Neurotoxins from Snake Venoms and α-Conotoxin ImI Inhibit Functionally Active Ionotropic γ-Aminobutyric Acid (GABA) Receptors*

Background: Different snake venom three-finger toxins interact with various receptors, channels, and membranes. Results: Here, we demonstrate that GABAA receptors are inhibited by α-cobratoxin, other long chain α-neurotoxins, nonconventional toxin from Naja kaouthia, and α-conotoxin ImI. Conclusion: Some toxin blockers of nicotinic acetylcholine receptors also inhibit GABAA receptors. Significance: Three-finger toxins offer new scaffolds for the design of GABAA receptor effectors. Ionotropic receptors of γ-aminobutyric acid (GABAAR) regulate neuronal inhibition and are targeted by benzodiazepines and general anesthetics. We show that a fluorescent derivative of α-cobratoxin (α-Ctx), belonging to the family of three-finger toxins from snake venoms, specifically stained the α1β3γ2 receptor; and at 10 μm α-Ctx completely blocked GABA-induced currents in this receptor expressed in Xenopus oocytes (IC50 = 236 nm) and less potently inhibited α1β2γ2 ≈ α2β2γ2 > α5β2γ2 > α2β3γ2 and α1β3δ GABAARs. The α1β3γ2 receptor was also inhibited by some other three-finger toxins, long α-neurotoxin Ls III and nonconventional toxin WTX. α-Conotoxin ImI displayed inhibitory activity as well. Electrophysiology experiments showed mixed competitive and noncompetitive α-Ctx action. Fluorescent α-Ctx, however, could be displaced by muscimol indicating that most of the α-Ctx-binding sites overlap with the orthosteric sites at the β/α subunit interface. Modeling and molecular dynamic studies indicated that α-Ctx or α-bungarotoxin seem to interact with GABAAR in a way similar to their interaction with the acetylcholine-binding protein or the ligand-binding domain of nicotinic receptors. This was supported by mutagenesis studies and experiments with α-conotoxin ImI and a chimeric Naja oxiana α-neurotoxin indicating that the major role in α-Ctx binding to GABAAR is played by the tip of its central loop II accommodating under loop C of the receptors.

Ionotropic receptors of ␥-aminobutyric acid (GABA A R) regulate neuronal inhibition and are targeted by benzodiazepines and general anesthetics. We show that a fluorescent derivative of ␣-cobratoxin (␣-Ctx), belonging to the family of three-finger toxins from snake venoms, specifically stained the ␣1␤3␥2 receptor; and at 10 M ␣-Ctx completely blocked GABA-induced currents in this receptor expressed in Xenopus oocytes (IC 50 ‫؍‬ 236 nM) and less potently inhibited ␣1␤2␥2 ≈ ␣2␤2␥2 > ␣5␤2␥2 > ␣2␤3␥2 and ␣1␤3␦ GABA A Rs. The ␣1␤3␥2 receptor was also inhibited by some other three-finger toxins, long ␣-neurotoxin Ls III and nonconventional toxin WTX. ␣-Conotoxin ImI displayed inhibitory activity as well. Electrophysiology experiments showed mixed competitive and noncompetitive ␣-Ctx action. Fluorescent ␣-Ctx, however, could be displaced by muscimol indicating that most of the ␣-Ctx-binding sites overlap with the orthosteric sites at the ␤/␣ subunit interface. Modeling and molecular dynamic studies indicated that ␣-Ctx or ␣-bungarotoxin seem to interact with GABA A R in a way similar to their interaction with the acetylcholine-binding protein or the ligand-binding domain of nicotinic receptors. This was supported by mutagenesis studies and experiments with ␣-conotoxin ImI and a chimeric Naja oxiana ␣-neurotoxin indicating that the major role in ␣-Ctx binding to GABA A R is played by the tip of its central loop II accommodating under loop C of the receptors.
Type A ␥-aminobutyric acid receptor (GABA A R) 4 is abundantly expressed across the nervous system. Most of the GABA A Rs in the brain are constructed from ␣1-6 and ␤1-3 subunits co-assembled with ␥1-3 or ␦/⑀// subunits. Interestingly, some ␤ subunits, usually found in heteropentameric GABA A Rs could also form constitutively active homopentameric ligand-gated ion channels. Channels with such properties were found in cultured neurons (1), and in fact the ␤3-homopentamer at the moment is the only GABA A R with known three-dimensional structure (2).
GABA A Rs are targeted by diverse toxins of low molecular weight, such as the channel blocker picrotoxin (3), a polyacetylenic compound oenanthotoxin (4), the GABA site antagonist bicuculline (5), and by GABA site agonists such as ibotenic acid and muscimol. Benzodiazepines are positive allosteric modulators of GABA A R and were the world's most prescribed drugs in the 1980s (6). It is almost impossible to introduce fluorescent labels into such molecules, and therefore, they cannot be used in live cell imaging studies. In contrast, peptide toxins can easily be tagged with fluorescent or radioactive labels. Many of them, for example ␣-conotoxins from Conus snails and three-finger toxins (TFTs) from snake venoms, target nicotinic acetylcholine receptors (nAChRs) and show high levels of selectivity between the different nAChR types (7). For example, ␣-bungarotoxin (␣-Bgt) from multibanded krait venom binds with nanomolar affinity to the muscle-type and ␣7 receptors but not to the heteropentameric neuronal nAChRs (8,9). Such peptide toxins also could be quite helpful in structural studies. The x-ray crystal structure of the extracellular domain of ␣1 (10) and ␣9 (11) nAChR subunits were obtained in a complex with ␣-Bgt.
TFTs form an abundant group of proteins found in Colubridae, Elapidae, and Psammophiinae snake venoms. Members of this group possess no enzymatic activity and bind to receptors from different groups as follow: nAChRs (short and long chain ␣-neurotoxins), muscarinic acetylcholine receptors (African mamba-derived muscarinic toxins), L-type calcium channels (calciseptine from black mamba venom); and ASIC1, acid-sensitive ion channel (mambalgines). There are also three-finger cytotoxins (cardiotoxins) that interact with the cell plasma membrane (12).
