Mimicry between receptors and antibodies. Identification of snake toxin determinants recognized by the acetylcholine receptor and an acetylcholine receptor-mimicking monoclonal antibody.

In several instances, a monoclonal antibody raised against a receptor ligand has been claimed to mimic the ligand receptor. Thus, a specific monoclonal antibody (Mα2-3) raised against a short-chain toxin from snake was proposed to mimic the nicotinic acetylcholine receptor (AChR) (1). Further confirming this mimicry, we show that (i) like AChR, Mα2-3 elicits anti-AChR antibodies, which in turn elicit anti-toxin antibodies; and (ii) the region 106-122 of the α-chain of AChR shares 66% primary structure identity with complementarity-determining regions of Mα2-3. Also, a mutational analysis of erabutoxin a reveals that the epitope recognized by Mα2-3 consists of 10 residues, distributed within the three toxin loops. Eight of these residues also belong to the 10-residue epitope recognized by AChR, a result that offers an explanation as to the functional similarities between the receptor and the antibody. Strikingly, however, most of the residues common to the two epitopes contribute differentially to the energetic formation of the antibody-toxin and the receptor-toxin complexes. Together, the data suggest that the mimicry between AChR and Mα2-3 is partial only.

Antibodies are well recognized for their ability to bind foreign antigens and hence to inhibit their biological properties. However, recent methodological progress associated with the preparation and selection of immunoglobulins has revealed that antibodies can exert much more complex functions. Thus, catalytic antibodies, also called abzymes, are capable of transforming substrates (2)(3)(4). Moreover, the abzymes possess catalytic groups whose spatial arrangements resemble those observed in enzyme structures (5), supporting the view that they might mimic enzymes. Antibodies are also capable of mimicking non-enzymatic antigens (6) and in particular receptors (7,8). However, our understanding as to the molecular basis associated with the concept of mimicry between antibodies and receptors remains obscure and has even been questioned (7)(8)(9). The goal of this paper is to shed further light on this concept, which is anticipated to have numerous implications, including the use of receptor-mimicking antibodies as molecular decoys for diagnosis or therapeutical purposes.
Understanding of the molecular analogies and differences between a receptor and a receptor-mimicking antibody requires the elucidation and comparison of the molecular contacts that are established between each of these different entities and a common ligand. As judged from x-ray studies, protein-protein interactions involve multiple contact points. The buried surface area currently ranges around 1600 Ϯ 350 Å 2 with 15 or more interacting residues in each (10,11). However, free energy calculations suggested that only a subset of the observed contacts contributes actively to the binding (12). In agreement with this view, mutational analyses of contacting areas in antibody-antigen or receptor-hormone complexes demonstrated the critical role of few residues only (13)(14). The present paper aims at comparing the amino acids of a ligand that are recognized by its receptor and a receptor-mimicking antibody.
The ligand chosen for the study is erabutoxin a (Ea), 1 a short-chain curaremimetic toxin from sea snake venom. It is a small protein that recognizes the peripheral nicotinic acetylcholine receptor (AChR) with high affinity (K d ϭ 0.07 nM), provoking flaccid paralysis (15) and death as a result of respiratory failure. AChR is a ligand-gated ion channel with four different subunits assembled in the molar stoichiometry ␣ 2 ␤␥␦ (16 -17). The site by which the toxin binds to AChR has been previously identified, on the basis of mutational analyses (18 -20). This site includes at least 10 residues, among which 4 amino acids seem to play a predominant binding role. The monoclonal antibody that is considered in the present study is M␣2-3, which recognizes the toxin with high affinity (K d ϭ 1.4 nM). Preliminary studies based on the use of chemically modified toxin derivatives revealed that three toxin residues involved in the recognition of M␣2-3 were also implicated in the AChR binding site, indicating that the determinants recognized by AChR and M␣2-3 do overlap (1). It was therefore suggested that M␣2-3 may mimic AChR.
