Snake neurotoxin α-bungarotoxin is an antagonist at native GABAA receptors

The snake neurotoxin α-bungarotoxin (α-Bgtx) is a competitive antagonist at nicotinic acetylcholine receptors (nAChRs) and is widely used to study their function and cell-surface expression. Increasingly, α-Bgtx is also used as an imaging tool for fluorophore-labelling studies, and given the structural conservation within the pentameric ligand-gated ion channel family, we assessed whether α-Bgtx could bind to recombinant and native γ-aminobutyric type-A receptors (GABAARs). Applying fluorophore-linked α-Bgtx to recombinant αxβ1/2γ2 GABAARs expressed in HEK-293 cells enabled clear cell-surface labelling of α2β1/2γ2 contrasting with the weaker staining of α1/4β1/2γ2, and no labelling for α3/5/6β1/2γ2. The labelling of α2β2γ2 was abolished by bicuculline, a competitive antagonist at GABAARs, and by d-tubocurarine (d-Tc), which acts in a similar manner at nAChRs and GABAARs. Labelling by α-Bgtx was also reduced by GABA, suggesting that the GABA binding site at the receptor β–α subunit interface forms part of the α-Bgtx binding site. Using whole-cell recording, high concentrations of α-Bgtx (20 μM) inhibited GABA-activated currents at all αxβ2γ2 receptors examined, but at lower concentrations (5 μM), α-Bgtx was selective for α2β2γ2. Using α-Bgtx, at low concentrations, permitted the selective inhibition of α2 subunit-containing GABAARs in hippocampal dentate gyrus granule cells, reducing synaptic current amplitudes without affecting the GABA-mediated tonic current. In conclusion, α-Bgtx can act as an inhibitor at recombinant and native GABAARs and may be used as a selective tool to inhibit phasic but not tonic currents in the hippocampus.


Introduction
The snake venom neurotoxin, a-bungarotoxin (a-Bgtx), binds as an inhibitor with high affinity to nicotinic acetylcholine receptors (nAChRs), including the heteromeric muscle receptors composed of abgd or abdε subunits, and homomeric neuronal subtypes, comprising a7, a8, or a9 subunits (Olsen and Sieghart, 2008;Pirker et al., 2000;Whiting et al., 1995). The g-aminobutyric acid (GABA) type-A receptors (GABA A Rs) are Cl À permeable ligand-gated ion channels from the same pentameric Cys-loop superfamily of receptors as nACh, 5-HT3 and glycine receptors (Smart and Paoletti, 2012). Members of this family share a common structural architecture, including an N-terminal extracellular ligand-binding domain and four a-helical transmembrane-spanning domains (Corringer et al., 2012;Miller and Smart, 2010;Thompson et al., 2010). It has long been of interest that two antagonists at nAChRs, d-tubocurarine (d-Tc; Caputi et al., 2003;Simmonds, 1982;Wotring and Yoon, 1995) and trimethapan (Wotring and Yoon, 1995) are also inhibitors at GABA A Rs. In addition, a-Bgtx can also inhibit homomeric b3 subunit-containing GABA A Rs by binding at the subunit interfaces (McCann et al., 2006). Whilst it is unclear whether b3 homomers constitute a defined physiological population of GABA A Rs, if a-Bgtx can bind at the b3eb3 subunit interface, it is plausible that more physiological ab subunit-containing By studying recombinant GABA A Rs expressed in HEK-293 cells, we reveal that from a selection of axb2g2 heteromers, a-Bgtx inhibited a2b2g2 receptors to the greatest extent. Furthermore, fluorescent a-Bgtx coupled to Alexa-Fluor 555 (a-Bgtx-AF555) yielded robust staining of a2b2g2 receptors. This was abolished by d-Tc and by the competitive antagonist at GABA A Rs, bicuculline, as well as by GABA, suggesting that the a-Bgtx-binding site on the GABA A R heteromers is most probably located at the bea interface. We also found that a-Bgtx inhibited GABA currents in hippocampal neurons, reducing the amplitudes of synaptic currents. Overall, a-Bgtx is an inhibitor at GABA A Rs displaying some selectivity for the a2 subunit-containing isoform.
Cells were transfected with equimolar ratios of cDNAs encoding for a1e6, b1e3 and d or g2 GABA A receptor subunits along with eGFP using a calcium phosphate method (Mortensen et al., 2011) 15e45 min after plating.