Typical members of the long-chain group of TFTs, such as ␣-Bgt from Bungarus multicinctus as well as ␣-cobratoxin (␣-Ctx) from Naja kaouthia cobra, neurotoxin NT I from Naja oxiana, and LsIII from marine krait Laticauda semifasciata, have an additional disulfide bond in loop II (Fig. 1, bottom left). This disulfide has been identified as a crucial feature for binding at the neuronal nAChR intersubunit sites (13,14). Short neurotoxins, such as Naja oxiana neurotoxin II (NT II), acting exclusively on muscle nAChRs, do not have this additional disulfide (Fig. 1, bottom center). Another group, closely related to ␣-neurotoxins, is the group of so-called nonconventional or weak neurotoxins. In contrast to the previous two groups, weak toxins have an additional disulfide bond in the first loop (Fig. 1, bottom right), and show both weaker affinity toward nAChRs and some cross-reactions with other receptors (15). The toxins listed above block nAChRs with affinities that range from nanomoles to tens of micromoles/liter, and ␣-Bgt is the most active (Table 1). Interestingly, evidence for ␣-Ctx inhibition of T-type calcium channels through muscarinic receptors was recently provided (16), which may indicate that ␣-Ctx also has biological targets other than nAChRs.
Fluorescent ␣-Bgt is widely used in applications concerning nAChR location and expression (17)(18)(19). Moreover, the "bungarotoxin-binding site" comprising a fragment of the ␣1 nAChR could be incorporated into the extracellular loops of different receptors allowing imaging of the receptors on the surface of living cells with the fluorescently labeled ␣-Bgt, which in particular also has been done for the GABA A R (20,21). Using fluorophore-conjugated ␣-Bgt, about a decade ago the first evidence of ␣-Bgt binding to GABA A R was obtained (22). However, it was suggested to be limited to the peculiar ␤3/␤3binding sites, whose existence in vivo in the wild-type organisms has not been unambiguously demonstrated. This work aroused our interest in searching for novel GABA A R ligands among snake venom components structurally similar to ␣-Bgt, FIGURE 1. Structural diversity of three-finger neurotoxins. Top set represents sequence alignments. All TFTs have three loops stabilized by four disulfides, but some neurotoxins have one additional disulfide bond. Long (names higlighted green) and weak or nonconventional (names highlighted pink) neurotoxins have additional disulfide bonds in the second and the first loops, respectively, although short neurotoxins (name highlighted yellow) do not have additional disulfides. Bottom set represents typical spatial structures of three toxin groups as follows: long neurotoxin ␣-Ctx (PDB code 1CTX), short neurotoxin NT II (PDB code 1NOR), and nonconventional or weak toxin, here represented by the published structure of candoxin (PDB code 1JGK) that shares homology with WTX (N. kaouthia nonconventional toxin) and OWT (N. oxiana nonconventional toxin). Sequence alignments were produced via Clustal W algorithm; signal peptide sequences were removed manually, and top line shows numbering of ␣-Ctx residues. Backbones are shown in gray; ␤-strands are shown as arrows; cysteine residues, forming disulfide bonds are shown as yellow sticks on structures and highlighted orange in sequences, loop II arginine and valine residues in sequences are shown in blue and green, respectively.

TABLE 1
Affinities of different three-finger toxins toward nAChRs and acetylcholine-binding proteins ␣7-nAChR is the human neuronal homopentameric nAChR, Tca-nAChR is the Torpedo californica muscle-type nAChR; Ac-and Ls-AChBP are L. stagnalis and Aplysia californica acetylcholine-binding proteins, respectively. ND means not determined. namely among various TFTs, as well as among ␣-conotoxins, another class of nAChR inhibitors.

Neurotoxin
Here, we report specific labeling of the ␣1␤3␥2 subtype of GABA A R with a fluorescent derivative of ␣-Ctx, as well as inhibition of GABA-induced currents in a series of GABA A Rs by well known TFTs belonging to different structural groups and exhibiting anticholinergic properties; interaction of ␣-conotoxin ImI with ␣1␤3␥2 GABA A R was found as well.

Experimental Procedures
Three-finger Toxin Preparations-Ls III was from Latoxan (Valence, France). WTX, ␣-Ctx, and NT II were obtained as described previously (23)(24)(25) and OWT as described previously (26). NT I was obtained as described previously (27) with additional purification by reversed phase HPLC. ␣-Bgt was purified from B. multicinctus venom by combination of gel filtration, ion-exchange, and reverse-phase HPLC, and ␣-Bgt Val-31 analogue was used in this work. The structures of toxins obtained were confirmed by mass spectrometry. Chimeric NT II/I toxin with the tip of the second loop from NT I grafted to the NT II structure was obtained as described previously (14).
Peptide Chemical Synthesis-Peptide WCDAFCSIRGKR from the ␣-Ctx loop II and ␣-conotoxin ImI was synthesized as described previously (28).
Electrophysiology-Oocytes were obtained from healthy mature (Ͼ10 cm) Xenopus frogs. After manual separation, oocytes were injected with 2-3 ng of laboratory PCI vector containing DNA sequences encoding mouse GABA A R subunits. Oocytes were incubated at 18°C in ND96 solution (5 mM HEPES/NaOH, pH 7.6, 96 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 2 mM MgCl 2 ) supplemented with ampicillin and kanamycin. Recordings were performed 36 -72 h after injections using turbo TEC-03X setup (npi electronic GmbH, Germany). During a typical experiment, agonist (␥-aminobutyric acid) was applied every 5 min. To evaluate the activity of the tested TFTs, they were pre-applied before the agonist and then co-applied with it. Incorporation of the ␥2 subunit was confirmed by application of diazepam. Incorporation of ␦ subunit was confirmed using DS2 compound as described previously (29). Data are presented as mean Ϯ S.E. and/or 95% confidence interval in parentheses.
