Dimeric α-Cobratoxin X-ray Structure

Background: α-Cobratoxin (αCT) dimer (αCT-αCT) has recently been discovered and found to bind both with α7 and α3β2 nicotinic receptors (nAChR). Results: αCT-αCT x-ray structure and intermolecular disulfides were established, intramolecular disulfides in central loops II were reduced, and interaction with distinct nAChRs was analyzed. Conclusion: Loop II disulfide is necessary for αCT-αCT binding to α7 but not α3β2 nAChR. Significance: Dimeric α-neurotoxins provide new means for distinguishing distinct nAChRs. In Naja kaouthia cobra venom, we have earlier discovered a covalent dimeric form of α-cobratoxin (αCT-αCT) with two intermolecular disulfides, but we could not determine their positions. Here, we report the αCT-αCT crystal structure at 1.94 Å where intermolecular disulfides are identified between Cys3 in one protomer and Cys20 of the second, and vice versa. All remaining intramolecular disulfides, including the additional bridge between Cys26 and Cys30 in the central loops II, have the same positions as in monomeric α-cobratoxin. The three-finger fold is essentially preserved in each protomer, but the arrangement of the αCT-αCT dimer differs from those of noncovalent crystallographic dimers of three-finger toxins (TFT) or from the κ-bungarotoxin solution structure. Selective reduction of Cys26-Cys30 in one protomer does not affect the activity against the α7 nicotinic acetylcholine receptor (nAChR), whereas its reduction in both protomers almost prevents α7 nAChR recognition. On the contrary, reduction of one or both Cys26-Cys30 disulfides in αCT-αCT considerably potentiates inhibition of the α3β2 nAChR by the toxin. The heteromeric dimer of α-cobratoxin and cytotoxin has an activity similar to that of αCT-αCT against the α7 nAChR and is more active against α3β2 nAChRs. Our results demonstrate that at least one Cys26-Cys30 disulfide in covalent TFT dimers, similar to the monomeric TFTs, is essential for their recognition by α7 nAChR, although it is less important for interaction of covalent TFT dimers with the α3β2 nAChR.

In Naja kaouthia cobra venom, we have earlier discovered a covalent dimeric form of ␣-cobratoxin (␣CT-␣CT) with two intermolecular disulfides, but we could not determine their positions. Here, we report the ␣CT-␣CT crystal structure at 1.94 Å where intermolecular disulfides are identified between Cys 3 in one protomer and Cys 20 of the second, and vice versa. All remaining intramolecular disulfides, including the additional bridge between Cys 26 and Cys 30 in the central loops II, have the same positions as in monomeric ␣-cobratoxin. The three-finger fold is essentially preserved in each protomer, but the arrangement of the ␣CT-␣CT dimer differs from those of noncovalent crystallographic dimers of three-finger toxins (TFT) or from the -bungarotoxin solution structure. Selective reduction of Cys 26 -Cys 30 in one protomer does not affect the activity against the ␣7 nicotinic acetylcholine receptor (nAChR), whereas its reduction in both protomers almost prevents ␣7 nAChR recognition. On the contrary, reduction of one or both Cys 26 -Cys 30 disulfides in ␣CT-␣CT considerably potentiates inhibition of the ␣3␤2 nAChR by the toxin. The heteromeric dimer of ␣-cobratoxin and cytotoxin has an activity similar to that of ␣CT-␣CT against the ␣7 nAChR and is more active against ␣3␤2 nAChRs. Our results demonstrate that at least one Cys 26 -Cys 30 disulfide in covalent TFT dimers, similar to the monomeric TFTs, is essential for their recognition by ␣7 nAChR, although it is less important for interaction of covalent TFT dimers with the ␣3␤2 nAChR.
