NMR Structural Analysis of α-Bungarotoxin and Its Complex with the Principal α-Neurotoxin-binding Sequence on the α7 Subunit of a Neuronal Nicotinic Acetylcholine Receptor*

We report a new, higher resolution NMR structure of α-bungarotoxin that defines the structure-determining disulfide core and β-sheet regions. We further report the NMR structure of the stoichiometric complex formed between α-bungarotoxin and a recombinantly expressed 19-mer peptide (178IPGKRTESFYECCKEPYPD196) derived from the α7 subunit of the chick neuronal nicotinic acetylcholine receptor. A comparison of these two structures reveals binding-induced stabilization of the flexible tip of finger II in α-bungarotoxin. The conformational rearrangements in the toxin create an extensive binding surface involving both sides of the α7 19-mer hairpin-like structure. At the contact zone, Ala7, Ser9, and Ile11 in finger I and Arg36, Lys38, Val39, and Val40 in finger II of α-bungarotoxin interface with Phe186, Tyr187, Glu188, and Tyr194 in the α7 19-mer underscoring the importance of receptor aromatic residues as critical neurotoxin-binding determinants. Superimposing the structure of the complex onto that of the acetylcholine-binding protein (1I9B), a soluble homologue of the extracellular domain of the α7 receptor, places α-bungarotoxin at the peripheral surface of the inter-subunit interface occluding the agonist-binding site. The disulfide-rich core of α-bungarotoxin is suggested to be tilted in the direction of the membrane surface with finger II extending into the proposed ligand-binding cavity.

The nicotinic acetylcholine receptor (nAChR) 1 (1) is a ligandgated ion channel that mediates excitatory transmission at the neuromuscular junction and at synapses in the central and peripheral nervous systems. It is the most intensely studied member of the ligand-gated ion channel superfamily and serves as a model for understanding the structure and function of related ion conducting channels including glycine, ␥-aminobutyric acid type A, ␥-aminobutyric acid type C, and type 3 serotonin receptors. nAChRs are pentameric complexes that assemble in the membrane with 5-fold symmetry. Each subunit contains an N-terminal extracellular domain about 200 amino acids long followed by four membrane-spanning segments (M1-M4) with an intracellular loop of variable length between M3 and M4. The second transmembrane region from each subunit contributes to the formation of the wall lining the channel pore. In muscle and Torpedo electric organ, the subunit composition is (␣1) 2 ␤␥␦ and (␣1) 2 ␤⑀␦ in embryonic and adult tissue, respectively (for review, see Ref. 1). Neuronal nAChR subunits (␣2-␣10 and ␤2-␤4) apparently can assemble in various combinations giving rise to multiple receptor subtypes. In heterologous expression systems, particular subunit combinations form functional hetero-pentamers (e.g. (␣4) 2 (␤2) 3 ) whereas the ␣7-␣9 subunits form functional homo-pentamers (2,3).
Snake venom-derived ␣-neurotoxins bind the muscle-type and, in some cases, homo-pentameric neuronal nAChRs with high affinity (4,5) and are grouped into two families (4). Short chain neurotoxins have 60 -62 amino acids. Long chain toxins have 66 -74 amino acids and a fifth disulfide at the tip of the second loop. Solution NMR and x-ray crystallographic studies (e.g. Refs. 6 -8) show that all ␣-neurotoxins share a tertiary structure known as the three-finger fold, a four-disulfide globular core from which emerge three loops or fingers and a C-terminal tail. An NMR dynamics study of a short ␣-neurotoxin reveals "disorder" at the tips of fingers I and II reflecting localized mobility on the picosecond to nanosecond time scales (9). The significance of this mobility is highlighted by evidence that mutation of residues at these locations in other related toxins produces large effects on binding (11).
␣-Bungarotoxin (Bgtx), a long neurotoxin from the venom of Bungarus multicinctus, has been an important tool in many biochemical and functional studies of nAChRs including the homo-pentameric nAChRs. The three-dimensional structure of Bgtx, determined by x-ray crystallography (12), differed from solution and x-ray structures of related short and long toxins particularly in the length of the highly conserved central ␤-sheet in finger II (6 -8). In contrast, an NMR-based investigation of the secondary structure in Bgtx indicated that the structure of Bgtx in solution most likely resembles that of other ␣-neurotoxins (13). Although this study paved the way to a better understanding of the solution structure of Bgtx, an adequately high resolution structure of Bgtx has not been available until now. The high resolution three-dimensional structure of unbound Bgtx presented here greatly facilitates the investigation of conformational changes that Bgtx undergoes as it binds nAChR-derived and engineered sequences, the subject of recent structural studies (14 -20).
Photolabeling and mutagenesis studies indicate that the ligand-binding site on the nAChR is formed at subunit interfaces in the N-terminal extracellular region of the receptor (1). The major structural determinants of ligand binding are found on the ␣ subunit with additional contributions made by residues of adjoining subunits. Strongly conserved residues from three discontinuous regions of the ␣ subunit (designated loops A-C) contribute to the binding pocket. In the crystal structure of the acetylcholine-binding protein (AChBP), a snail homologue of the extracellular domain of a homo-pentameric nAChR, loops A-C contribute to a relatively hydrophobic cavity at the subunit interface (21). Earlier studies (22) have shown that the major determinants of Bgtx binding lie between positions 173 and 204 on the ␣1 subunit, coincident with loop C. This region includes the highly conserved aromatic residues Tyr 190 and Tyr 198 and, in addition, Cys 192 and Cys 193 (in ␣1 numbering) which form an uncommon disulfide. Synthetic peptide binding studies have suggested that the Bgtx-binding site on the ␣7 subunit is in the homologous region (␣7 178 -196) (23). The importance of this region in toxin binding was further highlighted by the observation that six residues from this region of ␣7 could confer Bgtx sensitivity when placed into the corresponding positions of the Bgtx-insensitive ␣3 subunit (24). Additionally, mutation of loop C residues, Tyr 187 and Tyr 194 of ␣7, to Phe reduced Bgtx blockade of ACh-evoked currents in heterologously expressed receptors (25). This is consistent with a 60-fold reduction in Bgtx binding observed in synthetic peptides where Tyr 190 of ␣1 (which corresponds to Tyr 187 of ␣7) was mutated to Phe, showing the important role of Tyr 190 in binding Bgtx (26).
