The Solution Structure of the Complex Formed between a -Bungarotoxin and an 18-mer Cognate Peptide Derived from the a 1 Subunit of the Nicotinic Acetylcholine Receptor from Torpedo californica *

The region encompassing residues 181–98 on the a 1 subunit of the muscle-type nicotinic acetylcholine receptor forms a major determinant for the binding of a -neurotoxins. We have prepared an 15 N-enriched 18-amino acid peptide corresponding to the sequence in this region to facilitate structural elucidation by multi-dimensional NMR. Our aim was to determine the structural basis for the high affinity, stoichiometric complex formed between this cognate peptide and a -bungaro-toxin, a long a -neurotoxin. Resonances in the complex were assigned through heteronuclear and homonuclear NMR experiments, and the resulting interproton distance constraints were used to generate ensemble structures of the complex. Thr 8 , Pro 10 , Lys 38 , Val 39 , Val 40 , and Pro 69 in a -bungarotoxin and Tyr 189 , Tyr 190 , Thr 191 , Cys 192 , Asp 195 , and Thr 196 in the peptide participate in major intermolecular contacts. A comparison of the free and bound a -bungarotoxin structures reveals significant conformational rearrangements in flexible regions of a -bungarotoxin, mainly loops I, II, and the C-terminal tail. Furthermore, several of the calculated structures suggest that cation- p interactions may be involved in binding. The root mean square deviation of the polypeptide backbone in the complex is 2.07 rate the results from the three-dimensional HNHA experiment. Conformation Calculations— The cross-peak volumes in the two-dimensional NOESY spectra were integrated by the Gaussian fitting protocol using SPARKY. The cross-peaks were classified into three categories: strong, medium, and weak, with corresponding distance ranges of 1.8–3.0, 1.8–4.0, and 1.8–5.0 Å, respectively. The H N -H a 3 J coupling constants of the a 18-mer peptide were obtained from the three-dimensional HNHA experiment (37). The 3 J coupling constants were converted to dihedral angle restraints using previously described methods (38). For 3 J , 6 Hz, the dihedral angle restraint was assigned to 2 60° 6 30°; for 3 J . 8 Hz, the dihedral angle restraint was 2 120° 6 40°. All structures were calculated with distance geometry and simulated annealing protocols using the dg_sa.inp script of the NMR struc- ture calculation program, CNSsolve (39). The following is the potential energy function used in these calculations: F F F cdih , where F bon relates to bond length, F ang and F imp to bond angles, F vdw relates to the van der Waals repulsion term, F noe relates to NOE distance constraints, and F cdih relates to dihedral an- gles. Pseudoatoms were used for protons that could not be stereospecifically assigned. The pseudoatom correction feature of CNSsolve was used to adjust the NOE distance constraint range automatically. In each batch of calculations, a different random seed number was used to initiate the calculation of a set of 50 structures. From the pool of calculated structures, only those structures lacking any NOE violation ( . 0.5 Å) were selected as “acceptable” for further analysis. As a result of the weighting of the F noe term in CNSsolve, none of the other energy terms were as critical as F noe in determining an acceptable structure. In total, 120 acceptable structures of the Bgtx z a 18-mer complex were obtained from 6 batches of independent calculations ( i.e.

The nicotinic acetylcholine receptor (nAChR) 1 (1) has long been a prototype for ligand-gated ion channels. This receptor is involved in excitatory synaptic transmission at the neuromuscular junction and also plays an important role in the nervous system. The nAChRs are pentameric complexes composed of homologous subunits with subunits arranged around the central channel in a symmetrical manner. The muscle-type nAChR contains two ␣1 subunits and one each of the ␤1, ␥(⑀), and ␦ subunits. The ligand binding sites are situated at the ␣␥(⑀) and ␣␦ subunit interfaces. The muscle-type nAChR serves as an important model for the study of the structures and functions of related ligand-gated ion channels (for review, see Ref. 1).
The snake venom-derived ␣-neurotoxins fall into two categories, short and long neurotoxins, and act as high affinity competitive antagonists at the nAChR. Short neurotoxins (e.g. erabutoxin a) contain 60 -62 amino acid residues and 4 conserved disulfide bridges. Long neurotoxins have 66 -74 residues and 5 disulfide bonds, including four in a core region that are homologous in position to those found in the short neurotoxins. ␣-Bungarotoxin (Bgtx), obtained from the snake venom of Bungarus multicinctus, is a long ␣-neurotoxin that over the years has provided a powerful tool for the study of muscle-type nAChRs and which has come to be viewed as somewhat of a gold standard among the ␣-neurotoxins. A number of the ␣-neurotoxins have been heterologously expressed in recent years, allowing for investigations using site-directed mutagenesis (2)(3)(4)(5).
From its x-ray structure, Bgtx is a relatively flat, slightly concave, disc-shaped protein with a characteristic three-finger folding motif consisting of three loops of structure (6). Previous NMR structural studies indicate that the solution structure of Bgtx, although generally consistent with the x-ray structure, does reveal some differences (7). Notably, the side chain of Trp 28 in the two structures resides on opposite sides of the major plane of the molecule. In the solution structure, the Trp side chain is on the concave surface, as seen with most other ␣-neurotoxins containing this highly conserved residue. In contrast, the Trp side chain is located on the opposite face in the crystal structure (6). The structures of several other snake venom ␣-neurotoxins have been studied with NMR techniques (8 -12), and all exhibit the characteristic three-finger structure.