Fluorescent ␣-Ctx-Coupling of Alexa Fluor 546 NHS reagent (Invitrogen) to ␣-Ctx was conducted according to the manufacturer's protocol. This fluorophore has been chosen due to similar spectral properties to commercially available Alexa Fluor 555 ␣-Bgt (Invitrogen) also used in this work.
Live Cell Imaging with Fluorescent Ligand-Mouse neuroblastoma Neuro2a cells grown in 96-well plates were transfected with laboratory PCI plasmids encoding GABA A R subunits and GFP-encoding plasmid (Evrogen, Russia) following the Lipofectamine transfection protocol (Invitrogen). Transfected cells were grown in DMEM (Paneco, Russia) supplemented with 10% fetal bovine serum (PAA Laboratories, Austria) at 37°C in a CO 2 incubator for 48 h. The growth medium was removed; cells were rinsed with a buffer (20 mM HEPES, 140 mM NaCl, 2.8 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose, pH 7.4) and incubated with 50 nM Alexa Fluor 555labeled ␣-Bgt (Invitrogen) or Alexa Fluor 546-labeled ␣-Ctx for 20 min at room temperature. In control experiments, 50-fold molar excess of nonlabeled ␣-Ctx was added to the incubation medium. The cells were then washed twice with the buffer and were visualized by an epifluorescent microscope (Olympus, Japan) using the appropriate filter combination. In competition experiments, 1-4000 M muscimol or 400 M diazepam were added to the incubation medium (140 mM NaCl, 2 mM CaCl 2 , 2.8 mM KCl, 4 mM MgCl 2 , 20 mM HEPES, 10 mM glucose, 1% BSA, pH 7.4) as described previously (19). Data were analyzed using the ImageJ program with the standard set of plugins (30). Fluorescence was quantified as follows: field of view was photographed in two channels as follows: one to detect fluorescent ligand and another to detect GFP. Background of images was subtracted, and maximum points were found on the GFP image and selected. Selection was enlarged to fit cells on the image and transferred to an image containing fluorescent ligand detection (same field of view); then the mean fluorescence was measured and normalized to the difference between total fluorescent binding and unstained cells fluorescence. In each well, 2-4 random fields of view were photographed, and each individual experiment included several wells; at least three independent experiments were performed.
Molecular Modeling-The extracellular domains of ␣ and ␤ subunits forming the ␤ϩ␣Ϫ interface of GABA A R were modeled using the atomic coordinates of the ␤3 extracellular domain taken from the recently published ␤3-homopentameric x-ray structure (2) and the extracellular domain of the ␣1 subunit, which was constructed via the Swiss model homology modeling service using the crystal structure of the ␤3 subunit from homo-oligomeric receptor as a template (31). GABA A R subunits share some homology with the Lymnaea stagnalis acetylcholine-binding protein (AChBP). Assuming that ␣-Ctx binds to GABA A Rs in a way similar to the crystal structure of the ␣-Ctx-bound AChBP, we constructed models of ␣-Ctx complexes with GABA A R extracellular domains by aligning the ␤3 subunit extracellular domain with the principal AChBP subunit and ␣1 subunit extracellular domain with the complementary AChBP subunit from its x-ray structure with ␣-Ctx (PDB 1YI5) using UCSF Chimera (32). Resulting structures were energy-minimized with GROMACS 5.0 (33) and then treated with molecular dynamics procedure using OpenMM libraries (34) with the aid of AMBER force field parameters (35) and GBSA implicit solvent model to simulate 100 ns of complex behavior. Resulting coordinates were analyzed with VMD (36) and UCSF Chimera software.
Single-point Mutagenesis-The S67K and K67S mutations (in the ␣1 and ␣2 subunit, respectively) were obtained by PCR amplification of ␣1 and ␣2 subunits in the vector with the following primers: forward primer 5-gtgtttttccgccaaaagtggaaggatgaaagattaaaattc-3 and reverse primer 5-catccttccagctttggcggaaaaacacatctattg-3 (for S67K); forward primer 5-gctcgtctttccagctttgtcgaaagaaaacatctattg-3 and reverse primer 5-ctttcgacaaagctggaaagacgagcggttaaaatttaaag-3 for (K67S) by using Phusion High-Fidelity DNA polymerase. Amplification product was analyzed after 25 cycles of the following thermal steps: 2 min of initial denaturation at 98°C; 20 s of denaturation at 98°C; 1 min of annealing at 65-70°C; 3 min of extension at 72°C, and 10 min of final extension at 72°C. The product obtained was incubated with DpnI enzyme to restrict the methylated DNA and then transformed into XL1-Blue supercompetent cells. The mutations were confirmed by DNA sequencing.

Results
Long-chain ␣-Ctx and Naja oxiana ␣-Neurotoxin NT I Compete with Fluorescent ␣-Bgt for GABA A R-binding Site-First of all, we decided to check whether ␣-Ctx could displace a fluorescent ␣-Bgt derivative from ␣1␤3 GABA A R. We reproduced ( Fig. 2A) fluorescent Alexa Fluor 555 ␣-Bgt binding to (presumably) the ␤3/␤3-interface of ␣1␤3 receptors first reported by McCann et al. (22). ␣-Ctx at 5 M concentration competed with fluorescent ␣-Bgt and displaced it almost completely from the receptor. Interestingly, Naja oxiana NT I (see Fig. 1 for sequence alignments) was also an effective competitor; it diminished Alexa Fluor 555 ␣-Bgt binding by 91 Ϯ 2% (experiments not shown).
