Only Snake Curaremimetic Toxins with a Fifth Disulfide Bond Have High Affinity for the Neuronal α7 Nicotinic Receptor*

Long chain and short chain curaremimetic toxins from snakes possess 66–74 residues with five disulfide bonds and 60–62 residues with four disulfide bonds, respectively. Despite their structural differences all of these toxins bind with high affinity to the peripheral nicotinic acetylcholine receptors (AChR). Binding experiments have now revealed that long chain toxins only, like the neuronal κ-bungarotoxin, have a high affinity for a chimeric form of the neuronal α7 receptor, with K d values ranging from about 1 to 12 nm. In contrast, all other toxins bind to the chimeric α7 receptor with a low affinity, withK d values ranging between 3 and 22 μm. These results are supported by electrophysiological recordings on both the wild-type and chimeric α7 receptors. Amino acid sequence analyses have suggested that high affinities for the neuronal receptor are associated with the presence of the fifth disulfide at the tip of the toxin second loop. In agreement with this conclusion, we show that a long chain toxin whose fifth disulfide is reduced and then dithiopyridylated has a low affinity (K d = 12 μm) for the neuronal α7 receptor, whereas it retains a high affinity (K d = 0.35 nm) for the peripheral AChR. Thus, a long chain curaremimetic toxin having a reduced fifth disulfide bond behaves like a short chain toxin toward both the peripheral and neuronal AChR. Therefore, functional classification of toxins that bind to AChRs should probably be done by considering their activities on both peripheral and neuronal receptors.

Long chain and short chain curaremimetic toxins from snakes possess 66 -74 residues with five disulfide bonds and 60 -62 residues with four disulfide bonds, respectively. Despite their structural differences all of these toxins bind with high affinity to the peripheral nicotinic acetylcholine receptors (AChR). Binding experiments have now revealed that long chain toxins only, like the neuronal -bungarotoxin, have a high affinity for a chimeric form of the neuronal ␣7 receptor, with K d values ranging from about 1 to 12 nM. In contrast, all other toxins bind to the chimeric ␣7 receptor with a low affinity, with K d values ranging between 3 and 22 M. These results are supported by electrophysiological recordings on both the wild-type and chimeric ␣7 receptors. Amino acid sequence analyses have suggested that high affinities for the neuronal receptor are associated with the presence of the fifth disulfide at the tip of the toxin second loop. In agreement with this conclusion, we show that a long chain toxin whose fifth disulfide is reduced and then dithiopyridylated has a low affinity (K d ‫؍‬ 12 M) for the neuronal ␣7 receptor, whereas it retains a high affinity (K d ‫؍‬ 0.35 nM) for the peripheral AChR. Thus, a long chain curaremimetic toxin having a reduced fifth disulfide bond behaves like a short chain toxin toward both the peripheral and neuronal AChR. Therefore, functional classification of toxins that bind to AChRs should probably be done by considering their activities on both peripheral and neuronal receptors.
Animal venoms produce a wide diversity of toxins that bind to a variety of different receptors, ion channels, and enzymes. Thus, they block some physiological pathways of prey and/or predators that, as a result, are usually rapidly immobilized and often killed. In addition, however, animal toxins are capable of exerting more subtle activities. For example, mamba venoms contain an unexpectedly large number of muscarinic toxins that exert a spectrum of fine activities toward a diversity of muscarinic receptor subtypes, acting on them as agonists or antagonists (1). Certainly, identification of subtle pharmacological activities constitutes a real challenge in the domain of toxin discovery.