Previously, we reported NMR solution structures for two ␣1 subunit-derived peptides, an ␣12-mer (␣185-196) and an ␣18mer (␣181-198), in complex with Bgtx (14 -16). Here we extend our analysis of the Bgtx-binding site to the corresponding 19amino acid segment on the neuronal nAChR ␣7 subunit (␣7 178 -196) with an original solution NMR structure of the ␣7 19-mer in complex with Bgtx. Our primary goals are to determine an energetically favorable conformation for a region of the neuronal nAChR important in agonist and antagonist binding and to understand the structural basis for the strong interaction between Bgtx and the ␣7 nAChR. In addition, we present a new high resolution analysis of the secondary structure-rich core of Bgtx, significantly extending the initial NMR structural studies of this toxin (13).

EXPERIMENTAL PROCEDURES
Expression and Purification of the ␣7 19-mer-We prepared a synthetic gene encoding residues 178 -196 (IPGKRTESFYECCKEPYPD) of the chick neuronal nAChR ␣7 subunit (27) using mutually priming oligonucleotides. The oligonucleotides were designed according to the specifications of the pET-31 Peptide Expression System (Novagen) with 3Ј overhangs encoding methionine (28). The sequences of the two oligonucleotides are 5Ј-ATTCCGGGCAAACGTACCGAAAGCTTCTATGA-ATGCTGCAAAGAACCGTATCCGGATATG-3Ј and 5Ј-ATCCGGATAC-GGTTCTTTGCAGCATTCATAGAAGCTTTCGGTACGTTTGCCCGGA-ATCAT-3Ј. Two copies of this expression cassette were ligated in tandem into pET-31b(ϩ), and the construct was authenticated by DNA sequencing in the forward and reverse directions. The ␣7 19-mer was expressed as a ketosteroid isomerase fusion protein with a C-terminal polyhistidine tag in Escherichia coli BL21(DE3)pLysS cells (Novagen). Isotopically labeled ␣7 19-mer was produced in E. coli using M9 minimal medium with ( 15 NH 4 ) 2 SO 4 and 13 C 6 -glucose (Cambridge Iso-tope Laboratories) as the sole sources of nitrogen and carbon, respectively (29). The medium was supplemented with vitamins according to Ref. 30. Detailed methods for expression and purification of the ␣7 19-mer have been reported previously (16) for a similar nAChR recombinant peptide. Briefly, ketosteroid isomerase-␣7 19-mer fusion protein was isolated from resolubilized inclusion bodies by nickel affinity chromatography. The fusion protein was then treated with CNBr and the reaction products separated by RP-HPLC. ␣7 19-mer was isocratically eluted from a semi-preparative C 18 RP-HPLC column (Vydac) at 5 ml/min in 20% acetonitrile, 0.1% trifluoroacetic acid. Mass spectrometric analysis of HPLC fractions identified purified ␣7 19-mer (HHMI Biopolymer/Keck Foundation Biotechnology Resource Laboratory, Yale University School of Medicine). The redox status of the adjacent cysteine residues, Cys 189 -Cys 190 , was determined by RP-HPLC analysis of N-ethylmaleimide-treated ␣7 19-mer in its condition after purification and following pretreatment with dithiothreitol. Typical yields of isotopically labeled ␣7 19-mer were 2-4 mg/liter. Purified peptide was lyophilized and stored at Ϫ20°C.
Binding Experiments-The K D value for the ␣7 19-mer-Bgtx interaction was determined by measurement of inhibition of the initial rate of Bgtx binding to nAChR-enriched Torpedo membranes following preincubation of Bgtx with ␣7 19-mer (26). 125 I-Labeled Bgtx (2.5 nM) was incubated over a wide range of concentrations of ␣7 19-mer in 0.2% bovine serum albumin, 30 mM sodium phosphate, pH 7.4, for 18 h at room temperature. 100 l of the peptide/toxin mixture was added to microtiter plate wells coated with 2 g of Torpedo membranes that were pre-blocked with 200 l of 2% bovine serum albumin for 1 h. Following a 5-min incubation, the binding reaction was aspirated, and wells were washed 4 times with 200 l of 0.2% bovine serum albumin, 30 mM sodium phosphate, pH 7.4. Torpedo membrane-bound 125 I-Bgtx was measured in a ␥ counter. All measurements were done in triplicate. The effect of pH was also assessed by preparing samples in 30 mM sodium phosphate, pH 5.5, to replicate the buffer conditions of the NMR sample (see below).
NMR Sample Preparation-Hydrated Bgtx (Sigma) was used to resuspend the ␣7 19-mer in order to facilitate formation of a 1:1 peptidetoxin complex. Uniformly 15 N-labeled and uniformly 15 N, 13 C-doublelabeled ␣7 19-mer⅐Bgtx samples were prepared at a concentration of 1.5-2.0 mM. These samples contained 30 mM sodium phosphate and 50 M sodium azide in 95% H 2 O, 5% D 2 O at pH 5.5. Sodium 3-trimethylsilylpropionate (Cambridge Isotope Laboratories) was added at 50 M as an internal calibration standard. For deuterium exchange experiments, the complex was lyophilized and reconstituted in 99.9% D 2 O at pH 5.5 (isotope effect unaccounted). Preparation of the free Bgtx samples was described previously (13).
NMR Experiments-NMR experiments were carried out at 1 H frequencies of 400 and 600 MHz on Bruker Avance spectrometers at Brown University and at 500 MHz on a GE spectrometer equipped either with a GN console with a Nicolet computer or an Omega console with a Sun 3/160 computer at the University of California, San Francisco.
For free Bgtx, the following spectra in H 2 O were acquired at 15, 25, and 35°C, and pH 5.79: DQF-COSY (31), TOCSY (70 ms MLEV-17 spin-lock sequence) (32), and NOESY spectra (160-ms mixing time) (33) using the water suppression scheme described in Basus (34). No significant shifts in free Bgtx proton resonances were observed between samples at pH 5.5 and 5.79, suggesting that the chemical environment is comparable at both pH conditions. In addition, NOESY (160-ms mixing time) and E-COSY spectra (35) at 25 and 35°C in D 2 O were acquired.
