Enhanced macroscopic desensitization shapes the response of α4 subtype-containing GABA_A receptors to synaptic and extrasynaptic GABA

Abstract

Up-regulation of the GABA_A receptor α4 subunit subtype has been consistently shown in multiple animal models of chronic epilepsy. This isoform is expressed in both thalamus and hippocampus and is likely to play a significant role in regulating corticothalamic and hippocampal rhythms. However, little is known about its physiological properties, thus limiting understanding of the role of α4 subtype-containing GABA_A receptors in normal and abnormal physiology. We used rapid GABA application to recombinant GABA_A receptors expressed in HEK293T cells to compare the macroscopic kinetic properties of α4β3γ2L receptors to those of the more widely distributed α1β3γ2L receptors. These receptor currents had similar peak current amplitudes and GABA EC₅₀ values. However, α4β3γ2L currents activated more slowly when exposed to submaximal GABA concentrations, had more fast desensitization (τ = 15–100 ms), and had less residual current during long GABA applications. In addition, α4β3γ2L currents deactivated more slowly than α1β3γ2L currents. Peak currents evoked by repetitive, brief GABA applications were more strongly attenuated for α4β3γ2L currents than α1β3γ2L currents. Moreover, the time required to recover from desensitization was prolonged in α4β3γ2L currents compared to α1β3γ2L currents. We also found that exposure to prolonged low levels of GABA, similar to those that might be present in the extrasynaptic space, greatly suppressed the response of α4β3γ2L currents to higher concentrations of GABA, while α1β3γ2L currents were less affected by exposure to low levels of GABA. Taken together, these data suggest that α4β3γ2L receptors have unique kinetic properties that limit the range of GABA applications to which they can respond maximally. While similar to α1β3γ2L receptors in their ability to respond to brief and low frequency synaptic inputs, α4β3γ2L receptors are less efficacious when exposed to prolonged tonic GABA or during repetitive stimulation, as may occur during learning and seizures.

GABA_A receptors are pentameric cys-loop receptors composed primarily of two α subunits, two β subunits, and either a δ or a γ subunit selected from six α, three β, one δ, and three γ subunit subtypes. The distribution of specific subtypes is highly brain region and cell type specific, and varies during development and in certain disease states. The presence of a specific subunit subtype confers different pharmacological and physiological properties to receptor isoforms. For example, α subtypes strongly influence GABA_A receptor pharmacology. When assembled with β and γ subunits, GABA_A receptors containing α1, 2, 3 or 5 subtypes are highly diazepam sensitive. However, α1 subtype-containing receptors are much more sensitive to zolpidem than receptors containing α2 or α3 subtypes, and those containing α5 subtypes are completely insensitive to this drug. In contrast, GABA_A receptors containing α4 or α6 subtypes are insensitive to both diazepam and zolpidem. Furthermore, the imidazobenzodiazepine Ro 15-4513, which is an inverse benzodiazepine receptor agonist at GABA_A receptors containing α1, α2, α3 or α5 subtypes, actually enhances currents from GABA_A receptors containing α4 or α6 subtypes.

Unlike the pharmacological properties of GABA_A receptors, relatively little is known about the kinetic properties of different α subtypes. This limits our understanding of GABA_A receptor physiology, as receptor kinetics play an important role in shaping the postsynaptic response to GABA. For example, in synapses, GABAergic inhibitory postsynaptic current (IPSC) time courses are shaped by the rates of activation, desensitization and deactivation. During IPSCs, GABA_A receptor channels must activate rapidly and deactivate slowly to provide significant charge transfer during the very brief (≤ 1 ms) pulses of GABA present in the synaptic cleft. GABA_A receptor subtypes thought to be expressed in synapses are also often highly desensitizing, which may be linked to the slow deactivation that is crucial for effective synaptic neurotransmission (Jones & Westbrook, 1995). In contrast, extrasynaptic GABA_A receptors should be highly sensitive during prolonged exposure to low levels of GABA. As long as they maintain a steady-state level of charge transfer, they need not be rapidly activating, highly desensitizing, or slowly deactivating. Between these two extreme examples, subsets of GABA_A receptors may have different rates of activation, desensitization and deactivation, thereby allowing for maximal responses to specific frequencies, durations and concentrations of local GABA.

As a result, altered expression and distribution of certain GABA_A receptor isoforms has the potential to profoundly affect inhibitory neurotransmission. Perturbed expression of GABA_A receptors has been particularly well studied in epilepsy. Several animal models have consistently found an up-regulation of α4 subtype protein expression in animals with experimental epilepsy (Schwarzer et al. 1997; Sperk et al. 1998; Brooks-Kayal et al. 1998). Although the predominant GABA_A receptor isoform in the central nervous system is the α1β2γ2 isoform, and α4βγ receptors comprise only a small minority of all native GABA_A receptors, the α4 subtype is relatively abundant in brain regions involved in both partial and generalized epilepsies, including cortex, hippocampus and thalamus (Pirker et al. 2000). Furthermore, this subtype may be involved in both synaptic and extrasynaptic neurotransmission, depending on the brain region and physiological state of the animal (Hsu et al. 2003; Belelli et al. 2005; Cope et al. 2005). Nonetheless, the role of increased α4 subtype expression in the pathophysiology of epilepsy remains unclear. Without knowing the kinetic properties of GABA_A receptors containing the α4 subtype, it is difficult to predict whether increased expression of this GABA_A receptor subtype causes epilepsy or is simply a compensatory response to seizures. Thus, in the present study we compared the physiological and pharmacological properties of recombinant α4β3γ2 and α1β3γ2 receptor isoforms using GABA application protocols that mimicked synaptic and extrasynaptic conditions.

Methods

Expression of recombinant GABA_A receptors

GABA_A receptor α1, α4, β3 and γ2L subtype cDNAs were individually subcloned into the mammalian expression vector pCMV-neo. Deletion of an extraneous genomic sequence in the 5′ untranslated region of the α4 subtype cDNA resulted in improved expression, as previously described (Wallner et al. 2003). All cDNAs were sequenced by the Vanderbilt University Medical Centre sequencing core to confirm that they matched the published sequences for mature rat peptides corresponding to accession numbers NP_899155, NP_542154, P63079 and NP_899156 for the α1, α4, β3 and γ2 proteins, respectively.

Human embryonic kidney (HEK293T) cells were plated at a density of 200 000–400 000 cells per 60 mm culture dish and maintained in Dulbecco's modified Eagle's wedium (Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum and 100 IU ml⁻¹ each of penicillin and streptomycin (Invitrogen) at 37°C in 5% CO₂–95% O₂. On day one, cells were transfected using a previously established calcium phosphate precipitation technique (Angelotti et al. 1993). A total of 12 μg GABA_A receptor subunit-containing DNA, either with 4 μg of each subunit plasmid (ratio 1 : 1 : 1) for αβ3γ2L receptors or with 6 μg of each subunit plasmid (ratio 1 : 1) for αβ3 receptors. Two micrograms of pHook-1 (Invitrogen) was also added so that immunomagnetic bead selection could be performed on day 2 (Greenfield et al. 1997). Following selection, the cells were plated on 35 mm dishes, and recordings were made on day 3, approximately 18–36 h after selection. The cells used for macropatches were treated using the same protocol but were plated on 35 mm dishes that had been previously collagenized.

Electrophysiological recording and drug application

Whole cell voltage-clamp recordings were performed on transfected HEK293T cells. All experiments were performed using at least two separate transfected batches of cells from at least two separate days of recording. Cells were bathed in an external solution consisting of (mm): NaCl 142, CaCl₂ 1, KCl 8, MgCl₂ 6, glucose 10, Hepes 10 (pH 7.4, ∼320–340 mosmol l⁻¹) throughout the duration of the experiment. All recordings were done at room temperature. Glass micropipettes were formed from thin-walled borosilicate glass with a filament (World Precision Instruments, Sarasota, FL, USA) with a P2000 laser electrode puller (Sutter Instruments, San Rafael, CA, USA) and fire polished with a microforge (Narishige, East Meadow, NY, USA). Microelectrodes used for lifted cell recording had resistances of 1–2 MΩ when filled with an internal solution consisting of (mm): KCl 153, MgCl₂ 1, Hepes 10, EGTA 5, Mg²⁺-ATP 2 (pH 7.3, ∼300–310 mosmol l⁻¹). This combination of external and internal solutions produced a chloride equilibrium potential (E_Cl) of approximately 0 mV. Electrodes used for macropatch recording were fire polished to achieve resistances of 1.5–4 MΩ when filled with the same internal solution.

