Conductance of Recombinant GABA Channels Is Increased in Cells Co-expressing GABAA A Receptor-associated Protein*

High conductance γ-aminobutyric acid type A (GABAA) channels (>40 picosiemens (pS)) have been reported in some studies on GABAA channels in situ but not in others, whereas recombinant GABAA channels do not appear to display conductances above 40 pS. Furthermore, the conductance of some native GABAA channels can be increased by diazepam or pentobarbital, which are effects not reported for expressed GABAA channels. GABARAP, a protein associated with native GABAA channels, has been reported to cause clustering of GABAA receptors and changes in channel kinetics. We have recorded single channel currents activated by GABA in L929 cells expressing α1, β1, and γ2S subunits of human GABAA receptors. Channel conductance was never higher than 40 pS and was not significantly increased by diazepam or pentobarbital, although open probability was increased. In contrast, in cells expressing the same three subunits together with GABARAP, channel conductance could be significantly higher than 40 pS, and channel conductance was increased by diazepam and pentobarbital. GABARAP caused clustering of receptors in L929 cells, and we suggest that there may be interactions between subunits of clustered GABAA receptors that make them open co-operatively to give high conductance “channels.” Recombinant channels may require the influence of GABARAP and perhaps other intracellular proteins to adopt a fuller repertoire of properties of native channels.

Some, but not all, native ␥-aminobutyric acid type A (GABA A 1 ) channels have been reported to have maximum conductances above 40 picosiemens (pS) and many subconductance states (1)(2)(3), whereas recombinant GABA A channels formed by a wide range of subunits do not have such high conductances. Furthermore, the conductance of some native channels is increased by drugs such as diazepam, pentobarbi-tal, and propofol (4 -6), which are effects not reported for expressed GABA A receptors. A possible explanation for these differences is that native receptors are not identical to expressed receptors perhaps because of specific interactions that occur in native but not heterologous cells.
A GABA A receptor-associated protein, GABARAP, is an intracellular protein that can interact with the ␥ 2 subunit of GABA A receptors (7). When co-expressed with GABA A subunits in QT-6 quail fibroblasts, GABARAP causes clustering of GABA A receptors accompanied by changes in whole cell current kinetics (8). Although this protein was not found in close proximity to mature native GABA A receptors in the plasmalemma of cortical neurons in a study using fluorescent antibodies to GABARAP (9), there is compelling evidence that co-expression of GABARAP with subunits of GABA A receptors does result in increased clustering of receptors in many cells (8). These observations suggest that GABARAP may be involved at an early stage in the clustering of GABA A receptors even though it appears not to be present as part of the anchoring mechanism for the receptors.
Because it has been suggested that the clustering of GABA A receptors may allow co-operative opening of channels and hence a higher apparent single channel conductance (see Ref. 10 for review), we have co-expressed GABARAP with subunits of GABA A receptors in an attempt to cause clustering of the receptors. We report here that co-expression of GABARAP with subunits of GABA A receptors causes the clustering of receptors in L929 fibroblasts and also causes the appearance of high conductance channels that are only rarely seen in control cells transfected in the same way with the same GABA A subunits but without GABARAP.

EXPERIMENTAL PROCEDURES
Plasmid cDNAs and Antibodies-Full-length cDNAs for the human GABA A subtypes ␣ 1 , ␤ 1 , and ␥ 2S were subcloned individually into the pcDNA3.1ϩ (Invitrogen) mammalian expression vector. Green fluorescent protein was expressed off of the pEGFP-N1 vector (Clontech, Palo Alto, CA). Full-length human GABARAP cDNA, cloned into pCDNA3 (Invitrogen), was a gift from Dr. R. W. Olsen (UCLA). The monoclonal antibody bd24 was purchased from Chemicon International, and fluorescein isothiocyanate conjugated rabbit anti-mouse IgG was purchased from DAKO.
