblock) was minimal, and the slope of the fit was constrained to zero, producing an intercept of 2.7 ± 0.2 s⁻¹ (Fig. 7E). A simple bimolecular binding mechanism implies that the slope of the blocking rate curve is the binding rate (k_block) and that the intercept of the blocking rate curve is the unbinding rate (k_unblock; Newland & Cull-Candy, 1992; Fig. 7F).

The linear dependence of 1/τ_block on concentration is consistent with several mechanisms of inhibition, including physical obstruction of the channel pore (open channel block) and stabilization of a new or existing closed state (allosteric block). However, the inhibition of I_hold produced by 100 μm 17α-MeT exhibited a small but significant dependence on voltage, where the block at +60 mV was 88 ± 3% of the block at −60 mV (P = 0.03; n = 4). This observation is not consistent with a mechanism in which 17α-MeT acts to physically obstruct the channel pore, since reversal of ion flux might be expected to disturb such an obstruction even for an uncharged molecule. To differentiate between an open channel block and an allosteric block as the mechanism of AAS inhibition more fully, the effects of 17α-MeT on single channel properties were assessed.

Properties of single channel currents mediated by recombinant α₂β₃ɛ-subunit-containing receptors

Effects of AAS on open dwell times of spontaneous events

Open channel block mechanisms reduce channel open dwell time by providing an additional non-conducting state beyond normal closing (Sakmann et al. 1980). In contrast, inhibition of channel activity characterized by a decrease in channel frequency with no change in time constants of burst length is consistent with enhanced occurrence of a desensitized state or an allosterically blocked state (Newland & Cull-Candy, 1992). Ideally, an AAS-dependent change in open dwell time could be detected with single channel analysis; however, the mean open times of spontaneous events mediated by α₂β₃ɛ-subunit-containing receptors were quite brief (Fig. 8A), making a direct single channel assessment of AAS effects on spontaneous events difficult. Consequently, in order to discriminate between an allosteric versus an open channel block mechanism, the effect of a maximally inhibitory concentration of 17α-MeT (100 μm) on spontaneous current noise was examined by spectral analysis.

Noise spectra, both under control conditions and in the presence of 17α-MeT, were well described by two Lorentzian components with time constants of τ_I= 0.39 ± 0.08 ms (21 ± 3%) and τ_s= 2.8 ± 0.3 ms (79 ± 3%; Fig. 8C), suggesting that the 0.39 ms component of the noise spectrum reflects the apparent mean channel open time for spontaneously gated receptors. Values of τ_f are in agreement with the time constant reflecting the brief component of the open dwell distribution (0.388 ± 0.057 ms; 57%) for α₁β₃ɛ-subunit-containing receptors gated by a low (1 μm) concentration of GABA (Neelands et al. 1999) and the briefest component of the open and burst duration distributions (τ₁) we report here for currents elicited by 1 mm GABA from α₂β₃ɛ-subunit-containing receptors (Table 1).

Table 1.

Open dwell and burst duration characteristics of α₂β₃ɛ-subunit-containing receptors

17α-MeT (μm)	τ1 (ms)	τ2 (ms)	τ3 (ms)	τ4 (ms)	%τ1	%τ2	%τ3	%τ4	Cells	Events or bursts
Open dwell components
0	0.84	5.5	33	—	83	10	7	—	4	140 500
10	0.61	4.0	18	—	65	18	17	—	7	74 577
31.6	0.61	5.2	20	215	52	37	10	0.03	2	4264
100	0.93	6.2	—	—	26	74	—	—	4	6352
Burst dwell components
0	0.58	28	219	—	27	52	21	—	4	2974
10	0.66	27	187	—	31	45	24	—	7	3417
31.6	0.84	10	129	—	41	36	23	—	2	947
100	1.0	8.6	39	—	16	72	12	—	4	1411

Responses elicited by exposure to 1 mm GABA in cell-attached patches. Values represent means of 2–7 cells at each concentration. Assessments of the effects of 17α-MeT were made both for apparent open durations and also for burst durations, since bursts are believed to reflect the activity of a single receptor within the patch.

