Scanning Mutagenesis of ω-Atracotoxin-Hv1a Reveals a Spatially Restricted Epitope That Confers Selective Activity against Insect Calcium Channels*

We constructed a complete panel of alanine mutants of the insect-specific calcium channel blocker ω-atracotoxin-Hv1a. Lethality assays using these mutant toxins identified three spatially contiguous residues, Pro10, Asn27, and Arg35, that are critical for insecticidal activity against flies (Musca domestica) and crickets (Acheta domestica). Competitive binding assays using radiolabeled ω-atracotoxin-Hv1a and neuronal membranes prepared from the heads of American cockroaches (Periplaneta americana) confirmed the importance of these three residues for binding of the toxin to target calcium channels presumably expressed in the insect membranes. At concentrations up to 10 μm, ω-atracotoxin-Hv1a had no effect on heterologously expressed rat Cav2.1, Cav2.2, and Cav1.2 calcium channels, consistent with the previously reported insect selectivity of the toxin. 30 μm ω-atracotoxin-Hv1a inhibited rat Cav currents by 10-34%, depending on the channel subtype, and this low level of inhibition was essentially unchanged when Asn27 and Arg35, which appears to be critical for interaction of the toxin with insect Cav channels, were both mutated to alanine. We propose that the spatially contiguous epitope formed by Pro10, Asn27, and Arg35 confers specific binding to insect Cav channels and is largely responsible for the remarkable phyletic selectivity of ω-atracotoxin-Hv1a. This epitope provides a structural template for rational design of chemical insecticides that selectively target insect Cav channels.

The first peptide neurotoxins isolated from the venom of Australian funnel-web spiders (genera Atrax and Hadronyche) and shown to have selective activity against insects were members of the -atracotoxin-1 (ACTX) 1 family (1-3). These toxins comprise 36 -37 residues with six strictly conserved cysteine residues that are paired to form three disulfide bridges (4). The best studied family member, -ACTX-Hv1a, is one of the most potent insecticidal peptide toxins discovered so far (4,5); it has proved lethal to all insect orders that have been tested, including coleopterans, dictyopterans, hemipterans, orthopterans, and refractory lepidopteran pests such as the tobacco budworm Heliothis virescens and the cotton bollworm Helicoverpa armigera (1,2,5).
The phyletic specificity of these toxins appears to reside in their ability to block insect, but not vertebrate, voltage-gated calcium channels (2). Although nanomolar concentrations of -ACTX-Hv1a are sufficient to block high voltage-activated (HVA) calcium currents in the central nervous system of the fruit fly Drosophila melanogaster and the American cockroach Periplaneta americana (2,5), 1 M toxin does not block HVA calcium currents in vertebrate neurons (2) and the toxin is harmless when injected subcutaneously into newborn mice (1).
With the exception of the organophosphate and carbamate insecticides (which target acetylcholinesterase), the vast majority of synthetic insecticides are directed against voltagegated sodium channels, nicotinic acetylcholine receptors, or GABA/glutamate-gated chloride channels. This limited group of targets has promoted the evolution of cross-resistance to different families of insecticides (6) and stimulated interest in the development of new insecticides that act on targets outside of this established ion channel triumvirate (7). Although blockers of HVA calcium currents have evolved independently in the venoms of cone snails, snakes, and spiders (4,8,9), no commercially available insecticide exploits this target (7,10). Thus, -ACTX-Hv1a is a valuable lead for the development of a new class of insecticides that target insect calcium channels.
The three-dimensional structure of -ACTX-Hv1a comprises a disulfide-rich globular core (residues 4 -21), with residues 22-37 forming a finger-like ␤ hairpin that protrudes from this globular region (2). In a previous structure-activity relationship study, we demonstrated the functional significance of several residues in the toxin's ␤ hairpin (11). In the present study, we extended our scanning mutagenesis of -ACTX-Hv1a in several ways: first, we employed a complete panel of alanine mutations representing all solvent-exposed side chains of the toxin structure; second, we tested these mutations using lethality assays performed in two distantly related insects, the orthopteran Acheta domestica and the dipteran Musca domestica; third, by performing competitive binding assays using a neuronal membrane preparation from the head of P. americana, we obtained information about the toxin-target interaction that is not complicated by issues of bioavailability and/or pharmacokinetics. The data indicate that a spatially contiguous epitope composed of residues Pro 10 , Asn 27 , and Arg 35 is responsible for the specific interaction of this toxin with insect calcium channels. A panel of follow-up mutants was used to ascertain the key chemical features of these residues that are critical for the toxin-channel interaction. This information should facilitate the rational design of chemical insecticides based on the structure of -ACTX-Hv1a.