Previous work indicates that the main determinants for Bgtx binding to the muscle-type nAChR lie between residues 173 and 204 of the ␣1 subunit (13), a region that coincides with one of three segments of the ␣ subunit that have been implicated in agonist binding (for review, see Ref. 1). Tyr 190 , along with Cys 192 , Cys 193 , and Tyr 198 are selectively cross-linked with a variety of site-directed photoaffinity reagents (14). This region, termed segment C, contains a conserved pair of adjacent Cys residues, Cys 192 -Cys 193 , that form an unusual disulfide. Studies of synthetic peptides with sequences matching those in segment C have identified several peptides that bind Bgtx with affinities in the micromolar to submicromolar range (15,16). A peptide fragment (␣18-mer) with a sequence corresponding to amino acid residues 181-198 (␣-Y 181 -RGWKHWVYYTCCPDTPY 198 ) from the Torpedo californica nAChR binds Bgtx with an apparent K D of ϳ65 nM (17). Replacing the Tyr at position 190 with a Phe leads to a 60-fold decrease in Bgtx binding affinity for the altered peptide, suggesting an important role for this aromatic residue in complex formation (17). Mutations of Tyr 190 , when assessed in heterologous expression systems, also result in large decreases in ␣-neurotoxin binding (3,18,19). In addition, ligand gating is also dramatically affected by mutations in Tyr 190 (20,21). Studies with recombinant receptor fragments corresponding to the ␣ subunit from the mongoose nAChR, which is resistant to ␣-neurotoxins, suggest two subsites in the binding domain for Bgtx; one is a proline subsite consisting of Pro 194 and Pro 197 , and the other is an aromatic subsite involving positions 187 and 189 (15).
We previously described some of the structural features revealed by an NMR analysis of Bgtx complexed with a 12-amino acid peptide fragment (␣12-mer) of the Torpedo nAChR ␣1 subunit (␣-K 185 HWVYYTCCPDT 196 ), which has an apparent K D of ϳ1.4 M (17,22). We now describe a more expansive NMR structural analysis of the higher affinity complex formed with the ␣18-mer. The structure of this complex may provide valuable information on the orientation of the contact residues in the native nAChR and may help in elucidating the essential interactions that direct the ability of the ␣-neurotoxins to recognize receptor sequences with remarkable affinity and specificity.

EXPERIMENTAL PROCEDURES
Expression Construct-We designed an oligonucleotide sequence to encode for residues 181-198 (YRGWKHWVYYTCCPDTPY) of the ␣1 subunit from the nAChR of T. californica. Three copies of this expression cassette were inserted downstream of a 125-amino acid ketosteroid isomerase gene and upstream of a His-tag sequence in plasmid pET-31b(ϩ) (23). The final construct, ketosteroid isomerase-Met-(␣18-mer-Met) 3 -His-tag, contained single Met residues separating the three cassettes from each other, from ketosteroid isomerase, and from the C-terminal His tag. The oligonucleotide sequence of this construct is available upon request. The plasmid, whose insert sequence was confirmed by DNA sequence analysis, was used to transform cells of the expression strain BL21 (DE3) (Novagen).
Cell Growth-Cell cultures were grown in standard M9 medium except that 15 NH 4 Cl was used as a replacement for normal NH 4 Cl. All cultures were supplemented with 100 g/ml ampicillin (M9/Amp). A single colony from a fresh agar plate containing ampicillin was used to inoculate 100 ml M9/Amp medium and grown overnight at 37°C. The overnight culture was added to 2 liters of M9/Amp medium in a VirTis benchtop fermentor. This culture was grown at 37°C with stirring at 600 rpm until the A 600 was 0.7-0.8. Isopropyl-␤-D-thiogalactoside was added to a concentration of 1 mM to initiate induction of the fusion protein. After 3 h, cells were harvested by low speed centrifugation.
Nickel Affinity Chromatography-After resuspension in 40 ml of 20 mM Tris-HCl buffer (pH 7.9), cells were passed through a French pressure cell (SLM Instruments) at 15,000 p.s.i. Inclusion bodies were isolated by centrifugation, resuspended in "binding buffer" (6 M guanidine-HCl, 0.5 M NaCl, 5 mM imidazole, and 20 mM Tris-HCl (pH 7.9)), and applied to a column containing Ni 2ϩ -charged His-Bind resin (Novagen) prepared according to the manufacturer's specifications. After washing the resin with 10 column volumes of binding buffer and 5 column volumes of wash buffer (6 M guanidine-HCl, 40 mM imidazole, 0.5 M NaCl, and 20 mM Tris-HCl pH 7.9), the ketosteroid isomerase fusion protein was eluted in 5 column volumes of elution buffer (6 M guanidine-HCl, 0.3 M imidazole, 0.5 M NaCl, and 20 mM Tris-HCl (pH 7.9)).