Application of 100 μm 17α-MeT did not induce a significant change in either τ_f or τ_s (P = 0.41, n = 5), but did decrease the power of both components, S_f (to −57 ± 7% of control values, P = 0.001) and S_s (to −46 ± 9% of control values, P = 0.007). Exposure to this AAS also decreased the relative proportion of the fast (%S_f) component (to 83 ± 5% of control values, P = 0.03) while increasing the contribution of the slow (%S_s) component (to 105 ± 1% of control values, P = 0.02, n = 5; Fig. 8D). These findings suggest that AAS do not readily interact with the open state of the receptor.

Effects of AAS on apparent dwell times and burst durations of GABA-evoked single channel events

Cell-attached recordings of single channel activity elicited by 1 mm GABA were made in the presence and absence of three concentrations of 17α-MeT (10, 31.6 and 100 μm). Single channel openings appeared in relatively long bursts of activity separated by long periods of inactivity (Table 1 and Fig. 9A). Intraburst open dwell distributions could be described adequately with three exponential components, suggesting three distinct open states (Fig. 9B and E). The distribution for 31.6 μm was better fitted by four components, but the area of the slowest components was 0.03%, accounting for ∼12 unusually long events in approximately 30 min of recording (Table 1). Anabolic androgenic steroids had little if any effect on the time constants of channel open dwell times (Table 1 and Fig. 9E), but did produce a concentration-dependent redistribution of the relative contribution of each component to total channel activity (Table 1 and Fig. 9E). Specifically, 17α-MeT selectively diminished the proportions of the fastest and slowest open dwell times, while causing an increase in the proportion of the intermediate component. As with spectral analysis, this finding is not consistent with an open-channel block mechanism, where it would be predicted that longer events should be preferentially blocked (Colquhoun & Hawkes, 1995), but is consistent with an allosteric mechanism, in which interactions with AAS favour the receptor-occupying conformations from which the channel is more likely to open to the intermediate-duration open state over the longer and shorter open states.

Figure 9. Anabolic androgenic steroid does not alter channel open dwell times, but shortens burst durations.

A, representative single channel openings (downward deflections) that would constitute a burst that were recorded in the cell-attached configuration (V_pip=+80 mV) elicited by 1 mm GABA in the presence and absence of 17α-MeT. B, open dwell distribution histograms from single channel currents evoked by 1 mm GABA with three concentrations of 17α-MeT (10, 31.6 and 100 μm). C, burst dwell duration histograms show a dramatic reduction in length with increasing concentrations of 17α-MeT. D, closed dwell distributions show evidence that a long-lived closed state becomes more frequent at higher concentrations of 17α-MeT. sqrt. frequency, square root of the frequency. E, time constants and associated areas for the exponential components fitted to the open dwell histograms in B. Time constants show very little change with varying concentrations of 17α-MeT, but their relative proportion changes, favouring residence at higher concentrations of 17α-MeT in the state described by the intermediate time constant (τ₂; see also Table 1). Note that no data point is shown for dwell times corresponding to τ₃ at 100 μm 17α-MeT because these longest events were not detected at this high concentration of the AAS (B and Table 1). F, time constants and associated areas for the exponential components fitted to the burst dwell time histograms in C. Time constants for burst dwell durations show little change with increasing concentrations of 17α-MeT, indicating that changes in burst durations are achieved by general shifts in gating mode favouring the intermediate mode (τ₂). G, the change in mean burst duration has an IC₅₀ of 22.512 ± 0.007 μm, suggesting that changes in burst duration can account for a large portion of the inhibitory effect of 17α-MeT on GABA-evoked currents.

The effects of 17α-MeT on bursts of openings were also assessed. As with single channel open dwell times, 17α-MeT did not appreciably alter the time constants of the burst durations, but did promote a concentration-dependent decrease in overall duration with an IC₅₀ of 22.512 ± 0.007 μm (Table 1 and Fig. 9F and G), attenuating the relative contributions of the longest and shortest components while enhancing the intermediate component (Table 1 and Fig. 9C). As with the observed effects of 17α-MeT on open dwell distributions, the effects on burst durations are not consistent with an open channel block mechanism (Colquhoun & Hawkes, 1995), but suggest an allosteric effect in which the AAS may act to trap receptors in a gating mode with intermediate efficacy.