In this respect, the well known snake curaremimetic toxins and their current classification as long chain and short chain toxins constitute a most puzzling situation. These monomeric toxins possess 66 -74 residues with five disulfide bonds and 60 -62 residues with four disulfide bonds, respectively (2). Well known prototypes of long chain and short chain toxins are, respectively, ␣-bungarotoxin (Bgtx) 1 from venom of the krait Bungarus multicinctus (3) and toxin ␣ from the African cobra Naja nigricollis (4 -6). The amino acid sequences of long chain and short chain toxins can be aligned readily, using four invariant half-cystines (2). Not surprisingly, therefore, x-ray crystallographic studies and NMR analyses (7)(8)(9)(10)(11)(12) have demonstrated that the polypeptide chains of both types of toxins adopt the same overall fold, which consists of three adjacent loops, rich in ␤-pleated sheet, which are connected to a globular fold where the four invariant disulfide bonds are located. However, a closer inspection of both amino acid sequences (see Table I) and spatial structures (2) reveals that toxins from both categories display quite precise structural differences. Thus, virtually all long chain toxins have a longer COOH-terminal tail, an additional small loop cyclized by the fifth disulfide at the tip of the central loop, and a slightly shorter first loop. Notwithstanding these clear structural differences, toxins from both categories bind to similar sites and with comparable high affinities on peripheral nicotinic acetylcholine receptors, such as AChR from Torpedo marmorata (13). It has been claimed that long chain toxins may bind more irreversibly than short chain toxins to AChR (14,15). Therefore, one may wonder why a large number of Elapidae and Hydrophiidae produce simultaneously two categories of toxins (2) which apparently exert the same function.
In the present study, binding experiments have revealed that long chain and short chain toxins present marked biological differences. Thus, long chain toxins recognize a chimeric construction of the neuronal ␣7 nicotinic acetylcholine receptor with an affinity that is nearly 4 orders of magnitude higher than that of short chain toxins. Electrophysiological experiments performed with wild-type (WT) and chimeric (␣7-V201-5HT3) ␣7 receptors support these observations. Finally, we demonstrate that this discriminating property is associated with a specific structural feature of long chain toxins, i.e. the presence of an additional cyclized loop at the tip of their central loop. Our findings suggest that toxins should be regarded with more scrutiny concerning the expression of their possible biological activities.

EXPERIMENTAL PROCEDURES
Materials-The cDNA of the chimeric ␣7-V201-5HT 3 receptor was kindly provided by Prof. J. P. Changeux (Institut Pasteur, Paris). * This work was supported by the National Swiss Science Foundation and the Office Federal de l'Education et des Sciences (to D. B.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
␣7 Expression in Human Embryonal Kidney (HEK) Cells-Chimeric cDNA (␣7-5HT 3 ) was transfected into HEK 293 cells (ATCC CRL 1573) by calcium precipitation with a careful control of the pH (6.95), CO 2 level (3%), and cDNA amount (15 g) as described previously (17). Two days after transfection, the cells were harvested in phosphate-buffered saline (PBS) ϩ 5 mM EDTA, rinsed twice in PBS, and finally resuspended in this buffer (3 ml/plate). This high efficiency transfection protocol allowed direct binding experiments on intact cells for 2 days.
Site-directed Mutagenesis-The chimeric receptor (␣7-V201-5HT 3 ) was constructed in a pMT3 vector (18,19) which permits transient expression in HEK 293 cells. Site-directed mutagenesis of erabutoxin a was performed using the mutagene M13 in vitro mutagenesis kit from Bio-Rad as described previously (20). Each mutant was submitted to amino acid analysis and sequencing.
Selective Reduction and Alkylation of Cbtx-1.2 mM Cbtx was incubated with 2.5 mM dithiothreitol in 0.2 M Tris buffer, pH 8.5, 1 mg/ml EDTA for 90 min at 4°C. After this selective reduction, 15 mM 2,2Јdithiopyridine was added in methanol for 1 h at room temperature. After purification by gel filtration (Bio-Gel P-2) and high pressure liquid chromatography (Vydac C8) the Cbtx-dithiopyridine obtained was analyzed by mass spectrometry.
Binding Assays-The affinity of 125 I-Bgtx was tested by incubating the labeled toxin (final concentration, 5 nM) with 350 l of cells in suspension in PBS. The nonspecific binding was determined in the presence of 1 mM (Ϫ)-nicotine. After a 40-min incubation, the sample was diluted in 3 ml of cold PBS, filtered through a GF/C filter (Whatman) previously dipped in non-fat dry milk (1%), and rinsed with 3 ml of cold PBS. The radioactivity of the filter was determined on a ␥-counter.