The goal of this work is 3-fold. First, we further investigated the putative mimicry between M␣2-3 and AChR by searching for additional common functional and structural properties. Thus, we studied the capability of the antibody to bind various AChR ligands. Also, we investigated whether M␣2-3 can induce AChR-specific antibodies, as AChR does (21)(22)(23). This being the case, the anti-idiotypic antibodies were studied for their ability to induce, in turn, toxin-specific anti-anti-idiotypic antibodies. Finally, we searched for amino acid sequence similarities between the complementarity-determining regions (CDRs) of M␣2-3 and the ␣-subunit of AChR, as it sometimes occur between receptors and receptor-mimicking antibodies (8). Second, having in hand data indicating functional and structural similarities between M␣2-3 and AChR, we delineated the epitope recognized by M␣2-3 on the surface of Ea, on the basis of a mutational approach, using a large set of toxin mutants. We then compared this epitope with the previously identified de-terminant recognized by AChR on the same toxin (20). The data reported in the present study clarify the molecular basis associated with the notion of mimicry of a receptor by an antibody.

EXPERIMENTAL PROCEDURES
Construction, Expression, and Characterization of Site-directed Mutants of Erabutoxin a (Ea)-Probes for site-directed mutagenesis were synthesized using an Applied Biosystems model 381A synthesizer. Sitedirected mutagenesis was performed using the Muta-Gene M13 in vitro mutagenesis kit from Bio-Rad. DNA technologies applied to Ea were performed as described previously (19,20,24). Briefly, cDNA encoding native (25) or mutated Ea were cloned into the expression/secretion vectors pRIT5 (24) or pEZZ 18 (19,20). Both vectors were purchased from Pharmacia Biotech Inc. and led, respectively, to a toxin fused to either the complete protein A from Staphylococcus aureus (26) or the two synthetic IgG-binding domains (ZZ) derived from protein A (27,28). The bacterial host used for expression was Escherichia coli HB 101 (29). Cleavage by CNBr and purification of Ea mutants were done as described previously by Boyot et al. (30). Biochemical characterization of each mutant included an amino acid sequencing, an amino acid analysis, an isoelectric focusing on gel, and a circular dichroic analysis, as described previously (19).
Specific binding of toxin mutants to acetylcholine receptor-rich membranes was determined from competition experiments (31) using 3 Hlabeled toxin ␣ as a radioactive tracer (32). Equilibrium dissociation constants were deduced from competition binding experiments according to Ishikawa et al. (33).
Specific binding of Ea mutants to the toxin-specific monoclonal antibody (M␣2-3) was determined by radioimmunoassays, in solution using 3 H-labeled toxin ␣ and purified M␣2-3 as described previously by Boulain et al. (34). Calculated Gibbs free energy changes (⌬G) accompanying noncovalent complex formation of native or mutated Ea with AChR or M␣2-3, were determined according to Novotny et al. (12). ⌬⌬G values (13) correspond to the differences existing between the free energy of the mutated Ea (⌬G Ea(mut) ) and the wild type toxin (⌬G Ea(wt) ), calculated respectively from the standard equation: ⌬G°ϭ ϪRTln1/K d , with K d ϭ dissociation constant, T ϭ 293 K, and R ϭ 1.99 cal/mol/K.
cDNA Cloning and Nucleotide Sequencing of M␣2-3 Precursor Chains-A cDNA library was constructed using the SuperScript Plasmid System (pSPORT1 plasmid; Life Technologies, Inc.) according to the specifications of the supplier. Screening of immunoglobulin heavy and light chain cDNAs was performed using 5Јend-labeled oligonucleotide probes specific for C␥2a (5Ј-GCCAAAACGACACCCCCATCTG-3Ј) and C (5Ј-GTCTAGATTAACACTCATTTC(G/T)GTTGAA-3Ј) chains, deduced from Kabbat et al. (37). Selected cDNA clones were completely sequenced on both chains using the dideoxy chain termination method (38).
Immunization Procedures-1 mg of purified M␣2-3, emulsified in complete Freund's adjuvant, was injected intradermally into five rabbits. Then animals received the same dose following the same procedure every 2 weeks for up to 8 months. Rabbits were bled before the first immunization and then every week.