Nicotinic acetylcholine receptors (nAChRs) 4 are among the most thoroughly investigated ligand-gated ion channels. There are two main groups of nAChRs: muscle-type and neuronal receptors. Muscle-type nAChRs consist of five subunits (two ␣1, one ␤1, one ␥ (or ⑀), and one ␦ subunit) and contain two binding sites for agonists/competitive antagonists that are located at the interfaces of two ␣ subunits with the adjacent ␥ and ␦ subunits. Neuronal nAChRs consist of subunits of two different types: ␣ (␣2-␣10) and ␤ (␤2-␤4), and are hetero-(formed by different combinations of ␣ and ␤ subunits) or homo-(formed by ␣ subunits of ␣7-␣10 subtypes) pentameric proteins (for review, see Ref. 1). It has recently been shown that neuronal nAChRs are also present in nonneuronal cells, for example in the immune system, where they play an important role in the regulation of inflammation processes (2). Elucidating the roles of different nAChR subtypes requires fine and precise biochemical tools to discriminate among different receptor activities. Toward this end, ␣-neurotoxins from snake venoms have proven to be valuable instruments. These ␣-neurotoxins belong to the family of so-called three-finger toxins (TFTs). Three protruding loops (fingers) of their polypeptide chain are fastened by four conserved disulfide bridges, strengthening the core of the molecule, thus forming a "three-finger" spatial motif. ␣-Neurotoxins interact with nAChRs through their loops and can discriminate between different nAChR subtypes. So-called short chain ␣-neurotoxins (59 -63 amino acid residues, four disulfide bridges) block only muscle-type nAChRs whereas the long chain ␣-neurotoxins (66 -74 residues, with an additional disulfide bond within the central loop II) display an expanded target choice by also blocking the homopentameric ␣7 subtype nAChR (3,4). This additional disulfide bridge in loop II is crucial for the interaction with the homopentameric ␣7 nAChR. To date, however, neither short nor long chain ␣-neurotoxins have been shown to block heteropentameric neuronal nAChRs. TFTs from another group, namely -bungarotoxins, are able to block heteropentameric ␣3␤2 nAChRs and, to a lesser extent, ␣4␤2 and ␣7 nAChRs. Structurally, they resemble long chain ␣-neurotoxins, but they have been characterized as noncovalent dimers based on solution or crystal structures (5). Another group of TFTs, known as weak or nonconventional neurotoxins, comprising 62-68 amino acid residues and five disulfide bridges, has also been described in the literature. The main structural feature of these nonconventional toxins is the presence of a fifth disulfide bridge in the N-terminal loop I. These toxins interact with the ␣7 and muscle-type nAChRs as well as with muscarinic acetylcholine receptors (6).
Recently, we have isolated a disulfide-bound dimer (␣CT-␣CT) of ␣-cobratoxin (␣CT), a long chain neurotoxin, from the Naja kaouthia cobra venom (7). The peculiarity of this dimer is its capability to block heteropentameric ␣3␤2 nAChRs, in addition to blocking muscle-type and ␣7 nAChRs, whereas monomeric ␣CT is not effective against ␣3␤2 nAChRs. We could not establish by chemical means the positions of interchain disulfides in the ␣CT-␣CT dimer. In addition, although the circular dichroism data suggested the preservation of the three-finger fold, in agreement with more or less preserved activity against distinct nAChR subtypes (1), no reliable conclusions could be drawn about the relative spatial disposition of the protomers, and a reliable three-dimensional model of the covalent ␣CT-␣CT dimer could not be generated.
Here, we describe crystallization and the x-ray structure of the ␣CT-␣CT dimer which provides information on the disposition of the intermolecular disulfide bridges, the conformation of protomers, and their assembly into a dimeric neurotoxin. To study the relationship between the structures and the binding modes of monomeric and dimeric ␣CTs, we modified the ␣CT-␣CT by selectively reducing the additional disulfide bond in loop II (Cys 26 -Cys 30 ) either in one or in both protomers and tested the biological activity of the resulting ␣CT-␣CT derivatives on nAChR preparations. Complementary data were obtained from the analysis of a covalent heteromeric dimer, composed of ␣-cobratoxin and cytotoxin 3 (␣CT-CTX).

EXPERIMENTAL PROCEDURES
Isolation of Dimers-␣CT-␣CT and ␣-CT-CTX were purified as in Ref. 7.