Cell Culture-Mouse L929 fibroblasts (American Type Tissue Culture Collection, Rockville, MD) were maintained in Dulbecco's modified Eagle's Medium plus 10% heat-inactivated fetal bovine serum (CSL), 10 mM HEPES, and 4 mM L-glutamine. Cells for passaging were lifted using a cell scraper or following a 2-min incubation in phosphatebuffered saline (pH 7.3) containing 0.5% trypsin and 0.2% EDTA.
Transfection-Cells were transfected with plasmids encoding GABA A subunits ␣ 1 , ␤ 1 , ␥ 2S , and green fluorescent protein, with or without GABARAP. Cells were plated onto glass coverslips in multiwell dishes at a density of 1.0 -1.5 ϫ 10 5 cells/ml. Twenty-four hours later cells were washed with Optimem (Invitrogen) and then transfected with a total of 20 g of plasmid DNA (in equal amounts) plus 10 l of Lipofectin reagent (Invitrogen) in accordance with manufacturer's instructions. After a 5-h incubation at 37°C the mixture was replaced by fresh medium, and 48 -72 h later the cells were used for electrophysiological experiments.
Western Blotting-Lysates from untransfected and transfected cells were run on 4 -12% gradient gels (Invitrogen) in MES buffer. Proteins were transferred to nitrocellulose using a semidry buffer system. Following blocking in milk powder the blot was incubated in the anti-GABARAP polyclonal antibody (7) (1/3000) at 4°C overnight. After thorough washing the blot was incubated in secondary antibody (antirabbit horseradish peroxidase) for 2 h at room temperature and then visualized by ECL (West-Pico, Pierce). All blots included purified GABARAP as a positive control.
Immunofluorescent Labeling-Cells were transfected (as described above) with plasmids encoding GABA A subunits ␣ 1 , ␤ 1 , and ␥ 2S with or without GABARAP. Staining and fixation were carried out at 4°C 48 h post-transfection. Cells on coverslips were washed with phosphatebuffered saline containing 1% bovine serum albumin then incubated for 1 h with the anti-␣ 1 monoclonal antibody bd24 (1:50). Following fixation with 1% paraformaldehyde, cells were washed with phosphate-buffered saline, incubated for 1 h with fluorescein isothiocyanate-conjugated secondary antibody (1:50), and then washed again prior to mounting onto slides in Slowfade (Molecular Probes). Cells were imaged on a Radiance 2000 confocal microscope with a 60ϫ oil-immersion objective. Non-transfected cells treated as described above did not show any immunofluorescence.
Electrophysiology-Electrophysiological experiments were performed using conventional patch clamp techniques in cell-attached or outside-out patches. Cells were viewed using a light fluorescent attachment (Olympus BH2-RFL) so that only those cells successfully transfected with green fluorescent protein were selected for patching. Electrodes made from borosilicate glass (Clark Electromedical) and coated with Sylgard (Dow Corning) had resistances between 10 and 15 megohms.
Cell Outside-out Patches-The bath solution contained (in mM): 135 NaCl, 5 KCl, 2 MgCl 2 , 2 CaCl 2 , 10 glucose, and 10 TES, pH 7.4. Patch electrodes were filled with a solution containing (in mM): 50 NaCl, 80 KCl, 2 CaCl 2 , 2 MgCl 2 , 5 EGTA, and 10 TES, pH 7.3-7.4. With these solutions the predicted chloride equilibrium potential was close to 0 mV. Drugs were made up to final concentration in bath solution (as described below) and were applied to outside-out patches directly via gravity-fed flow tubes. The flow tubes had a diameter of 300 m and were placed directly in front of the outside-out patch. The presence of GABA A receptors in outside-out membrane patches was confirmed by an initial application of GABA (1 mM) applied via the flow tube.
Testing Drugs-Where indicated, pentobarbital (Sigma) was dissolved in bath solution. Diazepam (Hoffman-La Roche) was initially dissolved in dimethyl sulfoxide (Me 2 SO, Sigma) before being dissolved in bath solution. The final concentration of Me 2 SO in the bath solution was 0.01%. Both drugs were applied by gravity-fed perfusion through the bath. For application of bicuculline (Sigma) to a cell-attached patch, a sudden bolus of drug was ejected from the end of a fine polyethylene tube inserted in the electrode to within 0.5-1.0 mm of the tip. Because the increases in channel conductance caused by diazepam and pentobarbital were more prominent the lower the predrug (control) conductance (4, 6) was, patches containing low conductance (Ͻ30 pS) channels were normally used for testing the effects of these drugs on channel conductance.