Purification of Anti-idiotypic (Ab2) and Anti-anti-idiotypic (Ab3) Antisera-Ab2 was purified following a two-step procedure according to Pillet et al. (39). First, rabbit antisera were chromatographed onto a mouse IgG column equilibrated in phosphate buffer, 0.01 M, pH 7.4, NaCl 0.15 M, as described previously (40) until complete disappearance of anti-normal mouse IgG activity. Effluents were collected, and the column was washed with 0.1 M glycine/HCl buffer, pH 2.4, NaCl 0.15 M to elute bound immunoglobulins. Effluents were chromatographed several times through the same column until disappearance of anti-normal mouse IgG activity. Second, the resulting solution was loaded on a M␣2-3-Sepharose 4B column in which the purified M␣2-3 was covalently coupled to Sepharose, using a ratio of 10 mg of antibody/g of gel. Buffers were identical to those indicated above. Anti-anti-idiotypic (Ab3) fraction was collected from rabbit sera by submitting them to affinity chromatography on a column with toxin ␣ covalently linked to Sepharose. Identification of Ab2 and Ab3 fractions as well as determination of their binding properties were made according to the procedures previously described by Pillet et al. (39) for the highly homologous long-chain curaremimetic toxins.

On the Mimicry of AChR by M␣2-3
In the absence of structural data, mimicry between a receptor and an antibody is generally associated with the capacity of both proteins to recognize the same ligands (7), to share related immunological properties, and/or to possess amino acid sequence similarities (8,(41)(42)(43). The nicotinic AChR and the antibody M␣2-3 possess these three characteristics in common, at least to some extent.
M␣2-3 and AChR Both Recognize Short-chain Curaremimetic Toxins-AChR is recognized by a variety of agonists, which cause an opening of the channel, resulting in a macroscopic current and hence to the depolarization of postsynaptic cells (44). It is also recognized by a number of antagonists that maintain the channel in a closed state. Among the most potent antagonists are short-chain curaremimetic toxins from snakes (45). They are small proteins of 60 -62 amino acids and 4 disulfides, which bind to AChR with high affinities, their apparent equilibrium dissociation constants currently ranging between 10 Ϫ10 and 10 Ϫ12 M (46). Previously, a toxin-specific monoclonal antibody (IgG2a/kappa isotype) was raised by immunizing mice with toxin ␣ from Naja nigricollis and by screening antibodies with both toxin ␣ and Ea, two short-chain toxins with little cross-reactivities toward their reciprocal polyclonal antisera (1,47). This antibody was named M␣2-3. Table  I shows the relative affinities of various short-chain toxins toward both AChR and M␣2-3. Clearly, the antibody and the receptor recognize all these toxins, suggesting that they both bind to similar determinants. However, all toxins have similar high affinities for the receptor, with K d values around 6 -70 pM, whereas they display a wider range of affinities for the antibody, with K d values ranging from 1 M to 1.4 nM. These differences indicate that if the toxin determinants recognized by the antibody and the receptor are similar to each other they are unlikely to be identical. As a consequence, the complementary binding regions of the receptor and the antibody are also unlikely to be identical, excluding the possibility of a genuine mimicry between the two proteins. This conclusion is further supported by the observations that M␣2-3 failed to recognize any other antagonists of AChR, including long-chain curaremimetic toxins or curare alkaloids (1).

M␣2-3 and AChR Share Immunological
Properties-It is well documented that AChR triggers, in various animals, the production of antibodies that are capable of (i) binding to AChR (48), (ii) blocking access to ligands of AChR (49), and (iii) inducing signs of myasthenia gravis (50). In a previous report, we showed that a murine monoclonal antibody raised against the long-chain ␣-cobratoxin recognizes all other long-chain cu-raremimetic toxins (66 -71 residues and 5 disulfides) and binds to an epitope that overlaps the site by which the toxin recognizes the receptor (39,51). Strikingly, this antibody was capable of eliciting anti-AChR antibodies, which competed with long-chain toxins for binding to AChR. In turn, anti-anti-idiotypic antibodies were elicited and shown to neutralize these toxins (39). We repeated the same experiments using purified M␣2-3, a murine short-chain toxin-specific monoclonal antibody. This antibody was injected into rabbits according to the procedure described under "Experimental Procedures." After a period of 8 months and 16 injections, the sera were collected and purified by successive affinity chromatographies on columns of mouse IgGs and M␣2-3 covalently linked to Sepharose. The resulting antisera raised against M␣2-3 inhibited the binding of tritiated toxin to M␣2-3 ( Fig. 1A) and to AChR (Fig. 1B), in dose-dependent manners. These inhibitions were not due to residual amounts of M␣2-3 since no murine antibody was detected in the purified antisera, as revealed using assays with anti-mouse antibodies. Moreover, the antisera could not precipitate the labeled toxin itself, indicating that the antisera raised against M␣2-3 acted in fact as a competitor of the labeled toxin for binding to M␣2-3 or to AChR. Additionally, non-immune antisera were unable to inhibit the binding of the labeled toxin to either the antibody or the receptor. In aggregate, these data suggest that M␣2-3 can elicit neutralizing anti-AChR antibodies, as AChR does (50).