For competition experiments, we examined the effect of ligands on the initial rate of 125 I-Bgtx binding. The competitors were preincubated, at different concentrations, for at least 30 min with the cell suspension and filtrated 6 min after the addition of 5 nM 125 I-Bgtx. The protection constant calculated by fitting the competition data by the Hill equation was shown to correspond to the K d value (6). For the -Bgtx, the concentration and affinity constant were determined by assuming that the toxin interacts as a dimer (21). Competition experiments made with AChR from T. marmorata were performed at equilibrium, using 3 Hlabeled toxin as a radioactive tracer. Varying amounts of toxins were incubated with 2 nM AChR and 27 Ci/mmol, 5 nM 3 H-␣-toxin for 18 h at 20°C, and the mixture was filtered through Millipore filters (HAWP) which had been soaked in Ringer buffer. The filters were washed with 10 ml of Ringer buffer, dried, and after the addition of 10 ml of scintillation solution (Lipoluma) counted on a Rackbeta counter (LKB). Equilibrium dissociation constants were determined from competition experiments according to Ishikawa et al. (15).
Oocyte Preparation-Oocytes were isolated and prepared as described previously (22). Briefly, ovaries were harvested from Xenopus laevis females anesthetized with 0.2% 3-aminobenzoic acid ethyl ester (Tricaïne from Sigma) and sacrificed by section of the spinal cord. Ovaries were then rinsed with physiological solution and a portion cut into small pieces with a razor blade. Remaining portions of the ovary can be kept for several days in the refrigerator in a clean physiological solution for later use. Oocytes were isolated by gentle mechanical and enzymatic treatment using 0.2% collagenase (Sigma type I). Healthy oocytes were then selected manually under the microscope and injected intranuclearly the next day with 10 nl of solution (injecting buffer, 88 mM NaCl, 1 mM KCl, 15 mM Hepes, pH 7) containing 2 ng of the desired cDNA.
Electrophysiology-Electrophysiological recordings were performed 1-4 days later. Cells were impaled with two glass microelectrodes filled with 3 M KCl and maintained in voltage clamp at Ϫ100 mV using a GENECLAMP amplifier (Axon Instruments). Cells were superfused throughout the experiment with OR2 (oocyte Ringer) containing 82.5 mM NaCl, 2.5 mM KCl, 2.5 mM CaCl 2 , 1 mM MgCl 2 , 5 mM Hepes, at pH 7.4 (adjusted with NaOH) at a rate of about 6 ml/min. To prevent the possible contribution of endogenous muscarinic current all solutions were supplemented with 0.5 mM atropine. Acetylcholine was applied by a rapid switch of the perfusion with a system of electromagnetic valves (type III, General Valve, Fairfield, NJ) controlled by a personal computer. Data were converted on-line by an analog to digital converter (AT-MIO16, National Instruments) and stored for later analysis. Mathematical analysis and curve fitting were performed on Macintosh using a personal software.
Toxin Incubation-␣-Bgtx sensitivity of oocytes, expressing significant currents in response to acetylcholine, was determined in 8 ϫ 12 microplate wells (Nunc). To prevent toxin adsorption on plastic, all solutions were supplemented with 20 mg/ml bovine serum albumin. When incubated in the microplate, cells were tested as follows. (a) Currents evoked by a short pulse of acetylcholine were recorded, and electrodes were removed. (b) The cell was placed in a microwell, and the extracellular medium was exchanged completely for a solution containing the desired toxin concentration. (c) 30 min later oocytes were poked again, and the remaining acetylcholine-evoked current was measured.
Inhibition Dose-Response Curves-Fraction of the unblocked current was computed as the ratio of the acetylcholine current recorded after toxin incubation to the acetylcholine-evoked response measured in the control. Values determined for several cells were then averaged, plotted as a function of the logarithm of the toxin concentration, and fitted with the empirical Hill equation using a least squares minimizing method (Simplex).