Crystallization and X-ray Analysis-The lyophilized ␣CT-␣CT homodimer was resuspended in water to a concentration of 5 mg/ml and crystallized using the vapor diffusion method by mixing the toxin in a 1:1 ratio with the reservoir solution composed of 60% (v/v) 2-methyl-2,4-pentanediol and 0.1 M glycine-HCl buffer at pH 2. Crystals were flash frozen in liquid nitrogen prior to data collection. The ␣CT-␣CT dimer crystallized in space group P3 1 21 with parameters described in Table 1. Data were collected at 100 K on beamline ID14-EH1 at the ESRF (Grenoble, France) and were processed using XDS and scaled with XSCALE (8,9). Monomeric ␣-CT (Protein Data Bank (PDB) code 2CTX) (10) was used as model in molecular replacement trials using Phaser (11) as implemented in the CCP4 software suite (12). Automatic model building was performed using ArpWarp (13). Iterative rounds of manual structure building and refinement were carried out using COOT (14) and REFMAC (15). The first 67 residues of the toxin were well defined in electron density whereas the last 4 could not be built and are likely to be flexible. The final structure ( Fig. 1) was validated using the Molprobity server (16). Superpositions were performed using SSM (17), and figures were generated using PyMOL.
In the crystal structure, two protomers, chains A and B, are present in the asymmetric unit, but these are from two different dimer molecules and share only 183 Å 2 at their interface. Each chain makes a dimer with a protomer that is related by a 2-fold crystallographic axis. Hence, chains A and AЈ, or B and BЈ related by the symmetry operation (Ϫx, Ϫxϩy, Ϫzϩ1/3), make up ␣CT-␣CT dimers, which share an interface of ϳ1400 Å 2 . The structure of the ␣CT-␣CT dimer has been deposited in the PDB with accession code 4AEA.
Selective Reduction and Alkylation-10 nmol of ␣CT-␣CT were dissolved in 50 l of 50 mM Tris-HCl buffer, pH 8.0, and stirred with 5.5 l of 2 mM tris(2-carboxyethyl)phosphine (Sigma) under nitrogen for 20 min, then 1 mg of iodoacetamide (Merck) was added under nitrogen for 2 h. The reaction was stopped by direct application on a Phenomenex C18 Jupiter (4.6 ϫ 250 mm) column with subsequent separation of products in a gradient of 15-45% acetonitrile in 30 min in the presence of 0.1% trifluoroacetic acid, at a flow rate of 1.0 ml/min (Fig. 2). Each isolated product underwent rechromatography under the same conditions. Incorporation of amidoethyl groups was checked by MALDI-TOF MS and tandem MS as in Ref. 7.
Radioligand Analysis-Competition binding assays (Fig. 3) were carried out using suspensions of nAChR-rich membranes  (18), using 125 Ilabeled ␣-bungarotoxin as reported previously (19). Electrophysiological Experiments-All experiments were conducted at human nAChRs expressed in Xenopus oocytes. cDNAs encoding for the ␣7 or ␣3 and ␤2 subunits were injected into the oocyte nucleus using a roboinject (Multichannel Systems, Reutlingen, Germany). cDNA concentrations and injection procedures were standard and as described previously (7). Two to 3 days after injection, recordings were carried out with a two-electrode voltage clamp system using a HiClamp (Multichannel Systems). This system offers the advantage of minimal solution sample and multiple recordings from the same solutions.
Molecular Modeling-The refined x-ray structure of dimeric toxin was equilibrated in GROMACS 4.0.3. All molecular simulations were performed in an OPLS-AA force field, and molecular dynamics simulations were calculated at a temperature of 300 K with a dielectric permittivity ⑀ ϭ 1. Integration procedures were performed with time steps as small as 1 fs, for simulation times of 20 ns. After molecular dynamics simulations of the energy-minimized ␣CT-␣CT x-ray structure, the obtained molecule was docked to L. stagnalis AChBP (LsAChBP) (20). Hex 6.0 was used for docking, and subsequent visual analysis in the SPDBViewer allowed us to reject false positive solutions.

