Changing the Structural Context of a Functional -Hairpin SYNTHESIS AND CHARACTERIZATION OF A CHIMERA CONTAINING THE CURAREMIMETIC LOOP OF A SNAKE TOXIN IN THE SCORPION α/β SCAFFOLD

An approach to obtain new active proteins is the incorporation of all or a part of a well defined active site onto a natural structure acting as a structural scaffold. According to this strategy we tentatively engineered a new curaremimetic molecule by transferring the functional central loop of a snake toxin, sequence 26-37, sandwiched between two hairpins, onto the structurally similar β-hairpin of the scorpion toxin charybdotoxin, stabilized by a short helix. The resulting chimeric molecule, only 31 amino acids long, was produced by solid phase synthesis, refolded, and purified to homogeneity. As shown by structural analysis performed by CD and NMR spectroscopy, the chimera maintained the expected α/β fold characteristic of scorpion toxins and presented a remarkable structural stability. The chimera competitively displaces the snake curaremimetic toxin α from the acetylcholine receptor at 10M concentrations. Antibodies, elicited in rabbits against the chimera, recognize the parent snake toxin and prevent its binding to the acetylcholine receptor, thus neutralizing its toxic function. All these data demonstrate that the strategy of active site transfer to the charybdotoxin scaffold has general applications in the engineering of novel ligands for membrane receptors and in vaccine design.

The design of proteins with novel functions represents an exciting potential of protein engineering. Simple protein architectures or scaffolds (1)(2)(3)(4)(5)(6) have been obtained by de novo protein design, but only a few have been harnessed with a function, such as in the case of the heme-binding and redox-active proteins (7,8) and of the DNA-binding proteins (9,10). Genuine successes in the generation of functionally useful novel proteins, however, have been obtained by a more conservative approach which exploits the structure of some natural proteins as appropriate structural scaffolds and manipulates surface loops and insert new functional sequences in permissive regions. As an illustration of this strategy, a human-mouse chimera was obtained by the transfer of the complementaritydetermining regions of a mouse antibody into the corresponding one of a human myeloma protein, with concomitant reproduction of antibody specificity into the human protein (11). The resulting "humanized" antibodies are not recog-nized as foreign by the immune system and then possess great therapeutic potential. Other examples include the insertion of elastase inhibition activity in interleukin 1␤ by transfer of the protease inhibitor loop into a structurally compatible loop of the cytokine scaffold (12), the recruitment of RNase activity in angiogenin (13) or angiogenic activity in RNase A (14) by exchanging surface loops, the recruitment of carboxypeptidase activity into RTEM-␤ lactamase by introduction of a 28-amino acid segment of carboxypeptidase (15), the substitutions of few key residues in human prolactin to generate binding to the receptor of the homologous growth hormone (16), the stabilization of the Bacillus subtilis neutral protease obtained by the insertion of the 10-residue calcium-binding loop of thermolysin into the corresponding region of the bacterial protease (17), and the stabilization of subtilisin BPNЈ by incorporation of a Ca 2ϩbinding loop from the homologous thermophilic thermitase (18).
Small disulfide-rich structures are interesting candidates as structural scaffolds for protein engineering studies, since in many cases the disulfide bridges provide most of the stabilizing energy, leaving a large part of the sequence available for substitutions or insertions. Kunitz trypsin inhibitor was transformed into a powerful neutrophil elastase inhibitor by replacing the active loop with a new sequence deduced on the basis of phage display and selection methodology (19); a highly efficient inhibitor of human leukocyte elastase was engineered on the scaffold of human pancreatic trypsin inhibitor (20) and a bisheaded inhibitor of trypsin and carboxypeptidase A on the scaffold of a squash inhibitor (21) on the basis of a rational structure-based design.
Recently, we proposed the small disulfide-stabilized structure of the scorpion charybdotoxin as a basic scaffold, able to present guest sequences in a defined and well ordered conformation as a means to engineer novel proteins (22). The scorpion toxin structure (Fig. 1), containing only 37 amino acids, consists of a short ␣-helix and a triple stranded ␤-sheet, with three disulfide bridges forming most of the interior core. This structural motif is common to all scorpion toxins, irrespective of whether they contain 60 -65 amino acids, as in long scorpion toxins, active on Na ϩ channels or contain 30 -40 amino acids, as in short scorpion toxins, active on K ϩ channels and Cl Ϫ channels (23,24). The same structural motif is also present in insect defensins (24), expressing antimicrobial activity through membrane permeabilization and in plant ␥-thionins (25). The only amino acid residues common to all these proteins are the six cysteines forming the three disulfide bridges (23,24). This simple, compact, and well organized structural motif seems to have been naturally selected for its high sequence permissiveness (as shown by its compatibility with hundreds of different sequences from scorpion toxins and insect defensins) and func-* 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.