Production and Characterization of the Chimeric ␣7 AChR-
The extracellular domain of the ␣7 nicotinic AChR has been fused previously to the membrane region of the 5HT 3 receptor (␣7-V201-5HT 3 ) (18) displaying thus all of the pharmacological properties of the natural ␣7 receptor and allowing an efficient expression in HEK 293 cells, which failed in the case of the entire receptor (18,23). In agreement with previous results obtained from pharmacological or electrophysiological experiments with the entire ␣7 receptor (24, 25), we found that 125 I-Bgtx binds to the chimeric receptors with an equilibrium dissociation constant (K d ) equal to 0.96 Ϯ 0.12 nM, as deduced from both the ratio of the dissociation and association rate constants and competition experiments (data not shown). The yield of receptor production was approximately 0.3 pmol of 125 I-Bgtx binding site/10 6 cells. Methyllycaconitine, a well known ␣7-specific ligand (26,27), completely inhibited the binding of radioactive ␣-Bgtx with a binding affinity characterized by a K d value of 1 nM, as deduced from the binding inhibition curve (data not shown). This value is in agreement with those reported in previous studies describing the interaction of methyllycaconitine with entire ␣7 receptors produced in other types of cells (26 -28).
Binding of Snake Toxins to Chimeric ␣7 and Torpedo AChRs-The family of curaremimetic snake toxins includes a large number of toxin members (2). However, only ␣-Bgtx has been shown to bind with high affinity to the ␣7 receptor (29). We now investigated the capacity of other snake toxins to inhibit the binding of 125 I-Bgtx to the recombinant chimeric ␣7 receptor. Thus, we examined the inhibitory activity of (a) various long chain curaremimetic toxins from snakes, including Cbtx, the ␣-cobratoxin from Naja kaouthia; L. semifasciata III, N. naja oxiana I, a new toxin from L. colubrina, 2 and -Bgtx a neuronal toxin from B. multicinctus; and (b) some short chain curaremimetic toxins from snakes, including N. nigricollis toxin ␣ and erabutoxins a and b (Ea and Eb) from L. semifasciata. The amino acid sequences of these toxins are shown in Table I. In general, short chain toxins possess 60 -62 residues and four disulfide bonds, whereas long chain toxins possess 66 -74 residues and five disulfide bonds (2). L. colubrina toxin appears therefore as an exception since, although it possesses 69 residues, it lacks the 2 cysteines 29 and 33 that usually form a disulfide bridge in other long chain toxins. Therefore, L. colubrina toxin is an unusual long chain toxin with four disulfide bonds, like short chain toxins. Fig. 1 shows the capacity of varying concentrations of different toxins to inhibit the binding of a constant amount of 125 I-Bgtx to the chimeric ␣7 receptor. In all cases, a complete dose-dependent inhibition was observed, with IC 50 values ranging from approximately 9 ϫ 10 Ϫ10 M for ␣-Bgtx to 2 ϫ 10 Ϫ5 M for Ea and Eb. Clearly, two groups of competitors emerged. The most potent group is constituted exclusively of long chain toxins, whereas the weak competitors include all short chain toxins and L. colubrina toxin, whose chemical characteristics are hybrid between long chain and short chain toxins. To quantify these data better, we calculated the Hill coefficients and the protection constants that were then assimilated to K d values, as described previously (6). Results from these calculations are summarized in Table II. Thus, ␣-Bgtx, Cbtx, N. naja oxiana I, and L. semifasciata III display a high affinity for the ␣7 chimeric receptor with K d values equal to 1, 4.5, 12, and 3 nM, respectively. Similarly, the neuronal -Bgtx is characterized by a K d value equal to 5 nM. In sharp contrast, the other toxins are much weaker ligands for the chimeric receptor, the K d values of toxin ␣, Ea, Eb, and L. colubrina toxin being equal to 3, 21, 22, and 7 M, respectively. Snake curaremimetic toxins have been initially recognized for their capacity to bind to peripheral AChR from Torpedo electric organs