Relative Spatial Position of a Snake Neurotoxin and the Reduced Disulfide Bond α(Cys192-Cys193) at the αγ Interface of the Nicotinic Acetylcholine Receptor*

We determined the distances separating five functionally important residues (Gln10, Lys27, Trp29, Arg33, and Lys47) of a three-fingered snake neurotoxin from the reduced disulfide bond α(Cys192-Cys193) located at the αγ interface of the Torpedo nicotinic acetylcholine receptor. Each toxin position was substituted individually for a cysteine, which was then linked to a maleimido moiety through three different spacers, varying in length from 10 to 22 Å. We estimated the coupling efficiency between the 15 toxin derivatives and the reduced cystine α(192–193) by gel densitometry of Coomassie Blue-stained gels. A nearly quantitative coupling was observed between αCys192 and/or αCys193 and all probes introduced at the tip of the first (position 10) and second (position 33) loops of Naja nigricollis α-neurotoxin. These data sufficed to locate the reactive thiolate in a “croissant-shaped” volume comprised between the first two loops of the toxin. The volume was further restrained by taking into account the absence or partial coupling of the other derivatives. Altogether, the data suggest that αCys192and/or αCys193, at the αγ interface of a muscular-type acetylcholine receptor, is (are) located in a volume located between 11.5 and 15.5 Å from the α-carbons at positions 10 and 33 of the toxin, under the tip of the toxin first loop and close to the second one.

ever, only recently has some light been shed on the molecular interaction between snake toxins and the nAChR. At least 10 residues of a sea snake toxin are involved in binding to the receptor from Torpedo marmorata, i.e. Gln 7 , Ser 8 , Gln 10 , Lys 27 , Trp 29 , Asp 31 , Arg 33 , Ile 36 , Glu 38 , and Lys 47 (1,2). Receptor regions that interact with snake toxins have also been investigated in various studies. Some lines of evidence indicate that snake toxins can bind to two sites located at the interfaces of the ␣␥and ␣␦-subunits (3)(4)(5). Among the different receptor regions that may be implicated in toxin binding, the domain 180 -200 of the ␣-subunit is clearly an important determinant (for a review see Ref. 6). It is also involved in various other binding functions, because it is recognized by small organic agonists (7) and small antagonists (8 -12). More recently, double mutant cycle analyses have revealed a number of contacts that may occur between nAChR and a cobra curarimimetic toxin (13)(14)(15). Thus, Arg 33 of NmmI toxin is coupled to the receptors ␣Val 188 and ␥Leu 119 , whereas the toxin Lys 27 interacts with ␥Glu 176 , and to a lesser extent with ␣Tyr 190 , Pro 197 , and Asp 200 . Although double mutant cycle analyses have evaluated the distances between various charged residues of this short-chain toxin and some residues of the receptor ␥-subunit (15), the relative positioning of the toxin with respect to the receptor ␣-subunit still remains unclear. To address this question, we have estimated the distances that separate several functionally important toxin residues from Cys 192 and/or Cys 193 , two residues that belong to the critical binding region 180 -200 of the two ␣-subunits. More precisely, we used experimental conditions that allowed us to perform this study at a single binding site, namely, the site at the ␣␥ interface.
It is known that the disulfide bond ␣(Cys 192 -Cys 193 ) can be selectively reduced without affecting antagonist binding (16) and hence specifically labeled with irreversible antagonists (8,17,18) or agonists (19). Treatment of the receptor with bromoacetylcholine or [4-(N-maleimido)benzyl]trimethylammonium iodide (MBTA) in the presence of a disulfide reducing agent leads to the covalent modification of one or both ␣-subunits, depending on the concentration of affinity label (20 -22). Even more interestingly perhaps, when the reducing agent is eliminated before the alkylation step, only one site is labeled by both probes, independent of their concentrations (22). This site has been demonstrated to correspond to the high affinity binding site of d-tubocurarine (23), which was later located at the ␣␥ interface of the receptor (11,24). Using radioactive MBTA, it was further demonstrated that ␣Cys 192 and ␣Cys 193 are the exclusively labeled residues (8).