Membrane voltages were clamped at –20 mV using an Axopatch 200A amplifier (Axon Instruments, Union City, CA, USA) amplifier. GABA was applied to the lifted cells via a three-, four-, or six-barrelled square glass (Friedrich and Dimmock, Millville, NJ, USA). These multibarrelled pipettes were pulled on a P-87 Flaming-Brown (Sutter Instruments, San Rafael, CA, USA) electrode puller with custom made platinum–iridium filament and sanded to a final diameter of 200–400 μm for each barrel. The multibarrelled pipettes were attached to a Warner SF-77B Perfusion Fast-Step (Warner Instrument Corporation, Hamden, CT, USA), allowing for rapid solution changes. All GABA application protocols began with a cell positioned in the flow of external bath solution from which the multibarrelled array was repositioned such that the unmoved cell and electrode were now exposed to GABA. The drug application was initiated by an analog pulse triggered by the pCLAMP 9 software (Axon Instruments) that caused the motor of the Warner Fast-Step to reposition the multibarrelled array from one barrel to another (e.g. external solution to GABA). Exchange times were routinely measured and always found to be 0.3–0.7 ms at an open electrode tip by stepping from control to dilute external solution. However, all GABA concentrations > 1 mm were applied using barrels with exchange times < 0.45 ms. The exchange around an intact cell was measured in a subset of cells by stepping into 10 μm GABA and then another step into 10 μm GABA in external solution in which NaCl was replaced by NaSCN. The resulting current had a 10–90% rise time of 0.9 ± 0.1 ms (n = 9) using a drug application pipette with 0.35 ms open tip exchange time.

For generation of concentration–response relationships, peak GABA_A receptor currents evoked by randomly sequenced concentrations of GABA were recorded with at least 45 s of wash between each application. This time was empirically determined to be sufficient for complete recovery from desensitization. The preapplication concentration–response curves were generated by exposing the cells to 45 s of 1 μm GABA between briefer (1–4 s) applications of higher concentrations of GABA. To assess possible changes in the transmembrane chloride ion concentration gradient, pCLAMP 9 generated a 500 ms ramp voltage step from –50 mV to +50 mV. The current–voltage relationship was determined at the beginning and end of each prolonged exposure to 1 μm GABA, as well as at the end of the wash in external solution. The use of six-barrelled drug application pipettes allowed us to test several concentrations of GABA with each sweep of the pClamp protocol. The responses during concentration–response determinations and preapplication studies were normalized to the current elicited by 1 mm GABA after a prolonged wash in external solution during each sweep. Data were excluded if there was a greater than 10% rundown of the maximal response between sweeps.

Data analysis

Currents were low-pass filtered at 2 kHz, digitized at 5–10 kHz, and analysed using the pCLAMP 9 software suite. For those cells with very small (< 50 pA) currents, rise time, desensitization and deactivation were not determined. Current amplitudes and 10–90% rise times were measured using the Axon Instruments Clampfit 9 software package. The desensitization and deactivation time courses of GABA_A receptor currents were fitted using the Levenberg-Marquardt least squares method with up to six component exponential functions of the form ∑a_ne^(–t/τn)+C, where t is time, n is the best number of exponential components, a_n is the relative amplitude of the nth component, τ_n is the time constant of the nth component, and C is the residual current at the end of the GABA application. Additional components were accepted only if they significantly improved the fit, as determined by an F-test automatically performed by the analysis software on the sum of squared residuals. The time course of deactivation was summarized as a weighted time constant, defined by the following expression: .

Repetitive stimulation experiments applied 10 ms of 1 mm GABA at 10 Hz four times. The ratio of the fourth peak response to the first peak response was determined for each cell. Recovery from desensitization was studied as previously described (Overstreet et al. 2000). In brief, pairs of 5 ms applications of 1 mm GABA with variable wash intervals between the first and second GABA applications were used. Recovery from desensitization was defined as: , where Residual is the remaining current immediately before the second application of GABA, and Peak₁ and Peak₂ refer to the peak current during the first and second GABA applications, respectively. GraphPad Prism 4 (GraphPad Software Inc, San Diego, CA, USA) was used to fit the concentration–response results to a sigmoidal function using the equation:

where I is the peak current at a given GABA concentration, and I_max is the maximal peak current. Numerical data were expressed as means ± s.e.m. Statistical analysis was performed using GraphPad Prism 4. Data were compared using a Mann-Whitney test for pairs of data, or a Kruskal-Wallis test for comparing three or more groups. Statistical significance was taken as P < 0.05.

Kinetic modelling

Using QuB version 1.4 (http://www.qub.buffalo.edu), a modified version of the α1β3γ2L GABA receptor model proposed by Haas & Macdonald (1999) was fitted to representative α1β3γ2L and α4β3γ2L whole cell currents. For each GABA concentration used in the fitting process, a representative current was generated by averaging the responses of three to eight cells whose peak currents were normalized to the 1 mm GABA peak current. Before using this averaged current for fitting, it was verified against the average kinetic properties of individual cells at the same GABA concentration.

Since mono-liganded states should have negligible occupancy in the presence of saturating concentrations of GABA, we first fitted an averaged current elicited by 10 mm GABA to a version of the model including only the di-liganded states. This reduced the number of free parameters, thus decreasing the time required for each fitting iteration. To further reduce the number of free parameters, the exit rates from di-liganded open states were fixed so that mean open times would be consistent with previously published single channel data for these receptor isoforms (Haas & Macdonald, 1999; Akk et al. 2004). Once an adequate fit was obtained (i.e. when the log likelihood ratio changed by less than 0.5 between iterations), the currents were refitted to a model with scaled rate constants to obtain open time distributions consistent with the published single channel data. Closed time distributions were not considered during the fitting process, as the longest components tend to be artificially shortened due to the presences of multiple channels in most patches. Nonetheless, it should be noted that the di-liganded portion of our model contains five closed states. This gives rise to a closed time distribution with five components, a number consistent with published reports for both receptor isoforms.

After fitting the model to currents elicited by 10 mm GABA, all di-liganded rate constants were fixed, and mono- and unliganded non-conducting states were added to generate two GABA binding steps. The two GABA binding sites were assumed have equal affinity. To obtain an estimate of the unbinding rates, this model was first fitted to the time course of current deactivation following a brief GABA application (5 ms, 1 mm). Mono-liganded open and desensitized states were then added, and the model was refitted to currents elicited by the lowest GABA concentration that permitted significant activation and macroscopic desensitization (3 and 10 μm for α4β3γ2 and α1β3γ2 receptor isoforms, respectively). The final kinetic model for each receptor isoform was verified by generating theoretical currents with the differential equation solving program Berkeley-Madonna 8.0 (http://www.berkeleymadonna.com) using the fourth-order Runge-Kutta method and time intervals of 10–100 μs.

Results

To investigate the role of α4 subtypes in determining the physiological properties of αβγ GABA_A receptor currents, we used whole cell voltage clamp recording and a rapid drug delivery system to apply GABA to lifted HEK293T cells that had been cotransfected with β3, γ2L, and either α1 or α4 subtypes. We chose the combination of α4, β3 and γ2L subunits because rodent brain α4 subtypes coprecipitate with β2/3 subunits (Bencsits et al. 1999) and are among the subtypes coexpressed in thalamus and hippocampus (Pirker et al. 2000). Moreover, multiple models of epilepsy have found a consistent up-regulation of α4 subtype expression in the hippocampal dentate gyrus, where β3 is the predominant β subunit subtype. Furthermore, low levels of endogenous β3 subtypes have been detected by RT-PCR in HEK293 cells (Kirkness & Fraser, 1993; Davies et al. 2000). We therefore used the β3 subtype to prevent possible contamination with multiple β subtypes. Although α1β2γ2 receptors are the most abundant native receptors in mammalian brain, we used α1β3γ2L receptors to permit comparison of the effects of different α subtypes on receptor current kinetic properties, without the possible confound of different β subtypes.

α1,4β3 receptors yielded small currents with altered kinetic properties

We confirmed that cells transfected with α, β, and γ cDNAs actually form ternary (i.e. αβγ) complexes. Previous work has demonstrated that transfection of fibroblasts with the α, β, or γ subunits alone does not produce currents and neither does cotransfection of cells with β and γ subunits (Zezula et al. 1996). There have been inconsistent reports of immortalized cells expressing αγ GABA_A receptors (Verdoorn et al. 1990). The kinetic properties and zinc sensitivity of these currents, however, were very similar to cells expressing αβγ GABA_A receptors (Verdoorn et al. 1990; Draguhn et al. 1990), suggesting that these receptors probably included the endogenous β3 subunit expressed in HEK293 cells (Davies et al. 2000). However, α1β3 subunits have been shown to form functional currents (Angelotti & Macdonald, 1993; Angelotti et al. 1993), and there is evidence from immunoprecipitation studies that α4βx receptors may be present in rat brain (Bencsits et al. 1999). We found that application of a synaptically relevant GABA concentration (1 mm) consistently evoked currents from cells expressing either α1β3 or α4β3 receptors (Fig. 1); however, these currents activated more slowly, and the maximal current amplitudes were five to 10 times smaller than those recorded from cells expressing α1β3γ2L or α4β3γ2L receptors (Table 1, I_max). Furthermore, α1β3 receptor currents desensitized much more rapidly than α1β3γ2L receptor currents, although the overall residual current at the end of 4 s GABA application was not different (Fig. 1A; Table 1). In contrast, α4β3 receptor currents desensitized more slowly and less extensively than α4β3γ2L receptor currents (Fig. 1B; Table 1).