with high affinities, irrespective as to whether they are long chain or short chain toxins (13). In agreement with this general view, competition experiments made between 3 H-␣-toxin or 125 I-Bgtx and various snake toxins toward AChR from Torpedo (6, 30 -33, this study) revealed that both short chain and long chain curaremimetic toxins, including L. colubrina toxin, had high affinities for Torpedo receptor, with K d values ranging between 10 Ϫ9 and 10 Ϫ11 M. Only the neuronal -Bgtx had a lower affinity for this receptor. To appreciate better the differential affinities of the toxins for the ␣7 receptor over Torpedo receptor, we calculated the ratios of their respective affinities. As shown in Table II, the ratios vary between 2 and 35 for long chain curaremimetic toxins, except L. colubrina toxin, whereas they are close to 3 ϫ 10 5 for both short toxins and L. colubrina toxin. Therefore, long chain curaremimetic toxins exert similar binding affinities for both ␣7 and Torpedo AChRs, whereas short chain toxins and L. colubrina toxin have a marked preference for Torpedo AChR. Not unexpectedly, the ratio of affinities of -Bgtx is unique, with an affinity approximately 100-fold higher for the ␣7 chimeric receptor compared with Torpedo, further indicating the particular character of the ␣7 receptor, which is capable of binding both long chain curaremimetic and neuronal toxins with high affinities. Recently, it was observed that ␣-Bgtx and two new short chain toxins from the Australian snake Oxyuranus scutellatus bind to homogenates from chick optic lobes, a tissue that is likely to contain various neuronal receptors, including the ␣7 receptor. How-

Short chain toxins
Loop Data obtained with this experimental procedure for the ␣-Cbtx yielded an IC 50 of about 0.3 nM (see Fig. 2A), a value in good agreement with the binding experiments. It is important to notice that, as for ␣-Bgtx, little or no recovery from the blockade could be observed within 1 h of superfusion with the control medium (data not shown). When the same experiments were performed with Eb, an inhibition of the acetylcholineevoked current was observed only when the toxin concentration was raised to the micromolar range ( Fig. 2A). Comparison of the inhibition curves measured with ␣-Cbtx and Eb reveals a comparable apparent cooperativity for both toxins. In addition, as for the ␣-Cbtx, no significant recovery of the acetylcholineevoked current could be measured over long washing periods with the control solution (data not shown).
Previous experiments done with the chimeric ␣7-5HT 3 subunit have shown that the receptor reconstituted with this fusion protein displays a pharmacological profile resembling that of the chick ␣7 WT receptor (18). Because this fusion protein comprises only the receptor NH 2 -terminal domain, which is thought to contain the acetylcholine binding site, it was important to examine whether the blockade by ␣-Cbtx or Eb could be modified in the hybrid construct. Inhibition dose-response curves shown in Fig.  2B were made as for the WT receptor. Comparison of the ␣-Cbtx IC 50 values for the ␣7 WT and chimera reveals a reduction of the apparent affinity for the chimera, as deduced from the shift of the inhibition curve from 0.3 to 10 nM. Similarly, the chimera displays also a slightly lower affinity for the Eb. However, despite this shift, experiments made on WT ( Fig. 2A) and chimeric (Fig.  2B) ␣7 receptors show the same difference in the affinities of ␣-Cbtx and Eb, confirming the similar pharmacological profile of these two receptors.
A further illustration of the difference in apparent affinity of the short chain and long chain toxins is given by the recordings obtained in a single cell exposed first to 300 nM Eb and then to 3 nM ␣-Cbtx (Fig. 2C). The data clearly illustrate that a 30minute incubation with 300 nM Eb induces only a small reduction of the acetylcholine-evoked current, whereas incubation with a 100-fold lower ␣-Cbtx concentration practically abolishes any further agonist-evoked current.
