Solution Conformation of αA-conotoxin EIVA, a Potent Neuromuscular Nicotinic Acetylcholine Receptor Antagonist from Conus ermineus

We report the solution three-dimensional structure of an alphaA-conotoxin EIVA determined by nuclear magnetic resonance spectroscopy and restrained molecular dynamics. The alphaA-conotoxin EIVA consists of 30 amino acids representing the largest peptide among the alpha/alphaA-family conotoxins discovered so far and targets the neuromuscular nicotinic acetylcholine receptor with high affinity. alphaA-Conotoxin EIVA consists of three distinct structural domains. The first domain is mainly composed of the Cys3-Cys11-disulfide loop and is structurally ill-defined with a large backbone root mean square deviation of 1.91 A. The second domain formed by residues His12-Hyp21 is extremely well defined with a backbone root mean square deviation of 0.52 A, thus forming a sturdy stem for the entire molecule. The third C-terminal domain formed by residues Hyp22-Gly29 shows an intermediate structural order having a backbone root mean square deviation of 1.04 A. A structurally ill-defined N-terminal first loop domain connected to a rigid central molecular stem seems to be the general structural feature of the alphaA-conotoxin subfamily. A detailed structural comparison between alphaA-conotoxin EIVA and alphaA-conotoxin PIVA suggests that the higher receptor affinity of alphaA-conotoxin EIVA than alphaA-conotoxin PIVA might originate from different steric disposition and charge distribution in the second loop "handle" motif.

The fish-hunting cone snail Conus ermineus is the only known piscivorous Conus species in the Atlantic Ocean (1,2). In aquaria, the species captures its prey by extending its highly distensible amber-colored proboscis from which ejected is a hollow harpoon-shaped tooth that functions as a hypodermic needle to inject venom into the fish. Venom injection results in a complete inhibition of neuromuscular transmission in the injected prey. A major molecular component of this physiolog-ical strategy is to prevent neurotransmitter (acetylcholine) binding to the major postsynaptic receptor, the muscle subtype of the nicotinic acetylcholine receptor.
In most fish-hunting cone snail venoms, the competitive nicotinic antagonists of the skeletal muscle subtype are peptides with two disulfide bonds belonging to the ␣-conotoxin family (3). However, in C. ermineus, there are two different nicotinic receptor antagonists: 1) ␣-conotoxin EI with two disulfide bonds (4) and 2) ␣A-conotoxin EIVA with three disulfide bonds (5). The ␣A-conotoxins have been found in only two Conus species: 1) C. ermineus from the Atlantic and 2) the eastern Pacific purple cone, Conus purprascens (6). These species are believed to have a different evolutionary history from the Indo-Pacific fish-hunting Conus (7,8).
Highly selective antagonistic activity against particular subtypes of nicotinic acetylcholine receptors (nAChRs) 1 (Table I) has rendered ␣/␣A-conotoxins an important tool for studying molecular function of nAChR (3, 9 -11). In an effort to discover novel modulators of nAChR by identifying critical residues within the toxin for receptor contact, we have been applying a "reverse mapping strategy" using atomic resolution structures of various ␣/␣A-conotoxins (12)(13)(14)(15). Although considerable structure-function work has been reported on the ␣-conotoxins (16 -25), there has been only one report on the ␣A-subfamily (12) since their original discovery (6). ␣A-Conotoxin EIVA is different in its pharmacological specificity from the best studied ␣-conotoxins (such as ␣-conotoxin MI) in that it will inhibit both ligand-binding sites on the muscle nicotinic receptor with an approximately equal affinity (5) instead of being highly selective either for the ␣ 1 /␦ or for the ␣ 1 /␥ interface of heteropentameric neuromuscular nAChRs. In this article, we report the high resolution NMR structure of ␣A-conotoxin EIVA from C. ermineus that exhibits the highest (subnanomolar) receptor affinity among all of the known nAChR-targeting conotoxins.

EXPERIMENTAL PROCEDURES
Peptide Preparation-␣A-Conotoxin EIVA, originally purified from the venom of C. ermineus, was synthesized and purified as described previously (5). Disulfide-bonding patterns were assigned by mass spectrometry as well as by NMR. When ␣A-conotoxin EIVA was analyzed by liquid-secondary ionization mass spectrometry, a monoisotopic mass of 3095.2 was obtained (5), suggesting the presence of three disulfide bonds. In addition, the formation of three specific disulfide bonds in ␣A-conotoxin EIVA was unambiguously determined by using a well established NMR protocol that relies on the fact that NOEs involving relevant ␣ and ␤ protons of pairing cysteines are observed for a disulfide bond (26) (for details see "Structure Calculations").