Our strategy was inspired from this site-directed alkylation procedure. First, we chemically engineered analogues of a cu-rarimimetic snake toxin, by automated peptide synthesis. More precisely, we individually replaced each of the five functional residues Gln 10 , Lys 27 , Trp 29 , Arg 33 , and Lys 47 by a cysteine residue. The reason for the choice of these residues was that they are both important binding contributors to toxin-receptor formation and are widely spread over the functional topography of the snake toxin (1,2). Second, we linked each of them to three spacers of different lengths (10, 14, and 22 Å) that always ended in a maleimido group. Third, each of the 15 monomodified toxin derivatives was mixed with the receptor whose ␣(Cys 192 -Cys 193 ) disulfide bond was selectively reduced and affinity-labeled under conditions where only the ␣␥ interface was modified (23). Fourth, the coupling yield was estimated from gel electrophoresis experiments. Fifth, the data were exploited to build a model in which the toxin was positioned relative to cysteines 192 and/or 193 of the ␣-subunit at the ␣␥ interface.
Membrane Preparations-nAChR-rich membranes from the electric tissue of T. marmorata were prepared, as described previously (25), in the presence of 20 mM N-ethylmaleimide to block all free cysteines. They were further purified by alkali treatment (26). The concentration of acetylcholine binding sites was measured at equilibrium with [␣-125 I]Bgtx (27).
Synthesis of Toxin Analogues-N. nigricollis ␣-toxin analogues were synthesized as described earlier (28 -30) and refolded with 1.5-15 mM oxidized dithiothreitol (DTT) overnight, under inert atmosphere. The refolded toxins were purified by reverse-phase HPLC as described above, and checked by electrospray mass spectrometry and circular dichroism.
To conjugate the single free cysteine (Q10C, K27C, W29C, R33C, or K47C) of each toxin mutant with dimaleimide reagents, we first solubilized 0.7 mg (0.1 mol) of toxin analogue in 200 l of 250 mM sodium acetate buffer (pH 5.5)/2.5 mM EDTA and 100 l of H 2 O. The solution was then poured on 20 equivalents of dimaleimide-containing reagent dissolved in either a mixture of dimethylformamide (40 l)/acetone (160 l) for the two shorter reagents (Mal10 and Mal14) or in 200 l of sodium acetate buffer for the longest (Mal22) reagent, with rapid stirring. The mixture was left at room temperature for 30 min, except for the K27C derivatives, where the incubation time was prolonged to 8 h. Excess reactant was removed by exclusion chromatography (Biogel P2 from Bio-Rad), with the toxin migrating in the void volume. Each maleimido-toxin derivative was purified by HPLC as described above, and its mass was determined by electrospray mass spectrometry.
Reduction of ␣(Cys 192 -Cys 193 )-Selective reduction of the ␣(Cys 192 -Cys 193 ) bond was carried out using a slight modification of a previously described procedure (22). Briefly, the nAChR (3.3 nmol of toxin binding sites) was incubated under argon in degassed sodium phosphate buffer (100 mM, pH 7.6, 2.5 mM EDTA) containing 1 mM DTT, for 20 min at room temperature. The sample was centrifuged at 17,600 ϫ g for 15 min at 4°C. The supernatant was removed, and the pellet was resuspended in 600 l of degassed buffer A (250 mM sodium phosphate, pH 6.9, 2.5 mM EDTA) under argon. The last step was reproduced twice to eliminate excess DTT, and the pellet was resuspended in 1.1 ml of the same buffer, under argon.
Affinity Labeling-For affinity-labeling experiments, the reduced nAChR (1 M toxin binding sites) was mixed with each purified maleimido-toxin derivative (1 M final concentration) in buffer A (200-l final volume). After 15-or 90-min incubation at room temperature, under argon, each sample was centrifuged as described above. The supernatant was removed, and the pellets were resuspended in 70 l of Laemmli buffer.
For protection experiments, the reduced receptor was first incubated with native erabutoxin a (70 M) for 2 h at room temperature, under argon, before reaction with the maleimido derivatives.
Protection with Bromoacetylcholine-After reduction, the receptor was incubated with 10 M bromoacetylcholine for 10 min at room temperature, under argon, in the presence of 5 M tacrine (inhibition of acetylcholinesterase). The receptor was then reacted as described above with R33C-Mal22 for 15 min.