On the Epitopes by Which Snake Toxins Recognize the Chimeric ␣7 and Torpedo Receptors-That short chain toxins have a much lower affinity for the ␣7 receptor than for the Torpedo receptor raised the possibility that the sites by which they bind to these two receptors are different. To address this question, we examined whether residues critical in the affinity change of a short chain toxin for the Torpedo receptor (20, 34) also altered its affinity for the chimeric receptor. Thus, we selected (a) three mutants of Ea (S8T, R33E, and K47E) having a substantially lower affinity for the Torpedo receptor compared with the WT toxin (20,33) and (b) a mutant (I36R) which uniquely exhibited a higher affinity to Torpedo receptor. Fig. 3 shows the capacity of the four mutants to inhibit the binding of radioactive ␣-Bgtx to the ␣7 chimeric receptor. The WT Ea (open symbols) completely inhibits the binding of the radioactive tracer, at a concentration of approximately 3 ϫ 10 Ϫ4 M. In contrast, a concentration of 2.5 ϫ 10 Ϫ4 M of either EaR33E or EaK47E had practically no effect on the binding of radioactive ␣-Bgtx (Fig.  3A), indicating that the two residues Arg-33 and Lys-47 are important for Ea to bind to ␣7 receptor. EaS8T inhibits the binding of radioactive ␣-Bgtx with a lower potency compared with the WT toxin, the IC 50 value of the mutant being in the millimolar range (Fig. 3B). This finding also suggests that Ser-8 is involved in the binding of Ea to ␣7 receptor. Fig. 3C shows that the mutant EaI36R displays an approximately 5fold higher binding affinity for the receptor compared with the WT Ea. In the aggregate, the mutants globally behave similarly toward both the ␣7 chimeric receptor and Torpedo receptors (20,33), indicating that the epitopes by which Ea recognizes the two receptors do overlap. The observation that Ea recognizes them with so different affinities (see Table I) suggests that the ␣7 receptor possesses specific mutations that make some functional residues of Ea unable to establish some of the stabilizing contacts, which probably occur with the Tor- How Do Long Chain Toxins Recognize ␣7 Chimeric Receptor with High Affinity?-To identify tentatively the feature(s) that may account for the observed functional difference between long chain and short chain toxins for the neuronal receptor, we examined the amino acid sequences of the toxins shown in Table I. We took advantage of the unexpected observation that the L. colubrina toxin has the length of long chain toxins and the low affinity of short chain toxins. More precisely, we looked for structural features that could differ between L. colubrina toxin and other long chain toxins and which could account for their differential affinities. Examination of Table I reveals that the L. colubrina toxin has a long COOH-terminal tail, the amino acid sequence of which is similar to that of other long chain toxins. Therefore, the long COOH-terminal tail is unlikely to be responsible for the low affinity of L. colubrina toxin for the ␣7 receptor. This conclusion is enhanced by the high affinity constants obtained with L. semifasciata III and -Bgtx, two long chain toxins characterized by a short COOH-terminal tail (see Tables I and II). Twelve other residues are present in L. colubrina toxin, but not in the other long chain curaremimetic toxins (underlined in Table I). Two of them (Leu-71 and Ala-72, using the numbering of ␣-Bgtx amino acid sequence) are located in the COOH-terminal tail, which is not involved in the receptor interaction. Ile-55 is replaced by Thr, Gln, Val, or Glu in other long chain curaremimetic toxins. Similarly, Ile-2, Ala-9, Arg-11, and Ile-14 in the loop I are replaced by various residues in the amino acid sequences of other long chain toxins. Therefore, it seems unlikely that these 7 residues are responsible for the low affinity of L. colubrina toxin on the neuronal receptor. In contrast, the 5 other residues, located at the tip of loop II (Asp-29, Gly-31, Ser-32, Gly-33, and Lys-36) are replaced by highly conserved residues in all long chain curaremimetic toxins. Especially the 2 cysteines (Cys-29 and Cys-33) which form the fifth disulfide bond and the vicinal aromatic residue (Trp or Phe-32) are conserved in all long chain toxins but L. colubrina toxin, suggesting that one or more of these substitutions is associated with the low affinity of L. colubrina toxin.
With the above conclusions in mind, one can examine, with a bird's eye view, all toxin sequences presented in Table I. A simple glance at this alignment suffices to reveal that the only feature that is present in all high affinity long chain toxins and missing in other toxins is the fifth disulfide bond 29 -33. To probe specifically the role of this region, we therefore decided to reduce selectively the fifth bond of a long chain toxin and to examine the consequence of this derivatization on the affinity of the toxin for both receptors. Thus, the fifth disulfide of Cbtx was reduced with 2.5 mM dithiothreitol, and each free cysteine was modified with 2,2Ј-dithiopyridine. The Cbtx-dithiopyridine was then purified, characterized, and tested in binding experiments using the Torpedo and ␣7 chimeric receptors. The results of these experiments (Fig. 4) show that Cbtx and Cbtxdithiopyridine are both potent competitors for the Torpedo receptor, their K d values being equal to 0.2 and 0.35 nM, respectively (Fig. 4A). Therefore, the derivatization did not alter the binding capacity of the toxin for the peripheral receptor. In sharp contrast, Cbtx and Cbtx-dithiopyridine display respectively high and low affinities for the ␣7 chimeric receptor, with K d values equal to 4.5 and 12,000 nM respectively (Fig. 4B). The selective reduction of the fifth disulfide has therefore altered the toxin binding capacity for the neuronal receptor, only.