‡ Supported by a grant from the Institut de Formation Supérieure Biomédicale.
§ To whom correspondence and reprint requests should be addressed. Tel.: 33-1-69087133; Fax: 33-1-69089137. tional versatility (as shown by its compatibility with Na ϩ , K ϩ , Cl Ϫ channel blockage activity in scorpion toxins and antibacterial activity in defensins). A major difference of this structural scaffold from the previously reported ones is that a large part of its molecular surface might be substituted by a novel sequence and not just the exposed loops.
On the basis of these considerations, we thought that it was possible to artificially engineer many more functions on this fold than those that are naturally selected. Thus, by transferring the carbonic anhydrase metal binding site into the ␤-sheet of the scorpion scaffold, we engineered a new metal binding activity with affinity and selectivity for transition metals (22). We here report the engineering of a novel binding activity toward a membrane receptor, the acetylcholine receptor (AchoR), 1 by re-designing the ␤-sheet molecular surface of the charybdotoxin scaffold and taking advantage of the knowledge of the structural similarity between the charybdotoxin ␤-sheet region and that of natural ligands of this receptor contained in snake venoms.
Snake venom contains toxins that bind to the AchoR at the postsynapic membranes of skeletal muscle, thus provoking flaccid paralysis. This activity is similar to that shown by the alcaloid curare and is of potential clinical interest in anesthesia. Curaremimetic neurotoxins constitute a family of proteins with a high sequence similarity that present a similar overall folding, consisting of three adjacent loops forming a ␤-pleated sheet, which emerge from a globular core containing four conserved disulfides (26,27). On the basis of chemical modification and mutational analysis studies, several amino acid residues implicated in the AchoR binding of one of these toxins, erabutoxin a, were identified (28,29). These residues define a continuous surface, forming the curaremimetic site of snake neurotoxins, centered around the central loop (residues 26 -37) of the concave face of the ␤-sheet platform, containing six of the ten most "active" residues. Toxin ␣, from the venom of Naja nigricollis, has high sequence similarity and the same residues in the concave side of the central loop as erabutoxin a, besides the substitution His 33 3 Phe. Its three-dimensional structure ( Fig. 1) has been recently solved on the basis of NMR data (27) and has been shown to be highly similar to the crystal structure of erabutoxin a (30).
The mutational studies of curaremimetic toxins suggest that being able to fix the sequence of the central loop of toxin ␣ in its native conformation should result in a molecule possessing some affinity for AchoR. On the basis of a backbone structural similarity between the loop sequence 26 -37 of toxin ␣ and the ␤-hairpin 25-36 of the charybdotoxin structure, we transferred the snake sequence into the scorpion framework. The newly designed chimeric protein was produced by solid phase synthesis. Its structure was analyzed by CD and NMR spectroscopy. Its function was investigated by competitive binding to AchoR and by analyzing the specificity of the antibodies produced by immunization of this chimeric construction in rabbits.

EXPERIMENTAL PROCEDURES
Materials-Solvents and reagents for peptide synthesis were obtained from SDS (Peypin, France) and Nova Biochem, respectively. Oxidized and reduced glutathione, trypsin, chymotrypsin, and bovine serum albumin were from Sigma Toxin ␣, from Naja nigricollis, was purified as described previously (31). Charybdotoxin (synthetic) was obtained as published (32). Denatured chimera, used in binding assays, was obtained by carboxymethylation (33) of the six cysteines after disulfide reduction of the chimera and purified by HPLC. Other chemicals were of the highest grade available.
Disulfide Formation and HPLC Purification-The crude peptide (50 mg) was dissolved in water (5 ml) and then added to the oxidizing buffer (500 ml), consisting of 0.05 M sodium phosphate, 0.2 M NaCl, 5 ϫ 10 Ϫ3 M reduced glutathione, and 5 ϫ 10 Ϫ4 M oxidized glutathione, pH 7.8. After 2 h at room temperature the solution was acidified to pH 3.0 with acetic acid and then directly applied to an Aquapore RP 300 column (25 ϫ 1 cm), at 3 ml/min flow rate. The peptide was then eluted at 6 ml/min with a 30-min 12-33% acetonitrile gradient in 0.1% trifluoroacetic acid. Fractions containing the correct oxidized peptide were analyzed by analytical HPLC; single peak fractions were pooled and lyophilized; impure fractions were pooled, lyophilized, and purified on a Vydac C18 column (25 ϫ 1 cm) by using a 40-min 15-25% acetonitrile gradient in 0.1% trifluoroacetic acid.
Chemical Characterization of the Chimera-Analytical HPLC was carried out on a Spectra-Physics system formed by a P2000 pump, UV2000 detector, and CromJet recorder. Preparative HPLC was carried out on a Merck-Hitachi system, consisting of an L6200 pump, L4000 UV detector, and D-2500 integrator.