CNBr Cleavage-The fusion protein prepared as described above was dialyzed against water, and the insoluble fusion protein was isolated by centrifugation. The pellet was then resuspended in 20 ml of 80% formic acid in a round-bottom flask and mixed with 1 g of cyanogen bromide (CNBr). After flushing the solution with helium, the flask was sealed, and the reaction was allowed to proceed in the dark for 24 h. The reaction mixture was then diluted 1:1 with water and applied to a C 18 Sep-Pak cartridge (Waters). The peptide was eluted with 4 ml of 35% acetonitrile in water and dried using a SpeedVac (Savant). The dry peptide was resuspended in 50 mM sodium phosphate buffer (pH 6.0) at 37°C for 2 days to deformylate the products.
Reverse Phase-HPLC Purification-The peptide sample prepared as described above was diluted 1:1 with 0.1% trifluoroacetic acid (buffer A) and applied to a C 18 reverse phase Discovery column (Supelco). The peptides were eluted isocratically at 1 ml/min with buffer B (25% acetonitrile with 0.1% trifluoroacetic acid). Peak fractions were collected and then dried using a SpeedVac (Savant). Isolated peptides were analyzed by mass spectrometry (Yale Cancer Center Mass Spectrometry Resource and W. M. Keck Foundation Biotechnology Resource Laboratory). The disulfide form of the peptide with a C-terminal homoserine lactone was chosen for further structural analysis.
NMR Sample Preparation-The 15 N-labeled disulfide homoserine lactone form of the Torpedo ␣18-mer was resuspended in 50 mM perdeuterated potassium acetate buffer (pH 4.0) with 5% D 2 O and 0.05% sodium azide. Bgtx (from Sigma) was prepared in the same buffer at a concentration of 5 mM. Bgtx from this stock solution was added to the ␣18-mer to form a 1:1 Bgtx⅐␣18-mer complex. The final concentration of the Bgtx⅐␣18-mer complex was 2.1 mM. The "free Bgtx" NMR sample was diluted from stock Bgtx into the same buffer to a final concentration of 2.0 mM.
NMR Experiments-All NMR spectra were recorded on a Bruker Avance 600-MHz NMR spectrometer at a temperature of 35°C. Chemical shifts at this temperature were calibrated with respect to internal 3-(trimethylsilyl) tetradeutero sodium propionate (0.0 ppm). The formation of the Bgtx⅐␣18-mer complex was followed in a mole-ratio titration using a two-dimensional 15 N heteronuclear single quantum correlation ( 1 H-15 N HSQC) (24 -26) experiment. Amino acid spin systems were identified by two-dimensional total correlation spectroscopy (TOCSY) (24,25,27) and three-dimensional TOCSY-HSQC experiments (27-31) with a mixing time of 60 ms. The assignments of the NH protons and C ␣ H protons of the amino acid spin systems of the peptide were further confirmed by a three-dimensional HNHA experiment (32,33). The three-dimensional HNHA experiment provides the correlation between the 15 NH proton and the C ␣ H proton of the same amino acid; these data help confirm the identification of the NH and C ␣ H protons. Nuclear Overhauser effect (NOE) correlations (sequential, mediumrange, and long range NOEs) were identified by two-dimensional nuclear Overhauser enhancement spectroscopy (NOESY) (24,25) and three-dimensional NOESY-HSQC experiments (28 -31) with a mixing time of 120 ms. Spectra from these experiments were also collected at 25°C to facilitate the assignment of resonances. All NMR spectra were processed and analyzed with XwinNmr (Bruker), NMRPipe (34), and SPARKY (35).
In comparing our results with earlier, more preliminary assignments involving an unlabeled ␣18-mer peptide bound to Bgtx (36), we found that most Bgtx assignments are the same or similar (chemical shifts change less than 0.05 ppm) after calibration. However, the assignments of Val 2 , His 4 , Ser 9 , Ile 11 , Lys 26 , Cys 29 , Cys 33 , Val 40 , Lys 51 , Lys 52 , Lys 70 , Gln 71 , Arg 72 , and Gly 74 were significantly different; in most cases no comparable resonances were observed in our two-dimensional NOESY spectra. On the other hand, new resonance peaks appear elsewhere in the spectra. All these new resonances involving these residues were re-assigned based on sequential NOE connectivity. We think it most likely that the observed chemical shift differences between the two samples are caused by a difference in the ionic strength of the two samples even though both were prepared at pH 4.0 and spectra were acquired at 35°C. Our present sample is dissolved in 50 mM potassium acetate, whereas the earlier sample was simply adjusted to pH 4.0 with the addition of HCl. A change in the ionic environment could have significant effects on electrostatic interactions between side chains of charged residues, leading to changes in the chemical environment of a subset of spins. Similarly, it was necessary to re-assign most of the ␣18-mer peptide resonances. Of the eight previously assigned peptide residues, only Asp 195 and Thr 196 are unchanged between the two samples. However, the C ␣ H proton and C ␤ H proton of Thr 196 were erroneously assigned previously (36). The new swapped assignments incorpo-rate the results from the three-dimensional HNHA experiment.