While analysis of closed dwell distributions can be insightful, in multiple-channel patches it can be easily confounded by the occurrence of longer closed dwell times. With this caveat in mind, the allosteric model under consideration suggests that as the concentration of 17α-MeT increases, the number of long closures produced by the steroid should increase in parallel, and examination of the closed dwell distributions supports this assumption (Fig. 9D).

Functional analysis of receptors containing a TM2 point mutation

Structural considerations

The activity of the AAS at the GABA_A receptor resembles that of a host of other compounds that have a rather promiscuous relationship with multiple receptor isoforms. Interestingly, a single conserved residue in TM2 of each α-, β- and γ-subunit (Fig. 10A) has been implicated in the activity of a variety of structurally dissimilar compounds, including specific anaesthetics (Belelli et al. 1997, 1999; Jenkins et al. 2001; Krasowski et al. 2001), loreclezole (Wingrove et al. 1994), ethanol (Mihic et al. 1997) and BZs acting at a distinct low-affinity composite site (Walters et al. 2000). It has been suggested that these conserved residues in TM2 do not face the channel pore, but rather face into the membrane, perhaps interfacing with adjacent residues in TM1 and TM3, to form a pocket capable of binding these structurally diverse compounds (Jenkins et al. 2001; Krasowski et al. 2001; Bera et al. 2002), A notable point of sequence divergence in this TM2 pocket between the ɛ-subunit and other receptor subunit isoforms is residue 302 (glycine in ɛ; serine in γ_1–3), which has been implicated in regulating gating, the sensitivity of receptor to GABA and control of spontaneous activity (Ueno et al. 2000; Scheller & Forman, 2002; Miko et al. 2004). To test whether this residue is important in conferring the positive modulation by 17α-MeT at α₂β₃γ_2L-receptors versus the negative modulation by this AAS at α₂β₃ɛ-receptors, AAS modulation was assessed for receptors containing a mutant (G302S) ɛ-subunit.

Figure 10. Potential sites and mechanism of AAS interaction.

A, cartoon illustrates the four proposed transmembrane (TM) domains of a single GABA_A receptor subunit and regions believed to form a promiscuous low-affinity site between TM2 and TM3 (blue). An alignment of portions of the TM2 and TM3 regions of the receptor shows conserved residues crucial for the interactions of other allosteric modulators at this putative site in blue along with the homologous residues in GABA_A receptor subunits that may play a role in AAS activity. While the exact beginning and end of the transmembrane regions and the nature of the linker region are uncertain, the overlapping brackets (See areas of overlap indicated beneath TM2-Linker-TM3) reflect the possible boundaries in these domains proposed by Bera et al. (2002). The black text reflects the linker proposed by comparisons with the nicotinic acetylcholine receptor structure (Miyazawa et al. (2003). B, GABA elicited concentration-dependent currents from wild-type α₂β₃ɛ- and mutant α₂β₃ɛ(G302S)-receptors. Maximal currents were elicited by 125–1000 ms pulses of GABA from 0.1 to 10 mm. For each cell, the I_peak of each response was normalized to the maximal response for that cell. The response of α₂β₃ɛ(G302S) to GABA was significantly different from that of wild-type receptors (P = 0.003). See Table 2 for parameters. C, peak GABA-evoked currents were not significantly different between the two types of receptors; however, D shows that spontaneous current flux through the mutant receptors was significantly (P = 0.007) greater than that through wild-type receptors. E, the concentration–response curve of α₂β₃ɛ(G302S)-receptors was significantly different from that of wild-type receptors (P = 0.007). For purposes of display only, concentration–response curves were normalized to the asymptotes of their respective Hill fits. F, linear regression analysis indicates that the binding rate curves for 17α-MeT to wild-type and α₂β₃ɛ(G302S)-receptors are significantly different (P = 0.009).