Figure 1.

GABA-evoked αβ currents were much smaller and had altered kinetic properties, compared to αβγ currents Aa and Ba, representative currents in response to 1 mm GABA. The current amplitudes in cells which had been transfected with α1β3 and α4β3 subunits were much smaller than cells transfected with α1β3γ2L and α4β3γ2L subunits. To allow comparison of the kinetic properties, these currents were scaled to the same peak amplitude. Note the extremely fast desensitization of α1β3 currents, although the overall desensitization at the end of 4 s of GABA was similar to α1β3γ2L currents. In contrast, α4β3 currents activated and desensitized more slowly, but also had less overall desensitization during 4 s of GABA. Ab and Bb, the same currents on an expanded time base.

Table 1.

Summarized responses to prolonged application of saturating GABA (4 s, 1 mM)

	α1β3	α1β3γ2L	α4β3	α4β3γ2L
Imax (pA)	580 ± 197 (n = 8) ††	3402 ± 418 (n = 34)	322 ± 67 (n = 12)†††	2976 ± 355 (n = 25)
Rise time (ms)	1.83 ± 0.09 (n = 7)†††	1.03 ± 0.06 (n = 31)	3.5 ± 0.44 (n = 11)††	1.40 ± 0.11 (n = 25)*
A1	0.49 ± 0.06 (n = 8)††	0.18 ± 0.03 (n = 21)	0.10 ± 0.07 (n = 9)†	0.25 ± 0.03 (n = 24)
τ1 (ms)	11 ± 1 (n = 8)††	6 ± 1 (n = 20)	11 ± 4 (n = 4)	7 ± 1 (n = 23)
A2	0.18 ± 0.05 (n = 8)†	0.07 ± 0.01 (n = 21)	0.25 ± 0.07 (n = 9)	0.13 ± 0.02 (n = 24)*
τ2 (ms)	48 ± 11 (n = 7)	56 ± 5 (n = 16)	54 ± 6 (n = 7)	58 ± 4 (n = 22)
A3	0.04 ± 0.02 (n = 8)†	0.17 ± 0.03 (n = 21)	0.16 ± 0.03 (n = 9)	0.17 ± 0.02 (n = 24)
τ3 (ms)	347 ± 204 (n = 3)	431 ± 45 (n = 21)	280 ± 48 (n = 8)	373 ± 41 (n = 21)
A4	0.06 ± 0.03 (n = 8)†††	0.33 ± 0.03 (n = 21)	0.15 ± 0.05 (n = 9)†	0.33 ± 0.03 (n = 24)
τ4 (ms)	7382 ± 1730 (n = 4)	2652 ± 250 (n = 20)	3333 ± 589 (n = 5)	1934 ± 161 (n = 24)
Residual current	0.23 ± 0.06 (n = 8)	0.25 ± 0.02 (n = 21)	0.37 ± 0.06 (n = 7)†††	0.12 ± 0.02 (n = 24)**
Deact τ (ms) (4 s GABA)	577 ± 53 (n = 7)	374 ± 30 (n = 25)	1054 ± 256 (n = 3)	563 ± 77 (n = 18)

Compared to α1β3γ2L currents, α1β3 currents were smaller in maximal amplitude with more fast desensitization. Compared to α4β3γ2L currents, α4β3 currents were smaller with a slower rise time and less overall desensitization.

^†P < 0.05 αxβ3 versus αxβ3γ2

^*P < 0.05 α1β3γ2L versus α4β3γ2L

^††P < 0.01 αxβ3 versus αxβ3γ2 l

^**P < 0.01 α1β3γ2L versus α4β3γ2L

^†††P < 0.001 αxβ3 versus αxβ3γ2L. As defined in the methods, A1-4 and τ1-4 refer to the magnitudes and time constants of desensitization, respectively.

With the exception of α1β3 currents, these currents generally required four components to accurately fit the desensitization time course. These complex kinetics might suggest that α1β3γ2L currents are actually mediated by a mixed population of GABA_A receptors. However, cells transfected with α1, βx, γx, βxγx, or α1γx subtype combinations do not generally produce current (Verdoorn et al. 1990; Angelotti & Macdonald, 1993; Tretter et al. 1997; Davies et al. 2000). Moreover, previous studies have found that in cells transfected with α, β and γ subunits there is a preferential assembly of all three subunits into functional pentamers with a stoichiometry of two α, two β and one γ subunits per receptor (Angelotti & Macdonald, 1993; Chang et al. 1996; Zezula et al. 1996; Tretter et al. 1997; Farrar et al. 1999). Finally, given the small amplitudes of the αβ currents, we concluded that the majority of the current from αβγ transfected cells was from ternary receptors.

α4β3γ2L currents desensitized more rapidly and extensively than α1β3γ2L currents

To characterize the effect of α subtypes on current desensitization, α4β3γ2L and α1β3γ2L currents were recorded during prolonged (4 s) applications of a high concentration of GABA (1 mm). The α4β3γ2L currents desensitized more rapidly and extensively than α1β3γ2L currents (Fig. 2A and B). All currents were fitted best with a three or four component exponential function (online Supplemental material, Figure S1) with time constants that fell into four discrete groups, τ₁ (< 15 ms), τ₂ (20–100 ms), τ₃ (110–800), and τ₄ (800–4800 ms) (Fig. 2C). The α4β3γ2L currents had a greater contribution of fast (< 100 ms) desensitization than α1β3γ2L receptors (Fig. 2C), although the actual time constants of desensitization were similar for both receptors. Moreover, α4β3γ2L receptors had more overall desensitization, as assessed by the residual current at the end of the 4 s GABA application.

Figure 2.

α4β3γ2L currents had more rapid desensitization than α1β3γ2L currents A, representative currents illustrating the rapid and extensive desensitization that occurred during a 4 s application of 1 mm GABA. Peak currents were normalized to allow for comparison of their kinetic properties. B, the same currents were presented on an expanded time scale. C, although the desensitization time constants were similar for α1β3γ2L and α4β3γ2L currents, there was a greater contribution of fast (τ < 100 ms) desensitization in α4β3γ2L currents. Moreover, α4β3γ2L currents had less residual current at the end of 4 s of GABA. **P < 0.01, ***P < 0.001 α1β3γ2L versusα4β3γ2L currents.

The rapidly changing membrane conductance during GABA application introduces a transient series resistance error that cannot be compensated with our recording system. In theory, this could cause us to underestimate the true peak amplitude and degree of rapid desensitization. However, if this were the case, we would expect those cells with larger currents to have a smaller fraction of fast desensitization. Similar to previous work from our laboratory (Bianchi & Macdonald, 2002), we found no such correlation between current amplitude and desensitization (data not shown). Moreover, since the peak amplitudes of α1β3γ2L and α4β3γ2L are the same, any transient series resistance errors should be the same in both groups.

α4β3γ2L currents deactivated more slowly than α1β3γ2L currents

Synaptic inhibitory neurotransmission involves very brief exposure to high concentrations of GABA. To characterize the response of these receptors to a more synaptically relevant application of GABA, cells were exposed to 1 mm GABA for 5 ms, the briefest application that we were able to achieve reproducibly with our drug delivery system. The currents were fitted to multiexponential functions (Fig. 3B). Both α1β3γ2L and α4β3γ2L receptor currents decayed slowly after a brief exposure to GABA (Fig. 3A). However, α4β3γ2L currents deactivated more slowly, with a weighted time constant of 371 ± 54 ms (n = 18) versus 200 ± 24 ms (n = 13, P < 0.05) for α1β3γ2L currents. Deactivation of α1β3γ2L current was relatively simple, with two (1 of 13 cells), three (6 of 13 cells), or four (6 of 13 cells) time constants, but the major part of the deactivation was due to a single component (τ₃, 100–300 ms). In contrast, α4β3γ2L current deactivation tended to be more complex, requiring three (7 of 18 cells), four (8 of 18 cells) or five (3 of 18 cells) components to accurately fit the deactivation time course. Furthermore, unlike α1β3γ2L current deactivation, no single component dominated the deactivation time course. Interestingly, in most published reports, only one or two component exponential functions are generally used to fit IPSC decay. The basis for this difference is uncertain, but may be due to the fact that our recombinant currents were significantly larger than most IPSCs, which may have allowed us to detect greater kinetic complexity than would be apparent from the smaller IPSC currents. Different fitting techniques are another potential source of confusion when comparing studies among laboratories. To address this issue, all of the currents in these studies were also fitted with a two-component exponential function. Although this approach provided less precise fits of the data, there was no change in the overall weighted time constant compared to the more complicated fits described above (data not shown).

Figure 3.