SDS-PAGE and Immunoblots-SDS-polyacrylamide gel electrophoresis (PAGE) was performed according to Laemmli (32) using a mini-protean apparatus (Bio-Rad). Briefly, about 30 g of each sample was electrophoresed on a 10% SDS gel of 1-mm thickness, which was then stained with Coomassie Blue. For immunoblots, 3-to 10-g protein samples were subjected to SDS-PAGE and then transferred to a polyvinylidene fluoride (PVDF) microporous membrane. The PVDF sheet was quenched for 2 h at room temperature with buffer B (10 mM sodium phosphate, pH 7.4, 150 mM NaCl) containing 2% bovine serum albumin. The sheet was incubated overnight at 4°C with monoclonal antibody 155 (33) (1/4000), or horse antiserum to the venom (1 mg.ml Ϫ1 ) solubilized in 10 ml of quench buffer. The PVDF membrane was washed (four times for 5 min) in buffer B containing 0.1% Triton X-100. Peroxidase-labeled antibodies GAR-PO or GAH-PO (1/4000) in 10 ml of quench buffer was added to the PVDF membrane for 1 h at room temperature. The PVDF sheet was washed again (four times for 5 min) in phosphate buffer containing 0.1% Triton X-100 and for 5 more min in the same buffer without Triton. The solution containing the peroxidase substrate was used when freshly prepared and consisted of 25 ml of phosphate buffer, 250 l of CoCl 2 solution (30 g.l Ϫ1 ), and 25 l of H 2 O 2 solution (30%). Development was stopped by rinsing the membrane in water.

Preparation and Characterization of Cysteine-containing
Toxin Derivatives-Toxin ␣ from N. nigricollis is amenable to peptide synthesis on the micromole scale by automated Fmoc solid-phase peptide synthesis (28 -30). We thus prepared five toxin derivatives in which the functionally important Gln 10 , Lys 27 , Trp 29 , Arg 33 , and Lys 47 (1, 2) ( Fig. 1) were individually replaced by a cysteine. The deprotected peptides were then reoxidized in the presence of oxidized DTT. The resulting proteins all had a wild-type toxin secondary structure, as indicated by the similarity of their CD spectra (Fig. 2). The unique exception was the W29C derivative, which displayed no positive peak at 230 nm, a finding that confirms that this band is associated with a contribution of the tryptophan residue (34). Each protein possesses a single cysteine, as assessed with Ellman's reagent (35), and is characterized by the expected mass determined by electrospray mass spectrometry (not shown). Up to 15 mg of each toxin derivative was obtained, the overall average yield of synthesis being about 5%. Therefore, none of the substitutions of any of the five original functional residues by a cysteine altered the capacity of the toxin to refold with its native-like secondary structure.
As expected for residues that are critical for toxin binding to nAChR, their replacement by cysteine caused binding affinity decreases (Table I). The wild-type toxin had a K i of 13 pM, consistent with its K d value (36), and this value decreased 4-, 18-, 145-, 12-, and 29-fold upon use with mutations Q10C, K27C, W29C, R33C, and K47C, respectively. Despite the affinity decreases, the analogues still retained relatively high binding affinities.
Introduction of Maleimide-containing Reagents into Cysteine-containing Toxins-Three dimaleimides called Mal10, Mal14, and Mal22 (Fig. 3) were reacted with the additional cysteine introduced into each of the five toxin mutants. To avoid formation of toxin dimers and to favor formation of mono-modified derivatives, we used a large excess of dimaleimides (20 eq). Excess reactant was subsequently eliminated by gel exclusion, and the 15 derivatives were purified by reversephase HPLC. All but the K47C-Mal10 derivative eluted as a single peak, which is strongly indicative of their homogeneity. The K47C-Mal10 derivative HPLC profile showed a slight but unexplained shoulder (not shown). All the derivatives displayed the expected mass (not shown) and were used as such for the labeling experiments.
All chemically modified derivatives but W29C displayed similar affinities for nAChR as compared with cysteine-containing derivatives (Table I). Due to a lack of material, we could not determine the IC 50 values for W29C maleimido analogues. Interestingly, the K27C-Mal22 derivative discriminated between the two toxin binding sites, one of them being recognized with almost the same affinity as for the wild-type toxin and the other with an affinity approximately two orders of magnitude lower.