Distance restraints were calculated from experimental NOESY intensities using the program MARDIGRAS version 2.0 (36, 37) which uses the complete relaxation matrix to produce an upper and lower distance bound for each experimental intensity. To make use of this procedure that determines accurate distances from the integrated intensities of the NOESY cross-peaks, a value for the rotational correlation time, c , must be defined. For free Bgtx the correlation time was determined by measurement of the 13 C T 1 and T 2 relaxation times at natural abundance, using a sample dissolved in 99.96% D 2 O. These experiments were carried out using a double-DEPT technique with proton detection for maximum sensitivity. For T 1 , we used the double-DEPT sequence with inversion recovery (38), and for T 2 , we used the double-DEPT sequence with a Carr-Purcell-Meiboom-Gill modification using a series of 180°pulses with a repetition rate of 1 ms to replace the single 180°refocusing pulse in the sequence of Nirmala and Wagner (39). The sum of the resonances at different portions of the spectra was used to determine the relaxation times. The data were analyzed by fitting to a single exponential decay function. Peak volumes from D 2 O and H 2 O NOESY spectra were obtained by fitting the peak or peaks to be integrated to a Gaussian line shape in Sparky. In the case of peaks with low signal to noise ratio, the points within a manually selected rectangular or elliptical area surrounding the cross-peak were summed using Sparky (40). In NOESY spectra the 3 J HN␣ coupling constants were determined by line fitting as described above, using Sparky. These values were used as the minimum values, because the apparent coupling constant in NOESY spectra will be smaller than the actual coupling constant. In DQF-COSY spectra, coupling constants were determined by line fitting the antiphase multiplets, followed by measurement of the separation between the simulated anti-phase multiplets (35). These values were used as maximum values, because the apparent coupling in DQF-COSY spectra is larger than the actual coupling constant due to the summation of anti-phase cross-peaks.
Exchange rates were measured at 25°C for Bgtx lyophilized from H 2 O and re-dissolved in D 2 O. Immediately following this procedure, the sample was placed in the spectrometer, and one-dimensional and TOCSY spectra (40 ms mixing time) were alternately acquired. The first one-dimensional spectrum was acquired 14 min after dissolving the lyophilized powder in D 2 O, and the first TOCSY spectrum was started 1 min later. Several spectra were obtained up to 36 h after starting the exchange, and further spectra were obtained 1 and 2 weeks later with the sample maintained at 25°C. The cross-peaks in the resulting spectra were integrated, and cross-peaks between ␣and ␤-protons were integrated for use as intensity references to eliminate variations of the spectrometer conditions after the sample was removed and re-inserted in the magnet for the last time points.
Structure Calculations-For free Bgtx, distance geometry calculations were performed on the Cray Y-MP at the San Diego Supercomputer Center using the distance geometry program VEMBED (47), a vectorized version of EMBED (48). Restrained molecular dynamics calculations were carried out with the four-dimensional modifications (49) to the GROMOS-87 programs (50) using a SparcIPX work station. The 37D4 force field was used, and all the calculations were performed in vacuo with all charged groups neutralized and with the mass of the hydrogens set to 12 atomic mass units. The distance restraint constant, K dis , was set to 10,000 kJ mol Ϫ1 nm Ϫ2 , with an initial temperature of 800 K. Three ps of dynamics was run followed by 5 ps during which the three-dimensional projection force constant K 3d was increased from 0 to 5,000 kJ mol Ϫ1 nm Ϫ2 . At the same time, the temperature was slowly lowered in an annealing procedure over the next 11 ps with the final temperature set to 0.1 K. The time constant tc for coupling to the thermal bath was set to 0.005 ps.
For the ␣7 19-mer⅐Bgtx complex, distance constraints were derived from a two-dimensional homonuclear NOESY experiment (120-ms mixing time; 35°C). NOEs were manually classified as strong, medium, or weak according to intensity. The distance ranges corresponding to the categories are as follows: 1.8 -3.0, 1.8 -4.0, and 1.8 -5.0 Å, respectively. When stereospecific assignment of methylene and methyl protons was not possible, a pseudoatom correction was employed. Slowly exchanging amide protons involved in a network of NOEs characteristic of ␤-sheets were identified as hydrogen bond donors and assigned the following constraints for residues (i,j): H N i -O j 1.6 -2.5 Å and N i -O j 2.5-3.3 Å. Dihedral angle constraints were based on 3 J HN␣ coupling constants calculated from HNHA data (42). For 3 J HN␣ Ͼ 8 Hz, was restrained to Ϫ120 Ϯ 40°; for 3 J HN␣ Ͻ 6 Hz, was restrained to Ϫ60 Ϯ 30°. The peptide bond between adjacent cysteines involved in a disulfide deviates from the typical trans configuration (51). Accordingly, the dihedral angle between Cys 189 and Cys 190 was unrestrained to enable calculation of an energetically favorable conformation. Structures were calculated starting from a random conformation model using the distance geometry/simulated annealing program in CNSsolve (52). Distance restraints were used with a square-well potential. The force constant on the NOE restraints was 50 kcal mol Ϫ1 and on the dihedral restraints 200 kcal mol Ϫ1 radians Ϫ2 . The van der Waals energy was represented by a repel function whose force constant varied during the cooling stage from 0.003 to 4 kcal mol Ϫ1 Å Ϫ4 . From multiple rounds of calculations, each initiated with a random seed number, acceptable structures were defined as those with no NOE violations exceeding a 0.5-Å cutoff. The 10 lowest energy structures from a pool of over 100 accepted structures were selected for calculation of an energy-minimized average structure using the accept.inp program of CNSsolve. The average structure was further energy-minimized using 100 steps of steepest descents in DISCOVER (Molecular Simulations, Inc.) to remove averaging artifacts.
The structures were visualized in MOLMOL (53). Intermolecular contact surface areas of the ␣7 19-mer⅐Bgtx complex were calculated by Contact of Structural Units software (54), and hydrogen bond analysis was performed by the DSSP program (55). All structure coordinates for Bgtx and the ␣7 19-mer⅐Bgtx complex were deposited in the Research Collaboratory for Structural Bioinformatics Protein Data Bank. The ensemble structures of Bgtx have been assigned the identifier 1KFH and the ensemble structures and minimized average structure of the ␣7 19-mer⅐Bgtx complex 1KC4 and 1KL8, respectively.