α4β3γ2L currents deactivated more slowly than α1β3γ2L currents A, α4β3γ2L currents deactivated more slowly after a brief application of saturating GABA (1 mm, 5 ms). B, summary of the deactivation kinetics. The weighted deactivation time constant of α1β3γ2L currents was 200 ± 24 ms (n = 13) and was largely determined by a single time constant (τ₃). In contrast, α4β3γ2L current deactivation was more complicated and prolonged, with an overall weighted deactivation time constant of 371 ± 55 ms (n = 18), which was significantly longer than α1β3γ2L currents (P < 0.05). *P < 0.05, **P < 0.01, ***P < 0.0001 α1β3γ2L versusα4β3γ2L currents.

To further improve the speed of GABA application, 1 mm GABA was applied to macropatches. Similar to data from lifted cells, the peak current amplitudes were similar (543 ± 212 pA versus 351 ± 138 pA for α1β3γ2L and α4β3γ2L currents, respectively) but the rise times were prolonged for α4β3γ2L currents (1.78 ± 0.21 ms, n = 9, P < 0.05) compared to α1β3γ2L (1.30 ± 0.44 ms, n = 8). Both combinations had greater extents of fast desensitization than those recorded using lifted cells, although α4β3γ2L currents still had more fast desensitization (fraction of τ₁ = 0.47 ± 0.04, P < 0.05) than α1β3γ2L currents (0.38 ± 0.04). Finally, α4β3γ2L currents still had less residual current during application of 1 mm GABA for 4 s (0.02 ± 0.01, P < 0.05) than α1β3γ2L currents (0.07 ± 0.02). Interestingly, unlike data from lifted cells, there was no difference in the rate of deactivation following application of 1 mm GABA for 5 ms (weighted τ = 143 ± 18 ms, n = 8 and 109 ± 5 ms, n = 6 for α1β3γ2L and α4β3γ2L currents, respectively). The reason for the difference between macropatch and whole cell currents is not entirely clear. One explanation might be that there is improved GABA exchange around a macropatch. The enhanced fast desensitization and briefer deactivation is consistent with that interpretation. However, we measured the solution exchange around a whole cell and found it to be consistently faster than 1 ms. Furthermore, the fastest kinetic feature, current activation, was not different between macropatches and lifted cells. Another possible explanation is that the smaller macropatch currents were less distorted by series resistance error, thereby producing faster desensitization and deactivation compared to currents from lifted cells. However, similar to the lifted cell data, we saw no correlation between current amplitude and desensitization kinetics. Moreover, neither improved drug application speed or reduced series resistance error explain why despite enhanced fast desensitization of α4β3γ2L receptor currents, currents from α4β3γ2L macropatches did not have prolonged deactivation time constants compared to α1β3γ2L currents, as seen in lifted cells. We favour the interpretation that the act of pulling macropatches alters the function of GABA_A receptors, possibly a consequence of losing modifying cytoplasmic proteins. Therefore, the remaining studies were performed using lifted cells.

α4β3γ2L currents desensitized more rapidly than α1β3γ2L currents during repetitive stimulation

Although α4β3γ2L receptors deactivated slowly after a brief GABA application, they desensitized rapidly during long GABA applications. GABA_A receptor desensitization is thought to involve entry into a long-lived, non-conducting state(s). We therefore tested the hypothesis that during repetitive applications of brief GABA (10 ms applications of 1 mm GABA at 10 Hz), GABA_A receptors accumulate in desensitized state(s), resulting in progressively attenuated responses. During repetitive GABA application, current attenuation was much more pronounced for α4β3γ2L currents than for α1β3γ2L currents (Fig. 4A). The ratio of the fourth to the first peak response was 0.52 ± 0.06 (n = 10) for α4β3γ2L receptors versus 0.75 ± 0.03 (n = 11) for α1β3γ2L receptors (P < 0.01). Similar differences were noted when overall normalized charge transfer was compared between these two GABA_A receptor combinations (data not shown).

Figure 4.

α4β3γ2L currents desensitized rapidly during repetitive GABA application A, currents from representative cells during repetitive GABA application (1 mm, 10 Hz). Peak currents have been normalized to allow comparison of their kinetic properties. B, α4β3γ2L currents desensitized extensively during repetitive GABA application. **P < 0.01 α1β3γ2L versusα4β3γ2L currents.

To determine the time courses for recovery from desensitization of α4β3γ2L and α1β3γ2L currents, GABA was applied in brief pairs (5 ms, 1 mm GABA) with variable time intervals between applications (Fig. 5A). While these brief GABA applications would not allow these receptors to achieve steady state desensitization, they more closely mimic the response to brief repeated bursts of GABA that are likely to occur in inhibitory synapses. The time course of recovery from desensitization was fitted best by a four-component exponential function for both α1β3γ2L and α4β3γ2L currents. The α4β3γ2L currents were more suppressed during paired GABA applications, and required a longer recovery period between paired applications than α1β3γ2L currents. The weighted recovery time constant for α4β3γ2L receptors was 2.29 s (τ₁ = 6 ms (37%), τ₂ = 65 ms (20%), τ₃ = 801 ms (25%), τ₄ = 11.12 s (19%)) compared to the weighted recovery time constant for α1β3γ2L receptors of 314 ms (τ₁ = 6 ms (55%) and τ₂ = 35 ms (22%), τ₃ = 240 ms (14%), τ₂ = 2.99 ms (9%)). In summary, the highly desensitizing α4β3γ2L currents were progressively attenuated during high frequency GABA application and also required a prolonged wash period to allow recovery from desensitization after each exposure to GABA.

Figure 5.

α4β3γ2L currents recovered more slowly from desensitization than α1β3γ2L currents A, an envelope of currents from a representative cell illustrated the paired stimulation protocol. GABA (1 mm, 5 ms) was applied twice during each sweep with intervals ranging from 10 ms to 90 s. As demonstrated here, the peak current evoked by the second GABA application was reduced compared to the first. B, recovery from desensitization was measured as the fractional availability (see Methods) of GABA receptors at various time points after the first brief application of GABA.

α4β3γ2L currents activated more slowly than α1β3γ2L currents

Synaptic GABA is thought to reach high concentrations (∼1 mm) for only a very brief time (≤ 1 ms). Given the brief availability of GABA in the synaptic cleft, the rate of activation might shape α4β3γ2L receptor currents within or near the synapse. We therefore determined if there was a concentration-dependent difference in the activation rates of α4β3γ2L and α1β3γ2L currents. Activation rate was defined as the inverse of the 10–90% rise time and plotted as a function of GABA concentration (Fig. 6A). Since the maximal rates of activation at GABA concentrations > 1 mm approach the solution exchange time around the cell, these results may underestimate the true rates. However, the maximal activation rates of α4β3γ2L and α1β3γ2L currents were not obviously different and, more importantly, the entire concentration–activation rate curves were shifted to the right for α4β3γ2L currents (EC₅₀ = 988 μm) compared with α1β3γ2L currents (EC₅₀ = 377 μm). Thus, at any given submaximal GABA concentration (≤ 3 mm), the slower activation of α4β3γ2L currents would favour incomplete activation, and therefore, a truncated peak current.

Figure 6.

α4β3γ2L currents activated more slowly than α1β3γ2L currents A, α4β3γ2L currents had slower activation rates than α1β3γ2L currents at all but the highest concentrations of GABA. B, brief application of GABA altered the peak amplitude and decay kinetics of α4β3γ2L currents. Using a cell transfected with α4β3γ2L cDNAs, 10 μm GABA was applied for the times shown on the graph. The 10–90% rise time for this concentration on this cell was 28 ms. Arrows at the peak currents for 5 and 25 ms GABA applications. As shown here, briefer GABA application produces smaller amplitude current with an accelerated rate of deactivation.

Despite the very brief availability of GABA within the synaptic cleft, synaptic currents usually last on the order of tens to hundreds of milliseconds. This is thought to occur because receptors rapidly enter long-lived, agonist bound states, from which GABA unbinds slowly. During extremely brief GABA applications, however, receptor populations would be unable to achieve the same occupancy of long-lived agonist-bound states as would occur following longer pulses. With fewer receptors in these long-lived states, more rapidly deactivating postsynaptic responses occur. Furthermore, the response to saturating concentrations of GABA is thought to involve binding of two molecules of GABA by each receptor, with the longest lived states being achieved only when both GABA binding sites are occupied (Macdonald et al. 1989). When GABA is applied either very briefly or at subsaturating concentrations, not all receptors would be fully di-liganded, potentially also allowing for more rapidly deactivating currents (Mozrzymas et al. 2003).