Affinity Labeling of the Reduced Receptor-At first, all free cysteines of a membrane preparation of nAChR were blocked using 20 mM N-ethylmaleimide. The nAChR possesses three types of disulfide bonds spread over the extracellular domains of the various subunits. These are Cys 128 -Cys 142 on all subunits, Cys 192 -Cys 193 on the two ␣-subunits, and the Cys 500 -Cys 500 on the ␦-subunits, linking two monomers of the receptor (37,38). Under mild reducing conditions, only the ␣(Cys 192 -Cys 193 ) disulfide bond is affinity-labeled (8), and if the reducing agent is eliminated before the alkylation step, only the ␣␥ interface is labeled (23). We therefore carefully considered this procedure in a tentative attempt to generate selective labeling of the reduced ␣(Cys 192 -Cys 193 ) bond at the ␣␥ interface.
Incubation of the 15 maleimido-containing derivatives with the reduced nAChR was analyzed by SDS-PAGE and revealed by Coomassie Blue staining (Fig. 4A). The control lane with no derivative (lane 1) shows the four expected nAChR subunits.   Clear changes in intensity patterns were only seen for all derivatives modified at Q10C and R33C and for the derivative W29C-Mal22 ( Fig. 4A; lanes 2, 3, 4, 10, 11, 12, and 13). In all these cases, the intensity of the ␣-subunit decreased. This change was concomitantly associated with an increased intensity in the ␤-subunit, whose apparent molecular mass was close to 46 kDa, which not only fits with the ␤-subunit but also with a complex consisting of an ␣-subunit molecule associated with one toxin molecule. No such clear-cut phenomenon was seen with the eight other derivatives. Neither was it seen in control experiments where the receptor was not reduced, nor when the reduced receptor was treated with the well-known thiol blocking reagents N-ethylmaleimide or 5,5Ј-dithiobis-(2-nitrobenzoic acid) (not shown). Moreover, when the reduced nAChR was pretreated with an excess of an analogous snake toxin, erabutoxin a, the effect vanished for all the derivatives, except for the Q10C analogues for which the protection from affinity labeling was around 80% (data not shown). These data indicate that the phenomenon (i) is associated with the presence of free cysteines 192 and 193, (ii) requires a maleimido-containing toxin derivative, and (iii) is specific to the introduced linkers located at positions 10, 29, and 33.
To demonstrate that the toxin moiety was covalently associated with the new 46-kDa protein band, we blotted an SDS gel previously incubated with antitoxin antibodies present in a snake antivenom (Fig. 4B). Curiously, nonspecific detection was seen weakly at the level of the ␣-subunit and even more weakly around the ␤-subunit (Fig. 4B, lane 1). Nevertheless, strong differential labeling was readily observed around 46 kDa with W29C-Mal22 and all the derivatives modified at Q10C and R33C. For all these analogues, therefore, the toxin (molecular mass of 7 kDa) migrates with an apparent mass of approximately 46 kDa. No such detection was seen with other derivatives.
To establish whether the ␣-subunit is also present in the new 46-kDa band, we performed a similar blot experiment with an ␣-subunit-specific monoclonal antibody (33) (Fig. 4C). The control revealed an intense band migrating at the expected position of the free ␣-subunit (40 kDa). An additional band was seen around 46 kDa with the three derivatives modified at Q10C and R33C and with the longest derivative at W29C, confirming that, when these derivatives were present, the ␣-subunit migrated with a larger apparent molecular mass. In other words, these particular toxin derivatives comigrated with the ␣-subunit, a finding that strongly indicated their covalent association. We observed a similar though much weaker band with some other derivatives, in particular with K27C-Mal22, the two shorter W29C analogues, and the three K47C derivatives.