Snake Venom Neurotoxins from Different Structural Groups Inhibit GABA A R-To test whether ␣-Ctx and other toxins could inhibit ␣1␤3␥2 GABA A R, we performed two-electrode voltage clamp experiments. It was found that ␣-Ctx, as well as Ls III (long-chain ␣-neurotoxin from Laticauda semifasciata), and nonconventional toxin WTX from Naja kaouthia venom inhibit the GABA A Rs in a dose-dependent manner (Fig. 3). Given at 10 M Ls III and WTX blocked GABA-evoked response by 83 Ϯ 12 and 31 Ϯ 8%, respectively. Long-chain NT I and ␣-Bgt at 10 M inhibited ␣1␤3␥2 receptors only by 17 Ϯ 5 Note that Alexa Fluor 546 ␣-Ctx binding to ␣1␤3-transfected cells was significantly (p Ͻ 0.05) better blocked by ␣-Ctx, whereas Alexa Fluor 555 ␣-Bgt binding was blocked with equal efficiency by both toxins. However, Alexa Fluor 555 ␣-Bgt (G) and Alexa Fluor 546 ␣-Ctx (H) binding to cells transfected with ␣1␤3␥2 were blocked more effectively by ␣-Ctx (p Ͻ 0.05). Concentrations of fluorescent derivatives were 50 nM; the mixtures were incubated for 15 min, and then unbound ligand was washed out with buffer. Competition experiments using fluorescent derivatives and native ␣-Ctx and ␣-Bgt were performed by incubation of cells with unlabeled toxins at 5 M along with addition of labeled analogues. The data were analyzed with ImageJ. Note that data are normalized to differences between mean stained and unstained cell fluorescence. Three independent experiments were performed for each data point. Data are presented as mean Ϯ 95% confidence interval; differences were interpreted as significant (p Ͻ 0.05, asterisk) if intervals are not intersecting (if intervals intersect, difference was interpreted as nonsignificant, n.s.). and 19 Ϯ 3%, respectively. No inhibition was observed for short-chain neurotoxin NT II and nonconventional toxin OWT, both from N. oxiana venom, at concentrations of 10 M. Interesting, ␣-Ctx (5 M) also inhibited ␣1␤3␦ receptor (Fig. 3).
To characterize more broadly the ␣-Ctx potency and selectivity, we tested it on GABA A R with different subunit combinations. First of all, we investigated the potency of ␣-Ctx for inhibition of GABA-induced currents at ␣1␤3 and ␣1␤3␥2 Electrophysiological recordings demonstrate that three-finger toxins can inhibit GABA A R. GABA was applied alone or co-applied with the indicated toxin to oocytes injected with cDNA encoding ␣1, ␤3, and ␥2 subunits. First peak of each set represents control (contr.) with a 1-s application of 10 M GABA alone, corresponding to EC 20 -50 ; the concentration of co-applied toxin is indicated above the recording. The traces marked as wash represent the response to GABA after 5 min of oocyte washing with the buffer solution. The lower right inset shows inhibition of ␣1␤3␦ receptor by ␣-Ctx at 5 M. The presence of ␥ subunit is confirmed by the effect of the benzodiazepine 1 M diazepam (BZD) and differential sensitivity to 50 M Zn 2ϩ (see insets). Absence of a ␥ subunit influence on ␣-Ctx inhibition is consistent with the assumption that the ␣-Ctx-binding site is located at an interface between ␣ and ␤ subunits. B, this toxin manifests modest subtype selectivity; it inhibits ␣1␤3␥2 receptor five times more potently than ␣2␤3␥2. The potency of ␣-Ctx against diverse sets of GABA A R subunits was examined using the two-electrode voltage clamp method. To measure the effect of different ␣-Ctx concentration, ␣-Ctx was applied alone to the oocyte for 5 min before control GABA application and then was co-applied with GABA (10 M, EC 20 -50 ). Current amplitudes obtained in the presence of ␣-Ctx were expressed as percentage of control amplitudes and plotted against logarithm of toxin molar concentration. C, GABA dose-response curve at ␣1␤3␥2 receptors without ␣-Ctx (upper curve) and in the presence of 250 nM ␣-Ctx (lower curve). GABA EC 50 is shifted to the right and the Hill slope is reduced, suggesting noncompetitive or some mixed competitive/noncompetitive type of inhibition. D, muscimol competes with Alexa Fluor 546 ␣-Ctx on ␣1␤3␥2 GABA A R suggesting ␣-Ctx-binding sites overlap with orthosteric sites located at the ␤/␣ interface. Fluorescence data were normalized to the difference between mean stained cell fluorescence and mean cell fluorescence blocked by 10 M ␣-Ctx (i.e. data represent specific ␣-Ctx binding). Data were fitted to a dose-response equation using the Origin 7.5 software.
receptors. As expected, GABA-induced currents at ␣1␤3 receptors were completely inhibited by 50 M Zn 2ϩ in contrast to those at ␣1␤3␥2 receptors (Fig. 4A, insets). The incorporation of a ␥2 subunit in ␣1␤3␥2 receptors was further demonstrated by their sensitivity to 1 M diazepam. Interestingly, ␣-Ctx was able to completely inhibit diazepam-stimulated GABA currents, again demonstrating that ␣-Ctx is able to also block the actions of ␣1␤3␥2 receptors (Fig. 4A, insets). As shown in Fig. 4A, ␣-Ctx exhibited a comparable potency for inhibition of these two receptor subtypes, indicating that the ␥2 subunit does not influence the potency of ␣-Ctx.