RESULTS
Determination of the Structure of Bgtx-Integrated volumes from cross-peaks in NOESY spectra of Bgtx were used in the program MARDIGRAS (36,37) to calculate accurate distance constraints taking into account spin-diffusion effects. These calculations were carried out using a correlation time of 3.3 ns and were repeated at 3.7 and 2.9 ns to include the range of uncertainty in the correlation time determination. Based on the dipole-dipole relaxation mechanism due to the directly attached proton, the correlation time was determined from 13 C T 1 and T 2 relaxation times. Indirect measurement of the 13 C T 1 relaxation time yielded an average value of 0.6 Ϯ 0.06 s for the ␣-carbons; the 13 C T 2 relaxation time was 73 Ϯ 7 ms. Extended starting structures were used in the distance calculations initially, and subsequent calculations began with structures obtained in the previous round of calculations. The largest distances thus calculated were used as the upper bound and the lowest as the lower bound for each distance restraint.
Initial structure calculations revealed a triple-stranded ␤-sheet whose hydrogen bonds were confirmed by H/D exchange experiments (Table I and see below). The nine most unambiguous hydrogen bonds were used as additional restraints in the final structure calculations, which included 588 NOE restraints. Stereospecific assignments were obtained for 10 diastereotopic pairs of ␤-protons. From the H N -H ␣ coupling constants, 43 angle constraints were obtained, and from E-COSY spectra ␣Ϫ␤-proton coupling constants were obtained to yield constraints for 17 1 angles. The relevant structural statistics are shown in Table II. From 20 distance geometry structures calculated, 13 were selected that had the correct global fold, as indicated by the lower energies and smaller number of distance violations. To overcome local barriers in the restrained molecular dynamics calculations, the four-dimensional version of GROMOS (49) (Fig. 1). Two regions with regular secondary structure are well defined, the ␤-sheet in finger I (residues 1-16) and the triple-stranded ␤-sheet formed by residues 22-28, 39 -45, and 56 -60. The poorly defined regions have few NOEs, and in particular, finger II has sequential NOEs that are smaller than expected, indicating the possibility of additional local motion in this region. Those regions poorly defined in free Bgtx are better defined in the ␣7 19-mer⅐Bgtx complex as indicated by the lower overall r.m.s.d. (Table II and see below).
H/D Exchange Rates for Free Bgtx-H/D exchange in a sample of lyophilized Bgtx dissolved in D 2 O at 25°C and pH 4.0 was followed using TOCSY and one-dimensional spectra. A total of 30 amide protons were observed to be in slow exchange (Table I). The experimentally determined set of slow exchanging amide protons that denote potential hydrogen bond donors was compared with the set of predicted hydrogen bonds calculated by DSSP from the Bgtx coordinates (Table I). DSSP predicts hydrogen bonding based on the distance between hydrogen-bonding partners and their orientation. In general, there is good agreement between the hydrogen bonds detected in the structures and the amide protons slowly exchanging with deuterium in D 2 O. The only slowly exchanging amide proton with a long exchange time that did not form a hydrogen bond in the structures is Thr 47 -H N . Only two hydrogen bonds detected in all structures with low hydrogen bond energy, which were not slowly exchanging, occur between Glu 56 and Met 27 and between Asp 30 and Gly 37 . Mobility at the tip of finger II may explain this observation as both are found at the end of the triple-stranded ␤-sheet in finger II.
Expression and Purification of Metabolically Labeled ␣7 19mer-Homonuclear NMR studies of Bgtx complexed with unlabeled synthetic peptides derived from the nAChR ␣1 subunit resulted in incomplete assignment of peptide proton reso-  nances because of signal overlap (14,15). This precluded a structure determination of the entire peptide sequence and a complete description of the peptide/toxin interface. To overcome this problem, we adopted a recombinant approach to produce the ␣7 19-mer uniformly labeled with the half-spin isotopes 15 N and 13 C for heteronuclear NMR studies. The ␣7 19-mer was expressed in E. coli as a fusion protein (see "Experimental Procedures"). The fusion protein precipitated in inclusion bodies that were isolated and resolubilized for nickel affinity purification using the C-terminal polyhistidine tag of the fusion protein. Purified fusion protein was treated with CNBr to liberate the ␣7 19-mer from its fusion partner and from the polyhistidine tag, taking advantage of engineered methionine residues flanking the ␣7 19-mer sequence. The ␣7 19-mer was purified by RP-HPLC. As expected of CNBr-digested proteins, two related and interconvertible peptides were obtained differing only at their C terminus (56). As confirmed by mass spectrometric analysis, one contains a homoserine and the other has the dehydrated homoserine lactone. In solidphase competition binding studies, there was no difference between the recombinant ␣7 19-mer and a synthetic ␣7 19-mer peptide lacking a C-terminal homoserine. Cys 189 and Cys 190 readily form a disulfide in the purified peptides as evidenced by resistance to N-ethylmaleimide alkylation and susceptibility to N-ethylmaleimide alkylation following treatment with dithiothreitol. Both ␣7 19-mer HPLC peak fractions were combined to prepare NMR samples as homoserine and homoserine lactone are in pH-dependent equilibrium (56).
Affinity of the ␣7 19-mer⅐Bgtx Complex-The affinity of the ␣7 19-mer for Bgtx was determined in a solution-based assay that is rooted in the finding that synthetic peptides derived from the nAChR compete with intact Torpedo nAChR for binding Bgtx (24). Following equilibration of 125 I-labeled Bgtx with varying concentrations of the ␣7 19-mer, the percentage of Bgtx that remained unbound was determined by the amount of binding to the nAChR under initial rate conditions. The concentration of ␣7 19-mer that diminished the initial rate of binding by 50% (K D ) was 30 M (data not shown). The K D value was identical at pH 7.4 and pH 5.5.
Stoichiometric Interaction between the ␣7 19-mer and Bgtx-CD studies of ␣7 19-mer alone in solution showed that the peptide structure is random coil (data not shown). Hence, we did not undertake NMR studies of free ␣7 19-mer and proceeded directly to determine its structure in complex with Bgtx. The stoichiometry of the ␣7 19-mer-Bgtx interaction was determined by titrating U-15 N-labeled ␣7 19-mer with Bgtx and monitoring chemical shift changes in 1 H-15 N HSQC spectra. These NMR experiments track changes in the chemical environment of peptide amide protons as Bgtx is incrementally introduced. In the absence of Bgtx, the peptide amide proton resonances lie within a narrow chemical shift range, indicative of a lack of secondary or tertiary structure ( Fig. 2A). Upon addition of Bgtx, virtually all the amide proton resonances undergo a change in chemical shift (Fig. 2, B and C). New peaks, corresponding to bound peptide, appear in previously unpopulated environments. The intensities of the "bound" peaks progressively increase as the added toxin approaches equimolar ratio leading to the conclusion that ␣7 19-mer and Bgtx form a 1:1 complex (Fig. 2, C and D).