EXPERIMENTAL PROCEDURES
Production of -ACTX-Hv1a Mutants-Individual point mutations were introduced into a synthetic gene encoding -ACTX-Hv1a as described previously (11) using either 1) mutagenic PCR using the C2 primer, a mutagenic version of the N1 primer (with pHWT1 as a template), or 2) overlap extension PCR using a pair of complementary mutagenic primers in combination with the original N1 and C2 primers (with pHWT1 as a template). The full-length product of the mutagenic PCR was then digested with EcoRI and BamHI and subcloned into the expression vector pGEX-2T as described previously for pHWT-1 (11). Double mutations were introduced by a similar process in which a version of pHWT-1 encoding one of the two desired mutations was used as the PCR template. Overexpression and purification of recombinant wild-type (WT) -ACTX-Hv1a and mutants thereof was performed as described previously using glutathione affinity chromatography followed by C 18 reverse-phase high pressure liquid chromatography (11). The identity of each toxin was confirmed using electrospray mass spectrometry.
Insect Toxicity Assays-Insecticidal activity was tested as described previously (11,12) by injecting peptides dissolved in insect saline (13) into house crickets (Acheta domestica) or house flies (Musca domestica). LD 50 values (i.e. the dose that is lethal to 50% of insects) were calculated as described previously (11), and the reported values are the mean of two to three independent experiments.
CD Spectroscopy-CD spectra of -ACTX-Hv1a or mutants thereof were collected and processed as described previously using a Jasco J-715 spectropolarimeter (11). The toxin concentration was 15 M.
Radiolabeling of Recombinant -ACTX-Hv1a-Lethality assays performed with crickets showed that although mutation of Tyr to Phe at residue 13 of -ACTX-Hv1a was not deleterious, other mutations at this residue (such as Tyr to Trp or Ala) can lower toxin activity (see "Results"). Hence, Tyr 13 is not a good target for iodination. Because the initial alanine scan of the ␤-hairpin region of -ACTX-Hv1a (11) revealed that an F24A mutation was tolerated, we decided to engineer a "YF swap mutant," in which Tyr 13 was mutated to Phe and Phe 24 was mutated to Tyr, for use as a radioiodination substrate. The swap mutant was fully functional in LD 50 assays and it contained a single, non-critical Tyr residue (Tyr 24 ) that allowed radioactive iodine to be conjugated to residue 24 rather than to residue 13. Radioiodination of the YF swap mutant was performed as described previously for -conotoxin GVIA (14) except for minor changes in the reverse-phase high pressure liquid chromatography method used to purify the radiolabeled toxin. The purification was performed using a Vydac C 18 analytical reverse-phase column (4.6 ϫ 250 mm; 5 m pore size) eluted at 1 ml/min using a linear gradient of 0 -28% buffer B (0.085% trifluoroacetic acid and 50% acetonitrile in water) in buffer A (0.1% trifluoroacetic acid in water) over 5 min, followed by a linear gradient of 28 -35% buffer B in buffer A over 15 min, followed by a linear gradient of 35-53% buffer B in buffer A over 12 min. Cold toxin was observed to elute at 19 min (i.e. during the last minute of the second gradient); monoiodinated toxin was observed to elute at 21-22 min. Hereinafter, the radiolabeled swap mutant will be referred to as 125 I--ACTX-Hv1a.
Preparation of Cockroach Neuronal Membranes-Neuronal membranes were prepared from the heads of adult P. americana as described previously (15), except that the final Ficoll centrifugation step was eliminated. The concentration of membrane proteins in the final preparation was determined using a Bradford protein assay (Bio-Rad) with bovine serum albumin as a standard.