RESULTS
Structure of ␣-CT Dimer-We have solved the three-dimensional structure of ␣CT-␣CT, a dimeric ␣CT, at a resolution of 1.94 Å. Within this dimer, protomers are covalently linked by disulfide bonds between Cys 3 in one protomer and Cys 20 in the other, and vice versa (Fig. 1A). The two dimers that we reconstituted from our structural data display a slight variation in the relative arrangement of the two protomers around the 2-fold crystallographic axis (Fig. 1B), likely reflecting the degree of flexibility of the protomers around this hinge region. FIGURE 1. X-ray structure of dimeric ␣CT. A, ␣CT-␣CT dimer is composed of two protomers represented in blue and purple, respectively. The ␤1 strand from the first protomer is extended away from the toxin core and swapped with the ␤1 strand from the second protomer. Disulfide bridges between Cys 3 from one protomer and Cys 20 from the second protomer stabilize the dimeric toxin. B, ribbon representation of two ␣CT-␣CT dimers, reconstituted from our structural data, superposed with respect to one of the protomers. The variability between the two dimers likely reflects the flexibility between the two toxin cores. C, ribbon representation of the superposition of monomeric ␣CT in orange on a protomer of the dimeric ␣CT (blue and purple). The boxed region is a zoomed in view rotated 90°around a horizontal axis in D. Pro 7 from monomeric ␣CT (in orange) or from either protomer of the ␣CT-␣CT dimer (blue and purple) is depicted as sticks. In the dimeric form, the Pro 7 is extended away from the toxin core whereas in the monomeric ␣CT, Pro 7 makes a turn.

JOURNAL OF BIOLOGICAL CHEMISTRY 6727
The three-dimensional structure of protomers in the dimeric ␣CT resembles the monomeric ␣CT (PDB code 2CTX) (10) and essentially preserves the TFT fold, except for a kink at Pro 7 ( Fig. 1, C and D). This results in a swap of the first N-terminal ␤ strand from one protomer with that of the second protomer, hence forming the ␣CT-␣CT dimer (Fig. 1A). The two ␣CT protomers are related by a 2-fold axial symmetry. Whereas in monomeric ␣CT the N-terminal ␤ strand is arranged in an antiparallel manner with respect to the second N-terminal ␤ strand through a turn induced by Pro 7 , in the dimeric ␣CT-␣CT this Pro 7 is not involved in generating a turn but allows the outward extension of the first N-terminal ␤ strand away from the toxin core. The extended ␤1 strand displays a large interaction interface with the ␤2 strand from the second protomer contributing to the dimeric ␣CT-␣CT. These two N-terminal ␤ strands are covalently linked through two disulfide bridges: Cys 3 of each chain is bound to Cys 20 of the other chain (Cys 3 -Cys 20 Ј and Cys 3 Ј-Cys 20 ). Interestingly, these cysteine residues are highly conserved in ␣-neurotoxins and are involved in an internal disulfide bond in monomeric ␣CT. However, reduction of these disulfides in -bungarotoxin did not considerably decrease toxin activity against ␣3␤2 nAChRs (21). Even such a drastic disulfide bond rearrangement as observed in the ␣CT-␣CT does not affect the three-finger motif in each of the polypeptide chains of the protomers. Indeed, a protomer of ␣CT-␣CT superposes on the monomeric ␣CT with a root mean square deviation of 0.9 Å (over residues 9 -67) (Fig. 1C). The first loops of the protomers are maintained through the ␤ strand swap between the two protomers constituting the covalent dimer (Fig. 1A). The geometry of all other disulfide bridges in both polypeptide chains of ␣CT-␣CT is similar to that observed in monomeric ␣CT. The interface between the two protomers of the covalent dimer extends over ϳ1400 Å 2 , with the tip of loops II contributing ϳ440 Å 2 , and involves 42 hydrogen bonds between the two polypeptide chains as characterized by the PISA server (22).