Functional characteristics

Maximal peak current density (Fig. 10B and C) for responses elicited by GABA was not significantly different between α₂β₃ɛ- and α₂β₃ɛ(G302S)-receptors; however, the G302S mutation resulted in a small increase in the Hill slope and a left shift for the concentration–response relationship (0.1 μm to 10 mm GABA; P = 0.003). Wild-type and α₂β₃ɛ(G302S)-receptors were found to activate at the same rate, but α₂β₃ɛ(G302S)-receptors exhibited diminished desensitization during prolonged applications of GABA. Moreover, spontaneous flux was significantly increased in α₂β₃ɛ(G302S)-receptors relative to wild-type receptors (P = 0.01; Fig. 10D and E and Table 2). To determine whether the G302S altered the modulation induced by 17α-MeT, the inhibition of spontaneous current by 17α-MeT (0.316–100 μm) was determined as for wild-type receptors. The inhibition of baseline current produced by 100 μm PTX on both the wild-type and mutant receptors was similar (79 ± 4 versus 71 ± 2%, respectively, P = 0.26, n = 8 and 22). However when the effect of 17α-MeT was considered in terms of percentage inhibition of baseline current without normalization, the G302S mutant was significantly less sensitive to inhibition than the wild-type receptor at subsaturating concentrations of the steroid (P = 0.008). After normalization to the peak inhibition produced by 100 μm PTX, the concentration–response curve of α₂β₃ɛ(G302S) was significantly different from wild-type receptors (P = 0.007), an effect that could be attributed to the mutation producing a right shift in the IC₅₀ for 17α-MeT of ∼1.8-fold (Fig. 10D).

AAS effects on binding rate

To determine whether the observed shift in the sensitivity to 17α-MeT produced by the ɛ(G302S) mutation could be explained in terms of changes in binding of the AAS, the concentration dependence of the blocking and unblocking rate for the spontaneous current was determined at 0.316, 1, 10 and 31.6 μm 17α-MeT. As with the wild-type receptor, the onset of inhibition was fitted as a single exponential (τ_block). Wild-type and α₂β₃ɛ(G302S) receptor blocking curves were significantly different (P = 0.009; Table 2 and Fig. 10F). The dissociation (K_d) rate calculated from the blocking and unbinding rate in the ɛ(G302S) mutant, was diminished 2-fold compared to wild type; consistent with the 1.8-fold left shift in IC₅₀ observed in the concentration–response relationship to 17α-MeT. Thus, it appears that the ɛ(G302S) mutant interferes with 17α-MeT inhibition of spontaneous current by reducing the affinity of the binding site for 17α-MeT. It is possible that if 17α-MeT cannot associate with the open state, the higher apparent spontaneous activity level observed in the ɛ(G302S) mutant might lead to slowed binding because the receptor would spend more time in the steroid-inaccessible open state. To rule out this possibility, the blocking rate of 100 μm PTX was compared at both wild-type and ɛ(G302S)-receptors. Picrotoxin binds preferentially to open receptors; consequently, its blocking rate increases proportionately with the level of channel activation (Dillon et al. 1995). The blocking rate of 100 μm PTX between wild-type and ɛ(G302S)-receptors was not significantly different (P = 0.48), with values of 16 ± 1 s⁻¹ for wild-type receptors (n = 5) and 17 ± 1 s⁻¹ for ɛ(G302S)-receptors (n = 6); likewise, the maximum efficacy of PTX at both receptor types is comparable (see above). From this observation, it seems likely that the shift in equilibrium state occupancy between wild-type and ɛ(G302S)-receptors is insufficient to produce the observed decrease in 17α-MeT binding rates, providing additional evidence for a specific perturbation in the 17α-MeT binding site produced by this point mutation in the TM2 domain.

Discussion

AAS are synthetic derivatives of testosterone originally designed for the treatment of hypogonadal dysfunction in men, initiation of delayed puberty and growth promotion (Basaria et al. 2001; Shahidi, 2001). Although AAS continue to be used clinically, currently most AAS are taken illicitly in patterns that constitute abuse and have been associated with a plethora of untoward physical and behavioural effects (Lukas, 1996; Bahrke et al. 1996, 1998; Franke & Berendonk, 1997; Friedl, 2000). Individuals who self-administer AAS do so at levels that are often greatly in excess of physiological or therapeutic levels of gonadal steroids and current estimates suggest that AAS reach micromolar levels in these subjects (Wu, 1997; Daly et al. 2001). Moreover, chemical modifications made to enhance half-life have also endowed these synthetic compounds with characteristics not evident in testosterone, including the profile of allosteric modulation at the GABA_A receptor (for review, see Clark & Henderson, 2003; Clark et al. 2004).