Electrophysiology Measurements Indicate That ␣-Ctx Inhibits GABA A R in a Mixed Competitive/Noncompetitive Manner-To shed light on the ␣-Ctx mode of inhibition, we compared the GABA dose-response dependence (for ␣1␤3␥2 subtype) in the presence of 250 nM ␣-Ctx, a concentration that is sufficient to inhibit receptor response to 10 M GABA by about 50% as compared with control (Fig. 4C). GABA EC 50 value was shifted from 29 Ϯ 7 to 128 Ϯ 25 M in the presence of ␣-Ctx. After looking at the individual data points, however, it seems that ␣-Ctx caused a parallel shift of the GABA doseresponse curve in at least two steps. Such behavior would be consistent with a competitive inhibition of GABA binding by ␣-Ctx, when binding of ␣-Ctx to one of the two orthosteric sites caused an allosteric reduction of the GABA affinity for the second orthosteric site.
At 1 mM GABA, however, the current induced in the presence of toxin is 40% smaller than in its absence (Fig. 4C). Although the data do not allow a firm conclusion because the GABA concentration-response curve in the presence of toxin is not yet saturated at this concentration, they are consistent with an additional noncompetitive inhibition of GABA currents by ␣-Ctx (see below).
Muscimol but Not Diazepam Competes with Alexa Fluor 546 ␣-Ctx-We investigated whether diazepam, a classical benzodiazepine that binds at the ␣/␥ interface (37), or the agonist muscimol that binds at the ␤/␣ interface can compete with Alexa Fluor 546 ␣-Ctx for binding to Neuro2a cells expressing ␣1␤3␥2 GABA A R. Indeed, 200 M muscimol reduced fluorescence by about 80% (Fig. 4D), whereas diazepam had no effect at concentrations up to 400 M (experiments not shown). These data support the conclusion that Alexa Fluor 546 ␣-Ctx to a large extent binds to the GABA-binding site of GABA A receptors. The remaining 20% of bound Alexa Fluor 546 ␣-Ctx cannot be displaced by muscimol, suggesting that this binding of ␣-Ctx occurs via a site different from the orthosteric binding site.
␣-Cobratoxin Loop II Peptide (WCDAFCSIRGKR) and ␣-Conotoxin ImI Selective for ␣7 nAChR but Not the ␣-Conotoxins SIA and MII Selective for Muscle Type nAChR and Heteromeric Neuronal nAChRs, Respectively, Inhibit ␣1␤3␥2 GABA A R-We hypothesized that other polypeptide nAChR ligands could also inhibit GABA A R and chose ␣-conotoxin ImI because it was reported previously that ImI shares some structural features of the ␣-Ctx loop II (28). ImI binds efficiently to ␣7 nAChR, i.e. to neuronal homopentameric nAChR, whereas ␣-Ctx in addition to ␣7 nAChR also binds to muscle type FIGURE 5. A, structural similarity of ␣-conotoxin ImI, selective for ␣7 nAChR, to the tip of the ␣-Ctx loop II (magenta). Note that two positively charged arginines (highlighted red) are exposed in a very similar manner with respect to aromatic residues (tryptophan or phenylalanine, highlighted green) in both ImI (shown wheat) and loop II of ␣-Ctx (shown blue). Two different views (1st and 2nd rows) are provided for convenience. Amino acid sequences of ␣-conotoxins SIA and MII differ greatly from that of ImI (bottom). B, chemically synthesized WCDAFCSIRGKR peptide that mimics tip of ␣-Ctx loop II and ␣-conotoxin ImI at 100 M inhibits 10 M GABA-evoked current (upper left and right traces, respectively). SIA specific for muscle type nAChR and MII specific for heteromeric neuronal nAChR have no effect on the currents at 100 M. nAChR. We decided to test two other conotoxins: SIA that binds to muscle nAChR and MII that binds to neuronal non ␣7 nAChRs. ␣-Conotoxin ImI (see Fig. 5A for structure) inhibited the ␣1␤3␥2 GABA A R at 100 M by 45 Ϯ 5% in terms of GABAevoked peak current (Fig. 5B). ␣-Conotoxins SIA and MII failed to show inhibition of GABA A R at 100 M (Fig. 5B, bottom). This experiment also proves specificity of inhibition, and because 100 M is a rather high concentration, we had to make some control experiments with peptides chemically similar to ImI at such a high concentration.
We also performed experiments with the synthetic peptide WCDAFCSIRGKR comprising the tip of ␣-Ctx loop II. Synthetic fragment (see Fig. 5A for structure) inhibited the ␣1␤3␥2 GABA A R at 100 M by 28 Ϯ 10% in terms of GABA-evoked peak current, respectively (Fig. 5B). The above results suggest that new peptide ligands of GABA A Rs could be found among toxins sharing common structural features with ␣-Ctx loop II.
␣-Cobratoxin at Concentrations of Up to 40 M Does Not Inhibit Homopentameric Glycine Receptor-Glycine receptor is another example of abundantly expressed Cl Ϫ -permeable Cysloop receptors. It also shares very high sequence homology with GABA A R subunits. To check whether ␣-Ctx possesses some activity against this receptor, we tested its activity against the human homopentameric ␣1 glycine receptor. At concentrations up to 40 M, ␣-Ctx failed to inhibit glycine-evoked currents (data not shown). Thus, ␣-Ctx blocks GABA A Rs but not the glycine receptor. However, it should be noticed that there are other subsets of glycine receptor (␣2-3 and ␤) that were not tested.
Model for GABA A R⅐␣-Ctx Complex-GABA A R spatial structure (2) has a very high degree of similarity to structures of AChBPs and extracellular nAChR domains for which the ␣-neurotoxin-binding sites were defined. Based on these data, we assumed that a complex of ␣-Ctx with the GABA A R might resemble published structures of ␣-neurotoxins bound to AChBP or to the extracellular nAChR domain. Under this assumption, we suggested that loop II of ␣-Ctx might find its place under the loop C of the GABA A R receptor subunit. To get a picture of such a hypothetical intersubunit binding site, we constructed homology models of extracellular domains of ␣1 GABA A R subunits using the recently published structure of GABA A R ␤3-homopentamer (2) as a template. Then a molecular model of the ␤3/␣1 dimer with ␣-Ctx introduced under loop C of the ␤3 subunit was constructed by superimposition and alignment of the ␤3 extracellular domain structure with that of the AChBP protomer (PDB code 1YI5). The structure obtained was subjected to 100 ns of molecular dynamics. The last 40-ns complex was stable, and its backbone root mean square deviation did not exceed 0.1 nm. By the end of simulation, the complex retained the overall "TFT to Cys-loop" binding mode known from the x-ray structures (Fig. 6A) with the toxin loop II buried under loop C of the receptor. Interestingly, ␣-Ctx residues Arg-33 and Arg-36 form a stable network of salt bridges with both Asp-27 and ␤3 Glu-155 (Fig. 6B), which was present during all simulation times. The toxin C-terminal region and tips of loop I and loop II also formed contacts with both dimer subunits.