Primary hippocampal cultures
Dissociated hippocampal neurons were prepared from embryonic day 18 rat pups as described (Hannan et al., 2012). Briefly, hippocampi were dissected in icecold Hank's Balanced Salt Solution (HBSS) (Ca 2þ /Mg 2þ free) before enzymatic dissociation in 0.1% (w/v) trypsin at 37 C for 10 min followed by serial washes in pre-warmed HBSS to remove trypsin prior to trituration. Cells were mechanically dissociated using fire-polished glass Pasteur pipettes in plating medium composed of minimal essential medium (MEM) supplemented with 5% (v/v) fetal calf serum (FCS), 5% (v/v) horse serum, penicillin-G and streptomycin (200 units/ml and 200 mg/ ml), 2 mM L-glutamine and 35 mM glucose. Cells were plated at a density of 10 6 per ml in plating medium on glass cover-slips previously coated with poly-D-lysine. The neurons were grown and maintained at 37 C in a humidified 95% air/5% CO 2 atmosphere.

Whole-cell patch-clamp electrophysiology
Whole-cell GABA-activated currents were recorded from transfected HEK-293 cells or hippocampal neurons in culture at 12e14 DIV using patch clamp electrophysiology. Patch electrodes had resistances of 4e5 MU and were filled with an internal solution containing (mM): 120 CsCl 1 MgCl 2 , 11 EGTA, 30 KOH, 10 HEPES, 1 CaCl 2 , and 2 K 2 ATP; pH e 7.2. HEK-293 cells were superfused with a saline solution containing (mM): 140 NaCl, 4.7 KCl, 1.2 MgCl 2 , 2.52 CaCl 2 , 11 Glucose, and 5 HEPES; pH 7.4. The saline for recording from primary neurons was supplemented with 2 mM kynurenic acid and pH adjusted to 7.4 to block all spontaneous excitatory post-synaptic currents (EPSCs). Membrane currents were filtered at 5 kHz (À3 dB, 6th pole Bessel, 36 dB/octave). HEK-293 cells were studied 48 h after transfection by voltage clamping cells at a holding potential of À20 to À40 mV with optimised series resistance (Rs, <10 MU) and whole-cell membrane capacitance compensation. Neuronal membrane currents were similarly recorded at a holding potential of À60 mV. Changes of Rs greater than 10% during the experiment resulted in the recording being excluded from analysis.
GABA concentrationeresponse curves were generated by measuring the current (I) at each GABA concentration, applied at suitable time intervals, and normalizing the current to the maximum GABA response (I max ), before fitting the concentration response relationship with: where A is the concentration of GABA, EC 50 is the concentration of GABA giving 50% of the maximum response and n is the Hill slope.
For studying inhibition, a-Bgtx was either co-applied, or pre-applied for 30e60 s, followed by co-application with sub-maximal doses of GABA. Spontaneous inhibitory postsynaptic currents (IPSCs) were recorded from dentate gyrus granule cells (DGGCs) with the same internal solution as above. Cells were voltage clamped at À60 mV and IPSCs were recorded using 5 kHz filtering with optimal series resistance and whole-cell capacitance compensation. IPSCs were detected and analysed using WinEDR and WinWCP (John Dempster, University of Strathclyde, UK). IPSC frequency was calculated using events detected over 60 s epochs of recording. For IPSC amplitudes, in excess of several thousand IPSCs were recorded and analysed as overall mean and then displayed as an amplitude distribution and fitted with a sum of 1e4 Gaussian functions of the form: Where A defines the amplitude and C is a constant defining the pedestal of the histogram. This function provided the Gaussian mean amplitude current (m) and standard deviation (s). All the distributions were fitted using this function in Origin (Ver 6). The accuracy of the fits was checked by repeating the iterative non-linear fitting procedure after substituting the best-fit parameters obtained for the control and a-Bgtx datasets with new values.
For tonic inhibition, to determine the average holding currents, a 60 s continuous current recording was sampled every 1 s, discarding epochs that coincided with IPSCs. Any effect of drugs on the holding current was defined by subtracting the average holding currents in control and during drug application.
The baseline noise (RMS) was calculated before and during drug treatment. This was estimated from a continuous (30 s) current recording, sampled every 100 ms. The median current was calculated every 5 s and values more than twice the standard deviation from the median (usually due to IPSCs) were eliminated. Baseline GABA-mediated current noise was defined by subtracting RMS values before and after drugs, e.g., a-Bgtx or bicuculline.