Competitive Binding Assays using Radiolabeled -ACTX-Hv1a-Equilibrium competition binding experiments were performed using increasing concentrations of unlabeled mutant toxin in the presence of a fixed concentration of 125 I--ACTX-Hv1a. Individual binding reactions were brought to a final volume of 100 l in binding buffer (75 mM NaCl, 5 mM CaCl 2 , 50 mM HEPES, and 0.1% bovine serum albumin, pH 7.4). Cockroach neuronal membranes were added to binding reactions to a final concentration of ϳ0.15 g of membrane protein/l. 125 I--ACTX-Hv1a was added to a final concentration of 1-1.5 nM. Unlabeled competitors were present at various final concentrations. After incubation at room temperature for 60 min, the reactions were terminated by dilution with 2 ml of ice-cold wash buffer (300 mM choline chloride, 5 mM CaCl 2 , 50 mM HEPES, and 0.1% bovine serum albumin, pH 7.4). Bound toxin was recovered from the diluted reactions using rapid vacuum filtration through GF/C filters (Whatman) preincubated in 0.3% polyethylenimine (Sigma-Aldrich). The filters were then rapidly washed three times with 2 ml of wash buffer. Nonspecific toxin binding was determined in the presence of 10 ⌴ unlabeled -ACTX-Hv1a and typically comprised 40 -50% of total binding. Counts of the washed filters were analyzed using KaleidaGraph (Abelbeck/Synergy, Reading, PA) to fit data to a non-linear Hill equation for IC 50 determination. K i values were calculated using the Cheng-Prusoff equation (which assumes pure competition at a single site), where L* is the concentration of radiolabeled toxin and K d is its dissociation constant (16). The K d value used for these computations was obtained by using the computer program LIGAND (Elsevier Biosoft) to perform iterative Scatchard analysis on data from competition binding experiments performed with 125 I--ACTX-Hv1a and unlabeled wild-type -ACTX-Hv1a.

RESULTS
Alanine-scanning Mutagenesis of -ACTX-Hv1a-We expanded a collection of described previously alanine mutants of -ACTX-Hv1a (11) to give a larger panel of 28 mutants that comprised individual alanine mutations at all positions corresponding to solvent-exposed non-glycine sidechains of the toxin. The six cysteine residues that form the toxin's cystineknot motif (2, 10) were excluded from the mutational analysis because their side chains are involved in disulfide bonds that are critical to the three-dimensional structure of the toxin. Ile 5 was excluded because it forms part of the buried hydrophobic core of the toxin (2) and is likely to be critical for toxin folding. Gly 8 and Gly 30 were also excluded because they are located in ␤-turns that are likely to be structurally important (2). Moreover, with only a hydrogen atom side chain, the Gly residues are unlikely to be functionally important. The P6A and S19A mutant toxins were difficult to overexpress and purify, and were therefore not used in the fly toxicity assays.
Structural Integrity of Mutant Toxins-Although the mutations that we introduced into -ACTX-Hv1a were chosen to avoid major structural perturbations, it was important to test this experimentally. We did this by acquiring CD spectra of the WT and mutant recombinant toxins. The CD spectrum of WT toxin exhibits a ␤ sheet signature (Fig. 1A) with pronounced minima and maxima at 210 and 196 nm, respectively (11); this is consistent with a 16-residue ␤ hairpin (residues 22-37) being the major secondary structure feature of -ACTX-Hv1a (2) (see Fig. 1C). The small positive band at 236 nm is presumably caused by the aromatic chromophore of Tyr 13 , since this feature is absent in CD spectra of Y13A and Y13L mutants but present (albeit slightly blue-shifted) in the spectrum of a Y13W mutant (Fig. 1B).
We showed previously that none of the alanine-scan mutants in the ␤-hairpin region of the toxin were structurally perturbed except for mutations at Lys 34 (11). Likewise, the CD spectrum of 13 of the 15 newly constructed alanine mutants in the disulfide-rich globular portion of the toxin superimposed closely on the CD spectrum of the WT recombinant toxin, as illustrated for the S1A and Q9A mutants in Fig. 1A. Only the CD spectra of the Y13A and N16A mutants differed significantly from that of the WT toxin. The CD spectrum of the Y13A mutant was similar to that of the WT toxin except that the Tyr 13 side chain signature at 236 nm was absent as expected (Fig. 1B). We conclude that the backbone fold of the Y13A mutant is similar to that of the native toxin.