For amino acid analysis, peptide samples (50 g) were hydrolyzed in ultrapure 6 N HCl for 16 h at 120°C in sealed evacuated tubes. The hydrolysates were then analyzed on an Applied Biosystems model 130A automatic analyzer equipped with an on-line model 420A derivatizer for the conversion of the free amino acids into their phenylthiocarbamoyl derivatives.
Mass analysis was performed in a Nermag R10-10 mass spectrometer, coupled to an Analitica of Branford electrospray source. The quadrupole was scanning over the range m/z 300-2000. A HP-ChemStation software (Hewlett-Packard) was used to drive the spectrometer and to acquire the data.
For peptide mapping, the folded oxidized purified product (70 g), dissolved in 100 l of 0.1 M Tris⅐HCl buffer, pH 7.6, containing 1 mM iodoacetamide (to prevent disulfide scrambling) was digested for 15 h at 30°C with a mixture of trypsin (3 g) and chymotrypsin (3 g). Digest was analyzed by reverse phase HPLC on a Vydac C18 column (25 ϫ 0.46 cm) using a 30-min linear gradient of 1-25% acetonitrile in 0.1% trifluoroacetic acid at a flow rate of 1 ml/min. The eluate was monitored at 214 nm. Peaks were collected, lyophilized and analyzed by electrospray mass spectrometry.
Structural Characterization-Peptide concentrations were determined by 280 nm absorbance readings by using a molar extinction coefficient of 11,740, calculated on the basis of the known amino acid content (38).
Circular dichroism spectra were obtained with a Jobin-Yvon CD6 dichrograph, driven by an IBM-PC operating with a CD6 data acquisition and manipulation program. Spectra were run at 20°C in 5 mM sodium phosphate, pH 7.0, and from 180 to 255 nm, with a 0.1-cm quartz cell and a protein concentration 1.5-2.0 ϫ 10 Ϫ5 M; from 250 to 320 nm with a 1.0-cm quartz cell and a 1.0 -1.5 ϫ 10 Ϫ4 M protein concentration.
All the structures were displayed, analyzed, and compared on a Silicon Graphics 4D/25 station using the SYBYL package (Tripos Associates, Inc.).
Functional Characterization-Binding of the chimera to AchoR was performed by competitive inhibition of labeled toxin ␣ binding to microtiter plates, coated with AchoR according to the published procedure (45). Torpedo marmorata electric organ membrane fragments were prepared according to standard procedures (46), and acetylcholine receptor was solubilized as described previously (47). 96-Well microtiter plates (Maxisorb, Nunc) were coated with 100 l of solubilized receptor at a concentration of 2 g/well in 0.1 M carbonate buffer, pH 9.5, for one night at 4°C and saturated with 200 l of 10 mM sodium phosphate buffer, pH 7.4, containing 0.15 M sodium chloride and 0.3% bovine serum albumin (PBSA) for 2 h at room temperature. Soluble inhibitors were diluted in 50 l of the saturating solution and were titrated into the wells. Starting concentrations ranged from 10 Ϫ7 M for the native toxin to 10 Ϫ4 M for the chimera and the denatured chimera. A constant dilution of horseradish peroxidase (HRP)-conjugated toxin ␣ (45) in 50 l was then added to the wells. This dilution corresponded to 50% binding in direct binding assays. The plate was then incubated for either 4 h at 20 or 24 h, 48 h at 4°C. After washing five times with 0.1 M Tris⅐HCl buffer, pH 7.5, containing 0.05% Tween 20, the binding of HRP toxin ␣ was detected by using the substrate 2,2Ј-azino-bis(3ethylbenzthiazoline-6-sulfonic acid) (ABTS, Pierce) and by measuring the absorbance at 414 nm, after 30-min incubation. Inhibition of binding of labeled toxin ␣ to AchoR was expressed as the concentration of the inhibitor giving 50% inhibition of binding (IC 50 ).
Antibodies directed against the chimeric construction were obtained by immunizing, subcutaneously and at multiple sites, two rabbits (blanc du Bouscat, Wiss) with 200 g of chimera in complete Freund's adjuvants. Rabbits were reimmunized three times in incomplete Freund's adjuvants with the same quantity of peptide at 21-day interval. Animals were bled 15 days after each boosting, and their sera were tested for specific antibody production by enzyme-linked immunosorbent assay (ELISA).
Elicited antibodies were titrated against either the chimera and toxin ␣, coated on microtiter plates (1.0 and 0.1 g/well, respectively). After overnight incubation at 4°C, wells were then saturated with PBSA (saturating buffer), 100 l of rabbit antiserum, diluted 1/50 in saturating buffer, were added to the well (serially diluted 1/3) and incubated for 2 h at room temperature. After washing, 100 l of goat anti-rabbit IgG conjugated to HRP (Sigma), diluted 1/6000, was added to each well for 1 h at room temperature. Bound antibodies were detected as above using ABTS as substrate.