Discrete sets of peaks corresponding to free and bound ␣7 19-mer indicate that the peptide-toxin complex is in slow exchange (57). This is further supported by the observation that the bound peak line widths and chemical shifts do not change upon incremental addition of Bgtx.
Assignment of ␣7 19-mer and Bgtx Resonances-We used U-15 N-labeled ␣7 19-mer to distinguish unambiguously peptide resonances from those of Bgtx in the NMR spectrum fingerprint region. Three-dimensional TOCSY-HSQC and NOESY-HSQC spectra were collected to assign 15 N-correlated resonances using the sequential resonance assignment strategy of Wü thrich (58). Additionally, correlation of ␣7 19-mer intraresidue amide and ␣-protons was confirmed in an HNHA experiment. Of the 20 amino acids of the peptide, only 16 were assignable by the 15 N-edited experiments because the N-terminal NH is rapidly exchangeable and three residues are prolines with no amide proton. The sequential NOE assignments are summarized in Fig. 3.
Assignment of ␣7 19-mer resonances by heteronuclear NMR methods greatly simplified the assignment of Bgtx as several peptide resonances overlapped with those of Bgtx in homonuclear spectra. Bound Bgtx resonances were assigned by the sequential assignment strategy of Wü thrich (58) using twodimensional homonuclear TOCSY and NOESY data (Fig. 3).
Structure Calculations-For the ␣7 19-mer⅐Bgtx complex, 671 NOE-derived distance constraints were used in distance geometry and simulated annealing calculations, including 548 intra-toxin, 95 intra-peptide, and 28 intermolecular constraints. Five HNHA-derived angle restraints for the ␣7 19-mer were introduced, as well as 20 hydrogen bond constraints within Bgtx based on assignment of hydrogen bond donors determined by deuterium exchange. The 10 lowest energy structures that fit the experimentally derived constraints  Structure of ␣7 19-mer-bound Bgtx-The overall ␣-neurotoxin three-finger fold is preserved in ␣7 19-mer-bound Bgtx, as shown in Fig. 4. This view of the complex presents the concave surface of Bgtx with finger I on the left and finger III on the right. The central three-stranded ␤-sheet that is conserved among ␣-neurotoxins is evidenced by a network of long range H ␣ -H ␣ and H ␣ -H N NOEs, as well as 10 slowly exchanging amide protons. Similarly, a two-stranded antiparallel ␤-sheet in finger I is observed in many calculated structures. Notably, ␣7 19-mer binding induces chemical shift perturbations greater than 0.2 ppm in Bgtx resonances for at least 15 residues consistent with their involvement in intermolecular contacts or conformational rearrangements (Table III). Fig. 5A illustrates the conformational changes that occur in Bgtx upon ␣7 19-mer binding. In comparison to unbound Bgtx, the tip of finger I in bound Bgtx is extended and parallel to the central ␤-sheet, owing to intermolecular contacts in this region (Table IV). A hydrogen bond between Ser 9 -O and Ile 11 -H N , absent in the uncomplexed form, stabilizes the tip of finger I in 8 of the 10 ensemble structures. The flexible tip of finger II undergoes two significant and stabilizing conformational changes upon binding. The fifth disulfide loop (Cys 29 -Cys 33 ) adopts a convex conformation, whereas Gly 34 -Lys 38 move closer to the ␣7 19-mer, placing Arg 36 and Lys 38 in position to make intermolecular contacts. These qualitative observations are quantitated in a comparison of the calculated per residue r.m.s.d. for free and ␣7 19-merbound Bgtx using the respective average structures (Fig. 5B).
Three-dimensional Structure of the ␣7 19-mer-The ␣7 19- mer exhibits no regular secondary structure upon binding Bgtx according to the NMR data as no long range intrapeptide backbone NOE networks nor slowly exchanging peptide amide protons characteristic of ␣-helical or ␤-sheet structure were observed. Rather, the ␣7 19-mer is primarily constrained by contacts with Bgtx over a 9-residue stretch from Phe 186 to Tyr 194 (see below) and long range intrapeptide NOEs involving backbone proton interactions between Ser 185 and Tyr 194 and between Phe 186 and Tyr 194 . We also observe long range intrapeptide ring proton NOEs between Tyr 187 and Tyr 194 . In addition, medium range NOEs involving Tyr 187 /Cys 189 and Glu 188 /Cys 190 constrain the ␣7 19-mer describing a hairpin-like turn about the vicinal disulfide. Only sequential NOEs are observed for residues flanking the Ser 185 -Tyr 194 region, leaving the two ends of the ␣7 19-mer unconstrained.
We find that the sequential NOEs connecting Cys 189 and Cys 190 are characteristic of a trans peptide bond conformation. We observe strong d ␣N and d NN NOEs typical of a trans conformation, and there is no evidence for d ␣␣ and d N␣ NOEs that would be expected for a cis conformation (58).