We were surprised to find that the presumed Tyr 13 side chain signature was also absent in the N16A mutant (Fig. 1A). This suggests that the conformation of this mutant is perturbed in a manner that significantly alters the chemical environment of the Tyr 13 side chain chromophore. Asn 16 lies within a rare triple turn motif formed by three overlapping type I ␤ turns (residues 13-16, 14 -17, and 17-20) (2). Asn 16 is found at the iϩ2 position of the second turn, where Asn and Asp are by far the most favored residues (18). In contrast, alanine is one of the least favored residues at this position in type I ␤ turns (18). Thus, an N16A mutation might abrogate this ␤ turn, thereby disrupting the entire set of interconnected ␤ turns in the triple turn system and substantially altering the environment of the Tyr 13 side chain. We therefore suspect that the small reduction in cricket toxicity observed for the N16A mutant (see below) is the result of a perturbation of the toxin structure in the region of the triple turn (residues [13][14][15][16][17][18][19][20]. However, for all other alanine mutants, we can rule out the possibility that functional defects are caused by major structural perturbations. Cricket Bioassays-The insecticidal potency of each mutant was initially examined by comparing its LD 50 in a model orthopteran (house crickets) with that of WT recombinant toxin (Fig. 1D). Based on the reasoning described previously (11), we considered mutated residues to be functionally significant if the LD 50 of the mutant toxin was more than 4-fold greater than that of WT toxin, potentially significant if the corresponding LD 50 was 2-4-fold greater than that of WT toxin, or otherwise insignificant. We were surprised to find, based on these criteria, that only Gln 9 and Pro 10 were identified as functionally significant residues for activity against crickets (Fig. 1D), in addition to the ␤-hairpin residues Asn 27 and Arg 35 that were identified in our previous structure-activity relationship study (11). Residues Tyr 13 and Asn 16 were identified as potentially significant for cricket toxicity (Fig. 1D).
Fly Bioassays-We expanded the toxicity assays to include a model dipteran (house flies) for several reasons. First, dipterans are the most pernicious insects from a human and animal health perspective, being responsible for the transmission of malaria, filariasis, trypanosomiasis, leishmaniasis, and numerous arboviral diseases (4, 7). Thus, it was of interest to establish whether -ACTX-Hv1a might serve as a lead for development of insecticides directed against dipteran vectors of human disease. Second, although some venomous animals such as cone snails are specialist predators, spiders are mostly generalists that prey on a wide range of invertebrates and sometimes small vertebrates. Thus, there would be selective advantage for spiders that evolved venoms capable of dealing with subtle variations in ion channel targets between different insects, in addition to the intraspecific complexity introduced by the high degree of differential splicing and editing of transcripts encoding channel subunits (19,20). One evolutionary response to this problem seems to have been lineage-specific gene duplication events that created a suite of closely related paralogous toxins (21). However, an alternative possibility is that a single toxin encodes both a core pharmacophore that is crucial for interaction with the target plus additional residues that are critical contributors to channel binding in some insect species but not others. To address this latter possibility, we sought to examine whether the insect-active epitope of -ACTX-Hv1a was invariant between two highly divergent insect orders (i.e. Diptera and Orthoptera) or whether certain residues might be important for activity in one of these orders but not the other.
Finally, we were surprised that the cricket toxicity assays seemed to indicate that the insect-active epitope of -ACTX-Hv1a comprises only a few residues, thus raising the possibility that variations in toxin bioavailability/pharmacokinetics might be masking variations in channel-blocking activity of the mutant toxins. That is, in the LD 50 assays, the activity of a toxin depends not only on its intrinsic affinity for the target channel but also on its in vivo stability and its ability to penetrate anatomical barriers. The latter point is crucial in that -ACTX-Hv1a targets the central nervous system of insects (4,5), whereas most other spider toxins act presynaptically at insect neuromuscular junctions. Thus, a mutant toxin might have reduced activity not because of a reduced affinity for the target channel, but rather because of increased susceptibility to proteolysis or more limited access to the central nervous system. We anticipated that these potential problems would be less critical in the smaller house flies as they responded more rapidly than crickets to toxin injection, suggesting that anatomical barriers are less critical, and they were about 3-fold more sensitive to the toxin (LD 50 ϭ 86.5 Ϯ 1.3 pmol/g) than crickets (LD 50 ϭ 269 Ϯ 13 pmol/g). Fig. 1E shows the insecticidal potency of each alanine mutant in house flies relative to the activity of the WT recombinant toxin. As for the cricket bioassays, residues Pro 10 , Asn 27 , and Arg 35 were found to be functionally critical; each individual mutation caused a 10 -20-fold reduction in LD 50 . In contrast, Gln 9 and all residues identified as potentially significant in the cricket assays (Tyr 13 , Asn 16 , Asn 29 , Lys 34 , and Asp 37 ) seemed unimportant for activity against flies. An N27A,R35A doublemutant toxin (11) that carried mutations in two of the three residues most critical for insecticidal activity had a 60-fold higher LD 50 in house flies than WT recombinant toxin (data not shown).