Analysis of Currents-Currents were recorded using an Axopatch 200A current-to-voltage converter (Axon Instruments), filtered at 5 or 10 kHz, digitized at 44 kHz using a pulse-code modulator (Sony PCM 501), and stored on videotape for analysis. In some experiments, currents were digitized at 10 kHz with a Digidata 1200 A-D converter and stored on computer disk. Currents were played back and digitized at a frequency of 10 kHz for analysis using a computer program, Channel 2. 2 Current amplitudes were measured from all-points amplitude proba-bility histograms or from the directly measured amplitudes of individual currents. Where single channel currents displayed subconductance states, the single channel current amplitude was taken as the most frequently occurring and highest state. Channel conductance was calculated by dividing current amplitude by the difference between the pipette potential and the reversal potential, the latter being close to 0 mV for the solutions used (see Fig. 2, B and D). "Mean currents" were measured over periods of 10 s or more by integrating all data points, normally sampled at 10 kHz, to obtain total current and then dividing by the number of data points. The base-line current level was set during periods when there was no channel activity, and current variance was at a minimum. The base-line level was monitored throughout a record to ensure that no shift had occurred. This measure is a rapid and observer-free method of monitoring changes in channel open probability when channel conductance is not changing. Average results are expressed as mean Ϯ 1 S.E., and the probability of significant differences between means was calculated using the Student's t test.

Effects of GABARAP on GABA A Receptor Clustering
In L929 cells transfected with GABA A ␣ 1 , ␤ 1 , and ␥ 2S cDNAs, the fluorescently labeled antibody bd24 that labels the ␣ 1 subunit was diffusely distributed over the surface of the plasmalemma (Fig. 1A). In contrast, when L929 cells were cotransfected with cDNAs for the same three subunits plus GABARAP a punctate distribution of fluorescent staining, as illustrated in Fig. 1B, was observed in 58 -69% of cells. This punctuate appearance was visible in cells taken from five different sets of cells cotransfected with GABARAP but was not seen in "control" cells that had been transfected with the GABA A subunits but not GABARAP. These results are similar to those obtained in QT-6 fibroblasts (8) and indicate that GABARAP causes clustering of GABA A receptors in L929 fibroblasts also.

Characteristics of Recombinant
Conductance-Typical single channel currents recorded in cell-attached patches on cells transfected with the three GABA A subunits are shown in Fig. 2A (GABA concentration in the pipette was 3 M). The currents reversed at a pipette potential (V p ) close to 0 mV, as expected for chloride currents with the pipette solutions used, and showed no rectification (Fig. 2B). The maximum conductance (␥ m ) of the channels shown in Fig. 2, A and B  Single channel currents were also recorded in outside-out patches. Typical currents activated by 1 M GABA in one of these patches can be seen in Fig. 2C (V p ϭ ϩ60 mV top trace, Ϫ60 mV bottom trace). The currents reversed at 0 mV and were non-rectifying (Fig. 2D). In the experiment illustrated in Fig. 2, C and D, channels had an average conductance of 29.2 Ϯ 0.008 pS (n ϭ 934 openings). In 33 outside-out patches, similar nonrectifying currents had an average maximum conductance of 30.5 Ϯ 0.41 pS.