Coupling Yields at the ␣␥ Interface Binding Site-To evaluate the coupling yield between toxin derivatives and the ␣-subunit at the ␣␥ interface, we monitored the band intensities at both 39 and 46 kDa by densitometry of the Coomassie Bluestained gel for experiments where the reduced receptor and the derivatives were incubated for 15 and 90 min. The densitometry data must obviously be considered with caution, as they only reflect rough estimates of coupling yields whose determination was further complicated by the overlap of the intensity of the ␤-subunit, which added to background errors. However, these coupling yields seem to be in good agreement with the sensitive ␣-subunit immunoblot detection (Fig. 4C). Though the novel band at 46 kDa was not detectable by immunoblot experiments for the two shorter derivatives at Lys 27 (Fig. 4C,  lanes 5 and 6), a small coupling yield (around 10%) was determined by gel densitometry (Fig. 5). We therefore considered this value as the threshold above which coupling was significant.
Irrespective of the position of the probe, no more than approximately half of the two ␣-subunits were labeled, showing that a single site was labeled under our experimental conditions, where the reducing agent was removed before reaction of the toxin (18,21,22,39) (Fig. 5). Furthermore, it is known that low concentrations of bromoacetylcholine label the reduced receptor exclusively at the ␣␥ interface (22,23). In agreement with the view that the same site was probed by our derivatives, we found that preincubation with a low concentration (10 M) of bromoacetylcholine fully inhibited labeling with R33C-Mal22 (Fig. 4D). Therefore, our data support the view that the observed coupling occurs selectively on the reduced disulfide at the ␣␥ interface. The average 50% maximal coupling yield that was observed with respect to both ␣-subunits therefore corresponds to 100% labeling at the ␣␥ interface (Fig. 5). Fig. 5 shows that coupling yields obtained with Q10C-Mal22, W29C-Mal22, and the three derivatives of R33C were approximately 100% after both 15-and 90-min incubations. This suggests that the maximal coupling was reached within 15 min, in good agreement with an affinity-labeling process (39). Fig. 5 also shows that each modified toxin position behaved in a unique manner. At position 33, quantitative coupling was observed irrespective of the size of the spacer. At position 10, a 70 -80% coupling yield was seen with the two shorter linkers and quantitative coupling with the Mal22 arm. At position 29, the coupling yield was 30% and 45% with the Mal10 and Mal14 spacers, respectively, and coupling was complete with the Mal22 spacer. Finally, the longest probe at position 27 and the three probes at position 47 led to at most 20 -30% coupling, whereas the two shorter arms at position 27 led to virtually no coupling.

FIG. 3. Products obtained by coupling the free cysteinyl residue of each toxin analogue with three different dimaleimides in excess.
One maleimide moiety has reacted with the introduced thiol on the toxin, leading to a stable succinimidyl bond. The second maleimide moiety is left for reaction with the reduced receptor. The given distances correspond to the maximal length between the cysteine ␣-carbon and the reactive carbon on the maleimide moiety.

DISCUSSION
We prepared 15 derivatives of a snake toxin, each of them possessing a single maleimido moiety. We used these derivatives to localize the toxin spatially relative to cysteines 192 and/or 193 of the reduced ␣-subunit of the nAChR. Two lines of evidence indicated that only cysteines at the ␣␥ interface of nAChR were labeled. First, previous data showed that, under the same experimental conditions as those used in this study, only half of the toxin binding sites were labeled by a small receptor antagonist (8,22,40,41). Accordingly, we observed a maximal coupling yield of the ␣-subunits of approximately 50%. Second, data from several laboratories (11,(22)(23)(24) demonstrated that the labeled site is at the ␣␥ interface, this site being selectively discriminated by low concentrations of bromoacetylcholine (22,41). In agreement with these findings, we observed that small amounts of the same reagent fully inhibit toxin labeling.
To interpret our coupling data in terms of relative positions between the two receptor cysteines and the toxin, a number of assumptions had to be made. First, because the two shorter spacers (Mal10 and Mal14) are highly rigid, the maleimides located at their tips were considered to describe a surface controlled by the movement of the spacers. We assumed that the movement of the reactive carbons of the maleimides encompassed the surfaces of two spheres with radii of 10 and 14 Å centered, respectively, on the ␣-carbon of the modified residue. Second, because we observed almost comparable full coupling with the two shorter arms, we assumed that at least one of the thiolates of cysteines 192 and 193 is present in a volume comprised between two spheres with radii of approximately 11.5 and 15.5 Å (the distance between the reactive carbon and thiols is assumed to be approximately 1.5 Å). We called these volumes the "reactive volumes." Third, bearing in mind the imprecision of the data shown in Fig. 5, we considered that the coupling yields with the two shorter arms at positions 10 correspond approximately to full coupling, like those at positions 33. Fourth, for partial but significant coupling, in the range of 30 -50%, as observed with W29C derivatives, we considered that the reactive carbon was in proximity but not directly involved in the reactive volumes. Fifth, because the longer arm is highly flexible, we assumed it encompassed a large volume, which is difficult to predict, separated at most by 22 Å from the C␣ of the considered residue.