Fluorescent a-Bgtx staining and imaging
Live transfected HEK-293 cells were studied 48 h after transfection and washed with Krebs to remove cell culture media and incubated in 400 nM a-Bgtx coupled with Alexa Fluor 555 (a-Bgtx-AF555; Life Technologies) for 10 min at room temperature (RT). Cells were washed and fixed in 4% paraformaldehyde (PFA; Sigma) for 10 min at RT. The cells were imaged immediately post-fixation in saline using a Zeiss LSM 510 Meta confocal microscope and an Achroplan x40 water DIC objective (NA 0.8) as described previously (Hannan et al., 2012). This involved choosing the optimal z-section and acquiring images as a mean of 4 scans in 16-bits using a 543 nm HeliumeNeon laser and a 560 nm long-pass filter for a-Bgtx-AF555 and a 488 Argon laser with a 505e530 nm band-pass filter for eGFP.
In experiments using permeabilisation, cells were fixed in 4% PFA for 10 min at RT followed by washes (x3) in phosphate buffered saline (PBS; Sigma) and 0.1% triton-X100 (Sigma) was added for 10 min at RT in 10% (v/v) FCS. Cells were washed to remove the detergent and 400 nM a-Bgtx-AF555 was added for 10 min at RT to label intracellular receptors.

Image analysis
Confocal images were analysed using ImageJ (version 1.410) as described previously (Hannan et al., 2013). For each cell the surface membrane was identified by drawing a region-of-interest (ROI) in the eGFP channel and this was transferred to the a-Bgtx-AF555 channel and the mean membrane fluorescence values were determined. Mean background fluorescence was determined from a region devoid of cells. This was subtracted from the mean membrane fluorescence providing a mean corrected fluorescence intensity value. These values, for different combinations of receptors and drugs, were graphically plotted using Origin.

Bungarotoxin inhibits GABA A receptors in hippocampal neurons
Heteromeric abg receptors are a predominant GABA A R subtype in the neocortex, including the hippocampus (Olsen and Sieghart, 2008;Pirker et al., 2000;Whiting et al., 1995). A smaller proportion of receptors in these areas are thought to be ab heteromers, but to date, there is little if any direct evidence to support the existence of b3 homomers in neurons (Mortensen and Smart, 2006). Before commencing recombinant receptor studies, we first examined a-Bgtx and two other nAChR antagonists for their ability to inhibit whole-cell GABA-activated currents in primary hippocampal neurons, which express heteromeric GABA A Rs.
Receptors were activated by GABA (EC 50 ¼ 1.23 ± 0.04 mM; n ¼ 10; Fig. 1A) in the presence of 2 mM kynurenic acid to block excitatory postsynaptic currents (EPSCs). The potent a7 nAChR specific antagonist methyllaconitine (MLA) did not affect currents activated by sub-maximal GABA (10 mM) concentrations, either when co-applied with GABA (<1% of control; data not shown) or when co-applied with GABA after a 1 min pre-incubation with 1 nM MLA (0.8 ± 1.7% inhibition; n ¼ 6; P > 0.05; Fig. 1BeC) indicating that MLA is not an antagonist at GABA A Rs.
We next introduced a-Bgtx, a non-selective competitive nAChR antagonist. Pre-applying 5 mM a-Bgtx to cultured hippocampal neurons for 1 min in the presence of 1 nM MLA (to inhibit any crosstalk with endogenous nAChRs) followed by co-application with 10 mM GABA resulted in significant inhibition of GABA currents (by 33 ± 9%; n ¼ 8; P < 0.01; Fig.1FeG). The effect of a-Bgtx was reversible with GABA currents returning to control levels after 60 s recovery ( Fig. 1G; P < 0.05). As a control, these receptors were also modulated by pentobarbital with 1 mM GABA currents being potentiated by 20 mM pentobarbital in the same neurons ( Fig. 1B, D, F).
These data indicated that a-Bgtx can antagonise GABA-activated currents in neurons, but the level of inhibition was surprisingly high given that we would expect only a very small proportion, if for 1 min before co-application with GABA. Bargraph data presented here and in succeeding figures are means ± S.E.M, n ¼ 5e8 cells; *P < 0.05, **P < 0.01, ***P < 0.001; One-way ANOVA.
any, of native GABA A Rs to contain a b3eb3 interface. These results therefore suggested that a-Bgtx may be an antagonist at native heteromeric GABA A Rs, and possibly target a binding site that is discrete from the beb subunit interface.