Relative affinities of the purified antibodies for the chimera and native toxin ␣ were tested by competitive inhibition ELISA. These tests were performed by coating the purified anti-chimera antibodies into the wells (0.5 g/well) and, after saturation and washing, by titrating the soluble antigens (50 l) into the wells. Starting concentrations were 10 Ϫ6 M for chimera and toxin ␣ and 10 Ϫ5 M for the denatured chimera. A constant dilution (corresponding to 50% binding in direct binding assays) of HRP-toxin ␣ in 50 l was then added to the wells, and the plate was incubated for 2 h at room temperature. After extensive washing the bound HRP-toxin ␣ was detected as described above.
Antibody neutralization of toxin ␣ binding to AchoR was performed by using anti-chimera antibodies. Microtiter plates were coated with solubilized AchoR as described above. Serial dilutions of 50 l of either the affinity-purified anti-chimera antibodies or the monoclonal antibodies M␣ 2-3 (48) or M 14-1-1 (49) were added to the wells. The starting concentration for all three antibodies was 5 ϫ 10 Ϫ7 M. A constant dilution of HRP-toxin ␣ in 50 l was then added to the wells, and the plate was incubated for 3 h at room temperature. Bound HRP-toxin ␣ was detected as described above. The specificity of the binding of anti-chimera antibodies to toxin ␣ was tested by incubating the antibodies with 10 Ϫ6 M chimera 2 h before adding the HRP-toxin ␣ conjugate.

RESULTS
Design of the Curaremimetic Chimera-Mutagenesis experiments (28,29) have shown that the concave face of the central loop of curaremimetic toxins is involved in AchoR binding. The structure of this loop in toxin ␣, residues 26 -37, was taken as the reference model for the curaremimetic site ( Fig. 1). This loop has the conformation of a ␤-hairpin, formed by the ␤-strand 26 -29, the ␤-turn 30 -33 followed by the ␤-strand 34 -37. Charybdotoxin, the structural scaffold, presents a ␤-hairpin in position 25-36 (Fig. 1B). Superposition of the snake ␤-hairpin (sequence 26 -37 of toxin ␣) to the scorpion ␤-hairpin (sequence 25-36 of charybdotoxin) results in a good fitting with a calculated 1.1 Å root mean square difference for the backbone atoms, suggesting that the ␤-hairpin sequence may be transferred from one protein to the other, with little perturbation of the guest original conformation and the host overall fold. However, structure superposition and sequence alignment (Fig. 2) show that Val 27 , Arg 29 , Thr 34 , and Ile 36 of toxin ␣ correspond to the structurally important (23, 24) Gly 26 , Cys 28 , Cys 33 and Cys 35 of the scorpion scaffold. Mutations of the four above residues in erabutoxin a (29) have no functional effect, as expected from the fact that the side chains of these residues point to the opposite side of the active concave face of the toxin. Consequently, the above four residues of toxin ␣ were mutated into the corresponding scorpion residues (Fig. 2). Furthermore, in the superposition analysis, the 1-6 segment of the scorpion charybdotoxin scaffold does not correspond to any segment of the toxin ␣ structure, and this segment can be removed in charybdotoxin without apparent perturbation of the remaining structure (32): this segment was thus simply deleted, in order to allow full solvent accessibility to the trans- ferred site. Accordingly, a new sequence was designed in which eight residues of the truncated 7-37 sequence of charybdotoxin have been mutated, Arg 25 3 Lys, Lys 27 3 Trp, Met 29 3 Asp, Asn 30 3 His, Lys 31 3 Arg, Lys 32 3 Gly, Arg 34 3 Ile, Tyr 36 3 Glu (Fig. 2). This sequence, 31 amino acids long, corresponds to a newly engineered molecule, the curaremimetic chimera, containing six of the ten functionally important residues of toxin ␣ (Lys 26 , Trp 28 , Asp 30 , Arg 32 , Ile 35 , Glu 37 ) (28,29), fixed by the host scorpion scaffold in a ␤-hairpin motif, and clustered in space close to each other, thus forming a molecular surface designed to reproduce the central part of the curaremimetic site interacting with AchoR.