While the AAS and the neurosteroids, especially androgenic compounds such as 3α-diol, share obvious similarities, key structural moieties believed to be critical for neurosteroid modulation are absent in the AAS (Gee et al. 1988; Turner & Simmonds, 1989; Purdy et al. 1990; for review, see Lambert et al. 1995, 2003). Moreover, the AAS and the neurosteroids do not share common subunit dependence with respect to their modulatory abilities (for discussion, see Yang et al. 2002, 2005). The presence of the ɛ-subunit has been shown to confer an unusual pharmacological profile with respect to allosteric modulation by high-affinity BZ site compounds, barbiturates and neurosteroids (Whiting et al. 1997; Thompson et al. 1998, 2002; Davies et al. 1997, 2001; Neelands et al. 1999; Kasparov et al. 2001; Maksay et al. 2003; Irnaten et al. 2002). The goal of the present study was to determine whether inclusion of the ɛ-subunit also conferred an unexpected modulatory profile for the AAS, and the repercussions that expression of this subunit might have with regard to AAS effects on GnRH neurons and neuroendocrine control.

Here we report that the commonly abused AAS 17α-MeT (for review, see Clark & Henderson, 2003) modulates both GABA-evoked and spontaneous currents through recombinant α₂β₃ɛ-receptors, but does so in a fashion that markedly contrasts with the pattern of modulation for receptors composed of α-, β- and γ- or α-, β- and δ-subunits (Yang et al. 2002, 2005). In brief, while this AAS acts as a positive modulator or is without effect at α₁β₃γ_2L-, α₁β₃γ₁-, α₂β₃γ_2L-, α₁β₃γ₁- and α₂β₃δ-receptors (Yang et al. 2002, 2005; Clark et al. 2004), it induced negative modulation of recombinant α₂β₃ɛ-receptors. Significant inhibition of both GABA-evoked (tonic and phasic) and spontaneous activity was evident with 1 μm 17α-MeT, a concentration evident in the CSF of volunteers given moderate and short-term administration of this AAS (Daly et al. 2001), which probably reflects an appreciably lower level than that found high-dose users who chronically and illicitly abuse these steroids. For responses evoked by 1 mm GABA, 1 μm 17α-MeT significantly reduced I_peak, diminished the 10–90% rise time, accelerated desensitization and slowed recovery in response to paired, brief applications. This concentration of 17α-MeT had no overall effect on macroscopic deactivation, but did significantly increase the percentage of current carried by the intermediate component of decay (τ₂), an effect that is consistent with its observed action to selectively increase the proportion of the intermediate single channel open dwell times, and one that became more prominent as the concentration of the AAS was increased above 1 μm (Fig. 9B and C).

Exposure to 17α-MeT also blunted the ability of cells to respond to a rapid train of 1 mm GABA pulses and significantly decreased tonic currents elicited by stationary application of 1 μm GABA. While the subcellular localization of receptors containing the ɛ-subunit has not been determined using immunolocalization approaches, physiological data suggest that ɛ-containing receptors are present in both synaptic and extrasynaptic regions of primary hypothalamic neurons (Sergeeva et al. 2005). With respect to synaptic effects, GnRH neurons are characterized by a high frequency of GABA_A receptor-mediated events (Sim et al. 2000; Nunemaker et al. 2003). We show here that 17α-MeT significantly dampened high-frequency responses mediated by recombinant α₂β₃ɛ-receptors and predict that this AAS would similarly dampen high-frequency synaptic responses in GnRH neurons expressing synaptic ɛ-subunit-containing receptors and thus may interfere with GABAergic control of pulsatile LH release. For tonic currents, the low EC₅₀ of α₂β₃ɛ-receptors for GABA (∼4 μm) suggests that they would be good candidates to mediate tonic and spillover conductances that are known to play a critical role in determining neuronal input–output relationships (Semyanov et al. 2004). Therefore, unlike δ-subunit-containing receptors, which appear to be solely extrasynaptic, 17α-MeT may have a significant impact in neurons expressing ɛ-subunit-containing receptors via both synaptic and extrasynaptic mechanisms.