Bungarotoxin binds to a subset of recombinant GABA A R heteromers
Having established that a-Bgtx is an antagonist at native GABA A Rs, we explored its ability to bind to recombinant GABA A R subtypes considered to be physiologically relevant. To avoid any confounds, we did not include b3 subunits because of the potential to form b3eb3 interfaces which would bind a-Bgtx as previously reported (McCann et al., 2006).
labelling with a-BgTx-AF555 (P > 0.05). The imaging data demonstrated that a-Bgtx binds to several combinations of physiologically important heteromeric GABA A Rs in HEK-293 cells and that binding does not require the presence of one or more b3 subunits.
To discount the possibility that the failure of a-Bgtx to bind to receptors containing a3/5/6 subunits was due to poor receptor expression, we used patch clamp recording to examine their responsiveness to GABA. Expressing a3/5/6 subunits with b2g2 subunits produced receptors that supported robust GABA currents ( Fig 2E) confirming that the lack of staining observed with a-Bgtx for these receptors was not due to a lack of expression.

Bungarotoxin selectively inhibits a2 subunit-containing receptors
To complement our imaging studies we next examined the effect of a-Bgtx on GABA A R function using whole-cell patch electrophysiology in HEK-293 cells expressing axb2g2 receptor subtypes.
We selected the following subunit combinations based on their relative abundance in the hippocampal stratum pyramidale: a1b2g2 and a2b2g2, reflecting their relative importance as non-b3-containing synaptic GABA A Rs; a4b2g2 and a5b2g2, chosen since they may represent forms of extrasynaptic GABA A Rs in the hippocampus that underpin tonic inhibition (Glykys et al., 2008).
For these receptor subtypes there were notable differences in the GABA current profiles as expected from their subunit composition (Mortensen et al., 2011;Picton and Fisher, 2007). Differences in GABA current profiles were also observed depending on whether a-Bgtx was co-applied with GABA or also pre-applied for 1 min (Fig. 3B). The level of block was increased by pre-application of a-Bgtx and the slow sag in the GABA current, evident from just coapplying a-Bgtx, suggested that the toxin binds to GABA A Rs with a slow on-binding rate. Therefore, to achieve a full steady-state block, a-Bgtx was pre-applied.
At lower concentrations of a-BgTx (5 mM) however, inhibition was only observed at a2b2g2 receptors (Fig. 3CeE). These results are consistent with our data from fluorescent a-BgTx labelling with a2b2g2 and correlates the highest levels of staining with a low concentration (400 nM) of a-BgTx-AF555 (Fig. 2) with the highest sensitivity to block by a-Bgtx (Fig. 3DeE).
To examine the nature of a-Bgtx inhibition at GABA A Rs, we applied 5 mM a-BgTx with 1 mM GABA, to study antagonism at saturating GABA concentrations. Saturating GABA currents were reduced by 13.0 ± 5.9% (n ¼ 3) in the presence of a-BgTx indicative of some mixed-non competitive antagonism (Fig. 3F).

b-a subunit interface forms a bungarotoxin-binding site in GABA A Rs
To identify the location of the a-Bgtx binding site on GABA A Rs, we pre-incubated HEK-293 cells expressing a2b2g2 with a range of ligands that act at GABA A Rs and nAChRs, for 5 min at RT followed by co-incubation with a-Bgtx-AF555 for 10 min at RT (Fig. 4A). These ligands were selected to 'protect by binding occupancy' known binding site domains on GABA A Rs and nAChRs.
To ascertain whether our abg receptors were assembled intact ensuring the absence of beb subunit interfaces, we studied the inhibition caused by Zn 2þ at a2b2g2 receptors. Receptors composed of ab subunits will be inhibited significantly more by Zn 2þ when compared to abg receptors (Hosie et al., 2003;Krishek et al., 1998). Consistent with this, we found that 10 mM Zn 2þ significantly inhibited submaximal (5 mM) GABA currents for a2b2 by 90.8 ± 4.44% (n ¼ 3; Fig. 4F; P < 0.01) compared to 37.4 ± 8.9 for a2b2g2 (n ¼ 3) suggesting that most of the receptors used in the imaging studies are likely to contain g2 subunits. Furthermore, given that our results so far suggest that a-Bgtx binds to the bea Fig. 5. a-Bgtx-AF555 binds to b3 but not b1 or b2 subunits. A, Images of HEK-293 cells expressing eGFP with either b1, b2 or b3 subunits, incubated with 400 nM a-Bgtx-AF555 for 10 min at RT, 48 h after transfection, washed to remove the excess a-Bgtx-AF555, fixed and imaged. B, Mean surface membrane fluorescence of cells expressing eGFP and b1e3 subunits. ***P < 0.001, n ¼ 6e9. C, Images of HEK-293 cells expressing eGFP with or without either b1, b2 or b3 subunits, after fixation in 4% PFA, 48 h after transfection, and permeabilised with 0.1% w/v Triton-X100, and incubated in 400 nM a-Bgtx-AF555 for 10 min at RT, washed and imaged. Arrowheads indicate intracellular structures labelled with a-Bgtx-AF555. Scale bars 5 mm (A) and 10 mm (C).
interface, the incorporation of the g2 subunit is unlikely to be the determining factor for a-Bgtx binding to GABA A Rs.