Synthesis and Purification of the Curaremimetic Chimera-The chimeric sequence (Fig. 3) was obtained by automated solid phase synthesis by using the already established protocol, well performing for small and disulfide-rich proteins (50) and used in the previously engineered molecule (22,36). The crude product showing a major peak in reverse phase HPLC (Fig. 3A) was directly oxidized in redox buffer to allow disulfide bond formation and purified by reverse phase HPLC on a semipreparative column. Fig. 3 (A-C) shows the different purification steps of the synthetic chimera. The chemistry approach allowed to obtain 10.5 mg of pure product, starting from 50 mg of crude synthetic material (21% yields) in a short as a week time period. Purity and identity of the purified product were assessed by analytical HPLC (Fig. 3C), amino acid analysis (not shown) and electrospray mass spectrometry (determined mass, 3551.5 Da; calculated mass, 3552.07 Da). Peptide mapping was used to identify the disulfide bridges. Tryptic and chymotryptic fragments were purified by HPLC (not shown) and identified by electrospray mass spectrometry. The isolated fragments, 7-9/28 -31 (retention time: 10 This analysis demonstrates that the synthetic reduced chimeric sequence was able to form the native disulfide bonds in high yields (Fig. 3B), similarly to charybdotoxin (32), suggesting that the introduced sequence mutations and deletions did not compromise the ability of the sequence to spontaneously form the native disulfides. We interpreted this finding as the first indication that the chimeric sequence may be folded similarly to charybdotoxin.
Structural and Stability Studies-CD spectroscopy was used to analyze the secondary structure of the chimera in solution. The far-UV CD spectrum of the chimera (Fig. 4A) shows two minima near 210 and 220 nm and a maximum near 190 nm, suggesting the presence in this molecule of an ␣-helical conformation, as expected. Charybdotoxin, which contains an ␣ ϩ ␤ conformation (23,24), shows a far-UV CD spectrum with a minimum at 218 nm and a maximum at 195 nm, typical of a ␤-sheet conformation, indicating that the CD contribution of this secondary structure dominates the spectrum. The observed differences in the two spectra of Fig. 4A are consistent with the fact that the chimera is predicted to contain a reduced ␤-sheet secondary structure, since it does not contain the Nterminal 1-6 sequence, forming the third strand of the antiparallel ␤-sheet in charybdotoxin. In the near-UV region (Fig.  4B), the spectrum of the chimera shows features similar to that of charybdotoxin, with a broad negative band in the 250 -300 nm region, representing the disulfide contribution, and a shoulder at 282 and 288 nm, attributable to the tryptophan contributions (51). The noticeable differences, as the absence of shoulders at 262 and 268 nm, and the even more marked bands at 282, 288, and 292 nm in the chimera, are fully consistent with the different aromatic amino acid composition, i.e. the absence in the chimera of phenylalanine and the addition of one extra tryptophan. This spectral analysis suggests that the chimera presents disulfide bond conformations similar to charybdotoxin and the even more marked tryptophan contribution suggests that Trp 27 , introduced in the curaremimetic site, is present in a well ordered conformation.
The chimeric protein is soluble in water at millimolar concentration, thus allowing 1 H NMR characterization. Assignment of proton resonances was achieved according to the standard method developed by Wü thrich and co-workers (52). The backbone proton chemical shifts are close to those of charybdotoxin: differences higher than 0.25 ppm concern only a few residues, located at the N terminus (Cys 7 , Thr 8 , Thr 9 ) or in the mutated part of the sequence (Gly 26 , Trp 27 , Asp 29 , Arg 31 , Gly 32 , Ile 34 , Ser 37 ). 2 The secondary structure of the chimeric protein was established by analyzing the backbone proton NOEs and the 3 J HN-H␣ coupling constants. The pattern of strong sequential and long range dNN connectivities for residues 12-21 indicates the presence of a helical conformation, confirming the CD analysis. Furthermore, the presence of strong sequential d␣N NOEs, the typical pattern of long range d␣␣, dNN, and d␣N connectivities, and the large values of the coupling constants indicate that residues 25-28 and 33-36 are involved in a two-strand antiparallel ␤-sheet. Thus, the secondary structure elements of charybdotoxin (23) are found at the same positions in the new protein. The similarity of the chemical shifts and the conservation of the secondary structure elements in the new protein suggests that the tertiary structures of the two proteins are very similar. Full description of NMR data, three-dimensional structure resolution, and analysis will be reported elsewhere. 2 Charybdotoxin is characterized by a remarkably high conformational stability: CD spectroscopy reveals no significant conformational changes when the protein is heated up to 90°C (Fig. 4C) or is dissolved in 5 M guanidine HCl. When the chimera is treated under the same strong denaturing conditions, the far-UV CD spectrum (Fig. 4C) indicates the protein is still substantially folded. Thus, in spite of the fact that all the surface of the ␤-sheet has been re-designed and the N-terminal segment of charybdotoxin deleted, the chimera exhibits an exceptional conformational stability, comparable with that of the original scorpion scaffold.