Effects of Diazepam-In five of the six patches tested, diazepam caused no significant increase in conductance of recombinant channels formed by the three GABA A subunits alone as illustrated in Fig. 3 Effects of Pentobarbital-Pentobarbital also increased channel activity but there was no increase in channel conductance, as illustrated in Fig. 3, C and D. Channels in a cell-attached patch (V p ϭ Ϫ60 mV) activated by 3 M GABA (Fig. 3C) had a maximum conductance of 25 pS (mean current 0.09 pA). Within 1 min of exposing the cell to 100 M pentobarbital (Fig. 3D) channel activity had increased, and currents from two channels were superimposing. Mean current had doubled (0.18 pA), but there was no significant change in channel conductance. Similar results were obtained in three cell-attached patches exposed to 100 M pentobarbital. Average channel conductance before exposure to pentobarbital was 21. The currents reversed at a V p of ϳ0 mV (Fig. 4C) as expected for chloride currents with the pipette solution used. An allpoints histogram constructed from a longer section of the record at a V p of Ϫ80 mV shows a base-line peak at 0 pA and another prominent peak at ϳ6 pA with little in between indicating that subconductance states were infrequent. The maximum current amplitude was 6.5 pA when V p was Ϫ80 mV and Ϫ4.8 pA when V p was ϩ80 mV. As the currents reversed close to 0 mV (Fig. 4, A and C) these currents correspond to conductances of about 80 and 60 pS, respectively. Such high conductances were not seen in cells expressing ␣ 1 ␤ 1 ␥ 2S channels but not co-expressing GABARAP. The current-voltage curve in Fig.  4C shows some outward rectification. In 16 of 25 cell-attached patches, channels activated by GABA had a significantly higher maximum conductance than channels recorded in the control cells not expressing GABARAP. In the 16 patches showing high conductance channels average ␥ m (V p ϭ Ϫ60 mV) was 60.7 Ϯ 4.3 pS, which is significantly higher (p Ͻ 0.0001) than in the 15 patches in cells not expressing GABARAP (22.3 Ϯ 1.2 pS). The average channel conductance in the nine patches not showing a high conductance was 29.1 Ϯ 1.94 pS. Application of bicuculline (100 M) to the outer membrane surface (facing the pipette solution) in two patches containing high conductance channels blocked channel activity.
Effects of Diazepam-Diazepam has been reported to increase the open probability (11) and conductance (4, 5) of native GABA A channels. Diazepam produced both effects on channels in 11 cell-attached patches on cells cotransfected with the GABA A subunits and GABARAP. An increase in ␥ m caused by 10 M diazepam is illustrated in Fig. 5. Before exposure to diazepam, the amplitude of single channel currents (V p ϭ Ϫ40 mV) activated by 3 M GABA (Fig. 5A) was 1.32 pA corresponding to a conductance of 33 pS. The traces in Fig. 5B, recorded 2.5, 26, and 80 s after the exposure of the cell to 10 M diazepam, and the associated all-points histograms to the right of each trace, show a progressive increase in channel conductance and open probability during the exposure to diazepam. The average channel conductance recorded in the 11 cell-attached patches activated by 3 M GABA was 32.3 Ϯ 4.74 pS and after exposure to 10 M diazepam was increased to 56.3 Ϯ 7.74 pS (p Ͻ 0.01). Conductance after diazepam application increased regardless of the initial conductance seen in the patch. In 3 of the 11 patches in which the initial GABA-activated conductance was on average 52.0 Ϯ 8.93 pS, conductance increased to 85.3 Ϯ 14.7 pS (p ϭ 0.05) after exposure to diazepam. In the eight patches containing lower conductance channels with an average conductance of 24.9 Ϯ 3.40 pS, conductance increased to 45.4 Ϯ 5.70 pS (p Ͻ 0.005) after exposure to diazepam. Similar results were obtained from four inside-out patches in which ␥ m increased from 52.3 Ϯ 8.93 to 84.3 Ϯ 6.30 pS (p Ͻ 0.05) after exposure to diazepam.