To illustrate this idea, GABA (10 μm) was applied to α4β3γ2L receptors for durations varying from 5 to 200 ms (Fig. 6B). For the cell shown in Fig. 6B, the 10–90% rise time in the presence of 10 μm GABA was 28 ms. When GABA was applied for brief periods, there was not only a truncation of the peak, but also more rapid deactivation. Although similar experiments were not performed with α1β3γ2L receptors, the relatively rapid deactivation following applications of 1 mm GABA for 5 ms compared to 4000 ms durations suggests that a similar coupling of GABA application duration and deactivation time course occurs with these receptors as well. At synapses containing α4β3γ2L receptors, very brief exposure to GABA would be predicted therefore to produce synaptic currents with small amplitudes and brief durations.

GABA sensitivities of α1β3γ2L and α4β3γ2L receptors were similar

In addition to the synaptic receptors mediating IPSCs, another population of GABA_A receptors found outside the synaptic cleft respond to tonic, low levels of GABA (≤ 1 μm). Despite the small size of these tonic currents, their longevity allows a large overall charge transfer that can significantly alter neuronal excitability. Effective inhibition by extrasynaptic receptors would be best served by highly sensitive GABA_A receptors that are able to respond during prolonged exposure to low GABA concentrations. Concentration–response curves were generated by applying varying GABA concentrations in random order with at least a 45 s wash between applications of each concentration (Fig. 7). Since there was substantial cell-to-cell variability in current size, a near-maximal (1 mm) GABA concentration was applied intermittently to allow normalization of peak currents and to assess current run-down during the experiments. The GABA EC₅₀ values were similar for both α1β3γ2L and α4β3γ2L receptors (Fig. 7; GABA EC₅₀ values were 14 μm (95% CI = 12–17 μm) and 15 μm (95% confidence interval (CI) = 14–17 μm) for pooled data from α1β3γ2L and α4β3γ2L peak currents, respectively).

Figure 7.

α1β3γ2L and α4β3γ2L peak currents had similar concentration-dependent sensitivity to GABA The GABA potencies, as determined by complete concentration response studies in individual cells showed that the peak current EC₅₀ values (12 ± 2 μm (n = 9) and 17 ± 4 μm (n = 12)) were not different for α1β3γ2L and α4β3γ2l currents, respectively. Similar results were obtained when the data from all cells were pooled and then fitted as a group (EC₅₀ = 14 μm (95% CI = 12–17 μm) and 15 μm (95% CI = 14–17 μm), for α1β3γ2L and α4β3γ2l currents, respectively).

Tonic GABA receptor currents measured in neurons are thought to be caused by low levels of extrasynaptic GABA that change much more slowly than GABA levels in the synaptic cleft. Therefore, we sought next to determine the pseudo-steady-state concentration-dependent response of α1β3γ2L and α4β3γ2L receptors to prolonged GABA application. Various concentrations of GABA were applied for 4 s, and the current at the end of the GABA application was measured (Fig. 8A, arrow). At GABA concentrations above 10 μm, the response to prolonged GABA was lower for α4β3γ2L than for α1β3γ2L receptors (Fig. 8B). It is worth noting that the concentration dependence of these pseudo-steady-state currents had remarkably limited dynamic ranges. A maximal pseudo-steady-state response was seen with 10 μm GABA, although higher concentrations were able to accelerate the rates of activation and desensitization (Fig. 8B). In the range of GABA concentrations likely to be present in the extrasynaptic spaces (1–2 μm), the responses of α4β3γ2L and α1β3γ2L receptors were similar.

Figure 8.

Desensitization attenuated the responses to prolonged GABA applications A, representative α4β3γ2L currents illustrating the effect of desensitization to differentially influence peak and pseudo-steady-state currents in a concentration-dependent manner. Unlike the peak current, the pseudo-steady state current measured after 4 s of GABA (arrow) was responsive to a very limited range of GABA concentrations, with maximal stimulation occurring at approximately 10 μm.B, prolonged GABA application to α1β3γ2L receptors strongly suppressed the maximal GABA response (I_max = 31% of 1 mm peak current; 95% CI = 29–32%) and also caused an apparent leftward shift of the concentration–response curves (EC₅₀ = 3.3 μm; 95% CI = 2.4–4.5 μm GABA). The highly desensitizing α4β3γ2L receptors had an even more limited response to prolonged GABA application (I_max = 15%; 95% CI = 15–16%) (EC₅₀ = 1.0 μm; 95% CI = 0.7–1.3 μm GABA) Comparing the response of α1β3γ2L versusα4β3γ2L receptors to 4 s of 1 mm GABA confirmed that the highly desensitizing α4β3γ2L currents were more suppressed during prolonged GABA application (P < 0.001). For comparison, peak current concentration response data from Fig. 7 were replotted here as dashed lines.

The foregoing concentration dependence studies would suggest that despite increased desensitization to high GABA levels, both α1β3γ2L and α4β3γ2L receptors might respond reasonably well in an extrasynaptic environment. However, the previous studies all used a complete washout between GABA applications, which is probably different from the nearly constant low ambient levels of GABA thought to exist in the extrasynaptic space in some brain regions. To more realistically mimic extrasynaptic or perisynaptic GABA_A receptor responses, we measured the GABA_A receptor responses when pre-exposed to a prolonged low concentration of GABA. Cells were incubated for 45 s in 1 μm GABA (preapplication) followed by application of higher concentrations of GABA. Following a washout, the response to 1 mm GABA was measured and used to normalize all of the pre-exposed responses (Fig. 9A). It is possible that prolonged GABA application could produce a sufficient chloride ion flux that the normal electrochemical gradient was altered. Current–voltage relationships at the beginning and end of each sweep (Fig. 9A, arrows) confirmed that there was no change in the GABA reversal potential (data not shown). For both receptor isoforms, the maximal response to GABA following preapplication with low GABA concentrations was greatly reduced, compared to cells in which GABA was allowed to washout between applications (Fig. 9B). However, the suppressive effect of low tonic GABA levels was particularly prominent for α4β3γ2L currents.

Figure 9.

α4β3γ2L currents were highly suppressed by tonic, low GABA A, representative α4β3γ2L current which was recorded during exposure to 45 s of 1 μm GABA before application of higher GABA concentrations. The maximal response for each sweep was determined by measuring the peak response to 1 mm GABA after a 45 s wash in external solution. To ensure a stable chloride ion gradient, ramp I–V relationships were performed during control and at the beginning and end of the prolonged exposure to 1 μm GABA (arrows). B, pre-application of low levels of GABA suppressed both the potency and maximal response of both GABA_A receptor subtypes. The suppressive effects of tonic low levels of GABA were especially prominent for α4β3γ2L current, which had a nearly 50% reduction in the maximal response following prolonged application of 1 μm GABA.

GABA receptor kinetic modelling

To make more quantitative predictions about the potential physiological roles of α1 and α4 subtype-containing GABA_A receptors, a kinetic model was generated for these receptor isoforms that accounted for not only the observed macroscopic properties described above, but also for the previously published single channel properties (Haas & Macdonald, 1999; Akk et al. 2004) (Fig. 10). This model was based largely on the previous work of Haas & Macdonald (1999), but has been expanded to account for several additional microscopic and macroscopic kinetic observations. Specifically, single channel recordings from α1βγ GABA_A receptors have consistently demonstrated three open states, with the relative contribution of the shortest open state (O₁) being concentration dependent at GABA concentrations below 10 μm (Fisher & Macdonald, 1997). Although this suggests the presence of an additional GABA binding step distal to O₁, a feature found in most of the existing models, it remains unclear why a residual component of O₁ persists at higher concentrations. One possible explanation is that two open states with similar mean open times exist, one of which is mono-liganded and the other di-liganded. We have therefore added two additional states to the Haas and Macdonald kinetic model: a di-liganded closed (C₃) and a di-liganded open state (O₂). The addition of the closed state is intended to preserve the known burst characteristics of these receptors such that only one type of opening is observed per burst (Twyman et al. 1990). As with α1βγ receptors, single channel recordings from α4βγ receptors also demonstrated the presence of three open states that are concentration independent at GABA concentrations above 10 μm (Akk et al. 2004). We therefore applied the same kinetic model to those receptors.

Figure 10.

Kinetic model for the α1β3γ2L and α4β3γ2L receptor isoforms A, a modified version of the kinetic model proposed for the α1β3γ2L receptor isoform by Haas & Macdonald (1999) is shown (O, open; C, closed; D, desensitized). B, rate constants for the α1β3γ2L and α4β3γ2L receptor isoforms were determined by fitting the time course of macroscopic currents evoked by saturating and subsaturating GABA concentrations to the kinetic model in A after constraints were imposed based on published single-channel data (see Methods). Units for all rate constants are s⁻¹ except for k_on (m⁻¹ s⁻¹). Ca and Da, the optimized responses (black line) of α1β3γ2L and α4β3γ2L receptors to a 5 ms square pulse of GABA (1 mm; arrow) are superimposed on averaged currents evoked under similar conditions (grey line). Cb and Db, the simulated responses of the α1β3γ2L and α4β3γ2L receptors to a 4 s square pulse of a saturating GABA concentration (10 mm; filled bars) are superimposed on the averaged data traces used for fitting the di-liganded rate constants (see Methods). Cc and Dc, the optimized responses of the α1β3γ2L and α4β3γ2L receptors to a 4 s square pulse of subsaturating GABA concentrations (10 and 3 μm, respectively; filled bars) are superimposed on the averaged data traces used for fitting of the mono-liganded rate constants (see Methods).