Competitive Binding Assays-We developed a competitive binding assay to study the interaction between -ACTX-Hv1a and its presumed calcium channel target in a manner not complicated by issues of pharmacokinetics and bioavailability. This assay measured the ability of cold -ACTX-Hv1a and alanine mutants thereof to displace the radiolabeled tracer 125 I--ACTX-Hv1a from neuronal membranes prepared from the heads of adult cockroaches.
The assay was complicated by what seemed to be very high and inconsistent nonspecific binding of 125 I--ACTX-Hv1a to the membrane preparation, a problem that could not be re-solved through any modification of the assay conditions tested (i.e. using more or less toxin per binding reaction, more or less membrane preparation per binding reaction, additional washes, etc.). We presume this problem stems from two issues: 1) the low total number of -ACTX-Hv1a binding sites (B max ϭ 0.5-1.0 pmol/mg), consistent with the expected low density of voltage-gated calcium channels in excitable membranes (22); 2) the relatively low affinity of the tracer (K d ϳ 5 nM, see below). In contrast with -ACTX-Hv1a, the scorpion toxin Bj-xtrIT binds to cockroach neuronal membranes with K d ϭ 148 pM and B max ϭ 3 pmol/mg (23). The low receptor density and relatively low affinity of -ACTX-Hv1a for these receptors necessitated the use of large amounts of membrane for each binding reaction and a relatively high concentration (1-1.5 nM) of radioactive tracer. Nevertheless, in most cases, we obtained binding curves that did not appear disrupted by inconsistent background and which could be analyzed to yield reliable K i values (see inset to Fig. 1F for examples of typical binding curves).
Scatchard analysis indicated that the K d of the toxin for cockroach neuronal membranes was ϳ5 nM (n ϭ 4). Although much lower than the affinity of Bj-xtrIT for cockroach neuronal membranes, the K d is similar to that estimated for the scorpion toxin Lqh III under similar conditions (15). The K d estimate was used with the Cheng-Prusoff equation under "Experimental Procedures" to calculate the fold increase/decrease in K i for each mutant (Fig. 1F). Mutation to alanine of any of the three residues critical for toxicity against both crickets and flies (i.e. Pro 10 , Asn 27 , and Arg 35 ) reduced the K i by more than 150-fold, with the R35A mutation causing a dramatic 280-fold reduction in binding affinity (Fig. 1F). The Q9A mutation, which reduced activity against crickets but not flies, caused a moderate 16fold increase in K i , as did the Y13A mutation (Fig. 1F). No other mutations significantly affected the binding of -ACTX-Hv1a to cockroach neuronal membranes. Thus, as observed in a previous study of scorpion ␣ neurotoxins (24), there seems to be a good correspondence between the results of the binding and lethality assays. Differential bioavailability of mutants does not seem to be a major complicating factor in the LD 50 assays, except perhaps for the R35A mutation, which was significantly more deleterious in binding studies than in lethality assays.
Chemical Features of the Toxin Pharmacophore-The alanine scanning mutagenesis studies indicated that Pro 10 , Asn 27 , and Arg 35 are the key components of the toxin pharmacophore; Gln 9 and Tyr 13 are of minor importance in orthopteroids (crickets and cockroaches) but unimportant in dipterans (see Fig. 1). By using a panel of non-alanine mutants, we showed previously that the ␥-NH 2 group of Asn 27 most probably donates a hydrogen bond to an acceptor on the target calcium channel, whereas the ␦-guanido group of Arg 35 most likely forms an ion pair with a carboxylate group on the channel (11). We decided to probe the chemical features of Gln 9 and Tyr 13 to ascertain the nature of their interaction with orthopteroid calcium channels. CD analysis of toxins with mutations at Tyr 13 (Fig. 1B) and Gln 9 (data not shown) indicated that their structures were similar to that of WT toxin.