With the view to further investigate a possible mimicry between M␣2-3 and AChR, we studied the capacity of anti-idiotypic antibodies (Ab2) to induce in the same rabbit, anti-antiidiotypic antibodies (Ab3) capable of recognizing the toxin. Immunizations were carried out as described under "Experimental Procedures." The resulting serum was tested for the presence of Ab3 by determining the capacity of the antisera to bind to toxin ␣ from N. nigricollis coated on a 96-well plate. A substantial binding was revealed with goat anti-rabbit antibodies labeled with peroxidase (data not shown), whereas labeled goat anti-mouse antibodies failed to detect any binding (data not shown). Therefore, the antiserum that recognizes the coated toxin is from rabbit and not from mouse, excluding the possibility that contaminating M␣2-3 could be responsible for the observed binding. As shown in Fig. 2, this binding can be inhibited in a dose-dependent manner by the free toxin, with an IC 50 of approximately 0.3 M, further demonstrating the specificity of the binding. One may notice that the inhibition curve covers a range of toxin concentrations of nearly 4 orders of magnitude, suggesting a marked heterogeneity among the generated anti-anti-Id antibodies. In conclusion, therefore, M␣2-3 shares immunological properties with AChR since they are The mixtures were centrifuged 20 min at 3000 rpm at 4°C. The pellets were resuspended in 0.75 ml of 0.05 N NaOH, supplemented with 10 ml of Aqualipoluma solution, and the radioactivity was counted. The calculated mean standard deviation was lower than 3%. AChR (3 nM) was incubated with labeled toxin (3.4 nM) in the presence of various amounts of anti-M␣2-3 antiserum overnight, at 4°C, in a Ringer's solution. The mixtures were ultrafiltrated on two HAWP filters, washed with cold Ringer's solution, and the radioactivity retained on the filters was counted. As control experiments, it was assessed that a non-immune antiserum, used in the same volume ranges, could inhibit the binding of labeled toxin to the antibody (A) or AcHR (B).

FIG. 2.
Binding inhibition of anti-anti-antibodies to coated toxin ␣, by increasing amounts of free toxin. Binding of the antiserum was revealed with goat anti-rabbit antibodies labeled with peroxidase. The mean standard deviation was lower than 3%.
both capable of successively eliciting anti-AChR and anti-toxin polyclonal antiserum.
AChR and M␣2-3 Share Amino Acid Sequence Similarities-To identify the residues that are involved in the antigen combining site of M␣2-3, we cloned the cDNAs encoding its two chains and determined their nucleotide sequences. The deduced amino acid sequences of the two chains have been recently reported (52). As judged from the large number of known antigen-antibody complexes described by x-ray crystallography, the CDRs of the antibody provide the majority of contacts between an antibody-combining site and its complementary epitope, although some contact residues may also be located within framework regions (53). Fig. 3A shows the deduced primary sequences of CDR1L, -2L, and -3L from the light chain and CDR1H, -2H, and -3H of the VH domain of the ␥2a heavy chain of M␣2-3, flanked by proximal frameworks residues. Fig. 3B shows a comparison of these CDRs with the amino acid sequences of the region 100 -128 of the ␣-subunit of AChR from mouse (54) and electric ray fish (55). Clearly the CDRs and the region 100 -128 of the receptor ␣-chain share amino acid sequence similarities. Particularly striking is the strong analogy found between residues 106 -122 of AChR ␣-subunit and residues from CDR2H and -3H, with four and six identical amino acids and one conservative substitution, respectively. As indicated in the literature (56), the CDR3H results predominantly from a combination between germline diversity (D) and junction (J) segments. In the case of M␣2-3, segments DF 1 16-2 and J H 2 have been selected. However, we noticed the presence of a doublet of leucine residues at positions 109 and 110 according to Fig. 3B numbering. Such an unexpected addition results from the imprecision of the joining between segments D and J during the assembly of the complete heavy variable domain. As a consequence, approximately 66% identity is observed along the 106 -122 segment of AChR ␣-chain with the combined CDR3H, -2H, and -2L of M␣2-3. Clearly, therefore, M␣2-3 and AChR share mimicry in terms of primary structures.