Electrophysiological Analysis of Chimeric NT II/I-To verify the model, we decided to test whether the loop II of the toxins plays a crucial role in functional inhibition of GABA A R. In previous studies, short-chain toxin NT II with grafted loop II from long neurotoxin I was used to study the role of loop II in ␣-neurotoxin interaction with neuronal nAChRs (14). We used the same chimera in electrophysiological experiments on GABA A R. Short-chain toxin NT II at 10 M did not show any inhibition of the receptor but gained the inhibitory activity upon the grafting of loop II from long neurotoxin NT I (Fig. 7A). Chimeric NT II/I toxin and NT I generated a comparable inhibition of ␣1␤3␥2 receptors (16 Ϯ 3 and 19 Ϯ 3%, respectively, Fig. 7B). Molecular dynamics simulations indicated that the model of NT II was unable to form a complex with the GABA A R orthosteric intersubunit site in the manner similar to ␣-Ctx ( Fig. 7C). In contrast, the model of NT II/I formed a stable complex with loop II resting under the C-loop of ␤3 extracellular domain model (Fig. 7D).
Single Point Mutations ␣1(S67K) and ␣2(K67S) Located at the Complementary Surface of the Respective ␣-Subunit Change Inhibition by ␣-Ctx-Because we observed modest ␣-Ctx selectivity toward ␣1␤3␥2 (Table 2), we decided to test our model by introduction of ␣13␣2 and ␣23␣1 point mutations. Loop II of the toxin carries a significant positive charge, and we hypothesized that addition or deletion of a positive charge at position 67 of ␣1 and ␣2 subunits, located in close vicinity to the hypothetical toxin-binding site and occupied by a serine residue (Fig.  8A), might diminish affinity to ␣-Ctx in the ␣1(S67K) mutant and increase affinity toward ␣2(K67S) mutant. Indeed, we found that the ␣-Ctx IC 50 value for ␣1(S67K)␤3␥2 GABA A R (703 Ϯ 39 nM) is closer to the ␣2-containing receptor (1099 Ϯ 57 nM, Table 2) than the ␣1-containing receptor (236 Ϯ 7 nM, Table 2), and vice versa the ␣-Ctx IC 50 value for ␣2(K67S)␤3␥2 GABA A R (442 Ϯ 54 nM) is closer to the ␣1-containing receptor than the ␣2-containing receptor (Fig. 8B). At the same time GABA EC 50 values were not affected by these mutations.

Discussion
There has been considerable progress in understanding the structure and function of GABA A Rs (38). Notwithstanding, there are still significant gaps in our knowledge on peptide and proteins interacting with GABA A Rs. In general, polypeptide ligands manifest higher selectivity to particular receptor subtypes as compared with low molecular weight organics. Several polypeptides interacting with GABA A R have been reported earlier that include ␣-Bgt (22), diazepam-binding inhibitor protein, and its processing fragments such as the octadecaneuropeptide (39,40). Given a plethora of different GABA A R subunits, it seems surprising that there is such a paucity of known proteins interacting with the receptor. For example, a great deal of polypeptide toxins is known for nAChR, closely related to GABA A R. Here, for a series of GABA A R subtypes, we report specific fluorescent labeling and inhibition of GABA-induced ion currents by several protein toxins from snake venoms.
As a starting point, we reproduced ␣-Bgt binding to the ␣1␤3 GABA A R reported in Ref. 22 to obtain a verified test system of binding. As expected, long-chain ␣-neurotoxin ␣-Ctx, closely related to ␣-Bgt, showed competition with ␣-Bgt binding. We then investigated whether a fluorescently labeled ␣-Ctx, could stain cells expressing ␣1␤3␥2 or ␣1␤3 GABA A R. We found that Alexa Fluor 546 ␣-Ctx, as well as Alexa Fluor 555 ␣-Bgt, were able to stain both types of cells (Fig. 2, A-D). Whereas ␣-Ctx blocks Alexa Fluor 555 ␣-Bgt binding at ␣1␤3-transfected cells (Fig. 2E) as effectively as ␣-Bgt, Alexa Fluor 555 ␣-Bgt or Alexa Fluor 546 ␣-Ctx staining of ␣1␤3or ␣1␤3␥2-transfected cells was blocked more effectively by ␣-Ctx (Fig. 2, F-H). This observation suggests that GABA A Rs with different subunit combinations bear binding sites with slightly different properties. A, loop II peptide fragment CDAWCGS of NT I was grafted to NT II structure, replacing the corresponding SDH sequence (see Fig. 1 for sequence alignment). B, normalized GABA-evoked currents (10 M of GABA, ␣1␤3␥2 GABA A R) in the presence of wild-type NT I, NT II, and chimeric NT II/I at 10 M. Wild-type NT II does not inhibit the receptor, whereas both wild-type NT I and NT II/I chimeric toxin show current inhibition by 19 Ϯ 3 and 16 Ϯ 3%, respectively. Asterisks indicate significant inhibition (p Ͻ 0.01, n ϭ 6, paired t test). C, last frame of a 100-ns run of molecular dynamics simulation of NT II complex with ␤3/␣1 extracellular domains. NT II failed to reproduce mode of binding shown by ␣-Ctx model. D, molecular model of chimeric NT II/I toxin complex with ␤3/␣1 extracellular domains. Note that the tip of the loop II (which was grafted to the short toxin NT II molecule from long NT I toxin) rests under the C loop of ␤3 subunit.