Conformation Calculations-The cross-peak volumes in the twodimensional NOESY spectra were integrated by the Gaussian fitting protocol using SPARKY. The cross-peaks were classified into three categories: strong, medium, and weak, with corresponding distance ranges of 1.8 -3.0, 1.8 -4.0, and 1.8 -5.0 Å, respectively. The H N -H ␣ 3 J coupling constants of the ␣18-mer peptide were obtained from the three-dimensional HNHA experiment (37). The 3 J coupling constants were converted to dihedral angle restraints using previously described methods (38). For 3 J Ͻ 6 Hz, the dihedral angle restraint was assigned to Ϫ60°Ϯ 30°; for 3 J Ͼ 8 Hz, the dihedral angle restraint was Ϫ120°Ϯ 40°. All structures were calculated with distance geometry and simulated annealing protocols using the dg_sa.inp script of the NMR structure calculation program, CNSsolve (39). The following is the potential energy function used in these calculations: where F bon relates to bond length, F ang and F imp to bond angles, F vdw relates to the van der Waals repulsion term, F noe relates to NOE distance constraints, and F cdih relates to dihedral angles. Pseudoatoms were used for protons that could not be stereospecifically assigned. The pseudoatom correction feature of CNSsolve was used to adjust the NOE distance constraint range automatically. In each batch of calculations, a different random seed number was used to initiate the calculation of a set of 50 structures. From the pool of calculated structures, only those structures lacking any NOE violation (Ͼ0.5 Å) were selected as "acceptable" for further analysis. As a result of the weighting of the F noe term in CNSsolve, none of the other energy terms were as critical as F noe in determining an acceptable structure. In total, 120 acceptable structures of the Bgtx⅐␣18-mer complex were obtained from 6 batches of independent calculations (i.e. 300 total structures), and 122 acceptable structures of free Bgtx were obtained from 8 batches of independent calculations (i.e. 400 total structures). The 20 lowest-energy structures out of the acceptable structures for free Bgtx and for the Bgtx⅐␣18-mer complex were selected to form an ensemble of representative final structures. The mean structure corresponding to each ensemble was calculated by a program written by Dr. Christian Rölz (40). The two mean structures (free Bgtx and Bgtx⅐␣18mer complex) were further partially energy-minimized using DIS-COVER (Molecular Simulations, Inc.) to create representative structures complete with side chains. All structures depicted in Fig. 7 have been deposited into the Protein Data Bank, Research Collaboratory for Structural Bioinformatics. The four files corresponding to the mean and ensemble structures for the Bgtx⅐␣18-mer complex and for free Bgtx have been assigned the identifiers 1IDG, 1IDH, 1IDI, 1IDL.
We used Rasmol (41), MOLMOL (42), and INSIGHT II (Molecular Simulations, Inc.) for the graphical analysis of the calculated structures. The surface charge potentials were calculated using MOLMOL. The contact surface areas of all the final individual Bgtx⅐␣18-mer complex structures were calculated by contacts of structural units (CSU) using CSU software (43). The energetically significant cation-interaction analysis of the Bgtx⅐␣18-mer complex structures was performed using the CaPTURE program (44).

Preparation of 15 N-peptide and Its
Purification-To facilitate the assignment of the ␣18-mer peptide resonances while complexed with Bgtx, the peptide corresponding to Torpedo nAChR ␣1 subunit residues 181-198 was expressed heterologously in Escherichia coli as part of an insoluble fusion protein under conditions where 15 NH 4 Cl was used as sole source of nitrogen. After isolation of the fusion protein, CNBr cleavage at engineered Met sites was used to release the desired peptide, which contained the 18 residues of ␣1 subunit with an additional C-terminal residue derived from the engineered Met. As expected for CNBr cleavage of Met sites, the peptides isolated are a mix of the C-terminal homoserine form of the peptide and its corresponding dehydrated homoserine lactone form. HPLC analysis revealed three major peptide peaks which were further characterized (Fig. 1). Preliminary solid-phase binding studies indicated that all three peptide fractions bind Bgtx to an extent comparable with that obtained with a similar synthetic ␣18-mer peptide lacking the C-terminal homoserine (data not shown). All three peptide fractions were resistant to thiol alkylation with N-ethylmaleimide except after prior incubation of the peptide with dithiothreitol. These results suggest that the adjacent cysteines, Cys 192 and Cys 193 , are in the disulfide state in the isolated peptides. Mass spectrometric analysis revealed that P 2 corresponds to the C-terminal homoserine lactone form of the ␣18-mer, whereas P 1 is the C-terminal homoserine form. The P 2 peptide was chosen for the production of a Bgtx⅐␣18-mer complex.
The Formation of a Stoichiometric Bgtx⅐␣18-mer Complex-The pure ␣18-mer (P 2 ) was resuspended in 50 mM perdeuterated potassium acetate buffer (pH 4.0) and analyzed with a two-dimensional 1 H-15 N HSQC experiment that is designed to acquire signal only from protons bound to 15 N ( Fig. 2A). An equimolar amount of Bgtx was then added, and the sample was again analyzed by HSQC (Fig. 2B). A comparison of these spectra (Fig. 2) clearly demonstrates the formation of a stoichiometric complex between the ␣18-mer and Bgtx. The free peptide ( Fig. 2A) appears to be largely unstructured; all the NH resonances are poorly dispersed in chemical shift and vary in intensity. In contrast, after binding to Bgtx (Fig. 2B), nearly all the NH resonances undergo large chemical shift changes, and there is little evidence of any free peptide remaining based on the disappearance of resonances seen in the free peptide. These observations suggest that the ␣18-mer adopts a defined structure upon binding to Bgtx. Furthermore, in mole-ratio titration studies with less than stoichiometric concentrations of Bgtx, NH resonances corresponding to both the bound and the free peptide are present, and the chemical shift of the NH resonances for the bound peptide are fixed and do not vary with Bgtx concentrations (data not shown). These results indicate that the Bgtx⅐␣18-mer complex is in slow exchange.