Recombinant receptors containing α₂-, β₃- and ɛ-subunits (ɛ encoded by the human cDNA) also exhibited a PTX-sensitive, bicuculline-insensitive current in the absence of agonist that we attribute to spontaneous opening of these receptors. Assessment of both macroscopic currents and fluctuation analysis provided estimates of P_o for these receptors of ∼0.15–0.20. This value is appreciably greater than that estimated for human recombinant α₂β₁ɛ-receptors (0.04; Wagner et al. 2005) and may reflect differences conferred by the β-subunits. Inhibition of spontaneous current elicited by 17α-MeT was concentration dependent (IC₅₀= 8.7 ± 3 μm) and was characterized by both a slow relaxation of inhibition during exposure to 17α-MeT and a rebound (tail) current following AAS removal. Both the extent of relaxation and the amplitude of the tail current increased with the duration of AAS. These two processes were described by nearly identical time constants, suggesting that they reflect the same population of channels that moves into a blocked state in the presence of the AAS and preferentially rebound into the open state upon its removal.

Further assessment of the temporal characteristics of AAS inhibition, relaxation and rebound at different AAS concentrations suggested that the mechanism of AAS modulation could be consistent with either an open channel or an allosteric mechanism of block. However, the voltage dependence of AAS inhibition of the spontaneous current, in conjunction with results from single channel and fluctuation analyses (that demonstrated that this AAS altered the proportion of channels within a given open dwell or burst distribution without changing the time constants of those distributions), suggest that this AAS does not readily interact with the open state of the receptor. Conversely, our results are consistent with a mechanism of allosteric block.

The observed effects of 17α-MeT on I_peak, desensitization and deactivation kinetics of GABA-evoked activity suggest an allosteric mechanism in which this AAS interacts with specific conformation(s) of the receptor to generate a stable AAS-bound closed state. For example, the observed effect on desensitization and recovery could be reconciled if the interaction of 17α-MeT with the receptor was more efficient in certain GABA-bound conformations than in the unliganded form of the receptor. In this mechanism, interactions with AAS would be the equivalent of introducing additional desensitized states to the gating of the receptor. Introduction of stable desensitized-like states would be expected to reduce the single channel open probability by shortening burst durations, but not change open dwell time constants.

The mechanism by which we envision the AAS elicit this allosteric block can be encapsulated as follows. In the absence of steroid, most receptors reside in a slowly equilibrating resting state (R), but make occasional forays to the closed state (C) from which they can shuttle rapidly back and forth to the open state (O) several times before transitioning back to state R. This flickering between C and O is the origin of the spontaneous current. When the receptors are exposed to 17α-MeT, which interacts preferentially with state C, they become trapped in the inactive AAS-bound state (B). State B is relatively stable compared to the other states and, in the presence of the AAS, a large number of receptors accumulate there. Initially, most of these receptors come from the rapidly equilibrating states C and O (thus the inhibition develops quickly as receptors in O and C are rapidly depleted), but as these states are depleted, receptors emerge from the slowly equilibrating state R, which can enter the C and O states, transitioning between them, before subsequently moving to the absorbing blocked state as well. This produces the relaxation of peak inhibition. Upon rapid withdrawal of 17α-MeT, receptors are released from state B back to state C. These receptors can then transition to states O or R, but because the rate of entry to state O is ∼3000 times faster than the rate of entry to state R (1300 versus 0.39 s⁻¹), most of the receptors open and close several times before their first sojourn to state R. These rapid activations produce the resurgent current that then decays as receptors percolate back to state R according to 1/τ_tail. An essential feature of this model is the preferential binding of the steroid to particular closed states. Similar results can be obtained by allowing blocking from state O, but this transition is not supported by experimental data.

With respect to the mechanism of inhibition of GABA-induced currents by 17α-MeT, we note that there are similarities with the effects of the neurosteroid pregnenolone sulphate (PS) on GABA-induced currents (Shen et al. 2000; Akk et al. 2001), but also some notable differences. Specifically, while Akk et al. (2001) found that PS decreased the duration of channel clusters elicited by 1 mm GABA without altering open dwell time constants (apparent mean openings and cluster analysis), they also report no changes in the proportion of openings corresponding to those open duration time constants. In contrast, while 17α-MeT also inhibits GABA-induced currents without altering open or burst duration time constants, it does cause a significant shift in the distribution of openings. This difference underscores the more general conclusion that the AAS and the neurosteroids, while sharing many obvious similarities, are nonetheless distinct modulators of the GABA_A receptor. For example, PS induces negative modulation of both γ- and ɛ-subunit-containing receptors (Akk et al. 2001; Maksay et al. 2003), while 17α-MeT is a positive modulator of γ-subunit-containing receptors (Yang et al. 2002, 2005; Clark et al. 2004) and a negative modulator of ɛ-subunit-containing receptors.