To investigate how ␣-Ctx and other anticholinergic TFTs might influence GABA A R function, we performed electrophysiology studies of ␣1␤2␥2, ␣1␤3␥2, ␣2␤2␥2, ␣2␤3␥2, ␣5␤2␥2, and ␣1␤3␦ GABA A R subunit sets expressed in Xenopus oocytes. It was found that ␣-Ctx blocks ␣1␤3␥2 with a relatively high potency (IC 50 ϭ 236 nM, Fig. 4, A and B) and other receptor subtypes with lower potencies, namely in submicromolar to low micromolar range (Fig. 4B). It is worth noting that ␣-Ctx manifests some selectivity toward ␣1 as compared with the ␣2 subunit only in the presence of the ␤3 subunit, whereas in the presence of the ␤2 subunit it does not show noticeable selectivity to any ␣ subunit investigated. This suggests that the ␣-Ctxbinding site might be located at the interface between the ␤ and ␣ subunit. Other tested TFTs were less active, and their potencies were not measured. However, it is clear that Ls III and WTX are active in the micromolar range, whereas NT II and OWT do not show any inhibition of ␣1␤3␥2 receptor at 10 M.
Electrophysiology measurements indicated that ␣-Ctx (for the ␣1␤3␥2 subtype) shifted GABA dose-response curve to the right, at the same time changes in the slope and maximally achieved current were observed (Fig. 4C). Such a picture is typical of noncompetitive antagonists. However, we found that muscimol, acting as agonist of GABA A R, but not the allosteric modulator diazepam, displaced Alexa Fluor 546 ␣-Ctx from ␣1␤3␥2-transfected cells (Fig. 4D) to the level of ϳ20%. So a large part of ␣-Ctx action seems to be competitive and to be mediated via binding to the GABA-binding sites of GABA A receptors. Because occupation of only one orthosteric site is sufficient for receptor activation, the ␣-Ctx could bind either to one orthosteric site (as a competitive inhibitor) and allosterically inhibit receptor activation through the second site or ␣-Ctx could bind to both orthosteric and allosteric types of sites, shifting GABA dose-response curve and diminishing maximal current amplitude. The second possibility is supported by the finding that about 20% of bound Alexa Fluor 546 ␣-Ctx could not be displaced by muscimol, supporting an additional interaction of Alexa Fluor 546 ␣-Ctx with a nonorthosteric binding site.
The tip of loop II is the most important part of ␣-Ctx and other TFTs for binding to AChBP and nAChRs (41). To test whether this region is also involved in the interaction with GABA A R, we applied mutant NT II with loop II grafted from NT I. This chimeric toxin (14) seems to be a suitable candidate because we have not detected any inhibition of GABA A R by NT II at 10 M, but we observed some inhibition by NT I at the same concentration. Upon grafting of the NT I loop II fragment bearing a disulfide loop to the short neurotoxin NT II, the latter became functionally active against the receptor, strongly supporting the key role of loop II in TFTs in the interaction with GABA A R. Thus, we conclude that the tip of the loop II of TFTs plays a crucial role in GABA A R binding and inhibition.
Previously, it was suggested that some structural similarity (see Fig. 5A) of ␣-conotoxin ImI to the ␣-Ctx loop II may contribute to its binding to the neuronal nAChRs (28). Electrophysiology revealed that ImI inhibited GABA A R at 100 M (Fig.  5B), whereas muscle nAChR-specific SIA and heteromeric nAChR-specific MII, lacking ␣-Ctx-similar features, were ineffective (Fig. 5B, bottom). Therefore, even molecules with structural similarity to the ␣-Ctx loop II could inhibit GABA A R. These data indicated that ␣-conotoxin ImI represents another type of polypeptide compounds able to interact with GABA A R. Moreover, synthetic peptide representing the sequence WCDAFCSIRGKR of the ␣-Ctx loop II tip also inhibited GABA A R (Fig. 5B). Despite the fact that we observe only weak inhibition at a relatively high concentration, short peptide GABA A R ligands on the base of loop II peptide or ImI could represent a good starting point for rational design of more specific ligands.
Our findings suggest that ␣-Ctx binds at the ␤/␣ interface and the loop II tip plays a crucial role in its binding. Given that, we constructed a molecular model of ␣-Ctx bound to the ␤3/␣1 dimer of extracellular domains and performed 100-ns molecular dynamics studies in implicit solvent. Loop II residues 25-36 rested stably at the site under the loop C of the ␤3 subunit (Fig.  6A). With respect to ␤3 subunit backbone, root mean square deviation of these residues have not exceeded 0.15 nm, suggest- Residues that are different in ␣1 and ␣2 subunits are shown blue (stick representation), whereas ␣1 structure conserved between these two subunits is colored orange. Note that one of the closest to agonist-binding site residues is Ser-67 (dark blue). Inset shows that, according to our molecular modeling, by being mutated to lysine (as in ␣2 subunit) this residue comes close to the lysine of the ␣-Ctx loop II (magenta, space-filling model). B, inhibition of mutant receptor ␣1(S67K)␤3␥2 by ␣-Ctx is characterized by a higher IC 50 ϭ 703 nM (632, 781 nM) value than wild-type ␣1␤3␥2. However, inhibition of mutant receptor ␣2(K67S)␤3␥2 by ␣-Ctx is characterized by a lower IC 50 ϭ 442 nM (352, 554 nM) value than wild-type ␣2␤3␥2 suggesting the importance of ␤/␣ interface. GABA was given at 10 M, which is slightly less than EC 50 . Data points are expressed as percentage of control amplitudes (mean peak current Ϯ S.E.).