For both receptor isoforms, we also modified the arrangement of the desensitized states. We found that an accurate fit of the currents evoked by low concentrations of GABA was only possible when a mono-liganded closed state (D₁) was added, as has been previously suggested (Jones & Westbrook, 1995; Mozrzymas et al. 2003). Furthermore, we did not find that a D state connected to the terminal preopen state (C₅) significantly improved our fits. Without very long GABA applications (> 10 s), however, we cannot rule out the possibility that there may be even slower phases of desensitization than those described here, which may necessitate additional states. It should be noted that while at least four phases of desensitization were observed in the presence of saturating GABA concentrations (Table 1), the model proposed here includes only two di-liganded desensitized states. The cause of this discrepancy is twofold. First, it reflects the fact that macroscopic current kinetic properties do not directly correspond to specific microscopic rate constants or any precise connectivity between states. Instead, macroscopic current properties represent a complex mixture of all rate constants in the underlying kinetic scheme, such that the number of desensitized states need not directly correspond to the number of time constants required for fitting of macroscopic desensitization. Indeed, our model with two desensitized states gives rise to currents that require four exponential components to adequately fit the time course of macroscopic desensitization (Table 2). Second, it raises the important question: ‘what is a desensitized state?’ From the standpoint of charge transfer, desensitized states are simply non-conducting states. While it is tempting to propose that the dwell time in a given state determines if it should be designated as ‘desensitized’ (as we have chosen to do here in the interest of clarity), it should be noted that both the α1 and α4 models continue to undergo fast macroscopic desensitization in the absence of these desensitized states (data not shown). This suggests that other non-conducting states such as C₄ and C₅, while much shorter lived, may also play the role of desensitized states.

Table 2.

Kinetic models accurately recapitulate the single channel and macroscopic current responses of α1β3γ2L and α4β3γ2L receptor isoforms to a variety of GABA application protocols

	α1β3γ2L		α4β3γ2L
Parameter	Measured	Simulated	Measured	Simulated
Open Intervals (1 mm GABA)
τ1 (ms)	0.30*	0.30	0.35†	0.35
Fraction O1	0.24*	0.24	0.34†	0.32
τ2 (ms)	1.92*	1.92	1.30†	1.30
Fraction O2	0.48*	0.49	0.63†	0.65
τ3 (ms)	3.47*	3.47	6.30†	6.25
Fraction O3	0.28*	0.27	0.03†	0.03
Mean open time (ms)	2.14*	1.97	1.13†	1.16
Po	0.11*	0.08	0.35††	0.06
Desensitization (4 s, 1 mm GABA)
τ1 (ms)	6 ± 1	3	7 ± 1	7
Contribution A1	0.18 ± 0.03	0.18	0.25 ± 0.03	0.25
τ2 (ms)	56 ± 5	36	58 ± 4	49
Contribution A2	0.07 ± 0.01	0.10	0.13 ± 0.02	0.16
τ3 (ms)	431 ± 45	373	373 ± 41	392
Contribution A3	0.17 ± 0.03	0.22	0.17 ± 0.02	0.19
τ4 (ms)	2652 ± 250	1848	1934 ± 161	1406
Contribution A4	0.33 ± 0.03	0.32	0.33 ± 0.03	0.30
Residual Current	0.25 ± 0.02	0.18	0.12 ± 0.02	0.09
Deactivation (5 ms, 1 mm GABA)
Weighted τ (ms)	200 ± 24	214	371 ± 55	359
EC50
Peak (μm)	14	14	15	15
4 s (μm)	3.0	1.8	1.0	0.6
Repetitive Stimulation (5 ms, 1 mm GABA; final/initial peak)
2 pulses – 12.5 Hz	0.84 ± 0.02	0.77	0.60 ± 0.05	0.63
2 pulses – 1 Hz	0.89 ± 0.01	0.85	0.68 ± 0.04	0.65
2 pulses – 0.05 Hz	0.99 ± 0.01	0.99	0.92 ± 0.02	0.96
4 pulses – 10 Hz	0.75 ± 0.03	0.62	0.52 ± 0.06	0.46

^*Haas & Macdonald (1999).

^†Akk et al. (2004). ‡Note that this value overestimates the true P_o as it only reflects intracluster P_o.

Our studies and those of Akk et al. (2004) used slightly different transfection techniques and GABA subunit combinations. Furthermore, without repeating the single channel recording of others, we were not able to directly fit our model to actual single channel data and cannot exclude the possibility that other gating schemes might describe our results equally well. We have therefore chosen not to make further mechanistic arguments about the microscopic kinetic properties of these receptors based on these kinetic models. Rather, we sought to use these models to help predict the responses of these receptor isoforms to GABA application protocols that were not experimentally testable with our current techniques. As shown in Figs 10 and 11, and Table 2, the kinetic models accurately fit the experimentally determined single channel, whole cell, and pharmacological data quite accurately. Thus, we used this model to explore the responses of α1 and α4 subtype containing receptors to GABA under a variety of physiological conditions.

Figure 11.

Kinetic model accurately predicts the concentration dependence of peak and steady state responses to GABA for α1β3γ2L and α4β3γ2L receptor isoforms Using the α1β3γ2L and α4β3γ2L kinetic models, simulations of receptor activation were performed using square 4 s GABA pulses. A, peak currents at various GABA concentrations are compared to the peak current evoked by a square pulse of 10 mm GABA, above which the peak current amplitude did not change. Although slight differences in the shape of the concentration response profiles are evident, the EC₅₀ values are virtually identical for these two receptor isoforms (α1, 14 μm; α4, 15 μm). B, as above, except that 45 s square pulses of GABA were used to allow for the occupancies of all states in the kinetic model to reach equilibrium. As with peak currents, the models predict that α1 and α4 receptors have similar steady-state current EC₅₀ values (α1, 1.8 μm; α4, 0.6 μm). Note that maximal steady-state current for α4 receptors occurs at ∼3 μm, reflecting a predicted lower O state occupancy of di-liganded receptors than of mono-liganded receptors for this isoform at very low GABA concentrations.

Predicted physiological significance of enhanced α4 subunit expression

We first used the kinetic models to predict the responses of α1 versusα4 subtype-containing GABA_A receptors to brief pulses of saturating GABA, as might be expected in synapses. During extremely brief (100 μs) GABA applications, the peak currents of both α1β3γ2L and α4β3γ2L currents were truncated, but this effect was more pronounced for α4β3γ2L currents (Fig. 12A and B). As GABA durations were prolonged within the synaptically relevant range (0.1–1.0 ms), there was a gradual increase in peak currents for both isoforms, although the α4β3γ2L receptor required longer GABA applications to reach its maximal peak current (Fig. 12C). The rate of deactivation after an extremely brief (100 μs) GABA application was similar for α1β3γ2L and α4β3γ2L currents (Fig. 12D). However, the deactivation rate of α4β3γ2L currents became progressively prolonged with longer GABA exposures while deactivation of α1β3γ2L currents was relatively constant following synaptically relevant applications of GABA and only became progressively prolonged after GABA exposures longer than 1 ms. Because charge transfer after a single synaptic event depends on the complex interplay between peak current and the time course of deactivation, inhibition mediated by α1β3γ2L currents depended linearly on the duration of GABA application. In contrast, since both peak current and deactivation time course of α4β3γ2L currents were very sensitive to the duration of GABA, the overall charge transfer was strongly shaped by small changes in the duration of synaptically relevant (< 1 ms) GABA (Fig. 12E).

Figure 12.

The kinetic properties of α1β3γ2L and α4β3γ2L receptor isoforms have a unique dependence on the duration of GABA application A, incrementally increasing the GABA pulse (1 mm; arrow) duration to α1β3γ2L receptors from 0.3 ms (continuous line) to 1.0 ms (short dashed line) slightly increased the peak current amplitude, with little effect on the subsequent deactivation rate. Further increasing the GABA application duration to 10 ms (long dashed line) did not further increase the peak amplitude, but did prolong the deactivation (see inset). B, unlike the α1β3γ2L receptor isoform, increasing the GABA pulse (1 mm; arrow) duration from 0.3 ms (continuous line) to 1.0 ms (short dashed line) elicited α4β3γ2L currents with both increased peak amplitude and prolonged deactivation. Like α1 containing receptors, a further increase in GABA duration to 10 ms (long dashed line) had no further effect on peak current but did continue to prolong deactivation. C, brief GABA pulse durations (< 1 ms) elicit a smaller normalized peak response from α4β3γ2L currents than from α1β3γ2L receptor isoforms, which is likely to be a consequence of the slower activation observed with α4β3γ2L receptors. D, deactivation is slower for α4β3γ2L receptors than for α1β3γ2L receptors at all GABA pulse durations. For pulse durations < 1 ms, however, deactivation for α1β3γ2L receptors remains relatively stable, while deactivation for α4β3γ2L receptors continues to accelerate as the pulse length shortens. E, due to the combination of accelerated deactivation and heightened sensitivity of peak current amplitudes to pulses < 1 ms, the charge transfer of α4β3γ2L receptors is more sensitive than α1β3γ2L receptor to changes in synaptically relevant durations of GABA exposure. Furthermore, for pulses longer than 1 ms, conditions analogous to bursts of high frequency stimulation, charge transfer is relatively stable for α4β3γ2L receptors while charge transfer continues to increase with increasing GABA durations for α1β3γ2L receptors.