NMR Assignments-Because only the peptide is 15 N-enriched, 15 N three-dimensional NMR experiments can be used to filter out Bgtx proton signals that are not correlated to 15 N. Making use of this enrichment, three-dimensional TOCSY-HSQC, NOESY-HSQC, and HNHA experiments were obtained to make preliminary amino acid assignments of the ␣18-mer in its bound form. Fig. 3 illustrates a representative strip analysis used to identify the resonances of Lys 185 . Three-dimensional TOCSY-HSQC analysis is used to identify the resonances correlated by through-bound scalar connectivity to the 15 NH (i.e. C ␣ H proton, C ␤ H proton, etc.). Using these three-dimensional NMR experiments, we assigned the observable resonances for all of the amino acid residues in the ␣18-mer except for the FIG. 1. C 18 -reverse phase HPLC purification of the ␣18-mer. The peptide peak labeled P 1 corresponds to the disulfide homoserine form of the ␣18-mer; P 2 is the disulfide homoserine lactone form of the ␣18-mer that was collected and used in this NMR study. P 4 is an unidentified form of the ␣18-mer, which slowly converts to P 2 . See "Experimental Procedures" for a detailed description of chromatographic conditions. N-terminal Tyr 181 , which has an exchangeable NH, the Cterminal homoserine lactone, whose mobility may cause its signals to be too weak to be identified, and the two prolines, which lack amide protons (Fig. 2B). These assignments of the peptide resonances greatly facilitated the assignment of the Bgtx resonances in the homonuclear two-dimensional NMR data obtained with the complex.
Based on our three-dimensional NMR assignments, our twodimensional NMR data (NOESY and TOCSY), and published Bgtx assignments (7, 36), we completed the assignment of the resonances obtained with the Bgtx⅐␣18-mer complex. Fig. 4 summarizes the C ␣ H i proton to NH iϩ1 proton NOEs (sequential NOE), the C ␤ H i proton to NH iϩ1 proton NOEs, and the NH i proton to NH iϩ1 proton NOEs in the Bgtx⅐␣18-mer complex used to complete the connectivity of the polypeptide backbone. In addition, 11 unambiguous intermolecular NOEs on the NH region of the peptide were assigned (Table I). Additional intermolecular NOEs in the C ␤ H proton and C ␥ H proton regions were observed but have not as yet been unambiguously assigned. We performed a similar analysis on free Bgtx to generate the structure important for comparison of the free and bound forms of Bgtx.
Three-dimensional Structure Calculations and Comparison-The distance constraints resulting from the NOEs and the dihedral angle restraints obtained from the H N -H ␣ 3 J couplings (Table II)  structure determination. These calculations utilized both distance geometry and simulated annealing protocols. Twenty structures of free Bgtx and of the Bgtx⅐␣18-mer complex with the lowest potential energy and no NOE violation larger than 0.5 Å are superimposed and shown as an ensemble in Fig. 5. In Fig. 6, the mean structure of each ensemble, partially energyminimized, is shown to illustrate some of the key structural features. The overall backbone atomic root mean square deviation (r.m.s.d.) between the individual structures and the mean structure of free Bgtx is 2.03 Å, whereas that of the Bgtx⅐␣18mer complex is 2.07 Å. The pairwise r.m.s.d. between the two mean structures is 6.05 Å.
We did not find any inter-residue NOE constraints involving the last two residues of Bgtx, Pro 73 and Gly 74 . If these two residues are omitted from the r. A comparison of free and bound Bgtx reveals that the proton resonances of a number of amino acids in Bgtx undergo large chemical shift changes upon peptide binding. These shift changes reflect significant alterations in the chemical environ-ment of those protons. We find that Ala 7 (C ␣ H), Ser 9 (C ␣ H, C ␤ H), Ile 11 (NH, C ␣ H, C ␤ H, C ␥ H), Trp 28 Fig. 5A. The changes in Ala 7 , Ser 9 , and Ile 11 suggest that the outer tip of loop I participates in binding the ␣18-mer. This is substantiated by the intermolecular NOEs observed between the C ␥ H and C ␦ H protons of Pro 10 and the NH protons of Asp 195 and Thr 196 in the peptide (Table I) (Table I). In contrast to these examples of shift changes correlated with intermolecular contacts, several other shift changes in Bgtx residues appear to be due to conformational changes involving possible reorientations about the central triple-stranded ␤-sheet common to all ␣-neurotoxins. We believe that the chemical shift changes in the Trp 28 -Asp 30 , Phe 32 , Cys 33 region and in Val 57 are caused by general movement about the ␤-strand involving loop II and the proximal portion of loop III. Such secondary effects could also explain the chemical shift change of the side-chain C ␤ H proton in Lys 52 of loop III. Finally the chemical shift changes involving the side chains of Lys 64 and Asn 66 at the beginning of the C-terminal tail may reflect a change in chemical environment due to the apparent relocation of the distal C-terminal region to make a contact with the ␣18-mer peptide bridging the gap between Bgtx loop I and loop II (Fig. 5).