The ␣7 19-mer⅐Bgtx Interface-An extensive network of intermolecular NOEs ranging from Phe 186 to Tyr 194 on the ␣7 19-mer positions the ␣7 19-mer between fingers I and II of Bgtx on the concave face of the toxin and forms a contact area of ϳ650 Å 2 ( Fig. 4 and Table IV). The aromatic ring of Phe 186 alone is involved in contacts with Ala 7 , Ser 9 , and Ile 11 of the first finger in Bgtx (Table IV). Interestingly, we observe a 0.6 ppm upfield shift of the Ile 11 ␦ methyl protons (Table III). Our structures indicate that these protons are close to the ␦ ring protons of the Phe 186 aromatic ring, suggestive of a ring current-induced shift. A similar shift in the Ile 11 (H ␦1 ) 3 signal involving the corresponding residue Tyr 190 in ␣1 was observed in the ␣12-mer⅐Bgtx structure (14) and is thought to suggest a hydrophobic contact. More expansive contacts are made between the ␣7 19-mer and the side of finger II proximal to finger I involving Phe 186 -Glu 188 of the ␣7 19-mer and Lys 38 (59). NOEs that are characteristic of cationinteractions (60) are not observed between Bgtx and the ␣7 19-mer. The positive charge of Arg 36 points away from the ␣7 19-mer, suggesting that it may make contact with residues elsewhere in the intact ␣7 receptor (e.g. Trp 149 ). No intermolecular NOEs are observed involving Cys 189 and Cys 190 of the ␣7 19-mer, consistent with biochemical evidence that Bgtx binding does not depend on these residues (61). Finally, the N terminus of Bgtx is distant from the intermolecular contact zone, consistent with the observation of identical apparent affinities of N-terminal polyhistidine-tagged recombinant Bgtx and venom-derived Bgtx for both Torpedo and mouse muscle nAChRs (62). DISCUSSION Structure of Free Bgtx-The three-dimensional structure of Bgtx presented here builds on an earlier NMR structural study aimed primarily at determining elements of secondary struc-  ture (13). The present structure provides a high resolution view of the disulfide core and ␤-sheet regions that are the structuredefining features of three-finger proteins (6 -8 1IDL). Additionally, the new structure now makes apparent the structural disorder at the tip of finger II that is thought to be important in recognition of the nAChR, as illustrated in the ␣7 19-mer⅐Bgtx complex (see below). The Bgtx NMR structure is most similar to the x-ray structure of the long chain toxin, ␣-cobratoxin (Cbtx) (8), with an average backbone atom r.m.s.d. between Cbtx and the 13 Bgtx structures of 1.33 Å, for the segment of the well defined regions of Bgtx that are best matched to the corresponding regions in Cbtx (residues 1-6, 11-16, 22-28, 39 -48, and 54 -68 in Bgtx; residues 1-6, [9][10][11][12][13][14][19][20][21][22][23][24][25][36][37][38][39][40][41][42][43][44][45][51][52][53][54][55][56][57][58][59][60][61][62][63][64][65]. A comparison of the sequences reveals that finger I in Bgtx is two residues longer than in Cbtx. These additional residues appear as an extra bulge in finger I without changing the overall conformation of the finger. The sequences of Bgtx and Cbtx at the tip of finger II (residues 29 -38 in Bgtx; 26 -35 in Cbtx) are nearly identical with only two residues different out of ten. Although the overall structure of Bgtx suggests this region is disordered (Fig. 5A), a match of Bgtx residues 29 -38 indicates that it has elements of autonomous structure. Furthermore, the local structure between residues 29 and 38 agrees well with the x-ray structure of Cbtx (Fig. 6) with an average backbone r.m.s.d. of 1.74 Å. The local structure of the corresponding loop of the long neurotoxin LSIII from Laticauda semifasciata (9) also shows a similar structure, with an average backbone r.m.s.d. of 1.6 Å between Cbtx and LSIII for residues 26 -35 in both toxins. Like the Bgtx and LSIII solution structures, the NMR solution structure of Cbtx (7) also exhibits a large r.m.s.d. for finger II with respect to the overall structure, with a local structure for residues 26 -35 similar to its crystal structure. This suggests that the tip of finger II may move as a rigid body relative to the triple stranded ␤-sheet to which it is connected.
The ␣7 19-mer-Bgtx Interaction-The structure of the ␣7 19-mer⅐Bgtx complex reveals several important intermolecular contacts that form the structural basis for the high affinity interaction between nAChRs and ␣-neurotoxins. Previously, indirect structural information as determined in mutagenesis studies has helped identify ␣-neurotoxin binding determinants on the neuronal nAChR (24,63). The results presented here provide the first direct structural evidence for the contact zone involving an ␣-neurotoxin and a neuronal nAChR sequence. Additionally, the structure obtained may represent an energetically favorable conformation for a region of the nAChR important in agonist and antagonist binding and provide an understanding of differences between ␣-neurotoxin recognition of muscle-type and neuronal nAChRs.
The ␣7 19-mer on its own lacks stable secondary structure as demonstrated by its CD profile (data not shown) and narrow dispersion of its H N resonances ( Fig. 2A). Bgtx binding induces a transition whereby the ␣7 19-mer takes on a hairpin-like structure (Fig. 4) that is temporally stabilized, meaning that on the NMR time scale, the ␣7 19-mer does not rapidly exchange between the free and bound states (Fig. 2B). Such slow exchange kinetics, although often associated with complexes of higher affinity than the 30 M K D of the ␣7 19-mer (57), has also been observed in the interaction between cyclin-dependent kinase 2 and a fragment of its inhibitor p21 Waf1/Cip1/Sdi1 , which is characterized by an affinity in the 10 Ϫ5 M range (64). The p21 fragment undergoes a binding-induced transition from its native unfolded state to a folded bound state. Disorder-to-order folding transitions have been observed recently in other proteins that are involved in protein-protein or protein-nucleic acid interactions such as fibronectin-binding protein (65) and the transcriptional activation domain of the herpesvirus protein VP16 (66). The binding domains of these proteins are unfolded in the native state and form ␣-helical or ␤-hairpin structures upon interacting with their respective target mole-  cules. In thermodynamic terms, disorder-to-order transitions are accompanied by a large, negative change in conformational entropy. It is likely that the large entropic cost involved in stabilizing the ␣7 19-mer upon complex formation comes at the expense of association energy. A decrease in conformational entropy has been associated with increased specificity and weaker affinity in protein-protein and protein-DNA interactions (64,67). Indeed, the affinity of Bgtx for the ␣7 19-mer is relatively weak in solution. Interestingly, when the ␣7 19-mer is tethered to a solid surface, we observed a 250 nM IC 50 value in competition studies between 125 I-labeled Bgtx and unlabeled Bgtx (data not shown). The increased affinity observed in solidphase studies may be explained by conformational constraint of the ␣7 19-mer that lowers the entropic cost of complex formation of an otherwise random coil peptide. Consistent with this reasoning, the ␣18-mer (Fig. 8), which binds Bgtx with a K D of 65 nM (24) and shows no discrepancy between solution-based and solid-phase assays (data not shown), exhibits significant ␤-sheet secondary structure in CD studies. 2 Comparison of nAChR Peptide⅐Bgtx Complexes-The ␣7 19mer binds Bgtx between fingers I and II like all nAChR-derived peptides studied by solution NMR. This is illustrated in a comparison of the ␣18-mer⅐Bgtx (Research Collaboratory for Structural Bioinformatics Protein Data Bank code 1IDH) and ␣7 19-mer⅐Bgtx complexes (Fig. 7). The contact zone on the ␣7 19-mer, involving chiefly Phe 186 -Glu 188 and Tyr 194 , is consistent with a two-site model of the Bgtx-nAChR interaction within the major determinant of Bgtx binding (68). We find that Ala 7 , Ser 9 , and Ile 11 at the tip of finger I in Bgtx make extensive contacts with Phe 186 of the ␣7 19-mer, a strongly conserved residue among Bgtx-sensitive ␣-subunits ( Fig. 8 and Table IV). Similar intermolecular NOEs are observed between finger I residues of Bgtx and the homologous aromatic residue of the ␣12-mer, a phage display library-derived 13-mer (named LLPep) and two high affinity engineered 13-mers based on the library lead (named HAPep and HAP; Fig. 8) (14,17,18,20). Interestingly, mutation of finger I residues in the related long toxin Cbtx does not significantly affect its affinity for the ␣7 nAChR (63), possibly because finger I of Cbtx is two residues shorter than that of Bgtx. The importance of the length of the first finger for making contacts with the muscle-type nAChR is also reflected in short ␣-neurotoxins that are distinguished from long toxins by the increased length of their first finger. Mutations at the tip of the first finger of the short toxins erabutoxin a and Naja mossambica mossambica I (NmmI) cause a decrease in affinity for the Torpedo nAChR (11,69).