Given that the b3eb3 interface forms an a-Bgtx binding site, and that GABA receptor b subunits are highly homologous, we investigated whether the b1eb1 or b2-b2 interfaces could also form a-Bgtx binding sites. However, for cells expressing b1, b2 or b3 homomers in HEK-293 cells for 48 h and incubated in a-Bgtx-AF555 with or without permeabilisation, only b3 expressing cells showed high levels of cell surface and intracellular staining with a-Bgtx-AF555 (Fig. 5AeC). Such staining was absent for b1 and b2 subunits discounting the possibility that beb subunit interfaces were the sites for a-Bgtx binding in b1 or b2 subunit-containing heteromeric GABA A Rs.

Bungarotoxin inhibits only phasic GABA currents in dentate gyrus granule cells
Having established the selectivity of micromolar a-Bgtx concentrations for inhibiting a2b2g2 receptors, we assessed the sensitivity of native GABA A Rs in adult (P115e125) mouse acute hippocampal slices to nAChR ligands and to a-Bgtx. Voltage-clamp recordings of spontaneous inhibitory postsynaptic currents (IPSCs) were performed in dentate gyrus granule cells (DGGCs), which express a2, b2 and g2 subunits, amongst others.
DGGCs receive inhibitory inputs from local interneurons originating within the dentate gyrus and we initially studied whether any endogenous nAChRs affected GABA release. Although the DGGC holding current was slightly reduced by 1 nM MLA, the frequency of IPSCs remained unaltered (Control: 4.12 ± 0.9 Hz; þMLA: 3.75 ± 1.2 Hz; n ¼ 5; P > 0.05 two-tailed t-test; Fig. 6A, C). The IPSC amplitudes were also unaffected by 1 nM MLA (Control: median IPSC À30.82 pA, n ¼ 5252; þMLA: À30.21 pA, n ¼ 6098; P > 0.05; Fig. 6AeB). Furthermore, in HEK-293 cells, 1 nM MLA did not affect the amplitude of GABA-activated currents of a2b2g2 receptors (Fig. 6DeE). These results indicated that the release of GABA onto DGGCs from interneurons is not subject to basal control by a7 subunit-containing nAChRs and that MLA does not affect the amplitude of IPSCs. Nevertheless, as a precaution, we included 1 nM MLA in all recording solutions to obviate any a7 nAChR-mediated effects that may have confounded the interpretation of our results.
The distribution of peak IPSC amplitudes in control and in the presence of a-Bgtx was best described by the sum of four Gaussian components. The mean values for these components in control ( Fig. 7D) were reduced by a-Bgtx (Fig. 7E). The leftward shifts of the first (from À17.12 pA to À13.66 pA) and second peaks (from À27.69 pA to À19.01 pA) were significant (<0.001 and P < 0.01, respectively). Although there was a tendency for the two higher means to also be reduced in a-Bgtx, no statistical significance was observed. This may possibly be because they represent currents mediated by receptors predominantly composed non a2containing receptors that are less sensitive to a-Bgtx. As a-Bgtx could inhibit postsynaptic GABA A Rs, it was plausible that extrasynaptic GABA A Rs, which contain bea interfaces, might also be blocked by a-Bgtx thus affecting tonic inhibition. However, there was no change in RMS noise in the presence (3.39 ± 0.59 pA, n ¼ 6) or absence of 5 mM a-Bgtx (3.55 ± 0.59 pA, n ¼ 6; P > 0.05, two-tailed unpaired t-test; Fig. 8AeB). Application of bicuculline (50 mM) reduced RMS noise significantly (2.74 ± 0.34 pA, n ¼ 6; P < 0.05, Fig. 8AeB) compared to a-Bgtx, indicating the size of the tonic GABAergic current. Similarly, when DGGC holding currents were compared, no change was observed following the application of a-Bgtx (1.27 ± 1.21 pA, n ¼ 7; Fig. 8A, C) although the cells had an average tonic current of 9.31 ± 2.52 pA (n ¼ 7) revealed by applying 50 mM bicuculline. This tonic current was significantly higher than change in holding current observed after the application of a-Bgtx (P < 0.05, Fig. 8C).