DISCUSSION
As a result of their similar capacity to bind to the peripheral nicotinic AChR, snake long chain and short chain curaremi-metic toxins, which all adopt a similar fold, are usually considered as forming a family of functionally homogeneous proteins. This is however surprising because their respective polypeptide chains display clear differences, in terms of both length and number of disulfides. We now show that toxins from these two subgroups exert marked functional differences regarding their capacity to recognize the neuronal ␣7 receptor; furthermore, we show that this differential functionality is related to a structural feature that is uniquely present in the long chain toxins, i.e. the fifth disulfide bond.
We used a recombinant chimeric receptor that can be produced readily and relatively abundantly into HEK 293 cells (18), making this construction more appropriate than the natural ␣7 receptor for molecular pharmacology studies. When it is produced at the surface of HEK cells, 125 I-labeled ␣-Bgtx and methyllycaconitine bind to the chimeric ␣7 receptor with high affinities, similar to those obtained on the entire ␣7 receptor. It has been recognized previously that ␣-Bgtx from venom of the krait B. multicinctus binds with high affinity to the native ␣7 receptor (25,29). This toxin contains 74 residues and five disulfide bonds and belongs to the group of long chain toxins that possess the same number of disulfides and 66 -74 residues (2). Competition experiments between labeled ␣-Bgtx and other snake curaremimetic toxins revealed that toxins classified as long chain toxins also bind with high affinities to the ␣7 chimeric receptor, including not only the curaremimetic toxins but also -Bgtx, a neuronal toxin also from venom of B. multicinctus. The binding of this neuronal toxin to the ␣7 receptor was reported previously in an electrophysiological study in which 100 nM -Bgtx was necessary to abolish the acetylcholine response completely (29). All other snake toxins bind with a low affinity to the ␣7 receptor. These toxins include the short chain curaremimetic toxins and the L. colubrina toxin. The L. colubrina toxin, like two other long chain toxins from L. colubrina venom (35), is rather unusual in that it has the expected length (69 residues) of long chain toxins but lacks their fifth disulfide. Therefore, the ␣7 receptor, which offers a single class of binding sites to the curaremimetic toxins, clearly discriminates between the "true" long chain toxins and other toxins. This is in sharp contrast with the Torpedo receptor, which recognizes with high affinities both short chain and long chain curaremimetic toxins from snake venoms. A differential selectivity between long chain and short chain toxins has been depicted previously toward the chick optic lobe (34), suggesting that our observations, made with a recombinant ␣7 receptor, may also occur in vivo. However, for this extrapolation to be valid, one needs to demonstrate that the recombinant chimeric receptor behaves just like the natural ␣7 receptor toward the toxins. This has been made on the basis of electrophysiology experiments.
Electrophysiological experiments made with the reconstituted ␣7 WT or chimeric receptors showed that both a long chain toxin (␣-Cbtx) and a short chain toxin (Eb) inhibit the activation of the receptor by acetylcholine. The persistence of blockade observed with either toxin suggests that their dissociation kinetics are too low to allow a significant recovery within the experimental time frame. However, as expected from binding experiments, the inhibition dose-response curves observed with the chimera indicated that Eb is more than a 1,000-fold less potent than ␣-Cbtx. Although slightly more sensitive to both toxins, the ␣7 WT receptor displays a comparable pharmacological profile. Therefore, although the composition of these receptors differs significantly in their transmembrane domains as well as in their COOH-terminal regions, both receptors are inhibited by similar toxin concentrations, the long chain toxin being always much more potent than the short chain toxin. These data not only confirm that the major determinant for toxin binding and selectivity must be contained within the first 201 amino acids of the ␣7 receptor (18) but also that the binding and electrophysiological experiments are in good agreement, supporting the notion that both types of measurements are assessing the same state of the receptor. Therefore, binding experiments made with the chimeric receptor can safely be considered as indicative of the functional behavior of the ␣7 WT receptor.