Alkylation of Additional Disulfides in ␣CT Dimer-The loop II additional disulfide in long chain ␣-neurotoxins is crucial for the interaction with ␣7 nAChR (4). We wondered whether this loop II disulfide bridge is important for dimeric ␣CT function and used the fact that it can be selectively reduced and alkylated (23,24). These loop II additional disulfides are exposed in the structure and are therefore accessible to selective reduction and subsequent alkylation, in contrast to other intramolecular or intermolecular disulfide bridges of the toxin. We have generated ␣CT-␣CT derivatives where this disulfide bond (Cys 26 -Cys 30 ) has been reduced and ethylamidated (Fig. 2). This reaction appears to be kinetically controlled, similarly to what is observed with ␣CT (24) and does not disrupt the dimeric structure of ␣CT-␣CT. Interestingly, under chosen conditions the loop II disulfide bridges from each protomer can be reduced simultaneously, and the reaction mixture contains derivatives with either a single reduced and alkylated disulfide (␣CT-␣CT-2AE), or with both disulfides modified (␣CT-␣CT-4AE). This is in agreement with our earlier isolation both of the 24 -33 tryptic peptide fragment of the loop II, containing two pyridylethyl groups, and of the same peptide with the intact disulfide bond (7). Fortunately, it was possible to separate the two different derivatives from each other and from the parent molecule (Fig. 2) for further analysis of their biological activity.
Competition Experiments with 125 I-Labeled ␣-Bungarotoxin-Disturbing the central loop II structure in both protomers by reducing the additional disulfides and alkylating the respective cysteine residues results in a significant decrease in affinity of the modified ␣CT-␣CT both for the ␣7 nAChR (Fig. 3A) and for AChBPs (Fig. 3, B and C), soluble structural homologs of the nAChR ligand binding domains (20). However, if only one of the loop II disulfide bridges in the ␣CT-␣CT is disrupted, the affinity remains virtually unchanged (Fig. 3, A-C). This suggests that preserving the native structure in only one of the central loops II of the dimer is sufficient for effective interaction of ␣CT-␣CT with these targets.
Nevertheless, as might be expected, reduction of one or both these disulfides in ␣CT-␣CT does not practically affect its interaction with the muscle-type nAChR (Fig. 3D), similar to what was observed for the interaction with this receptor following the reduction of the loop II disulfide bridge in ␣CT (4) Electrophysiology Studies-The activity of ␣CT-␣CT and its reduced derivatives was compared at human ␣7 and ␣3␤2 nAChRs expressed in Xenopus oocytes. Because naturally occurring dimers ␣CT-␣CT and ␣CT-CTX and especially individual reduced/alkylated derivatives of the former dimer were available only in minor amounts, we could not characterize the results of the electrophysiological experiments in terms of IC 50 as was done in the above described competition binding experiments. In view of the latter, all dimers were applied at 1 M concentration, and their inhibitory activity is illustrated by Figs. 4-7 representing a decrease in the amplitude of the acetylcholine-induced currents from 10 to 90%.
Similar to the results of binding studies, reduction of only one disulfide in ␣CT-␣CT practically does not decrease the inhibitory activity against ␣7 nAChR, but it drops considerably if both disulfides are reduced and alkylated (compare Fig. 4A with Fig. 5, A and B). Surprisingly, with the ␣3␤2 nAChR both ␣CT-␣CT derivatives have a very similar activity which is much higher than that of ␣CT-␣CT itself (compare Fig. 6, A-D, with Fig. 7, A and C). Interestingly, for the heterodimer ␣CT-CTX which showed quite potent inhibition of 125 I-labeled ␣-bungarotoxin binding to ␣7 nAChR (7), in the present electrophysiology experiments we found that this protein is not only slightly more active than ␣CT-␣CT in inhibiting ␣7 nAChR (compare Fig. 4, A and C, with Fig. 4, B and D), but considerably exceeds the ␣CT-␣CT homodimer in blocking the ␣3␤2 nAChR (Fig.  7).
Computer Modeling-Superposition of one protomer of the ␣CT-␣CT dimer onto monomeric ␣CT bound to LsAChBP (PDB code 1YI5) (20) (Fig. 8A) reveals that such an interaction would be energetically unfavorable due to a very close disposition of the loop II tips from the two protomers leading to clashes between the second ␣CT-␣CT protomer and LsA-ChBP. To identify potential conformational changes in the dimeric toxin likely to lead to binding to AChBP, we performed molecular dynamics simulation trials followed by docking to LsAChBP. Our results suggest that a rotation around the Pro 7 -Asp 8 bonds from both protomers could provide conformations (Fig. 8B) that are likely to be energetically more favorable for the complex formation (Fig. 8C) and which would resemble those of -bungarotoxin (Fig. 9B) or haditoxin (Fig. 9C).