To probe the extrasynaptic receptors further, we studied the interaction of recombinant a4b2d receptors by a-Bgtx in HEK-293 cells, chosen because DGGCs predominantly express a4 and d subunits, which, most likely, mediate the majority of the tonic inhibition in the DG (Pirker et al., 2000;Stell et al., 2003;Sun et al., 2004).
To ensure the HEK-293 cells expressed d subunit-containing receptors, the superagonist ability of THIP (300 mM) at abd receptors was confirmed by comparison with responses induced by maximal GABA concentrations (1 mM; Fig. 9A) (Brown et al., 2002;Mortensen et al., 2010). Surprisingly, a-Bgtx potentiated submaximal GABA responses (1 mM; Fig. 9B) at a4b2d receptors at 5 mM (46 ± 5%, n ¼ 3; p < 0.001; Fig. 9CeD) and 20 mM (40 ± 6%, n ¼ 3; p < 0.001, One-way ANOVA; Fig. 9C, E). This potentiation was observed when a-Bgtx was pre-applied to, but not when co-applied with, GABA (Fig. 9F). Interestingly, the leak current was clearly reduced by 5 mM (56 ± 3.7 pA, n ¼ 3) and 20 mM a-Bgtx (75 ± 13 pA, n ¼ 3; Fig. 9GeH) when a-Bgtx was pre-applied which may reflect a degree of spontaneous activity for a4b2d receptors in the absence of GABA similar to that described for a4b3d receptors (Tang et al., 2010). Preapplication of a-Bgtx may therefore shut spontaneously open a4b2d receptors. Interestingly though, GABA-activated currents were now potentiated by a-Bgtx, when we would have expected a block similar to that observed with abg receptors. Thus a-Bgtx inhibits spontaneously opening a4 and d subunit-containing receptors but with GABA binding, a-Bgtx is transformed into a potentiator. The mechanism for this is unclear, but may arise by the blocked spontaneously-active channels exhibiting a higher affinity for GABA resulting in a potentiated agonist-activated response.
Alternatively, once spontaneity is reduced and GABA is bound, a-Bgtx could bind to another site on the abd from which it acts as a positive allosteric modulator. The corollary of this is that a-Bgtx block of spontaneous channels is not equivalent to a-Bgtx block of GABA-activated abg receptors.
Given the reduction in the leak current in HEK-293 cells, why do we not see any effect of a-Bgtx on tonic inhibition in neurons? The reduction in the leak current by 5 mm a-Bgtx for HEK-293 cells expressing a4b2d is over 40-fold greater compared to any reduction by a-Bgtx of the leak in DGGCs. The reasons for this difference may reflect the extent of over-expression of recombinant receptors and differential post-translational processing in cell lines (Tang et al., 2010) that may cause significant levels of spontaneous activity that is not replicated under more physiological conditions. It could also reflect a role for the d subunit in limiting the binding of a-Bgtx to abd receptors in a neuronal setting. If tonic inhibition is mostly due to basal GABA-activated current rather from spontaneously opening receptors in the DGGC, on the basis of the recombinant receptor data, we would expect a-Bgtx to slightly potentiate rather than inhibit tonic current. Overall, these results suggest that a-Bgtx does not inhibit extrasynaptic GABA A receptors, but will inhibit synaptic axb2g2 containing receptors that are most likely populated by the a2b2g2 isoform.

Discussion
a-Bgtx is a widely recognised tool for studying the trafficking, expression and inhibition of nAChRs in the nervous system (Changeux et al., 1970;Harel et al., 2001). Its use for imaging has been extended to other receptors, e.g., GABA A Rs, by enabling a-Bgtx binding. However, there is little detailed information on whether a-Bgtx could inhibit GABA A Rs, though we do know it can affect the function of b3 subunit homomers (McCann et al., 2006). Here, we provide the first report that a-Bgtx can bind to and inhibit recombinant and native neuronal GABA A Rs in a mixed inhibitory manner.