As viewed in Fig. 6A, free Bgtx is oriented with loop I (blue) on the left, loop III on the right, and the tip of loop II (green) at the bottom. This view corresponds with the so-called "concave" surface of the ␣-neurotoxins (6,12). In the free form of Bgtx, the tip of loop II forms the lower rim of this concave surface, and the loop I and loop II of free Bgtx are well separated with no inter-loop NOE constraints. The distance between the N of Ser 34 and the N of Pro 10 in free Bgtx is ϳ31 Å. Upon peptide binding, both loop I and loop II interact with peptide residues (see below), as revealed by intermolecular NOEs (Table I). As a consequence of binding, loop I and II move closer to each other; the average distance between the N of Ser 34 and the N of Pro 10 narrows to ϳ24 Å. In addition, the tip of loop II switches to a convex conformation upon binding ( Fig. 6 and Fig. 7, A-D). The C-terminal tail region of free Bgtx is relatively unconstrained in free Bgtx but participates in peptide interaction in the bound state (see below). The rest of the structure appears to change little upon complex formation (Fig. 6B). To provide a full comparison of binding-induced structural changes in Bgtx, we have superimposed the mean backbone structures (Fig. 7, B and D) and the full ensemble of structures (Fig. 7, A and C) as stereo images. The red traces correspond to the backbone of free Bgtx, whereas the blue traces refer to the backbone of bound Bgtx. The ␣18-mer backbone is colored green. In the front view (Fig.  7, A and B), loop I and loop II move toward each other and toward the peptide, whereas loop III is largely unaltered. In the left-profile view, the ␣18-mer has been removed for a better view of the changes in the Bgtx backbone (Fig. 7, C and D). The C-terminal tail is shifted to the right (toward the peptide) in the bound state. The tip of loop II moves left, highlighting the change in general orientation in this region from a concave to a convex surface. We observed five pairs of C ␣ H-C ␣ H long-range NOEs in the two-dimensional NMR study of the ␣18-mer complex. These included the following pairs of residues: Cys 23 -Cys 44 , Tyr 24 -Cys 59 , Arg 25 -Leu 42 , Lys 26 -Val 57 , and Met 27 -Val 40 . These NOEs were previously reported as evidence for an antiparallel triple-stranded ␤-sheet in the solution structure of Bgtx (7), and similar ␤-sheet defining, inter-strand NOEs have been observed in the solution structures of other ␣-neurotoxins (12).
Binding Interactions-The structures indicate that Bgtx interacts with the ␣18-mer at three sites. These are the tip of loop I, the C-terminal tail region (intermolecular NOEs are obtained with Thr 8 , Pro 10 , and Pro 69 ), and loop II residues facing loop I including Val 39 , Val 40 , and Lys 38 . Although Loop III of some ␣-neurotoxins (3) has been reported to be involved in binding to the nAChR, we did not observe any direct intermolecular NOEs between the ␣18-mer and loop III of Bgtx. The chemical shifts of most residues in loop III change little upon binding of the FIG. 5. Structural comparison between free-Bgtx and the Bgtx⅐␣18-mer complex. A, twenty superimposed structures calculated based on distance constraints in free Bgtx. The red lines mark those Bgtx regions that undergo large chemical shifts upon ␣18-mer binding. B, twenty superimposed structures calculated based on distance constraints for the Bgtx⅐␣18-mer complex. The blue lines correspond to Bgtx, and the green lines signify the ␣18-mer. The black lines and the labels N and C mark the N and C termini of Bgtx. The labels NЈ and CЈ mark the N and C termini of the ␣18-mer. Only backbone traces are shown. The figures were prepared using the program MOLMOL (42).
␣18-mer, consistent with a lack of involvement of Bgtx loop III in ␣18-mer binding.