In addition to these assumptions, a number of elements had to be considered. First, the coupling yield is influenced by the distance that separates the thiols and the maleimides and by the accessibility of the sulfhydryl which, for example, could be favorably (or unfavorably) orientated for the attack of the thiolate perpendicularly to the plane of the maleimide ring. Second, at the molecular level, only one of the two cysteines can effec- tively react with the proposed maleimido group but, on average, a mixed population of labeled cysteines is likely to be generated, the relative proportion of labeled cysteines probably varying from one derivative to another. Bearing all these assumptions and considerations in mind, we interpreted our coupling data in terms of possible distances separating cysteines 192 and/or 193 from the ␣-carbon to which the reacting maleimido group was linked.
With the Q10C and R33C derivatives, the two short linkers led to virtually full coupling, so at least one of thiols 192 and 193 is likely to be at the intersection of the two reactive volumes centered on C␣10 and C␣33. The resulting volume adopted a "croissant-like" shape (partially visible in Fig. 6B) located in between the first and second loops almost perpendicularly to the plane of the large ␤-sheet defined by the three toxin loops and orientated at nearly 45°from the axis of the central loop (Fig. 6A). K47C-Mal22 showed almost no coupling, suggesting that the corresponding C␣ is more than 23.5 Å from the reactive thiol(s). We therefore amputated the croissant-like area by about one fourth of its volume, in the region close to the second loop of the toxin. Interestingly, the reactive carbon of the W29C-Mal14 derivative, which showed no more than 50% coupling, brushed against the remaining volume. In Fig. 6 we colored violet the region of the remaining reactive volume, which is located 15.5 Å from the C␣29. In conclusion, the thiol(s) of Cys 192 and/or Cys 193 at the ␣␥ interface of the reduced Torpedo nAChR, is(are) most likely to belong to the remaining volume (colored yellow), in close proximity to the violet zone, which is framed by the three maleimides displayed by Q10C-Mal10, W29C-Mal14, and R33C-Mal10 (Fig. 6). Because binding of the receptor antagonists does not seem to be affected by reduction of the disulfide bond ␣(Cys 192 -Cys 193 ) (16), we suggest that our data also reflect the positioning of the toxin when bound to native nAChR.
Considering the rather large distance that separates the thiol(s) occupying the identified region of the reduced receptor and the closest toxin residues, Gln 10 , Arg 33 , and Ile 35 , we suggest that the considered thiol(s) is(are) not fully hidden when the toxin is bound to its receptor. This situation could thus explain why the bound toxin does not fully protect labeling of the reduced receptor ␣-subunit with bromoacetylcholine (42) or N-ethylmaleimide. 2 Probably, therefore, the reactive thiol(s) is (are) not in direct interaction with the toxin. Possibly also, the low coupling yields observed for the K27C and K47C derivatives (Ͻ30%) may be due to the flexibility and/or the accommodation of the molecules during the interaction between the receptor and toxin.
It was previously suggested that the positively charged quaternary ammonium ion of the antagonist MBTA is located at about 10 Å from the ␣(Cys 192 -Cys 193 ) disulfide bond (8). Based on circumstantial evidence, it was also proposed that ammonium cations of small organic ligands may be mimicked by the positive charge of Arg 33 of curarimimetic toxins (43,44). Considering the present data, Arg 33 is indeed the only positively charged toxin residue that is consistent with the proposed mimicry, the positive charges of the two other candidates (Lys 27 and Lys 47 ) being too remote from the disulfide bond.