Our findings serve as a cautionary note. Firstly, using a-Bgtx to study nAChR function in the nervous system will not be specific unless attention is paid to the concentrations used. Secondly, the use of a-Bgtx as a tool to study receptor trafficking gained popularity because a 13-amino acid mimotope (WRYYESSLEPYPD;Harel et al., 2001) forms a high affinity a-Bgtx binding site (BBS) and this can be introduced into receptors relatively easily, enabling a-Bgtx binding. The BBS has been engineered into: GluA2 AMPA receptor (Sekine-Aizawa and Huganir, 2004), GABA A R a2 (Brady et al., 2014), b3 (Bogdanov et al., 2006;Saliba et al., 2007), g2 (Joshi et al., 2013), and d (Joshi et al., 2013) subunits: GABA B R1a (Hannan et al., 2011), R1b (Hannan et al., 2012 and R2 subunits (Hannan et al., 2011); mGluR2 (Hannan et al., 2012); Kv4.2 channels (Moise et al., 2010), and Ca 2þ channel a 2 d-2 subunits (Tran- Van-Minh and Dolphin, 2010). This approach has greatly improved our understanding of receptor trafficking and expression. However, for imaging studies with fluorophore-linked a-Bgtx, its ability to bind to native GABA A Rs could complicate the interpretation of results. Fig. 8. a-Bgtx and tonic inhibition. A, Representative IPSCs recorded from dentate gyrus granule cells from acute adult hippocampal slices (P115e125) in control aCSF followed by 5 mM a-Bgtx and 1 nM MLA (þa-Bgtx). Bicuculline (50 mM) was added after a-Bgtx to block GABA A Rs and assess the extent of the GABA tonic current. B, Root mean square (RMS) noise from DGGCs in control, and after a-Bgtx and then in bicuculline. RMS noise was only significantly reduced in bicuculline (n ¼ 6, P < 0.05, unpaired two-tailed t-test). C, Net changes in tonic current after application of a-Bgtx or bicuculline.
Several studies have employed strategies to avoid problems associated with a-Bgtx binding to principally nAChRs. For example, AMPAR (Sekine-Aizawa and Huganir, 2004), GABA A R (Brady et al., 2014) and GABA B R (Hannan et al., 2011(Hannan et al., , 2012 trafficking studies have used d-Tc to prevent binding of a-Bgtx to nAChRs. From our and other studies, it is clear that high doses of d-Tc can prevent a-Bgtx binding to GABA A Rs, but also cause direct inhibition of GABAactivated currents. Another strategy to avoid complications would be to limit the over-expression of GABA A Rs by using a controlled transfection of only specific subunits (Joshi et al., 2013). In this regard, over-expression of b3 subunits could result in the formation of homomers with b3eb3 interfaces, which bind to a-Bgtx without the need for an engineered BBS.
In addition, low concentrations of fluorophore-conjugated a-Bgtx can also be used to avoid labelling of GABA A Rs as the affinity of a-Bgtx for endogenous GABA A Rs is lower compared to that for the BBS. Under our experimental conditions, we do not observe any binding of fluorescent a-Bgtx (400 nM) during live cell confocal microscopy to E18 rat hippocampal cultured neurons.
These cultures express very low levels of a7 nAChRs and robust staining with a-Bgtx can only be observed at this concentration (400 nM) when the neurons are transfected with membrane expressing BBS-tagged receptors. Similarly, the over-expression of recombinant receptors in heterologous expression systems (HEK-293 cells) allows the detection of a-Bgtx fluorescence at low concentrations. Fig. 9. a-Bgtx does not inhibit GABA-activated a4b2d currents. A, Representative whole-cell GABA-activated currents from HEK-293 cells expressing a4b2d receptors in response to sub-maximal (1 mM) and maximal GABA concentrations (1 mM) in comparison to 300 mM THIP. B, GABA concentration response curve for a4b2d receptors expressed in HEK-293 cells. Experiments were performed 48 h after transfection. pEC 50 : 6.04 ± 0.12 (n ¼ 5) and EC 50 ¼ 0.91 mM. C, Representative whole-cell GABA-activated currents in response to submaximal (1 mM) concentrations of GABA in the absence (left-panels) and presence of 5 and 20 mM a-Bgtx (pre-applied for 30s) in HEK-293 cells. D, Potentiation of 1 mM GABAactivated current by 5 mM a-Bgtx compared to control (Con). E, Potentiation of 1 mM GABA-activated current by 20 mM a-Bgtx compared to control (Con). F, Representative wholecell GABA-activated currents in response to a maximal GABA concentration in the absence (left panel) and presence (right panel) of 5 mM a-BgTx with (bottom panel) or without (top panel) a 30s pre-application of a-BgTx. G, Representative GABA-activated maximal currents before, during, and after application of 5 mM a-Bgtx. Note the change in leak current during a-Bgtx). H, Changes to leak current in control (Con), þ5 mM and þ20 mM a-Bgtx *P < 0.05, **P < 0.01, ***P < 0.001, n ¼ 3e7, unpaired two-tailed t-test and One-way ANOVA.