The ␣18-mer residues responsible for the intermolecular NOEs with the Bgtx sites are Tyr 189 , Tyr 190 , Thr 191 , Cys 192 , Asp 195 , and Thr 196 . The contact zone is ϳ18 Å in length (from N of Tyr 189 to N of Thr 196 ), and the total van der Waals contact surface areas range from 500 to 690 Å 2 , with a mean of 580 Å 2 . It is noteworthy that there is a cluster of four positive charged Bgtx residues (Arg 36 , Lys 38 , Lys 70 , Arg 72 ) near the contact zone. In close apposition in the complex are found peptide residues Trp 187 , Tyr 189 , Tyr 190 (Fig. 8). This arrangement suggests the possibility of cation-interactions (44) contributing energetically to the formation of the Bgtx⅐␣18-mer complex. A cation-interaction analysis carried out on the 120 Bgtx⅐␣18mer-calculated structures revealed that a total of 24 individual structures, including two from the 20 best structures, have energetically significant cation-interactions as determined by CaPTURE (44). In this analysis, cation-interactions are selected as energetically significant if the electrostatic energy is ϽϪ2.0 kcal/mol or, alternatively, if the electrostatic energy is ϽϪ1.0 kcal/mol and the van der Waals interaction for the pair is ϽϪ1.0 kcal/mol (44). As shown in Table III pairing was observed in 12 of these 24 structures, and one of these structures is shown in Fig. 9. DISCUSSION NMR studies of complexes formed between peptides and peptide binding domains have been very instrumental in many other systems in elucidating mechanisms of molecular recognition (45). In seeking to understand the structural basis for the relatively high affinity (K D ϳ 65 nM) binding observed between Bgtx and the ␣18-mer, we identified a number of NOE distance constraints that reveal important information about the interaction between the ␣18-mer and Bgtx and that also define conformational changes in Bgtx upon complex formation. These constraints have been used to generate the structural features presented in this study. The disc-like shape of Bgtx and of the complex limits the total number of long-distance NOEs available (NOEs between residues distant in primary sequence), and this in turn limits the final resolution of the structures obtained (Table II). This limitation is applicable to all members of the ␣-neurotoxin family (11,12). Finally, we compare the structural features of the binding between Bgtx and the cognate ␣18-mer to the biochemical-and mutagenesis-based observations made with the intact nAChR.
One of our most striking observations is the binding-associated re-orientation in several segments of Bgtx that normally are not in contact with one another. Loop I and the C-terminal tail segment alter their configuration with respect to the main body of the protein, and the tip of loop II undergoes a change in its relative curvature and shape (Fig. 7). Together, these changes suggest a considerable coordinated reconfiguration of Bgtx upon interacting with receptor sequences. Such a reconfiguration is consistent with the extensive flexibility that has been noted in these and other protein toxins. It has been suggested that such flexibility may have evolved to serve important functional purposes (e.g. Ref. 46).
We previously reported the solution structure of the complex formed between Bgtx and a dodecapeptide corresponding to amino acid residues 185-196 from the ␣ subunit of the Torpedo nAChR (22). The apparent affinity of Bgtx for the ␣12-mer is about 15-20-fold lower than for the ␣18-mer. In both structures, Tyr 189 and Tyr 190 lie in a similar position, close to Val 39 and Val 40 of Bgtx. In addition, in both structures the peptides are relatively elongated. There are significant differences, however, in the orientation of the two peptides relative to Bgtx. In the Bgtx⅐␣12-mer complex, the polypeptide backbone of the five identified amino acid residues is between loop I and loop II, in a position roughly parallel to loop II. In the Bgtx⅐␣18-mer complex, the peptide is in a more perpendicular orientation with respect to loop II and the tip of loop I (Fig. 6).
The Bgtx⅐␣12-mer complex may represent an intermediate  stage in binding with the aromatics of the peptide, Tyr 189 and Tyr 190 , forming a nucleation site for further interactions. Although the Bgtx⅐␣18-mer complex is more energetically favored, the total contact surface areas are very similar, ϳ 600 Å 2 , for the two complexes. A library-derived peptide selected for its ability to bind Bgtx with high affinity also contains two adjacent Tyr residues, and these residues contribute the largest contact area in the complex (47). Although this 13-residue library-derived peptide adopts a more globular or sphere-like conformation, it is also found localized to the cleft formed between loop I and loop II with close apposition of the Cterminal region (47).
In the Bgtx⅐␣18-mer complex, the intermolecular NOEs between Pro 10 and both Asp 195 and Thr 196 and between Thr 8 and Cys 192 (Table I) demonstrate the involvement of Bgtx loop I in the formation of the peptide complex. This finding is consistent with mutagenesis studies of erabutoxin a, a short ␣-neurotoxin. Loop I mutations, S8T and Q10A, in erabutoxin a result in a very large reduction in binding affinity for the Torpedo nAChR (2). Recently, a double-mutant cycle analysis involving the related short ␣-neurotoxin, Naja mossambica mossambica I (NmmI), and the mouse nAChR has revealed an interaction between Ser 8 in NmmI and Tyr 198 on the ␣ subunit at the ␣␥ site (3). In contrast, no evidence of an interaction was observed between Ser 8 and Val 188 in the NmmI study. In addition, Val 188 can be energetically coupled to Arg 33 and Arg 36 in NmmI (19). Based on sequence alignment, the corresponding Bgtx residues would be Arg 36 and Lys 38 , respectively. These observations are entirely consistent with the structure of the Bgtx⅐␣18-mer complex; the tip of loop I is in closer proximity to the peptide residues C-terminal to Thr 196 (Tyr 198 is located at the extreme left in Fig. 6B), whereas Val 188 is removed from loop I and in close proximity to the loop II residue, Val 39 (Table I).
Mutational analysis of loop I residues in the long ␣-neurotoxin, ␣-cobratoxin from Naja kaouthia venom, failed to detect a significant role for this region in binding to the Torpedo nAChR (2). Because the sequence of the loop I region differs greatly between Bgtx and this ␣-cobratoxin and because loop I is two residues longer in Bgtx (12 versus 10 between the corresponding Cys residues delimiting loop I), it is possible that these two toxins differ in this region in their mode of interaction with the nAChR. In addition, recent comparisons of short and long ␣-neurotoxins suggest significant differences between these two families of toxins in their detailed mode of interaction with the nAChR (2,18).