Analogies or identities between the amino acid sequences of CDRs of antibodies and other proteins is not uncommon. Thus, sequential similarities have been previously observed between CDRs of receptor-specific monoclonal antibodies and a receptor ligand (42,(57)(58)(59)(60)(61)(62) and, as in our case, between CDRs of an antigen-specific monoclonal antibody and the receptor that is recognized by the antigen (8, 63). These similarities have even been exploited for identifying the binding area of a ligand on its receptor (8). Similarly, we wish to suggest that the region 106 -122 of AChR is involved in the recognition of the shortchain curaremimetic toxins. Previously, Ruan et al. (64) reported that a fragment of the receptor subunit that encompasses this region binds to radioactive erabutoxin b, an isoform of Ea. The region 106 -122 together with the stretch 170 -210 of the ␣ subunit of AChR (65-68) may provide appropriate contacts for a short-chain toxin to bind AChR with high affinity.

On the Epitope That is Recognized by M␣2-3 at the Surface of Ea
One approach to elucidate the determinants of a protein, here a toxin, that are recognized by an antibody and a receptor is to submit the protein to a mutational analysis with the view to identifying the residues whose mutations affect the stability of the protein-antibody and the protein-receptor complexes. Such a strategy has been followed in a number of cases, and the associated results nicely agree with those derived from structural analyses (69 -72). We tentatively mapped the epitope  3 H-Labeled toxin (5.2 nM; 25 Ci/mmol) to M␣2-3 (5 nM) were incubated overnight at 4°C in 0.05 M phosphate buffer, pH 7.2, with 0.5% bovine serum albumin and normal horse serum. The mixtures were precipitated by poly(ethylene)glycol 6000 (final concentration 12.5%) and centrifuged 20 min at 3000 rpm and 4°C. The pellets were suspended in 0.05 N NaOH supplemented with 10 ml of Aqualipoluma solution. Radioactivity was counted. recognized by M␣2-3 on the surface of Ea, a short-chain curaremimetic toxin (73) using a set of Ea mutants prepared as described previously (18 -20). In brief, preparation of the recombinant toxin as well as the mutants has successively re-quired the cloning of the cDNA encoding Ea (25), its expression as a fusion protein in E. coli (24) and the cleavage of the fusion protein by CNBr treatment (30). The resulting recombinant Ea was undistinguishable from the snake venom toxin regarding  (33). ⌬⌬G ϭ ⌬G mut Ϫ ⌬G WT ϫ ⌬G°was calculated from the equation ⌬G°ϭ ϪRTln1/K d , where K d is the equilibrium dissociation constant, T ϭ 293 K and R ϭ 1.99 cal/mol/K. Standard state is unit molarity. The residues for which mutations caused an affinity decrease greater than 10-fold (⌬⌬G Ͼ 1.3 kcal/mol) are indicated in bold. The functional residues that are commonly important for the recognition of both the antibody and receptor are underlined. Eb and Ec are two natural isotoxins differing from Ea by one and two residues, respectively. Asn-26 in Ea is replaced by His in Eb, and Asn-26 and Lys-51 in Ea are, respectively, replaced by His and Asn, in Ec. The monoacetylated residues have been prepared as described previously (1). The transversal dashed lines indicate, from top to bottom, the limits of the first loop, the large turn that joins the first to second loop, the limits of the second loop, and the limits of the third loop.    its functional (30) and structural properties (74). Using the same strategy, as many as 51 Ea mutants have been thus prepared. The apparent affinity constant (K d ) that characterizes the stability of the toxin-M␣2-3 complex was conveniently determined in solution, at equilibrium, on the basis of competition experiments using a sensitive and reproducible radioimmunoassay (34,75). Typical curves resulting from competition between a tritium-labeled toxin (32) and the wild type Ea or a number of mutants are shown in Fig. 4. To assess that the introduced mutations did not alter the overall structure of the toxin, each mutant was submitted to a circular dichroism analysis. The data indicated that the secondary structure was essentially preserved in all cases (19,20). 2 In addition, some mutants, including those that were substituted at position 8 (20), were submitted to x-ray analyses. These data clearly excluded any substantial change subsequently to mutation (76). Therefore, a substantial difference in binding affinities between the wild-type toxin and a mutant is unlikely to result from a major structural perturbation in the toxin; instead, it is likely to reflect the functional importance of the mutated residue.