As a next step toward understanding the potential role of α1 versusα4 subtype-containing receptors in synaptic physiology, we used the model to predict the responses of these receptors to repetitive, brief applications of GABA (1 ms, 1 mm) (Fig. 13A and B). The responses of α1β3γ2L and α4β3γ2L receptors to repetitive exposure of brief pulses of GABA were associated with a progressively diminished peak current (Fig. 13C) and charge transfer per pulse (Fig. 13D), thereby limiting the overall charge transfer that occurred during a long barrage of inhibitory synaptic input, as might occur during a seizure. The overall charge transfer of both α1β3γ2L and α4β3γ2L currents exhibited a steep dependence on stimulus frequency, but α4β3γ2L currents were attenuated by repetitive stimulation at lower frequencies than α1β3γ2L currents (Fig. 13E). Similar results were obtained by modelling repetitive 100 μs pulses of GABA (data not shown). Thus, assuming equal receptor densities and similar peak current open probabilities, synapses with α4 subtype-containing receptors would have a stronger inhibitory drive at low stimulus frequencies than those with α1 subtype-containing receptors given their longer deactivation time course. However, α4 subtype containing synapses would also be expected to respond to a narrower range of IPSC frequencies compared to α1 subtype-containing synapses.

Figure 13.

α4β3γ2L and α1β3γ2L receptors respond to different frequencies of stimulation during repetitive GABA exposure A and B, using the kinetic models for the α1β3γ2L and α4β3γ2L receptors, simulations of pulse trains were performed to evaluate the response of these isoforms to repetitive stimulation. 1 ms square pulses of 1 mm GABA were applied at the indicated frequencies for 10 s. For comparison, the responses are superimposed on the response to a continuous application of 1 mm GABA (dashed line). Note that the shape of the response to repeated stimulation for both receptor isoforms begins to approach that of the continuous application at higher frequencies. C, the peak current of the final pulse in the pulse train is compared to a control 1 mm pulse. Although the peak response of both isoforms is attenuated with high frequency stimulation, α4β3γ2L peak currents are more sensitive than α1β3γ2L peak currents. D, charge transfer of the final pulse in the pulse train is compared to a control 1 mm pulse. As with peak currents, the charge transfer following a GABA pulse applied to α4β3γ2L receptors is more sensitive to high frequency stimulation than the charge transfer of α1β3γ2L receptors. E, the total charge transfer of the 10 s pulse train is compared at each frequency to the total charge transfer of a continuous GABA application. Compared to α1β3γ2L receptors, α4β3γ2L receptors approach maximal charge transfer at much lower frequencies. Indeed, the entire frequency–response curve is shifted to the left for α4 containing receptors, suggesting that these receptors may act as low-pass filters compared to α1 containing receptors.

In normal rats, α4 subtype-containing receptors on hippocampal dentate granule cells occupy a primarily perisynaptic position. Activation of these receptors would be expected to respond to synaptic overflow of GABA, which may be largely influenced by changes in GABA reuptake. There is also evidence for synaptic crosstalk among synapses that appears to involve overflow from one synapse to another (Overstreet & Westbrook, 2003). Furthermore, there are complex changes in the expression of GABA transporters associated with epilepsy (Dalby, 2003) with the potential to modify the basal GABA levels in both the synaptic and extrasynaptic spaces. We therefore sought to explore how these receptors responded following exposure to low levels of GABA. We first used the model to predict the response to a synaptically relevant GABA concentration (1 mm) following prior exposure to low GABA concentrations. Exposure to low levels of tonic GABA greatly attenuated the response to 1 mm GABA (Fig. 14). For illustrative purposes, the duration of low GABA concentrations was varied to offset the subsequent responses to 1 mm GABA; identical results were obtained with a single fixed duration of low GABA. The responses of α4β3γ2L receptors were particularly sensitive to attenuation during application of very low concentrations of GABA (Fig. 14C). Alternatively, perisynaptic and extrasynaptic GABA receptors may be constantly bathed in low levels of GABA but still need to react to small changes in ambient GABA in response to changes in nearby neuronal activity. Consistent with our whole cell current data, prior exposure to a physiologically relevant low GABA concentration (1 μm) strongly suppressed the response to higher concentrations of GABA, with this suppression being especially apparent for α4β3γ2L receptors.

Figure 14.

α4β3γ2L receptors are more inhibited than α1β3γ2L receptors following exposure to low levels of GABA A, low concentrations of GABA (dashed lines) ranging from 100 nm to 6 μm (100 nm, 200 nm, 300 nm, 600 nm, 1 μm, 2 μm, 3 μm and 6 μm) were applied to the α1β3γ2L kinetic model for a minimum of 40 s prior to being exchanged for 1 mm GABA for the remainder of the 80 s trace (continuous lines). The length of the preapplications was varied for illustrative purposes only, and does not affect the results (equilibrium occupancy is reached by all states in the kinetic model before the earliest time point). The control 1 mm pulse is indicated with an asterisk. B, simulations were performed as in A, except using the α4β3γ2L kinetic model. C, peak currents evoked by 1 mm GABA following prolonged exposure to various GABA concentrations are compared to the control 1 mm GABA application. α4β3γ2L receptor peak current responses to 1 mm GABA were particularly sensitive to attenuation by prior exposure to (IC₅₀ = 0.8 μm) compared to α1β3γ2L receptors (IC₅₀ = 1.8 μm). D, after preapplication with 1 μm GABA, receptors were switched into various concentrations of GABA. The resulting peak currents were compared to peak currents evoked by a control 1 mm application. At all but the lowest GABA concentrations, α4β3γ2L receptors responded more weakly compared to α1β3γ2L receptors.

Discussion

Although α4 subtype up-regulation has been consistently found in multiple models of temporal lobe epilepsy, the actual physiological roles of α4 subtype-containing GABA_A receptors remain unknown. It is unclear whether α4 subtype up-regulation contributes to the generation of epilepsy or, alternatively, is a compensatory response to seizures. To date, this subtype has been most well studied in the thalamus and hippocampus, where it assembles with δ subunits. The high GABA sensitivity of α4βδ receptors allows them to respond to the persistent low levels of GABA present in the extrasynaptic space, thereby conveying a long-lasting tonic current with the ability to powerfully modify neuronal excitability in these brain regions (Nusser & Mody, 2002; Jia et al. 2005; Cope et al. 2005).

Much less is known about the physiological role of α4βγ receptors. Based on immunoprecipitation studies and analysis of benzodiazepine-insensitive Ro 15-4513 binding of native brain receptors, α4βγ receptors appear to be at least as abundant as α4βδ receptors (Benke et al. 1997; Bencsits et al. 1999). Furthermore, acute status epilepticus and chronic epilepsy can disrupt the normal patterns of GABA_A receptor subunit expression (Naylor et al. 2005; Mangan et al. 2005). The increased expression of α4 subtypes in epileptic animals is correlated with up-regulation of γ2 subunit expression, and also with down-regulation of δ subunit expression in the hippocampal dentate gyrus (Peng et al. 2004; Nishimura et al. 2005). These findings raise the intriguing possibility that there is a preferential assembly of α4βγ receptors at the expense of α4βδ receptors in some models of epilepsy. As a first step toward elucidating the role of α4 subtype expression in neuronal circuits, we used a variety of GABA application protocols to perform a direct comparison of the kinetic properties of recombinant α4β3γ2L and α1β3γ2L receptors. We found that α4β3γ2L currents activated more slowly, desensitized more rapidly and extensively, and deactivated more slowly than α1β3γ2L currents. In addition, α4β3γ2L currents had slower recovery from desensitization, greater loss of peak current amplitude in response to repetitive stimulation, and had more suppressed responses following exposure to low GABA concentrations than α1β3γ2L currents.