On finger II of Bgtx, at Val 39 and Val 40 , we find that multiple contacts are made with Phe 186 and Tyr 187 in the ␣7 19-mer. Our structural data confirm the importance of similar hydrophobic interactions observed between aromatic residues at ␣1 homologous receptor positions 189 and 190 and Val 39 and Val 40 of Bgtx in complexes with the ␣12-mer, ␣18-mer, LLPep, and HAPep (see Refs. 14 and 16 -18; Fig. 8). Taken together, these results indicate that Val 39 and Val 40 in Bgtx are important for binding multiple nAChR subtypes through interactions with these adjacent aromatic residues, highly conserved among Bgtx-sensitive ␣-subunits. As hydrophobic and, in particular, aromatic residues are directly involved in many protein-protein interactions (70), these residues may provide binding stability and specificity and may serve as a nucleating site for subsequent nAChR subtype-specific interactions.
The hairpin-like backbone trace of the structured region of the ␣7 19-mer (Ser 185 -Tyr 194 ) resembles the ␤9-␤10 hairpin in the homologous region of the AChBP (21). Whereas Bgtx-bound HAPep, HAP, and ␣p25mer, a peptide including ␣1 182-202 ( Fig. 8) (19), form a ␤-hairpin, we failed to observe any slowly exchanging amide protons or extensive long range intrapeptide NOE networks that would indicate a ␤-sheet component in the structure of Bgtx-bound ␣7 19-mer. In contrast with the abovementioned Bgtx-bound peptides, the turn that is observed in the corresponding region of the ␣18-mer (16) in its complex with Bgtx is more extended (Fig. 7). The ␣18-mer ends before Asp 200 (␣1 numbering), which is a residue that participates in formation of a ␤-sheet in the ␣p25-mer, apparently stabilizing the ␤-hairpin turn. The absence of Asp 200 in the ␣18-mer may help explain the lack of any NMR evidence to date for a ␤-sheet in the ␣18-mer. Interestingly, an unnaturally introduced constraint that generates a hairpin turn in the ␣18-mer results in no violation of the NMR-derived constraints for the ␣18-mer⅐Bgtx complex. 3 Differential Affinity of Short and Long Chain Neurotoxins for the ␣7 nAChR-Short neurotoxins bind the ␣7 nAChR with lower affinity than long neurotoxins, whereas both bind the muscle-type nAChR with equally high affinity (5). A key structural difference between the long and short toxins is a fifth disulfide located at the tip of finger II in long toxins. The necessity of a fifth disulfide for high affinity interaction with ␣7 was clearly demonstrated by disruption of this disulfide bridge in the long toxin Cbtx. Although reduced Cbtx bound the Torpedo nAChR with subnanomolar affinity, comparable with unmodified Cbtx, its affinity for the ␣7 nAChR was decreased 10 4 -fold (5). The importance of the fifth disulfide was later confirmed in mutational studies (63). Interestingly, NMR studies have shown that the end of finger II in ␣-neurotoxins is a dynamic entity (9, 10). Our structural analysis of the ␣7 19-mer⅐Bgtx complex reveals that the fifth disulfide loop (Cys 29 -Cys 33 ) moves as an independent unit upon binding to re-orient onto the convex face of Bgtx (Figs. 4 and 5A). The chemical shift perturbations from Cys 29 to Cys 33 and the appearance of at least four new long range intratoxin NOEs involving these residues support this conformational rearrangement. The conformational rearrangement induced by the ␣7 19-mer may therefore help explain the importance of the fifth disulfide in long toxin-␣7 interactions. Accompanying the movement of the fifth disulfide loop, there also is a change in curvature of the neighboring segment at the tip of finger II (Fig. 5A) (63).
A variety of mutations in Cbtx have been examined for effects on binding muscle-type and ␣7 nAChRs (63,71). It was concluded that Cbtx uses a common set of residues to recognize both receptor subtypes as well as non-overlapping sets for binding one or the other subtype. It was argued that Arg 36 in Cbtx (homologous to Lys 38 in Bgtx) is an ␣7-specific binding residue. This was somewhat surprising as Cbtx-Arg 36 lies on the convex face of the toxin, whereas toxin-nAChR interactions until then were thought to involve only residues on the concave toxin face. The ␣7 19-mer⅐Bgtx structure confirms the importance of the position occupied by Arg 36 in Cbtx and Lys 38 in Bgtx. Our structural data suggest a possible electrostatic interaction between Lys 38 on the convex face of Bgtx and Glu 188 of the ␣7 19-mer. Supporting the importance of this receptorsubtype interaction, a negatively charged residue at position 188 is conserved in ␣7 subunits across species but not in the muscle-type ␣1 subunit (Fig. 8).