Binding Activity on AchoR-Binding affinity of the chimera for AchoR from the electric fish T. marmorata was evaluated on the basis of competition experiments by using the solubilized receptor coated to the microwell plate and toxin ␣ conjugated with peroxidase (HRP) as a tracer. As shown in Fig. 5, the chimera inhibits the binding of the labeled toxin ␣ to AchoR, in a dose-dependent manner. Varying the incubation time from 4 to 48 h (not shown) did not affect the inhibition potency of the chimera, indicating no kinetics dependence in the experimental conditions used. Denatured chimera did not show any affinity for AchoR on the basis of competition experiments (Fig. 5): this indicates that the chimera, which has been shown by CD and 1 H NMR to contain the main features of the scorpion scaffold, shares some binding properties of toxin ␣. The determined IC 50 of the chimera is 5.10 Ϫ5 M, which has to be compared with that of toxin ␣, 3.10 Ϫ10 M (Fig. 5). The solid phase method used does not allow the determination of a real affinity constant, but is quite reproducible and provides an easy and sensitive estimation of a relative affinity. The affinity of toxin ␣ for T. marmorata AchoR is 2.10 Ϫ11 M (53); assuming a proportionality between IC 50 and K d , the affinity of the chimera for AchoR may fall in the micromolar range, which is 10 5 lower than that shown by toxin ␣. This low value of affinity prevents its evaluation by the more conventional filtration assay, previously used in the characterization of toxin mutants (28,29).
Charybdotoxin is a powerful blocker of K ϩ channels, activated by both intracellular Ca 2ϩ increase and membrane depolarization (54 -56). The chimera, which maintains a folding similar to charybdotoxin, has been tested on rat skeletal muscle Ca 2ϩ -activated K ϩ channels inserted in a lipid bilayer (57): no evidence of blockage was recorded up to a chimera concentration of 1 M. 3 Thus, the chimera, which presents a completely redesigned ␤-sheet sequence (and a N-terminal truncation), does not retain any affinity for the K ϩ channel, while it has acquired a new affinity for a different receptor, the AchoR.
Characterization of Anti-chimera Antibodies-In order to test if the transferred ␤-hairpin in the chimera indeed represents an image of the original curaremimetic loop of toxin ␣, we injected the chimera in rabbits, and we analyzed if the elicited antibodies could recognize native toxin ␣. The chimera was Engineering of a Curaremimetic Snake-Scorpion Chimera injected without being coupled to a carrier. After the third booster immunization, on day 100, the serum titer (in both immunized rabbits) was quite high and close to 10 6 . This high immunogenic response may look surprising from such a small molecule, but its high conformational stability, which may increase its t1 ⁄2 and maintain a high concentration in the blood, and the possibility that the scorpion scaffold may contain epitope inducing T cell help, as shown by the even higher immunogenicity of native charybdotoxin, 4 may explain this experimental finding.
When tested in direct binding assays, the anti-chimera-elicited antibodies were able to recognize coated toxin ␣ (not shown). In order to select a more specific antibody population recognizing uniquely the transferred ␤-hairpin, we purified the antiserum by affinity chromatography on a column containing immobilized toxin ␣. Furthermore, to avoid partial unfolding of the chimera during the coating procedure and to minimize artifacts due to the ELISA plates, we used a solution assay in which the chimera and the toxin ␣ were in solution free to compete with labeled toxin ␣ for binding to the coated antibodies (45). The results of this binding inhibition experiment (Fig.  6A) reveal that the chimera and the parent toxin are equally well recognized by those antibodies, indicating a full antigenic equivalence between the chimera and the toxin ␣ for the selected antibody population. Absence of reaction with denatured chimera and native charybdotoxin (Fig. 6A) demonstrates that the antibody subpopulation is indeed recognizing, in a conformation restricted manner, a surface region that is shared by both toxin ␣ and the chimera: this region is the part of the toxin ␣ curaremimetic site that has been transferred from the snake toxin to the scorpion scaffold.
Neutralization of Toxin ␣ Binding to AcChoR by Anti-chimera Antibodies-We investigated the ability of the anti-chimera affinity chromatography-purified antibodies to neutralize the binding of the toxin ␣ to AchoR by using the same solid phase colorimetric assay used for the biological activity studies. Two monoclonal antibodies were also used as controls: M␣ 2-3 , as positive control, which is an antibody raised against the native toxin and is able to neutralize its toxic activity (48) and M 14-4-4, which recognizes the class II major histocompatibility complex molecule (49), used as negative control. Fig. 6B demonstrates that the purified anti-chimera antibodies could efficiently inhibit the binding of toxin ␣ to AchoR in a dose-dependent manner. In addition, its effect was specific, since it was abolished by the presence of an excess of the chimera (note that the concentration of the chimera used, 10 Ϫ6 M, is not sufficient by itself to inhibit the binding of the toxin as shown in the biological activity assay (Fig. 5)). The neutralizing potency of anti-chimera antibodies was as high as the one of the monoclonal antibody M␣2-3, according to the measured IC 50 values (3.10 Ϫ8 M and 5.10 Ϫ8 M for sera and the monoclonal, respectively).