NMR Experiments-Samples for the NMR studies were prepared in 90% H 2 O, 10% 2 H 2 O or in 100% 2 H 2 O with a final concentration of ϳ5 mM at pH 3.6. The pH was measured as a direct reading from a combination microelectrode calibrated at two reference pH values. All of the NMR experiments were performed using a Varian UNITY 500 or UNITY INOVA 600 spectrometer at three temperatures to obtain unambiguous resonance assignment. For TOCSY experiments (27), mixing times of 65-87 ms were applied. All of the peaks were referenced to a residual water signal (4.76 ppm at 25°C). Spectral widths were 6 kHz in both dimensions. Typical two-dimensional data consist of 2048 complex points in the t 2 dimension with 256 complex t 1 increments.
Structure Calculations-Interproton distance restraints used for computation of structures were derived primarily from the NOESY spectrum recorded with a mixing time of 200 ms obtained at 15°C. Several long range H ␣ -H ␤ NOEs (residues 3-11, 11-3, 14 -24, 16 -2, and 24 -14) and H ␤ -H ␤ NOEs (residues 2-16, 3-11, and 14 -24) were clearly observed, confirming the presence of three disulfide bonds, Cys 2 -Cys 16 , Cys 3 -Cys 11 , and Cys 14 -Cys 24 . The FELIX program in the NMR Refine module of Biosym 95.0 software (Molecular Simulations Inc., San Diego, CA) was used for quantification of NOE volumes and for converting them into interproton distance restraints. As a distance reference, the NOE volumes of four non-overlapping geminal ␤-proton cross-peaks were averaged and correlated with the appropriate geminal distance of 1.8 Å. Volume integration errors and influence of possible conformational averaging were taken into consideration by adding 0.5 and 1.0 Å to distance restraints involving only backbone protons and to those containing at least one side chain proton, respectively (28).
Actual computation of structures was done in two major steps. The first involved the generation of 50 low resolution structures using DGII calculation based on the metric-matrix method (Molecular Simulations Inc.). For example, to refine the initial distance restraints and to obtain more accurate distance restraints by correcting for spin-diffusion effect, the RANDMARDI program (29) was run for the structures generated by DGII. The final restraint set included a total of 388 restraints, 357 of which were NOE-derived distance restraints that in turn were composed of 164 intraresidue distances, 152 short range (͉i-j͉ Ͻ 5) interresidue distances, and 41 long range (͉i-j͉ Ͼ 5) interresidue distances. Also included in the restraint file were 11 torsion angle restraints along with 11 chirality and 9 disulfide-bond constraints. Backbone torsion angle was set to Ϫ120°(Ϯ30°) when 3 J HNH␣ Ͼ 8 Hz, and backbone torsion angle is set to Ϫ60°(Ϯ30°) if 3 J HNH␣ Ͻ 6 Hz. The second step of structure calculation was refinement of the DGII-generated structures by restrained molecular dynamics (30) done in a biphasic manner such that distance restraints associated with the backbone conformation were applied during the first 22 ps of dynamics at 1000 K followed by application of the rest of distance restraints. Each dynamics run lasted for 84 ps, 52 ps at 1000 K and 32 ps of annealing period down to 100 K, finally followed by a short energy minimization using a conjugate gradient method. Force constants for the NOE restraints and for torsion angles were gradually increased to the final values of 30 kcal/mol.

RESULTS
NMR Spectroscopy-Complete 1 H resonance assignment for the ␣A-conotoxin EIVA was achieved using homonuclear twodimensional NMR methods according to the standard sequential resonance assignment strategy. Typically, classification of spin systems in a TOCSY spectrum was followed by the sequential resonance assignment procedure based upon sequential NOE connectivity (31). Fig. 1 shows the finger-print region in the NOESY spectrum of ␣A-conotoxin EIVA obtained in 90% H 2 O, 10% 2 H 2 O at pH 3.6, and summarized in Fig. 2 are shortand medium-range NOEs, 3 J HNH␣ , and chemical shift index (32) along the amino acid sequence of the ␣A-conotoxin EIVA. For some residues such as His 12 , Lys 17 , Arg 20 , and Arg 26 , the presence of unique "back-transfer" cross-peaks in the TOCSY spectrum (data not shown) helped to obtain unambiguous resonance assignment readily without relying on the NOE connectivity information.