These data are most conveniently analyzed in the frame of the three-dimensional structure of Ea. As previously reported (74,77), the polypeptide chain of Ea is organized into three adjacent loops forming a large ␤-pleated sheet, which emerges from a small globular core where are located the 4 disulfides. The toxin has two opposite faces (Fig. 5). The side that is closer to the C-terminal loop is slightly convex (Fig. 5B), whereas the other side is somewhat concave (Fig. 5A). Two complementary sets of evidences indicated the location of the epitope recognized by M␣2-3 on Ea surface. First, several residues could be mutated without affecting the binding of the toxin to the antibody. These amino acids, therefore, are unlikely to be involved in the epitope recognized by the antibody. These excluded residues are green in Fig. 5. They are located on the convex side of the toxin (Fig. 5B) and on the core region of the concave side (Fig. 5A). Second, 10 mutations markedly affected the toxin affinity for the antibody (Table II). These residues, colored in red, orange, or yellow, are Gln-7, Lys-27, Trp-29, Asp-31, Phe-32, Arg-33, Ile-36, Glu-38, Lys-47, and Ile-50. These results confirm and further extend the previous data based on the use of mono-modified derivatives of short-chain toxins, which indicated that modifications of Lys-27, Trp-29, and Lys-47 affected the toxin affinity for M␣2-3 (1). It is of particular interest that residues like Gln-28, Ser-30, or Thr-35, which are sequentially adjacent to critical residues, can be mutated without affecting the affinity of Ea for the antibody. This observation most clearly emphasizes that the epitope is only located on one of two faces of the toxin. Therefore, our data indicate that the epitope recognized by M␣2-3 forms an homogeneous area on the concave side of the toxin, with at least 10 residues spread on the three toxin loops, a feature that illustrates the topographical nature of the epitope. Thus, Lys-27, Trp-29, Asp-31, Phe-32, Arg-33, Ile-36, and Glu-38 are located on the second loop,  A and B, respectively. Residues whose mutations caused an affinity change for M␣2-3 lower than 3-fold are in green. They are considered to be excluded from the epitope. Residues whose mutations (at least one) caused an affinity change between 3-and 9-fold, 10-and 100-fold, and higher than 100-fold are, respectively, in yellow, orange and red.
whereas Gln-7 is on the first loop and Lys-47 and Ile-50 are on the third loop. The surface covered by the epitope is approximately 700 Å 2 , a value that is comparable with those observed in other protein-protein complexes (10,11,53,69,78).
Mutations of the residues indicated in red (Fig. 5A) caused affinity decreases higher than 100-fold. These are Q7L, W29H or W29F, D31H or D31N, R33E, E38K or E38L, K47E, and I50Q. Among these, the three mutations, Q7L, E38K or E38L, and W29H caused affinity decreases by factors higher than 500. Even more dramatic effects were seen with the mutation E38K or E38L, which caused affinity decreases of nearly 4 orders of magnitude. Certainly, the glutamate at position 38 plays a critical role in the formation of the Ea-M␣2-3 complex, possibly by establishing strong electrostatic contacts and/or hydrogen bonds with the antibody. Clearly, our findings agree with the general view that not all substitutions within an epitope equally affect the binding of the antigen toward a specific antibody (72, 79 -81). Thermodynamic analyses previously suggested that an epitope involving 10 -15 residues contains a smaller "energetic epitope" (12,82), which may contribute to most of the binding energy. Recently, a "hot spot of binding energy" was also observed at a hormone-receptor interface (14). Presumably, residues Gln-7, Trp-29, and Glu-38 form a predominant "energetic core" associated with the binding of Ea to the antibody. These three residues are spatially close to each other and form a compact cluster that is assisted by two separated groups of less important residues (Lys-27, Lys-47, and Ile-50; and Asp-31, Phe-32, Arg-33, and Ile-36, respectively) (Fig. 5A).