DISCUSSION
All of our results show that, by redesigning a part of the molecular surface of the natural scaffold of charybdotoxin, we were able to engineer a chimeric protein exhibiting a new function. Naturally, the structural scaffold used in our design has evolved to block the K ϩ channel (at nanomolar concentration). By systematic mutagenesis studies, the ␤-sheet region of charybdotoxin, comprising the residues 25-36, has been mapped as the molecular surface intimately interacting with the K ϩ channel mouth (58), thus physically obstructing the K ϩ ion flux through the ionic pore. Other scorpion toxins also seem to use this region to interact with other K ϩ channel subtypes, with different affinities. In any case, although the active molecular surfaces of the various scorpion toxins are different as defined by their amino acid sequences, their overall structure is similar, as shown by the published three-dimensional structure of charybdotoxin (23,24), iberiotoxin (59), PO5 (60), kaliotoxin (61), margatoxin (62), chlorotoxin (63), agitoxin (64), and noxiustoxin (65). We adopted charybdotoxin as a structural scaffold and replaced the amino acid side chains forming the exposed functional ␤-sheet surface with those of a loop from a curaremimetic neurotoxin presenting a structurally similar ␤-hairpin. This way we succeeded in generating a novel and artificial ligand, with micromolar affinity for AchoR, which is completely unrelated to the K ϩ channel.
The present results confirm previous examples reported in the literature (see Refs. 66 and 67 for two recent reviews), showing that novel activities can be generated on protein scaffolds by replacing exposed functional loops. However, two major differences characterize our engineering work. First, most of the literature examples include a loop transplantation between structurally homologous host proteins. In our case no overall structural homology exists between charybdotoxin and toxin ␣. Local structural similarity, however, exists between the two exchanged loops: they form two regular ␤-hairpins (Fig.  1), superimposable with little differences in the spatial disposition of their backbone atoms. Second, although Hynes et al. (68) and Wolfson et al. (12) showed that loops can be transferred from one structural context to another as structural 4 E. Drakopoulou and C. Vita, unpublished result. cassette without perturbating their own structural and functional properties, a notable difference exists between these two examples and our work: the transferred loop 26 -37 in our chimeric construction is fixed to the scaffold by three covalent disulfide bonds and not just held by the N and C termini. These bonds, involving cysteine residues within the transferred loop, are supposed to limit considerably the conformational space of the transferred loop. Even if we cannot disprove that solely the four residues of the turn, between Cys 28 and Cys 33 , are responsible of the observed functional properties of the chimeric construction, we believe the entire transferred sequence is responsible of that. The fact that no receptor binding or antibody recognition is exhibited by the unfolded linear chimera (Figs. 5 and 6) stresses, by itself, the importance of the disulfide bonds in determining the structural and functional properties of our chimeric construction. The two above differences are the direct consequences of two major properties of the scorpion scaffold, advantageous for protein engineering: (i) functional versatility and (ii) sequence permissiveness. These two characteristics are directly related and the second may explain the first: in fact, permissiveness in sequence mutation may allow this fold to adapt its structure to the different functions of blocking the different ionic channel types (in toxins) and also of forming holes in membranes (in defensins).
Insertion of exogenous sequences in protein structures most often results in a decrease of overall conformational stability of the host scaffold (68,14). This is one of the major difficulties, for example, in the engineering of the immunoglobulin scaffold and in rat antibody humanization. The ␣/␤ scorpion scaffold does not suffer from this weakness: even after multiple substitutions (8 residues upon 31), it maintains its exceptional conformational stability. The previously designed metal binding protein (22), containing the same scorpion fold, also maintains such a stability. This stability makes a powerful structural solution of the scorpion scaffold, which naturally evolved to be compatible with a great variety of sequences and functions. We believe that the presence of an interior core formed essentially by three disulfides (69) may explain not only the sequence permissiveness but also the structural stability of the scorpion scaffold.