Structural Description of ␣A-Conotoxin EIVA- Fig. 3 shows the ensemble of 18 final converged structures of ␣A-conotoxin EIVA. The global molecular fold is stabilized mainly by three disulfide bridges of Cys 2 -Cys 16 , Cys 3 -Cys 11 , and Cys 14 -Cys 24 without any distinct secondary structural elements. From a structural standpoint, ␣A-conotoxin EIVA consists of three distinct structural domains. The first domain is mainly composed of residues in the Cys 3 -Cys 11 loop and exhibits a high conformational disorder with a backbone r.m.s. deviation of 1.91 Å. The C␣ T 1 and T 2 relaxation times for some residues (Cys 2 , Pro 5 , Ala 9 , Ala 10 , and Cys 11 ) in the first domain are on the order of ϳ350 and ϳ100 ms, respectively, with the exception of the Pro 5 . These numbers do not differ significantly from the C␣ T 1 and T 2 values of the residues in the other domains, indicating that the conformational disorder observed for the first domain is most probably because of the absence of appropriate NOE restraints, in particular, near the Gly 4 -Pro 5 -Tyr 6 -Hyp 7 segment.
This first domain is connected to an extremely well defined loop domain formed by residues His 12 -Hyp 21 that has a backbone r.m.s. deviation of 0.52 Å and corresponds to the "handle" of an iron (see below). The third C-terminal domain consists of residues Hyp 22 -Gly 29 and shows an intermediate structural order with a backbone r.m.s. deviation of 1.04 Å. The presence of such three distinct structural domains appears to be the common feature of ␣A-conotoxins (12). In Fig. 4, GRASP images of ␣A-conotoxin EIVA (left) and ␣A-conotoxin PIVA (right) are shown. The overall shape of ␣A-conotoxin PIVA was previously shown to resemble an "iron" with the Gln 25 located at the forwarding tip of an iron (12). In the case of ␣A-conotoxin EIVA, however, Arg 26 assumes the corresponding position. The C-terminal additional four residues, Hyp 27 -Ser 28 -Gly 29 -Gly 30 , of ␣A-conotoxin EIVA seem to attach themselves as a mobile anchor to the molecular body (Fig. 4, left top). In Fig. 4, the three residues (Lys 17 -Asp 18 -Arg 19 ) in ␣A-conotoxin PIVA protrude toward the reader out of the "bottom plate" of an iron that is formed by the rest of the molecule. In contrast, four residues of ␣A-conotoxin EIVA, Lys 17 -Val 18 -Gly 19 -Arg 20 , form a rigid loop in the central stem of ␣A-conotoxin EIVA. Detailed structural comparison of these loops is shown in Fig. 5 where the protruding loop in ␣A-conotoxin EIVA shows more hydrophobic character than the corresponding loop in ␣A-conotoxin PIVA.   PROCHECK analysis (35). Note that the angular order parameters for the backbone torsion angles become very high when the structurally ill-defined N-terminal domain of the molecule is excluded. DISCUSSION Two nAChR antagonist peptides produced by the Atlantic cone shell C. ermineus, ␣-conotoxin EI (4, 15) and ␣A-conotoxin EIVA (5), are both peculiar. The former selectively antagonizes neuromuscular receptor despite the fact that it belongs to the ␣ 4/7 2 subfamily, all of the other members of which target neuronal subtypes (Table I). In addition, it is the only nAChRantagonizing conotoxin that shows a high affinity toward the ␣ 1 /␦ site of the Torpedo nAChR. All of the other neuromuscular toxins either show a high affinity for the ␣ 1 /␥ site or do not distinguish between the two sites. The second nAChR antagonist ␣A-conotoxin EIVA produced by C. ermineus, whose threedimensional structure is presented in the current report, is unique in other ways. ␣A-Conotoxin EIVA targets nAChR, not Na ϩ or Ca 2ϩ channels, as do the other three disulfide-constrained conotoxins such asor -conotoxins (3,4,6,9). Interestingly, ␣A-conotoxin EIVA is incapable of discriminating between the ␣ 1 /␦ site and the ␣ 1 /␥ site in contrast to the other neuromuscular conotoxins such as ␣ 3/5 -conotoxins or ␣-conotoxin EI.