Conformation of the Vicinal Disulfide Peptide Bond-A disulfide bridge between two adjacent cysteine residues in nAChR ␣ subunits is an unusual structural feature found in only a small number of other proteins, e.g. methanol dehydrogenase (72) and atracotoxin (51). The significance of this disulfide is underscored by its conservation among all nAChR ␣ subunits (1, 2). Moreover, biochemical evidence indicates that this disulfide is critical for channel activation. When the disulfide is reduced, affinity for acetylcholine is significantly diminished (73). Additionally, mutation in ␣1 of Cys 192 and Cys 193 to serine results in complete loss of acetylcholine binding, locking the channel in the closed state (74). Because of its proximity to the ligand-binding site and importance for proper channel function, the vicinal disulfide may act as a molecular switch, initiating conformational changes that lead to channel opening upon agonist binding (75). Interestingly, the disulfide is not necessary for Bgtx binding (61). The disulfide between adjacent cysteines forms a strained eight-membered ring. The energetic strain is shared in part by the peptide bond between the adjacent cysteines. The conformation of this peptide bond, cis or trans, has been the subject of numerous theoretical and experimental studies with conflicting results (e.g. Refs. 76 and 77). In a pentapeptide corresponding to positions 191-195 of the Torpedo ␣ subunit (TCCPD), cis and trans peptide bond conforma-tions, as well as cis-trans interconversions, were observed in NMR conformational and dynamics studies. It was proposed that a cis-trans interconversion may lower the activation barrier of the nAChR as it repositions backbone atoms in a region that may act as a molecular switch (78). Indeed, in the AChBP structure, the vicinal disulfide is located at the loop of the ␤9-␤10 hairpin, which is in position to direct agonist-induced conformational changes in loop C directly to the first transmembrane domain beginning just after ␤10 (21). As Bgtx acts to stabilize the receptor in the closed state (79), our structural data suggest that the peptide bond between the adjacent cysteines is trans in the closed state of the receptor. If interconversion of the vicinal disulfide peptide bond is involved in channel activation, the direction of conformational change is likely to be trans to cis. The structures of methanol dehydrogenase (72) and atracotoxin (51) both also indicate an overall trans conformation for the corresponding bond in these proteins.
Orientation of Bgtx Bound to the nAChR-The recent crystal structure determination of AChBP coupled with the structure of the ␣7 19-mer⅐Bgtx complex presented here permits an initial juxtaposition of Bgtx onto this structural model for the extracellular domain of the nAChR (21). As the AChBP is homologous to members of the nAChR family and binds nAChR agonists and antagonists, its structure serves as a model for the nAChR fold. Although an x-ray structure of a mimotope peptide⅐Bgtx complex was recently superimposed onto the AChBP structure (20), we predict the orientation of receptorbound Bgtx based on an authentic receptor sequence.
The most highly constrained region of the Bgtx-bound ␣7 19-mer (Ser 185 -Tyr 194 ) superimposes onto the corresponding region of the AChBP with a backbone r.m.s.d. of 1.93 Å (Fig.  9A). As a first approximation of the orientation of Bgtx about the nAChR extracellular domain, we anticipated and observed side chain clashes between the Bgtx and AChBP structures, but notably, the two backbone traces do not intersect (Fig. 9B). As the AChBP crystal structure is hypothesized to represent a "desensitized" state that is more compact than the resting state that Bgtx stabilizes (3), Bgtx would have even easier access to its binding site.
From a global perspective, our model shows Bgtx localized to the interface of two receptor subunits (Fig. 9C) with the concave surface facing the "plus side" of the subunit interface (21), similar to the mimotope/Bgtx superposition (20). The disulfiderich globular core of the toxin is proximal to the membrane with fingers I-III extended toward the AChBP-binding cavity FIG. 8. Sequence comparison of Bgtx-binding proteins. Amino acids are numbered according to the chick ␣7 sequence (27). Those residues shaded in light gray are conserved across species and between ␣1 and ␣7 nAChR subunits. They also have been localized to the binding site by chemical modification or photoaffinity labeling (1). A, alignment of nAChR and AChBP sequences. Glu 188 (dark gray shading), a residue that interacts with Lys 38 on the convex side of Bgtx, is conserved among ␣7 subunits of various species. B, alignment of Bgtx-binding peptides studied by NMR and x-ray crystallography. ␣12-mer (14), ␣18-mer (15,16), and ␣p25 (19) sequences are derived from the Torpedo ␣1 subunit. LLPep is a phage display library-derived peptide isolated for its ability to bind Bgtx (17), and HAPep and HAP are high affinity engineered peptides based on the library lead (18,20).
at about a 45°angle to the vertical axis of the receptor. Fingers I and II are near binding loops on the AChBP that are believed to form the principal ligand-binding site on the plus side of the subunit interface. This region includes binding loop C that overlaps with the continuous segment (␣1 173-204) that composes the major muscle receptor determinant for binding ␣-neurotoxins (22). Finger II, in particular, extends deep into the subunit interface approaching binding loops A and B. Fingers II and III are in position to interact with the complementary ligand-binding site on the adjoining subunit. For example, residues homologous to Trp 55 , Leu 119 , and Glu 176 of the nAChR ␥ subunit are in position to interact with finger II, consistent with mutational studies that demonstrated the significance of these nAChR residues in ␣-neurotoxin recognition. Double cycle mutant analysis of the short ␣-toxin NmmI and the nAChR ␥ subunit identified the interacting pairs ␥Trp 55 /NmmI-Arg 33 , ␥Trp 55 /NmmI-Lys 27 , ␥Leu 119 /NmmI-Arg 33 , and ␥Glu 176 / NmmI-Lys 27 (80). Additionally, Bgtx binding to the muscletype nAChR was blocked by chemical modification of the ␥Leu 119 3 Cys mutant and the homologous residues in the ␦ and ⑀ subunits with a methanethiosulfonate compound bearing a quaternary ammonium group (81).
This view of receptor-bound Bgtx suggests that ␣-neurotoxins approach their binding site from the receptor periphery, rather than from the vestibule. A similar superposition of ␣18mer residues forming the ␣18-mer⅐Bgtx contact zone (Tyr 189 -Thr 196 ) (16) also places Bgtx on the AChBP periphery, suggesting that the path Bgtx travels to bind the nAChR is similar for multiple receptor subtypes. Furthermore, these models provide a new perspective on the ␣-neurotoxin-nAChR interaction. Based on a number of previous studies, the prevalent model placed Bgtx at the crest of the synaptic side of the receptor with finger II extending into the vestibule (14). A strikingly different picture of the ␣-neurotoxin-nAChR interaction emerges from our model of Bgtx proximal to the membrane with its loops extended toward the binding cavity at the subunit interface. Based on structural data and consistent with recent biochemical findings, this predictive model forms a solid basis for future studies.