It was previously observed that Ea binds to M␣2-3 with an affinity 5-fold higher than toxin ␣, the original immunogen (1). As it is now identified, the epitope offers a possible explanation as to the heteroclitic behavior of Ea. In effect, one of the 10 functionally important residues of the epitope are different in Ea and toxin ␣; Phe-32 in Ea is H in toxin ␣. We have not yet introduced an histidine at this position in Ea; however, it is noticeable that mutation of Phe into Leu produces a nearly 13-fold affinity decrease.
Comparison of the toxin determinants recognized by M␣2-3 and AChR-In previous studies (18 -20), we have mapped the determinant by which Ea binds to AChR. In these studies, 36 Ea residues were substituted using 45 different mutations. These substitutions mostly concerned the concave side and the loop regions of the convex side, leaving the core area of the convex side not appropriately investigated. To better explore the convex side of the toxin, we now introduced six new individual substitutions, i.e. S22A, I37A, G42A, S57N, V59A, and N62A. In addition, we further investigated the core region by introducing two deletions (⌬Ser-18 and ⌬Pro-19) localized in the large turn that joins the loop I to the loop II. Finally, we further explored the functional role of lysine 47 (19 -20) by introducing a glutamic acid, which induces a charge reversion. Therefore, we have now probed individually as many as 44 positions of the toxin, using 53 mutations. In addition, Eb and Ec, two natural mutants of Ea (see Table II) are available (45), allowing us to probe the additional positions His-26 and Lys-51. Then if one excludes the 8 half-cystines whose mutation is anticipated to cause dramatic structural perturbations, the functional role of more than 80% of the remaining positions of Ea have now been explored individually, providing a clear picture as to how the toxin recognizes AChR. For sake of clarity, all the affinities and deduced ⌬⌬G values that have been determined either previously (18 -20) or in the present work are all indicated in the third and fourth columns of Table  II, respectively. Together with our previous data, the new mutants confirm that the convex side of the toxin, including both the core and loop regions, are excluded from the determinant recognized by the receptor. Also, in agreement with previously proposals that Lys-47 is functionally important, we now observed that introduction of a glutamic acid caused more than a 100-fold affinity decrease. As it is recalled in Table II, only individual mutations FIG. 6. Comparison between Ea epitopes recognized by the nicotinic acetylcholine receptor (AChR) (A) and the antibody M␣2-3 (B). The concave face of Ea is shown in both representations. Residues whose mutations caused affinity decrease for AChR or the antibody, lower than 3-fold are in green. Residues whose mutations (at least one) caused affinity decreases comprised between 3-and 9-fold, 10-and 100-fold, and higher than 100-fold are, respectively, colored in yellow, orange, and red. The residue for which mutation caused an affinity increase is in blue.
at Gln-7, Ser-8, and Gln-10 on loop I, Lys-27, Trp-29, Asp-31, Arg-33, and Glu-38 on loop II, and Lys-47 on loop III caused at least a 10-fold affinity decrease, and mutation at Ile-36 caused a substantial affinity increase. Thus, the functional determinant by which the toxin recognizes AChR includes 10 amino acids, just like M␣2-3. This is all the more interesting as the two complexes have different stabilities, with the toxin-antibody complex and the toxin-receptor complex being characterized, respectively, by K d values equal to 1.4 nM and 70 pM. Therefore, although two toxin determinants possess the same number of functional residues, they contribute differently to the stability of two toxin-protein complexes. Not only both determinants are composed of an identical number of residues, 80% of them are the same. These identical residues are Gln-7, Lys-27, Trp-29, Asp-31, Arg-33, Ile-36, Glu-38, and Lys-47, which cover an homogeneous surface of approximately 600 Å 2 (Fig. 6). That this core, conserved in nearly all short-chain toxins (45), is commonly recognized by the antibody and AChR offers a clear and simple explanation as to the functional properties of M␣2-3. Thus, it explains that the antibody (i) recognizes all short-chain toxins that possess this core; (ii) inhibits, by simple steric hindrance, the binding of Ea and other related toxins to AChR; and (iii) elicits anti-idiotypic antibodies that recognize AChR, and that these antibodies can in turn elicit anti-anti-idiotypic antibodies, which neutralize the toxin. The AChR-specific antibodies are likely to be raised against the complementary paratopic region of the antibody, which may, although it remains to be demonstrated, correspond to the region 106 -122 that is sequentially similar in CDR2H and -3H of the antibody and in ␣-subunits of the AChR.