DISCUSSION
Our 1.94 Å resolution x-ray structure of ␣CT-␣CT, a covalently linked dimer of ␣CT isolated earlier from the N. kaouthia cobra venom (7), reveals how the two protomers composing the toxin are bridged and how they are oriented relative to each other (Fig. 1A). It shows that two interchain disulfide bonds connect the interweaved loops I of the two monomeric ␣CT, namely Cys 3 of one monomer is bound to Cys 20 of the second monomer and vice versa. Each protomer essentially preserves a    8) suggests that binding to receptors would be energetically unfavorable. In this configuration, the interaction of one ␣CT-␣CT protomer with the receptor ligand binding domain would lead to steric clashes of the second protomer with the receptor (Fig. 8A). Indeed, for ␣CT-␣CT to interact with the receptor, a conformational change inducing a relative reorientation of the loops II of the two protomers would be required (Fig. 8B). Interestingly, the spectrum of ␣CT-␣CT activity toward distinct nic-  otinic receptors is different from what is observed with the monomeric ␣CT. The dimeric ␣CT-␣CT is able to bind to the muscle-type and ␣7 nAChRs, but in contrast to monomeric ␣CT it can also block the heteromeric ␣3␤2 nAChR (7). Because in this activity it resembles -bungarotoxin, a noncovalent dimer (27), it is possible that dimerization is a requirement for the binding of a TFT to heteromeric ␣3␤2 nAChRs.
After our discovery of the dimeric ␣-cobratoxin (7), two novel dimeric three-finger neurotoxins have been described, namely irditoxin from Boiga irregularis venom (29) and haditoxin from the king cobra (30). Haditoxin is a noncovalent dimer composed of two short chain ␣-neurotoxins and adopts a topological arrangement (Fig. 9C) reminiscent of that observed for -bungarotoxin, on which it superposes with a root mean square deviation of 2 Å. It also displays a protomer-protomer interaction surface of 570 Å 2 , comparable with that between the protomers of the -bungarotoxin dimer. Haditoxin can block not only muscle-type nAChRs, as typically observed for short chain ␣-neurotoxins, but unexpectedly also blocks homomeric ␣7 and heteromeric ␣3␤2 nAChRs. This confirms our earlier conclusion that dimerization resulting in noncovalently or in covalently linked TFTs is an important factor for recognition of heteromeric nAChRs (7). Haditoxin does not comprise an additional loop II disulfide, and its activity raises the question about the role of such a disulfide in recognition of neuronal nAChRs. Nevertheless, it should be kept in mind that blocking of neuronal nAChRs by haditoxin was only observed at very high toxin concentrations (30). The monomeric form of haditoxin has not been isolated and cannot be used for comparison purposes, contrary to the ␣CT-␣CT dimer where the activity of ␣CT is well characterized and overlaps considerably with that of the covalently bound dimer. Furthermore, the classification of haditoxin to the family of short ␣-neurotoxins is controversial since its homology to erabutoxin, a classical short chain ␣-neurotoxin, is only 50%, whereas it is of 75-80% with the muscarinic toxin-like proteins, which are of different origin (25).
Irditoxin, yet another recently described dimeric toxin, consists of two highly homologous "weak" or "nonconventional" toxin-like protomers linked through a disulfide bridge and is selective for bird muscle-type nAChR (29). Each of these two protomers contains one additional cysteine residue (absent in typical TFTs), which covalently link the two irditoxin protomers via an intermolecular disulfide bridge. In the first chain, the additional cysteine is located in loop I whereas it is in loop II in the second chain. The three-dimensional structure of irditoxin (29) (Fig. 9D) shows that loops II of the two protomers are oriented in a somewhat similar way as those of the ␣CT-␣CT dimer (Fig. 9A). In the case of irditoxin, however, a much smaller interaction surface is shared by the two protomers (ϳ430 Å 2 against ϳ1400 Å 2 in ␣CT-␣CT).