Effects of Pentobarbital-Barbiturates have also been reported to increase GABA A channel-open probability (11) and conductance (6) in native channels. In patches in cells cotransfected with the GABA A subunits but not with GABARAP, pentobarbital caused no increase in channel conductance as illustrated for a cell-attached patch in Fig. 3, C and D. In contrast, in 10 patches from cells expressing the GABA A subunits and GABARAP, pentobarbital did increase channel conductance from 41.9 Ϯ 5.65 to 63.9 Ϯ 5.52 pS (p Ͻ 0.01). The effect of pentobarbital on conductance was most marked in 5 of the 10 patches where the initial GABA-activated currents were small (corresponding to a ␥ m Ͻ 40 pS) increasing from 28.4 Ϯ 4.50 to 59.8 Ϯ 5.16 pS following application of 100 M pentobarbital (p Ͻ 0.001). This effect is illustrated in Fig. 6. Before exposure to pentobarbital (Fig. 6A), single channel conductance was 25 pS (V p ϭ Ϫ40 mV), and the all-points histogram contained multiple peaks that corresponded to the superimposed activity of several channels in the patch. The separation of 1.0 pA between adjacent peaks reflects the single channel conductance of 25 pS. Following exposure of the patch to 100 M pentobarbital, channel activity became greater and single channel current amplitude increased (Fig. 6B). There was still evidence of several channels in the patch, but the separation between

Outside-out Patches
Conductance-High conductance channels were seen in a lower percentage of outside-out patches than in cell-attached patches from cells transfected with the GABA A subunits plus GABARAP. GABA A channels with conductances above 40 pS were recorded in 6 of 20 outside-out patches activated with 1 M GABA. Examples of these high conductance channels recorded in one of these outside-out patches are shown in Fig. 7A. The traces are shown on a fast time base to illustrate that there were immediate (within 100 s) transitions between conductance levels, i.e. they were "single" channel openings and not superimposed non-synchronous openings. The average maximum conductance of the channels in the patch illustrated in Fig. 7A was 61.9 Ϯ 0.011 pS (n ϭ 475 openings). In the six patches containing the high conductance channels, the average ␥ m was 53.7 Ϯ 2.8 pS, which is significantly different from the ␥ m in cells not expressing GABARAP described above (p Ͻ 0.0001). In the other 14 patches, channel conductance was 28.5 Ϯ 0.85 pS, which is not significantly different from results obtained in cells not transfected with GABARAP.
Effects of Diazepam-In 5 of 13 outside-out patches tested with diazepam, channel conductance was increased by the drug. Two of these were exposed to 10 M diazepam, the other three were exposed to 1 M diazepam. Examples of this effect in two of the patches are shown in Fig. 7. In Fig. 7, B and C, single channel currents increased in amplitude after the patch was exposed to 10 M diazepam (V p ϭ Ϫ60 mV). The channels in Fig. 7B had a maximum conductance of ϳ30 pS (30.1 Ϯ 0.012 pS, n ϭ 193). After exposure to diazepam, ␥ m increased by a factor of 2 to 60.0 Ϯ 0.02 pS (n ϭ 255). In the two patches showing an increase in channel conductance when exposed to 10 M diazepam, ␥ m increased from 27.7 to 42.3 pS and from 30.6 to 58.7 pS. The effect of 1 M diazepam is illustrated in Fig.  7, D and E. The histograms show the distribution of channel conductances measured before (D) and after (E) exposure to 1 M diazepam. There is a clear shift in channel conductance to higher values. The average channel conductance before diazepam was 25.9 Ϯ 0.40 pS (n ϭ 100) and after diazepam was 40.7 Ϯ 0.52 pS (n ϭ 133). In the three patches exposed to 1 M diazepam, ␥ m increased from 28.0 Ϯ 0.71 to 48.8 Ϯ 1.73 pS.