That substitution of the ɛ- for a γ-subunit in recombinant α₂β₃x-receptors can invert modulation induced by 17α-MeT from positive (α₂β₃γ) to negative (α₂β₃ɛ) suggests that sequence divergence between these two subunits may be of critical importance in determining the effects of this AAS. Previous studies have implicated residues within the TM2 domain as providing a binding pocket for a wide range of low-affinity allosteric modulators of the GABA_A receptor, and we have hypothesized that residues within this site may also be important for AAS modulation (Yang et al. 2002, 2005). To test this hypothesis, we engineered an ɛ-subunit in which residue 302, which is glycine in the ɛ-subunit, was mutated to serine, the amino acid present at the comparable position in the γ-subunits. Effects of 17α-MeT were still inhibitory at these mutant receptors, indicating that this one residue in the ɛ-subunit does not, on its own, determine whether AAS modulation will be positive or negative. The G302S mutation did, however, diminish inhibition by 17α-MeT by decreasing apparent AAS binding (2-fold decrease in K_d) and promoting a 1.8-fold left shift in IC₅₀ in the concentration–response relationship to 17α-MeT. Thus, these data implicate this residue within TM2 as playing an important role in mediating AAS inhibition in ɛ-subunit-containing receptors.

What are the potential physiological ramifications of the profile of AAS modulation of ɛ-subunit-containing receptors, especially as it relates to the known effects of these abused steroids on reproductive behaviours? Expression of the ɛ-subunit is highly restricted in the mammalian brain, with appreciable levels being observed in regions of the forebrain and brainstem that subserve reproductive function in both rodents and primates (Whiting et al. 1997; Davies et al. 1997; Sinkkonen et al. 2000; Heikkilä et al.2001; Moragues et al. 2002, 2003). Particularly intriguing are recent studies that ɛ-subunit expression is enriched in peptidergic neurons of the forebrain and that expression of the ɛ-subunit mRNA is detected in virtually all GnRH-immunoreactive neurons (Moragues et al. 2002, 2003; Sergeeva et al. 2005). These forebrain peptidergic neurons act as the master regulators of the hypothalamic–pituitary–gonadal axis, with GnRH controlling the pulsatile secretion of LH that is essential for the onset of puberty, the establishment and maintenance of oestrous cyclicity in females and reproductive competence in males (for review, see Ojeda & Urbanski, 1994; Freeman, 1994). GABAergic transmission imparts significant control over GnRH release by modulating the characteristic action potential bursting patterns of these cells (Sim et al. 2000; Nunemaker et al. 2003; Han et al. 2004). Here we show that acutely dissociated GnRH neurons exhibit a PTX-sensitive, bicuculline-insensitive spontaneous current that is inhibited by 1 μm 17α-MeT to an extent that is comparable to that observed for spontaneous currents in HEK293 cells expressing α₂β₃ɛ-receptors. Furthermore, we show that rapid application of millimolar GABA elicits responses from GnRH neurons with deactivation kinetics comparable to those elicited by rapid application of millimolar GABA to recombinant α₂β₃ɛ-receptors, and that 1 μm 17α-MeT elicits a small but significant inhibition of these phasic responses as well. Taken together, our results suggest that native GnRH neurons express functional ɛ-subunit-containing receptors and that the AAS 17α-MeT may elicit its inhibitory effects in these primary neurons by selective actions on these receptors. While experiments remain to be performed to determine the role of AAS inhibition of steady state (tonic and spontaneous) versus phasic (synaptic) GABA_A receptor-mediated responses in these cells and the impact that such modulation may have on GnRH release, our results support the hypothesis that AAS effects on ɛ-subunit-containing receptors may contribute to the observed effects on reproductive status observed in AAS users.