The 10 4 -fold higher binding affinities of short chain toxins for the peripheral receptor from T. marmorata compared with the ␣7 neuronal receptor may indicate that these toxins recognize the two receptors by different "toxic" sites. To address this question we tentatively localized the site by which Ea, a short chain toxin, binds to the ␣7 receptor and compared it with the previously identified site by which it binds with high affinity to the Torpedo receptor (20,33). This was done using four toxin mutants of Ea harboring a single mutation at residues Ser-8, Arg-33, Ile-36, and Lys-47. The reason for this choice was 2-fold. First, mutations at these positions were shown previously to affect markedly the binding affinity of Ea for the Torpedo AChR (20,33), indicating that these residues are critical for the toxin to bind to the peripheral receptor. Second, three of them (Ser-8, Arg-33, and Lys-47) are spread at the periphery of the Torpedo AChR binding site, whereas Ile-36 is just in its center. Competition experiments made with these mutants revealed that although Ea has a low affinity for the ␣7 receptor, individual mutations at three positions (Ser-8, Arg-33, and Lys-47) further decreased the affinity for this receptor, whereas mutation at Ile-36 increased it. These observations are similar to those made previously with the Torpedo receptor (20,33). More precisely, mutations S8T, R33E, and K47E caused a decrease in affinity equal to or higher than 50-fold for the ␣7 receptor and, respectively, 780-, 310-, and 120-fold for the Torpedo receptor. Also, the mutation I36R caused an approximately 5-fold affinity increase, a value that is similar to what has been observed with the peripheral receptor (20). Therefore, although the binding of Ea to the ␣7 receptor is weaker than to the Torpedo receptor, it involves identical residues in both cases, indicating that the functional surfaces by which Ea binds to the two types of receptors largely overlap. Because short chain and long chain toxins interact with the Torpedo AChR by similar sites (11,20,37), it is likely that they all recognize both the peripheral and neuronal receptors by related though not necessarily identical sites.
How can we explain the differential affinities between the long chain and short chain toxins toward the ␣7 receptor? A comparison of the amino acid sequences of the different toxins revealed that all high affinity toxins possess a fifth disulfide at the tip of the central loop, suggesting that this region is associated with the remarkable capacity of long chain toxins to recognize the neuronal ␣7 receptor. A direct demonstration of the validity of this correlation was obtained by reducing the fifth disulfide of ␣-Cbtx, a long chain toxin. This modification caused virtually no effect on the affinity of ␣-Cbtx for the peripheral receptor, indicating in agreement with previous data (32,37) that residues around the fifth disulfide bond of the long chain toxins play little role in the binding to periperal Torpedo receptor. In sharp contrast, reduction of the fifth disulfide bond caused a nearly 10 4 -fold affinity decrease for the neuronal ␣7 chimeric receptor. In other words, by reducing the fifth disulfide of the long chain toxin, its behavior became quite similar to that of the short chain toxins and of the unusual L. colubrina toxin. Clearly, therefore, residues in the vicinity of the fifth disulfide bond must be directly related to the capacity of a long chain toxin to recognize the neuronal ␣7 receptor. This conclusion does not preclude that other unique elements of long chain toxins may also contribute to their discriminating behavior toward the ␣7 receptor.
Therefore, we have shown that structural differences between long chain and short chain toxins are directly related to distinct functionalities. The precise residues that, in the area of the fifth disulfide, provide the long chain toxins with their remarkable discriminating property remain to be identified. Because the curaremimetic toxins are amenable to synthesis by both recombinant (38 -40) and chemical 3 approaches, a systematic alanine-scanning experiment of this region can now be envisioned. Also, it would be worth identifying the residues of the ␣7 receptor which are uniquely recognized by the fifth bond toxin region. Finally, it would be of interest to investigate whether subtle differences in the primary or tertiary structures of toxins forming other large families, such as, for example, potassium channel-blocking toxins from scorpions, are also associated with fine functional selectivities.