How does our conclusion fit with previous experiments (13,15,(45)(46)(47) designed to position a short or a long-chain toxin with respect to nAChR? An NMR study of a complex between the long-chain ␣-Bgtx and a dodecapeptide of the Torpedo receptor (␣185-196) indicated that the region ␣186 -190 is located between the first and the second loops of the toxin (46). Another NMR study of a complex between ␣-bungarotoxin and a library-derived peptide sharing some similarity with the receptor peptide ␣187-199 showed that this 13-mer peptide also 2 P. Kessler, personal observation. binds roughly in between the first and second toxin loops. Both studies clearly agree with our findings that the receptor region comprising ␣(Cys 192 -Cys 193 ) is located in between the first and second loops of a snake toxin. In one study, the peptide residues that are homologous to cysteines 192 and 193 were too mobile to be located (46). In the other study (47), these residues do not fit in the reactive region, but are located much more to the back of the plane defined by the large ␤-sheet of ␣-bungarotoxin. More precisely, the distance that separates the region colored violet (Fig. 6) from those peptide residues is approximately 15 Å. This situation might result from differential binding between short (this study) and long toxins (46,47). Two lines of evidence support this view. First, a derivative of a long-chain toxin modified on the residue homologous to Lys 27 with a mercurial compound similar in length to our Mal14 derivative, slightly coupled to the reduced cysteines of the ␣-subunits (45), in contrast to our own findings. Second, recent data have indicated that the binding topographies of short-and long-chain neurotoxins for Torpedo AChR are not strictly identical (48). It could also be argued that the small peptides used in NMR studies might behave differently when free or integrated in the whole receptor. More work is needed to clarify this question.
Recently, double mutant cycle analyses have shown that when another short toxin, NmmI, binds at the ␣␥ interface of the non-reduced mouse muscular receptor, Arg 33 of the toxin is in proximity to ␣Val 188 and ␣Tyr 190 (13). This result fully agrees with our proposed location of the disulfide bond ␣(Cys 192 -Cys 193 ) at the same interface (Fig. 6). In addition, it was observed that Arg 33 is also in close proximity to Trp 55 and Leu 119 of the ␥-subunit of the muscular receptor (15). Assuming that the Torpedo and mouse receptors are comparable, our data combined with those of Osaka provide a molecular basis for the positioning of a short-chain toxin relative to residues of the ␣and ␥-subunit at the ␣␥ interface. It is currently assumed that the sequential homology of ␥Asp 174 and ␦Asp 180 is associated with an identical position of the two residues in their respective sites (49). Thus, a sitedirected labeling experiment has shown that ␦Asp 180 is located 1 nm from the reduced disulfide bond ␣(Cys 192 -Cys 193 ) (38,50). Furthermore, ␥Asp 174 was recently located near Lys 47 and Lys 48 on the third loop of NmmI toxin (15). Interestingly, our model (Fig. 6) shows that the homologous Lys 47 on N. nigricollis toxin ␣, and therefore the interacting ␥Asp 174 (15), are over 10 Å from the disulfide bond at the ␣␥ interface. This evidence thus suggests that the binding sites at the ␣␥ and ␣␦ interfaces are structurally distinct.  10,27,29,33, and 47 with their corresponding native residue label. A high temperature molecular dynamics simulation (1000 K) of a dummy atom restrained by two constraints, i.e. to be located between 11.5 and 15.5 Å from the C␣10 and C␣33 (full coupling with the two shorter linkers for Q10C and R33C derivatives), engendered a croissant-like volume. A filter, suppressing the area located at less than 23.5 Å from C␣47 (no coupling with K47C-Mal22) restricted this volume to the depicted dots. Another filter allowed us to visualize the area located at 15.5 Å from C␣29 (partial labeling with W29C-Mal14) (violet dots). All together, Q10C-Mal10 (labeled Gln10), W29C-Mal14 (labeled Trp29), and R33C-Mal10 (labeled Arg33) were drawn pointing at the most probable area of localization of the reactive(s) thiolate(s) of Cys 192 and/or Cys 193 . Because K27C and K47C derivatives did not lead to any coupling to the reduced receptor, the original lysinyl residues have been shown. A, view from the receptor face; B, view from the bottom of the toxin.
In conclusion, the data reported in this paper describe the respective spatial positioning of an element of the nAChR and a snake toxin, a potent receptor antagonist. Together with recent double mutant cycle analyses (15), it offers a suitable molecular basis for building up the remaining receptor elements that interact with the toxin.