Efforts to gain a clear understanding of molecular function of nAChR would be greatly aided by a high resolution threedimensional structure of nAChR. Unfortunately, attempts to Selected residues are labeled. An "iron-shaped" structure (top-down view) of ␣A-conotoxin PIVA with Gln 25 at the forwarding tip is clearly visible with the "handle" of an iron composed of Lys 17 -Asp 18 -Arg 19 protruding toward the reader. ␣A-Conotoxins EIVA also resembles an iron when considered only up to Arg 26 . However, an additional C-terminal residues present in ␣A-conotoxin EIVA stick out toward left (top left when viewed in the current orientation) from the forwarding tip (Arg 26 ) of the molecule. Notable is the fact that the "handle" portion of an iron in ␣A-conotoxin EIVA is hydrophobic because of Val 18 , whereas the corresponding position is occupied by Asp 18 in the case of ␣A-conotoxin PIVA. Also noteworthy is the fact that the forwarding tip of an iron is occupied by a positively residue (Arg 26 ) in the case of ␣A-conotoxins EIVA, whereas the same position in ␣Aconotoxin PIVA is occupied by a smaller and less hydrophilic residue, Gln 25 . A cavity is present in the lower portion of ␣A-conotoxin PIVA. However, the corresponding position in ␣A-conotoxin EIVA is filled since ␣A-conotoxin EIVA has a proline residue (Pro 5 ), whereas ␣Aconotoxin PIVA has a serine (Ser 5 ).  obtain such information have not been successful (36 -38) even though a low resolution electromagnetic image (39,40) and an x-ray structure of an acetylcholine-binding protein that resembles the extracellular domain of nAChR have been obtained (41). In the absence of a high resolution structure of nAChR, we have been determining high resolution structures of ␣/␣A-conotoxins and tried to "reverse-predict" the receptor residues that might interact with the receptor-contacting residues of ligands (12)(13)(14)(15). Such an approach is possible because although various ␣/␣A-conotoxins exhibit a great variety of receptor subtype specificity, they differ only slightly in their sequences with a similar overall fold exhibiting subtle local structural differences. Two useful features have emerged from structural studies on ␣/␣A-conotoxins. First, potential receptor contact residues within each toxin have been suggested (12,14,15). High resolution structures have also helped us to visualize how receptor residues that had been identified by other means such as affinity labeling and mutagenesis (42)(43)(44)(45)(46)(47) could be spatially positioned with respect to the receptor-contacting residues of ligands in order to be properly engaged in -cation or hydrophobic interactions that are known to be major driving forces for ligand-nAChR binding (48 -50).
The second feature that emerged from structural studies of ␣/␣A-conotoxins is the estimated size of the ligand binding pocket in nAChR (15). The suggested size of the ligand binding pocket of ϳ20 Å (height) ϫ 20 Å (width) ϫ 15 Å (thickness) is in good agreement with what has been proposed by electromagnetic images (38,39) or by a recent homology-modeling study (51). ␣A-Conotoxins are larger than typical ␣-conotoxins and hence may need to bend or deform themselves to fit into the same ligand binding pocket of nAChR. The presence of a flexible N-terminal loop in ␣A-conotoxins may help such a conformational change upon receptor binding.
␣A-Conotoxin EIVA blocks the acetylcholine-evoked response of Torpedo neuromuscular nAChRs expressed in Xenopus oocytes with an IC 50 of 17 nM (5). [Pro 7,13 ] ␣A-Conotoxin PIVA inhibits the binding of 125 I ␣-bungarotoxin to Torpedo membrane with an IC 50 of ϳ350 nM (6). Although affinities of the two toxins were not measured by an exactly same method, it is clear that ␣A-conotoxin EIVA is a more potent antagonist than ␣A-conotoxin PIVA. There are two prominent sequence differences between the two toxins that might account for such affinity difference. First, ␣A-conotoxin EIVA has an additional four residues, Hyp 27 -Ser 28 -Gly 29 -Gly 30 , at the C terminus. The second difference is found in the protruding loop region. It is unknown whether both differences contribute to the affinity difference between the two toxins. Ligand-nAChR interactions are believed to be mediated mainly by -cation or hydrophobic interactions (48 -50). Therefore, the differences in charge as well as hydrophobicity in the loop region shown in Figs. 4 and 5 are expected to contribute to the affinity difference between the two toxins.
The overall molecular size of nAChR-binding ligands increases in the following order: acetylcholine, nicotine Ͻ tubocurare Ͻ ␣-conotoxins Ͻ ␣A-conotoxins Ͻ ␣-bungarotoxin. An interesting correlation seems to exist, albeit not fully proven yet, between the size of nAChR ligands and their ability to discriminate two unequivalent ligand-binding sites in the neuromuscular nAChR. As the size of the ligands increases, they tend to lose the ability to discriminate two unequivalent ligandbinding sites. For example, ␣A-conotoxins and ␣-bungarotoxins seem to become incapable of discriminating two ligand-binding sites when compared with smaller ␣-conotoxins or curares, although mutant forms of NmmI (52) and ␣-cobratoxin (53) are certainly exceptions to such a trend. In the case of ␣-bungarotoxin, it is known that the most critical receptor-contacting residues are located in its second finger (51,54,55), which is approximately similar in size to ␣-conotoxins. However, being a large ligand, this snake toxin may have other receptor-contacting residues that are outside the second finger such as Cterminal residues (56,57). Obviously, ligand-nAChR interactions require more than one ligand-contacting "microsite" residue from the receptor with the possible number of such microsites increasing with the ligand size. The fact that ␣Aconotoxins cannot distinguish the two unequivalent ligand binding sites (␣ 1 /␦ versus ␣ 1 /␥) of neuromuscular receptor raises the possibility that binding of ␣A-conotoxins may involve a microsite of the nAChR that typical ␣-conotoxins might not touch.