We have determined the three-dimensional structure of the chimera on the basis of 1 H NMR data: detailed description of this structure will be reported elsewhere. 2 The analysis of this structure confirmed the presence of the ␣/␤ fold in the chimera, as suggested by CD data, and of a conformation in the trans-ferred loop similar to that of the central loop of toxin ␣, as suggested by the strong competition between the chimera and the toxin ␣, seen in the immunological characterization. However, structural differences exist at the level of some amino acid side chain orientations. It is well documented that single point mutations of biologically active surface may be enough to cause the loss of several order of magnitude in biological potency, e.g. in erabutoxin a the substitution Arg 33 3 Glu result by itself in a 318-fold lower affinity molecule (28), with consequent loss of 3.8 Kcal⅐mol Ϫ1 in receptor binding energy. Differences in the orientation of some critical side chains thus may be responsible for the low potency of the chimera as AchoR ligand. However, to account for the large difference in biological activity of the chimera (10 5 -fold difference in affinity, which accounts for a loss of 7.0 Kcal⅐mol Ϫ1 ) as compared with toxin ␣, other factors must play additional important roles. Recent mutagenesis data (29) have defined the functional site of erabutoxin a as an approximately 700-Å 2 surface, including residues from loops I, II, and III (Fig. 7). Our chimeric construction contains only the active residues of loop II, corresponding to the central part of the curaremimetic site (colored in red in Fig. 7). The absence in the chimera of the functional residues of loops I and III (in orange in Fig. 7), by itself, may explain its lower affinity. For example, from the data of Trémeau et al. (29), we calculated that the contribution to binding free energy of Ser 8 and Gln 10 of loop I and Lys 47 of loop III accounts for 3.1, 3.2, and 2.0 Kcal⅐mol Ϫ1 , respectively. Thus, the deletion of these residues is predicted to account for a total energy loss of 8.3 Kcal⅐mol Ϫ1 : this figure has to be compared with the 7.0 Kcal⅐mol Ϫ1 lower binding energy exhibited by the chimera. These calculations assume a simple additivity in free energy of binding (70). We did not experimentally prove the additivity of mutational effects in the case of curaremimetic toxins; however, this simple calculation may serve to indicate that, even if the structural resemblance of the transferred site to the original one might be improved in future designs, the biological activity of the actual chimera may be close to that expected by the central loop of toxin ␣ alone. This consideration suggests that a substantial improvement of the binding activity of the chimera may be obtained if, in addition to the modifications that might be suggested by its detailed structural analysis, we also combine a nonrational combinatorial approach. The substitution Ile 36 3 Arg in erabutoxin a has already been shown to increase the binding affinity by a factor of 7 (29): similar substitutions may be also selected, if single positions of the curaremimetic site of FIG. 7. Space-filling structure of the chimera (left) and the snake toxin ␣ (right). In red are the residues corresponding to the transferred site in the loop II of toxin ␣ (left) and in the chimera (right). In orange are the residues of loops I and III of the toxin ␣ active site (28,29) that contribute more than 1 Kcal⅐mol Ϫ1 to the energy of acetylcholine receptor binding. the chimera will be subjected to random substitution by combinatorial chemistry and the best binder to the AchoR selected on the basis of screening tests. A recent work on the minimization of the atrial natriuretic peptide hormone (71) has shown that indeed affinity loss due to deletion of active residues may be rectified by optimizing the remaining residues by means of phage display and selection methodology.
A close structural resemblance between the transferred site in the chimera and that present in the parent snake toxin is clearly apparent from the immunological characterization of the antibodies, elicited by immunization with the chimeric construction. This resemblance is the basis of the efficient recognition of the snake toxin and the prevention of its binding to AchoR by these antibodies. Clearly, the chimera may represent a less toxic, but equally effective, immunogen than snake toxins themselves in the production of toxin neutralizing antibodies. Faithful reproduction or molecular mimicry of a protein antigenic site is of fundamental importance in the design of novel constructs able to elicit high affinity antibodies for use as vaccines. Due to the high chain flexibility, simple peptides are not able to mimic the conformational aspects of protein epitopes. Insertions of antigenic epitopes in carrier proteins (72,73) or in de novo designed scaffolds (74,75) have been already used in an attempt to mimic discontinuous determinants for the purpose of activating B-cells specific for native proteins. The results presented here demonstrate that immunization with a highly constrained peptide immunogen is sufficient to induce high titer antibodies, with high affinity and specificity for a toxic protein antigen. Our immunogen, which is able to reproduce (or mimic) the secondary and tertiary conformation of a toxin protein epitope, has been constructed on the basis of an epitope transfer to a permissive, versatile, and stable host scaffold. Utilization of this approach and similar ones are significant in the rational design of synthetic vaccines.
In our previous work (22), we proposed the ␣/␤ scorpion fold as a natural structural scaffold for protein engineering. By transferring the carbonic anhydrase metal binding site in it, we engineered a novel metal-binding protein, exibiting affinities and selectivity for different metal ions. The present results confirm that this fold is a useful structural scaffold and host plate form for the construction of highly stable miniproteins, presenting transplanted sites in a fixed and well defined conformations, thus extending its scopes and potentialities to the generation of biologically active proteins involved in relevant protein-protein recognition processes.
Furthermore, the presence in this fold of structural motifs (␣-helix, ␤-sheet, loops) generally found in globular proteins, the tolerance for sequence mutations, and its retained stability after multiple substitutions appear as unique qualities of this fold and extremely valuable for protein engineering. In addition, in the creation of novel proteins, since apparently any region can be engineered with new sequences, a combinatorial and functional selection approach can be suitably used to improve the design. Furthermore, given the reduced size of the scaffold, the chemical methods can increase the structure diversity of novel proteins by using unnatural amino acids, peptide mimetics, or labeled compounds.