The intermolecular NOE between Pro 69 and the backbone NH of Cys 192 together with the chemical shift changes observed within the C-terminal tail region upon formation of the Bgtx⅐␣18-mer complex clearly indicate that the C-terminal tail region of Bgtx also plays a role in peptide binding. This observation is consistent with biochemical and mutagenesis studies examining the role of the C-terminal tail region in binding to native nAChRs where binding affinity was decreased by 7-15fold when C-terminal residues were removed (4,48).
The proximity of the positively charged Bgtx residues (Lys 38 and Arg 36 ) with the aromatics of the peptide (Tyr 189 and Tyr 190 ) in the Bgtx⅐␣18-mer complex suggests an important functional contact. Mutagenesis studies in Bgtx and related toxins also point to important roles for Arg 36 and Lys 38 . Alasubstitution of Arg 36 in Bgtx leads to a 90-fold decrease in Bgtx binding affinity as measured with heterologously expressed mouse nAChR (4), whereas charge reversal studies in ␣-cobratoxin demonstrate that R33E (position corresponds to Arg 36 in Bgtx) causes a 767-fold decrease in binding affinity for Torpedo nAChR (2). Ala-substitution at Arg 36 , the position corresponding to Lys 38 in Bgtx, reduces binding affinity by 7.4-fold (2).
An important contact role for Tyr 189 is consistent with chimeric analysis and toxin footprinting studies of nAChR ␣ subunits. The ␣3 subunit, which shows no sensitivity to Bgtx block, acquires a significant sub-micromolar affinity for Bgtx with a single point mutation, K189Y, involving the introduction of an aromatic residue at position 189 (49). In a footprinting protection study using Cys-substituted mutations in the heterologously expressed mouse ␣1 subunit, the introduced thiol of ␣F189C is protected by Bgtx from reaction with a hydrophilic biotinylmaleimide (18). Similarly, the functional importance of Tyr 190 in ␣-neurotoxin binding is supported by mutagenesis studies where large decreases in Bgtx and NmmI toxin binding affinity are observed (3,18,19). A double-mutant cycle analysis with NmmI toxin revealed pairwise contacts between Tyr 190 and R33E and R36E, with a greater coupling energy to R36E (3,19). In the present study, we document an intermolecular NOE constraint between Lys 38 (position corresponds to Arg 36 in NmmI toxin) and Thr 191 , one residue removed from Tyr 190 in the Bgtx⅐␣18-mer complex.
Cation-interactions are interactions between a cationic group, such as the side chain of Arg and Lys, with the electronegativity of an aromatic cloud (44). Cation-interaction pairs within the Bgtx⅐␣18-mer complex were identified using CaPTURE (44). Of the 120 acceptable structures, 24 showed evidence of cation-interactions including two of the 20 best structures (Table III). One of these latter two structures was chosen to illustrate one such candidate cation-interaction (Fig. 9). In this example, the NZ nitrogen of Bgtx Lys 38 is oriented within 6 Å of the aromatic ring of Tyr 190 . We speculate that cation-interactions may be involved in Bgtx binding to the nAChR just as cation-interactions involving ␣Trp 149 may be important in the binding of acetylcholine to the nAChR (50). Additional high resolution structures with better resolution of the side-chain positions would be needed to test this proposal further.
In the Bgtx⅐␣18-mer complex, Cys 192 and Cys 193 of the peptide are located in between loop I and loop II and adjacent to the C-terminal tail of Bgtx. This position correlates well with recent biochemical cross-linking data concerning the spatial orientation of the reduced disulfide bond between Cys 192 and Cys 193 at the ␣␥ interface of the Torpedo nAChR (51). Several Cys-substituted mutants of Naja nigricollis ␣-neurotoxin, a short-chain neurotoxin, were cross-linked to the Torpedo ␣ subunit with various efficiencies depending on the spacer length of the dimaleimide derivative used and the site of toxin mutagenesis. It was concluded that Cys 192 and Cys 193 were located under the tip of the first loop, ϳ11.5 Å from the ␣-carbon at toxin position 10, and close to the second loop, ϳ 15.5 Å from the ␣-carbon at toxin position 33 (51). This orientation is entirely consistent with the structure shown in Fig. 6B.
In summarizing the contact information, we find that Bgtx interacts with the ␣18-mer primarily through three contact regions. Arg  in loop II appears to be a core binding site that is common to both the Bgtx⅐␣18-mer complex and the earlier Bgtx⅐␣12-mer complex. We also find that the tip of loop I is involved in peptide binding. Lys 70 and Arg 72 of the Cterminal tail provide the third binding region that serves to function as a physical continuum between the two other binding sites. As described above, there is a remarkable correlation between the conclusions drawn from various mutagenesis and biochemical cross-linking studies and the structures presented here for the Bgtx⅐␣18-mer complex. Although it is clear that additional receptor contacts are required to achieve the Bgtx binding affinity observed with the native receptor, the evidence to date nevertheless suggests that the ␣18-mer complex with Bgtx may serve as a useful and accessible model for studying the structural and energetic basis of the high affinity toxinreceptor interaction.