Although the regions that are commonly recognized by the antibody and the receptor include identical residues, they cannot be considered as being "epitopically" identical, sensu stricto. There is a general propensity for mutations of epitope residues to cause higher affinity decreases in the toxin-antibody com-plex than in the toxin-receptor complex. This is better illustrated in Fig. 7, where ⌬⌬G values are represented as diagrams. In average, each individual mutation at positions 7, 27, 29, 31, 33, 38, and 47 is associated with an average ⌬⌬G value of approximately 3.1 kcal/mol in the toxin-antibody complex and 2 kcal/mol in the toxin-AChR complex. In other words, the introduced mutations tend to affect more the stability of the toxin-antibody complex than the toxin-AChR complex. This is particularly obvious with the mutation E38L, for which ⌬⌬G values are, respectively, more than approximately 5 kcal/mol and 1.87 kcal/mol, and with the mutation Q7L, where ⌬⌬G values are, respectively, 3.70 and 1.81 kcal/mol in the two complexes. This observation emphasizes the differential energetic contribution of the common residues to the stability of the two complexes (Fig. 6). As a consequence, one can anticipate that the interacting surfaces on the antibody and the receptor will not be identical in terms of structural and/or energetic complementarity.
In addition to these differences, each determinant possesses specific functional residues. Thus, Ser-8 and Gln-10 are uniquely part of the receptor-recognized epitope, whereas, conversely, Phe-32 and Ile-50 are uniquely part of the antibodyspecific epitope ( Fig. 6 and Table II). Clearly, aside from a number of functional and structural analogies, the two determinants present topological differences.

On the Mimicry between a Receptor and an Antibody
A few years ago, we searched for a monoclonal antibody that could share some binding properties with AChR (1). The rationale of the experiment was to immunize mice with one AChR-specific toxin and to screen the resulting clones with another toxin known to cross-react weakly with the immunizing toxin. It was anticipated that a monoclonal antibody that would bind to both toxins with high affinity would also recognize a surface commonly shared by the two toxins, i.e. the AChR binding region. One such antibody, M␣2-3, was obtained and, as demonstrated in the present paper, our anticipation was correct since 80% of the residues forming the antibody epitope belong to the determinant recognized by AChR. However, can we conclude that the selected antibody mimics the receptor?
Functionally speaking, M␣2-3 and AChR share some properties. They both recognize all tested short-chain toxins; however, only AChR recognizes the long-chain toxins, alkaloid antagonists and agonists. Also, they both elicit anti-AChR antibodies; however, whether the two immune responses are similar in terms of proportion of generated anti-AChR antibodies remains to be established. Structurally speaking, some CDRs of the antibody and the ␣-subunits of the receptor share approximately 66% amino acid sequence identity in their respective polypeptide chains. Although interesting, this observation does not constitute conclusive evidence that these sequences adopt similar conformations. In effect, the distinct contexts in which they occur in the antibody and receptor may influence them differentially in terms of adopted secondary structures (83). Finally, if the determinants recognized by the antibody and the receptor share 80% of critical residues, the relative energetic contributions of these residues are markedly different in the two complexes and furthermore 20% of the functional residues are different in the two epitopes. Therefore, in agreement with previous proposals (7), all our data aggregate to indicate that M␣2-3 is a partial and not an exact mimic of AChR, a conclusion that may be general for all receptormimicking antibodies.