While the microscopic basis for the altered kinetic properties is currently unknown, it is of interest that the single channel properties of α4β2γ2L receptors (Akk et al. 2004) are similar to those of α1β2/3γ2L receptors (Fisher & Macdonald, 1997; Steinbach & Akk, 2001), suggesting that the microscopic kinetic processes underlying activation of these two isoforms are analogous. Following agonist binding, α4β3γ2L receptors likely gain access to multiple open and closed states, many of which are long-lived. Progressive accumulation of receptors into long-lived non-conducting states, referred to as desensitized states, gives rise to the phenomenon of macroscopic desensitization (Jones & Westbrook, 1995; Haas & Macdonald, 1999). Since these states cannot release bound GABA (Bianchi & Macdonald, 2001a), receptors must transition though multiple intermediate closed and open GABA-bound states in order to unbind GABA, the macroscopic correlate of which is prolonged deactivation (Jones & Westbrook, 1995; Haas & Macdonald, 1999). Thus, one parsimonious interpretation of our results is that α4β3γ2L receptors have an increased steady-state occupancy of desensitized states, which would explain both the increase in the extent of macroscopic desensitization and the prolongation of deactivation observed with α4β3γ2L currents. It should be noted, however, that the complexity of GABA_A receptor gating makes it almost impossible to discern microscopic properties from macroscopic data alone, particularly as the number of agonist bound states increases. Indeed, stabilizing the occupancy of open states can also prolong deactivation, and depending on the precise connectivity of the underlying kinetic scheme, can lead to seemingly paradoxical consequences with regard to macroscopic desensitization (Bianchi & Macdonald, 2001b; Wagner et al. 2004). While outside the scope of the current results, obtaining single channel data that directly compares the microscopic properties of α1β3γ2L and α4β3γ2L receptors will be beneficial for ruling out other bases for the altered macroscopic currents reported here, such as altered opening or closing kinetics.

α4β3γ2L receptors as synaptic GABA_A receptors

Since high concentrations of GABA are thought to exist in the synaptic cleft for only very brief periods of time (< 1 ms), effective synaptic GABA_A receptors should activate very quickly and deactivate slowly. Within individual synapses, however, there is significant variability in the peak amplitude and duration of IPSCs that appears to be due to fluctuations in both peak GABA concentration and duration (Nusser et al. 2001; Mozrzymas, 2004). Similar to native receptors, the activation rates reported here were quite dependent on GABA concentrations (Maconochie et al. 1994) and were consistently slower for α4β3γ2L receptors during all but the highest GABA concentrations. Therefore, while the concentration time course of synaptic GABA may contribute to the shape of postsynaptic responses, our results suggest it may do so differently in α1β3γ2L and α4β3γ2L receptor-containing synapses. While multiple studies have found increased α4 subtype expression in epileptic animals (Schwarzer et al. 1997; Brooks-Kayal et al. 1998; Sperk et al. 1998), it is currently unclear whether this is associated with an increased number of α4 subtype-containing receptors that are actually within the synapses. One model of CNS hyperexcitability, chronic intermittent ethanol, produces animals with a lowered seizure threshold and an increased expression of α4 subunits in the hippocampus. In these animals, the subcellular localization of α4 subtype-containing receptors shifts from a primarily perisynaptic to a central synaptic location on dentate granule cells (Liang et al. 2006). If the same phenomenon proves true in epileptic animals, our results would predict that α4β3γ2L receptor-containing synapses would have attenuated peak currents during brief synaptic GABA. Our results also predict, however, that these responses would have prolonged decay times compared to α1β3γ2L currents, thus suggesting a possible compensatory role for up-regulation of α4 subtype-containing GABA_A receptors. In the hippocampal dentate region of epileptic animals, there is a loss of interneurons and an associated reduction in IPSC frequency (Kobayashi & Buckmaster, 2003; Cohen et al. 2003; Naylor et al. 2005). The prolonged responses conveyed by α4 subtype-containing synapses may help to normalize the charge transfer over time. However, because of the observed coupling between desensitization and deactivation, α4βγ receptors are rendered relatively insensitive to prolonged repetitive stimulation, especially at frequencies above 2–5 Hz. Therefore, α4βγ receptors may serve as low pass frequency filters, which respond best to brief low frequency barrages of synaptic activity, but poorly to high frequency bursts of synaptic input. Similar to our findings, studies using brain slices have shown prolonged IPSCs in dentate granule cells during status epilepticus or chronic epilepsy (Cohen et al. 2003; Shao & Dudek, 2005; Naylor et al. 2005). However it is unknown what role, if any, enhanced α4 subtype expression plays in altered inhibitory neurotransmission in epileptic rats. One of the few reports describing the physiological significance of increased α4 subtype expression used recordings from CA1 pyramidal neurons following neurosteroid withdrawal. Pharmacological characterization of the IPSCs and comparison with recombinant GABA_A receptors suggest that synaptic α1βγ GABA_A receptors may be replaced by α4βγ and/or α4βδ receptors (Hsu et al. 2003; Smith & Gong, 2005). Interestingly, in contrast to our studies, the IPSCs in neurosteroid withdrawn neurons are significantly shortened. This discrepancy may be related to several technical differences, but one likely explanation is that the different findings are due to differences in the β subtype. Work in other models of epilepsy have also found accelerated IPSC decay in CA1 pyramidal neurons (Mangan & Bertram, 1997; Morin et al. 1998), where the predominant β subtype is β2. However, the enhanced α4 subtype expression in epileptic animals is found in the dentate gyrus where the β3 subtype is highly expressed. Consistent with this, preliminary work has found that the deactivation of α4β2γ2L currents is very brief compared to α4β3γ2L currents (Lagrange & Macdonald, 2005). Thus, the physiological consequences of α4 subtype expression may be very different from region to region, depending on the other GABA_A receptor subtypes that are available for assembly.

α4β3γ2L receptors as extrasynaptic GABA_A receptors

In addition to brief IPSCs, many neurons also have a long-lasting ‘tonic’ current that is largely conveyed by extrasynaptic GABA_A receptors in response to ambient low levels of GABA. Although this tonic current is small in amplitude, its long duration allows the transfer of substantial charge and may play an important role is regulating neuronal excitability (Mody, 2001). The GABAergic tonic current is most commonly mediated by αβδ or occasionally α5βγ receptors (Nusser et al. 1998; Spigelman et al. 2003; Caraiscos et al. 2004; Cope et al. 2005; Belelli et al. 2005). Pharmacological experiments have suggested that α4βγ receptors may also convey part of the tonic current in some brain regions, and that the subcellular distribution of synaptic and extrasynaptic α1 and α4 subtypes may be altered in pathological conditions (Liang et al. 2004; Liang et al. 2006). We found that during prolonged GABA application both α1β3γ2L and α4β3γ2L receptors responded to a remarkably limited range of GABA concentrations, with a maximal response evoked by < 10 μm GABA. Therefore, small changes in GABA concentration (on the order of 1–2 μm) would have a relatively large effect on the current produced by extrasynaptic α1βγ or α4βγ receptors. However, persistent low levels of GABA renders GABA_A receptors largely insensitive to higher GABA concentrations (Overstreet et al. 2000), an effect that is much more pronounced with the highly desensitizing α4βγ receptors. As such, it is conceivable that drug treatments that increase the levels of extrasynaptic GABA may differentially suppress the ability of extrasynaptic α4βγ receptors to respond to variations in ambient GABA that may occur during status epilepticus (Naylor et al. 2005).

Future directions in understanding the response of GABA_A receptors to physiological GABA applications

These studies used recombinant GABA_A receptor proteins in an immortalized cell line to elucidate the kinetic properties of specific GABA_A receptor subunit combinations. It is possible that modulatory factors, such as protein phosphorylation, neurosteroids, or anchoring proteins differentially alter the kinetics of one GABA_A receptor isoform, but not another. Moreover, in some brain regions, a given GABA_A receptor population may be composed of receptors containing multiple α or β subtypes within the same receptor (Fritschy et al. 1992; Mertens et al. 1993). For example, using immunoprecipitation it has been shown that α4 subtypes may exist in combination with α1, α2, or α3 subunits (Benke et al. 1997). The kinetic properties of GABA_A receptors with a mixture of α subunits are currently unpredictable and will need to be specifically tested.

In summary, α4βγ GABA_A receptors produce rapidly desensitizing and slowly deactivating currents. Based on these findings, disease states in which α1βγ receptors are replaced by α4βγ receptors would be expected to be associated with impaired inhibitory neurotransmission. These changes would alter both synaptic and extrasynaptic inhibition in a frequency and GABA concentration-dependent fashion. Since epilepsy is a network phenomenon, further experiments in brain slices will be needed to specifically test whether altered α4 subtype expression in epilepsy is actually associated with the changes in CNS neurotransmission predicted by these studies. Nonetheless, our results provide a basis for testable hypotheses about the role of altered synaptic and extrasynaptic inhibitory neurotransmission in epilepsy.

Supplementary material

The online version of this paper can be accessed at:

DOI: 10.1113/jphysiol.2006.122135

http://jp.physoc.org/cgi/content/full/jphysiol.2006.122135/DC1 and contains supplemental material consisting of a figure:

Figure S1.

Fitting the desensitization of whole cell GABA-evoked currents required multiple exponentials

This material can also be found as part of the full-text HTML version available from http://www.blackwell-synergy.com