ing that such a complex might exist in principle. Interestingly, we found that both Arg-33 and Arg-36 residues in ␣-Ctx form salt bridges with ␣-Ctx Asp-27 and ␤3 Glu-155 (Fig. 6B). Arg-36 residue is unique to ␣-Ctx among all tested toxins, and its substitution by hydrophobic residues in other toxins (Fig. 1) could explain their weaker GABA A R inhibiting activity. Molecular modeling of NT II interaction with the ␤3/␣1 dimer revealed inability of this short toxin to form a complex with the orthosteric site of GABA A R (Fig. 7C). In contrast, a molecular model of the NT II/I chimeric toxin showed properties very close to that of ␣-Ctx complex model with the ␤3/␣1 dimer. In particular, loop II, grafted to the NT II scaffold from NT I, formed stable contacts with the C-loop of the ␤3 subunit extracellular domain (Fig. 7D).
It should be noted that previously we demonstrated that upon modification of ␣-Ctx with NHS ethers, the major product is the one with the label attached at Lys-23 (42). This lysine residue is in loop II, which according to our model takes part in receptor binding. However, it is not one of the deeply buried residues, and its modification does not necessarily disrupt ␣-Ctx binding, as indicated by the interaction of modified ␣-Ctx with the nAChRs (42).
In electrophysiological studies, we observed a somewhat more potent inhibition of ␣1␤3␥2 over ␣2␤3␥2 GABA A R. Because both ␣1 and ␣2 subunits share extremely high sequence homology, the respective receptors should differ only slightly in their extracellular parts. To test the above-described model, as a candidate for single-point mutagenesis we chose the Ser-67 residue (numbering of residues is given according to the homology model of extracellular domain) situated at the ␤/␣1 interface. It is substituted by a lysine residue in the ␣2 subunit, and this positively charged residue could confer the difference in the ␣-Ctx affinity toward these two receptors, because ␣-Ctx loop II bears significant positive charge itself. This mutation was done, and the ␣-Ctx potency at ␣1(S67K)␤3␥2 GABA A R was diminished, approaching that at ␣2␤3␥2 (Fig. 8).
When our manuscript was ready for submission, two papers (43,44) appeared, clearly demonstrating the perspective of TFTs for research on GABA A Rs. In Ref. 43, the authors analyzed in detail the interactions of ␣-Bgt and its fluorescent derivative with GABA A R in hippocampal neurons and with recombinant GABA A R subtypes, comparing their effects on ion currents and staining patterns in the presence of diverse ligands (agonists and antagonists) both with GABA A Rs and nAChRs. Their conclusion was that ␣-Bgt inhibits various functional GABA A R subtypes with different efficiency, with the ␣2␤2␥2 receptor being inhibited most potently. Here, we show that the preferred target of ␣-Ctx is the ␣1␤3␥2 receptor, indicating that diverse ␣-neurotoxins may have different affinity for distinct GABA A R subtypes and might be able to distinguish between various GABA A R subtypes. Concerning the mode of ␣-Bgt binding, Hannan et al. (43) described it as "a mixed inhibitory manner." It is quite clear that further work is necessary to obtain a high resolution picture for recognition of GABA A R by both ␣-Bgt and ␣-Ctx.
Interestingly, in Ref. 43 it was also found that ␣-Bgt inhibited spontaneous channel openings in GABA A R composed of ␣4 and ␦ subunits, but in the presence of GABA the toxin behaved as a positive allosteric modulator. Such effects were not observed in this study, supporting our observation (Fig. 2) that different toxins, dependent on the receptor composition, might exhibit different effects. However, ␣-Bgt was found to bind at the orthosteric site (43). It is thus possible that binding of toxins to one of two orthosteric sites could lead to an allosteric change in the conformation of the second site and that the functional consequences of these allosteric interactions might be influenced by the type of additional (in this case ␦ and ␥) subunits in the receptor (45).
The second above-mentioned paper (44) also deals with GABA A R potentiators by describing two TFTs from the coral snake venom that potentiate the GABA A R activity at low nanomolar concentrations. An undisputable advantage of these toxins is that they do not act on nAChRs and are the first, strictly speaking, GABA A R-specific TFTs. These toxins (MmTX1 and MmTX2) are not ␣-neurotoxins (like ␣-Ctx or ␣-Bgt) but belong to the nonconventional neurotoxins (additional 5th disulfide in loop I) like the GABA A receptor inhibiting toxin WTX identified in our study (Fig. 3). We and the authors of Ref. 44 hypothesized that TFTs should bind to GABA A R in a similar manner as ␣-neurotoxins bind to nAChRs. This is supported by their finding that the toxin interaction is decreased upon the H33S mutation in the central loop II of MmTX2 or upon double mutation G228E/Q231K in the receptor loop C. Concerning the binding interface of the toxin, this single mutation agrees with the conclusions we made based on the activity of the chimeric TFT. Our S67K and K67S mutations in the ␣1 and ␣2 subunits of GABA A R, respectively, although not being in the loop C, also affect binding, and it is clear that mapping of the toxin-binding interfaces in the GABA A R is a challenging task.
In conclusion, even such extensively studied objects as ␣-Ctx, which has been investigated since the 1970s, could hide some secrets, as demonstrated by its earlier overlooked inhibition of GABA A Rs. We showed that ␣-Ctx and NT I compete with ␣-Bgt for binding site(s) at the ␣1␤3 receptor, and we found that ␣-Ctx, LsIII, and WTX inhibit the functional activity of ␣1␤3␥2 GABA A R in the submicromolar to micromolar concentration ranges. In addition, the first peptide toxin ␣-conotoxin ImI inhibiting GABA A R was identified. These facts together with previously published observations demonstrate that GABA A Rs are the target for ␣-conotoxin and diverse three-finger toxins.