To test the possible link between channel clustering and conductance, cells were transfected with the three GABA A subunits plus a truncated GABARAP. A truncated form of GABARAP (deletion of the first 35 residues) loses the ability to bind to tubulin or microfilaments and to cause clustering (12). In contrast to the effects of full-length GABARAP, high conductance channels were not seen in cells expressing truncated GABARAP together with the three GABA A subunits. In seven cell-attached patches, ␥ m was 28 Ϯ 1.7 pS (V p ϭ Ϫ60 mV) (Table I), which is similar to the conductance of channels in the absence of GABARAP. In another series of experiments, we recorded GABA A single channel currents in cells expressing GABA A ␣ 1 and ␤ 1 subunits (no ␥ 2S ) together with full-length GABARAP. Because the GABARAP binding site is on the ␥ subunit, GABARAP would not bind to these receptors, and indeed there is no clustering of receptors under these conditions (8). A, examples of high conductance channels activated with 1 M GABA in one of these outside-out patches. The currents were recorded at Ϫ60 mV, and the calibration bar shows 50 pS. B, conductance of channels recorded in another outside-out patch (1 M GABA, V p ϭ Ϫ60 mV). C, conductance of channels recorded from the same patch after exposure to 1 M GABA plus 1 M diazepam. Vertical calibration bar shows 50 pS. D, channel conductance probability histogram from another outside-out patch exposed to 1 M GABA. E, channel conductance probability histogram from the same outside-out patch exposed to 1 M GABA plus 1 M diazepam. and ␤ 1 subunits plus GABARAP, no high conductance channels were seen (Table I). The average channel conductance was 18.9 Ϯ 0.91 pS. These observations suggest that the effects of GABARAP on channel conductance require binding of GABARAP to the GABA A ␥ subunit at some stage of trafficking or assembling of GABA A receptors. DISCUSSION Our results show that co-expression of GABARAP with ␣ 1 , ␤ 1 , and ␥ 2S subunits of GABA A receptors can cause changes in the properties of GABA A channels. When compared with recombinant GABA A channels in cells expressing ␣ 1 , ␤ 1 , and ␥ 2S subunits, channels formed by these subunits in cells also expressing GABARAP often had a higher maximum conductance (above 40 pS), and their conductance could be increased by diazepam or pentobarbital. These properties are strikingly similar to those of GABA A channels in cultured hippocampal neurons from neonatal rats (3,4,6,13).
GABARAP has been shown to cause the clustering of GABA A receptors (8). The cause of this clustering has not been established, and in particular whether GABARAP is closely associated with the receptors (14). Whether or not GABARAP anchors GABA A receptors to the cytoskeleton, it is clear that co-expression of GABARAP does produce aggregation of receptors in hot spots in Japanese quail QT-6 fibroblasts (8) perhaps by influencing their trafficking or insertion in the plasmalemma. We have confirmed that co-expression of GABARAP in the L929 fibroblasts we were using in the electrophysiological experiments caused the aggregation of receptors (Fig. 1). We did not detect endogenous GABARAP in L929 cells using Western blotting with the GABARAP antibody (7) (data not shown). Hence, if L929 cells do contain very low amounts of endogenous GABARAP below the sensitivity of Western blotting, it does not cause the same clustering as the co-expressed GABARAP.
We believe that it is the clustering of receptors produced by the co-expressed GABARAP that is responsible for the changes in "single channel" characteristics. It has been suggested previously that channels can open cooperatively (for reviews see Ref. 10). There is evidence for direct protein-protein interactions (cross-talk) between the intracellular domains of different channels in close proximity (15,16). It is possible that GABARAP, by aggregating receptors, promotes interaction(s) between the intracellular domains of the closely packed receptors so that several pentameric channels open cooperatively. This hypothesis would predict that any method of aggregating receptors might induce cooperative opening and closing of channels. This remains to be tested.
Not all cell-attached patches and fewer outside-out patches from cells co-expressing GABARAP showed high conductance channels. If the high conductance channels depend on the dense packing of receptors, it would not be surprising if some patches did not contain clustered receptors and high conductance channels. Although receptors are presumably in their native state in cell-attached patches, in forming an outside-out patch it is necessary to drag both membrane and receptors into the patch, and it would not be surprising if interactions between receptors and receptor clusters were often disrupted during formation of the patch. This may explain why high conductance channels were seen less frequently in outside-out than in cell-attached patches from cells expressing GABARAP with the three GABA A subunits.
Although the absence of endogenous GABARAP in our L929 expression system is able to explain why high conductance recombinant GABA A channels are not observed here, it is not clear why there appears to be a discrepancy in some native systems. Perhaps there too, the presence or absence of proteins involved in the clustering mechanism is responsible. Also, our data indicate that patch configuration affects the probability of recording high conductance channels. Whatever the mechanism of the changes caused by GABARAP, our results suggest that recombinant channels may not always be ideal preparations for characterizing the properties, or the responses to drugs, of native channels.