Considering the importance of the additional loop II disulfide bridge for ␣CT interaction with the ␣7 nAChR (4) and for -bungarotoxin binding to ␣3␤2 nAChR (27), we characterized the role of this disulfide in ␣CT-␣CT interaction with different targets, namely with several subtypes of nAChRs and with AChBPs. Toward this end, we alkylated ␣CT-␣CT using iodoacetamide, which does not significantly contribute to the molecular charge or to the molecular mass of the dimer. Our binding (Fig. 3) and electrophysiology data  show that the presence of an additional disulfide in only one of the two loops II of the dimer is sufficient for the interaction with both ␣7 nAChR and AChBPs. However, both disulfides can be reduced and alkylated without diminishing the efficiency of binding to the muscle-type nAChR (Fig. 3D). Thus, our binding and electrophysiology data on the interaction of ␣CT-␣CT with the ␣7 and muscle-type nAChRs correlate well with previous data on monomeric ␣CT concerning the role of the disulfide bridge in the central loop II. However, in contrast to what is known for the -bungarotoxin noncovalent dimer, loop II disulfides in ␣CT-␣CT are not essential for interaction with the ␣3␤2 nAChR. On the contrary, reduction and alkylation of one or both of those disulfide bridges considerably enhance the inhibitory activity of the toxin variant against the ␣3␤2 nAChR ( Figs. 6 and 7).
The opposite roles of the central loop II disulfides in the interaction of dimeric TFTs with the ␣7 nAChR compared with the ␣3␤2 nAChR are also illustrated by our results for the ␣CT- In the -bungarotoxin dimer the ␤ strands from the two protomers constitute an extended antiparallel ␤ sheet, with the loop II from each subunit oriented in opposite directions. A similar topological arrangement is observed for haditoxin. A disulfide bridge is formed between the two irditoxin protomers, involving a loop I cysteine from the first chain (blue) and a loop II cysteine from the second chain (orange). Contrary to -bungarotoxin or haditoxin, but similar to the ␣CT-␣CT dimer, the loops II from the two protomers are oriented toward each other.
CTX heterodimer composed of ␣-cobratoxin and cytotoxin 3. The cytotoxin in this heterodimeric toxin is a short type TFT and contains no additional disulfide bridge in the loop II. However, the presence of a single loop II disulfide bridge in the ␣CT protomer within the ␣CT-CTX heterodimer is sufficient for blocking the ␣7 nAChR at least as efficiently as does ␣CT-␣CT (Fig. 4, A and B). Moreover, this heterodimer is more potent than ␣CT-␣CT against the ␣3␤2 nAChR (Fig. 7). These results confirm and extend our earlier conclusion that dimerization is essential for the activity of TFTs against heteromeric neuronal nAChRs. Our data for the reduced derivatives of ␣CT-␣CT, as well as literature data for haditoxin (30), also show that ␣3␤2 nAChRs can be recognized by dimeric TFTs containing no additional disulfides in the loops II. ␣CT-␣CT, in which one of the additional disulfides is preserved, appears to be more active against ␣3␤2 nAChR than ␣CT-␣CT in which these both disulfides are disrupted (Fig. 6, C and D) and comparable with the heterodimer that has also only one such disulfide (Fig. 7, B and  D).
In summary, we have determined the x-ray structure of ␣CT-␣CT, a covalent ␣CT dimer, and found that the dimerization results from a ␤ strand swap between the two protomers, two covalent linkages arising between Cys 3 of one protomer and Cys 20 of the second, and vice versa. We also investigated how specific ␣CT-␣CT binding is to different nAChR subtypes and how effective the dimeric toxin is at blocking these receptors. Our data confirm that dimerization is an important prerequisite for TFT interaction with heteromeric neuronal nAChRs, and from comparison of our and literature data it follows that the receptor recognition is possible with different relative orientations of protomers in the TFT dimers. We further find that at least one loop II disulfide bridge is required for activity against the ␣7 nAChR, whereas the reduction of one or both loop II disulfide bridges potentiates the activity of the dimeric toxin against the heteromeric ␣3␤2 nAChR.