Although pentameric ligand-gated ion channel family members share many common structural features, they exhibit relatively distinct pharmacological profiles. However, there are circumstances where some ligands can affect more than receptor type, notably GABA A R b3 homomers binding of a-Bgtx (McCann et al., 2006). In the present study, we extend this observation to a-Bgtx inhibiting physiologically-important GABA A R heteromers. In addition to a7 homomers, a-Bgtx also inhibits abgd or abdε heteromeric nAChRs. This suggests that the a-Bgtx binding site on heteromeric GABA A Rs may share a similar architecture to the binding site found on heteromeric nAChRs whereas the b3 a-Bgtx binding could be similar to homomeric nAChRs. The b3-homomeric receptors are a pharmacologically distinct population of GABA A Rs that do not respond to GABA, but are sensitive to potentiation by bicuculline (Wooltorton et al., 1997). Therefore, the a-Bgtx binding site on these homomers is likely to be different from binding site on abg heteromers. For muscle-type nAChRs, a-Bgtx binding occurs at the interface between a-g and a-d subunits. From our study, the two bea interfaces of abg GABA A Rs (which are comparable to the a-g, a-d nAChR interfaces (Smart and Paoletti, 2012)), could therefore contain a similar a-Bgtx binding site. This deduction is based upon the abolition of fluorophore-linked a-Bgtx binding to GABA A Rs by competitive GABA receptor antagonists. This is in accord with a-Bgtx binding at the GABA-binding bea interface, enabling a-Bgtx to be used as a tool to study GABA A R function. However, for a4b2d receptors, GABA responses were not inhibited by a-Bgtx. This was surprising since these receptors will retain the bea interfaces, so we must presume the fifth subunit (d) does have some influence on whether a-Bgtx will bind and cause inhibition.
Profiling a-Bgtx binding (by fluorescence) and inhibition (by electrophysiology) at low concentrations demonstrated that the highest levels of both were achieved with a2b2g2 receptors. This preference for a2b2g2 is likely to reflect different binding affinities of a-Bgtx for various b2-ax interfaces, presumably because of divergent amino acid sequences between the different a subunits on the complementary (À) (as opposed to the principal side (þ)) side of the bea subunit interface (Smart and Paoletti, 2012). From the fluorescent a-Bgtx binding profiles of axb1g2 receptors, we would expect the a-Bgtx inhibitory profiles to be similar, with a2b1g2 being inhibited most and a5b1g2 least, by a-Bgtx. The receptor subtype selectivity of a-Bgtx is useful for studying native GABA A R function in acute hippocampal slices, without any dependence upon nAChRs. Application of MLA revealed that presynaptic a7-containing nAChRs are not regulating the basal release of GABA onto adult DGGCs. However, previous reports suggest that nAChR-activation can cause increased GABA release (Radcliffe et al., 1999), though, MLA had no effect on basal GABA release in mouse CA1 pyramidal neurons (Proctor et al., 2011) and layer V prefrontal cortex (Aracri et al., 2010). In addition, a-Bgtx neither affected the average tonic current nor the RMS noise suggesting that other subtypes of nAChRs do not control basal GABA release in the dentate. This further suggested that a-Bgtx, at concentrations specific for inhibiting a2 subunit-containing receptors, has little if any effect on a4 or a5 subunit-containing extrasynaptic receptors.
The inhibitory effect of a-Bgtx on IPSC amplitudes in adult slices was rapid in onset. The reduced IPSC amplitudes were more pronounced for smaller compared to larger events, and given the selectivity of a-Bgtx, implies that a2b1/2g2 receptors contribute largely to these sub-populations of events. Of course, a small proportion of IPSCs may be inhibited by the binding of a-Bgtx to b3 subunit-containing receptors. Nevertheless, the binding of a-Bgtx to heteromeric GABA A Rs together with the inhibition of GABAactivated currents in HEK-293 cells, primary hippocampal neurons and in slices, clearly indicates that a-Bgtx can act as an antagonist at native GABA A Rs to block synaptic inhibition, presumably by binding at the GABA A receptor bea interface.
Author contribution SH carried out the imaging and electrophysiology of slices. MM and SH carried out the electrophysiology of recombinant receptors and cultured neurons. SH and TGS designed the project and wrote the paper. All authors contributed to the writing of the paper.