A Y13F mutant was almost equipotent with native toxin in the cricket bioassay ( Fig. 2A), which suggests that the hydroxyl group is relatively unimportant and that this residue mainly contacts the channel via hydrophobic interactions involving the aromatic ring. Consistent with this hypothesis, replacement of Tyr 13 with leucine, a similarly sized hydrophobic, non-aromatic side chain, yielded a mutant toxin (Y13L) that had toxicity similar to that of the Y13F and native toxins. In contrast, replacement of Tyr 13 with either a much smaller alanine residue (Y13A) or a significantly larger tryptophan residue (Y13W) caused a 3-4-fold reduction in LD 50 (Fig. 2A). We conclude that Tyr 13 makes contacts with a hydrophobic pocket in the channel that cannot be readily accessed by significantly smaller or larger hydrophobic sidechains at this position.
A Q9E mutant was about 50% less toxic to crickets than a Q9A mutant and had a 6-fold lower LD 50 than WT toxin (Fig.  2B). It is impossible to determine from these data alone whether the loss of ␦-NH 2 group, the introduction of a negative charge, or a combination of these factors is responsible for the large reduction in activity caused by the Q9E mutation. The data suggest, however, that the Gln 9 side chain carbonyl group, which is retained in the Q9E mutant, does not make critical hydrogen bond interactions with the target channel. A Q9N mutation (2.4-fold reduction in LD 50 ) was significantly less deleterious than a Q9A mutation (4-fold reduction in LD 50 ) (Fig. 2B). We interpret these results to indicate that the ␦-NH 2 of Gln 9 makes hydrogen-bond interactions with acceptor groups on the channel and that this interaction can be retained to some degree when Gln 9 is replaced by the slightly smaller Asn, in which the side chain is ϳ1.5 Å shorter.
Effect of -ACTX-Hv1a on Vertebrate Calcium Channels-We tested the effect of -ACTX-Hv1a and a functionally defective N27A,R35A double-mutant toxin on whole-cell calcium currents recorded from HEK cells that transiently expressed rat Ca v 2.1, Ca v 2.2, or Ca v 1.2 calcium channels. At a concentration of 10 M, neither -ACTX-Hv1a nor the doublemutant toxin had any observable effect on calcium currents carried by these channels (data not shown). However, at a concentration of 30 M, the WT toxin partially antagonized all three channel subtypes; the strongest inhibition was observed against Ca v 1.2 channels (Fig. 3). Remarkably, the inhibition of Ca v 2.1 and Ca v 2.2 currents was slightly enhanced in the double-mutant toxin, whereas the current block was marginally reduced for the Ca v 1.2 subtype. The fact that the moderate current block at this extremely high toxin concentration is not abrogated by the double mutation suggests that the weak inhibition of rat HVA calcium channels relies on structural determinants that are uniquely different from those involved in block of insect calcium channels.

DISCUSSION
Calcium Channels as Insecticide Targets-Peptidic calcium channel blockers have been isolated from the venoms of numerous spiders, including plectoxin II from Plectreurys tristes, CSTX-1 from the subtropical wandering spider Cupiennius salei, several families of -agatoxins from the American funnelweb spider Agelenopsis aperta, and two families of -atracotoxins from various species of Australian funnel-web spiders (genera Atrax and Hadronyche) (7). The evolution of calcium channel blockers in spider venoms as part of a mechanism for incapacitating envenomated prey is not surprising because voltage-gated calcium channels (VGCCs) play critical roles in modulating synaptic transmission in insects. The essential role of insect VGCCs is highlighted by the fact that the D. melanogaster genome appears to encodes pore-forming ␣ 1 -subunits for only one Ca v 1 family member (Dmca1D) and one Ca v 2 family member (Dmca1A/cacophony) (25), and severe loss-of-function mutations in either of these genes is embryonic lethal (26 -28).
Despite the fact that insect VGCCs seem to be excellent insecticide targets, as first suggested almost a decade ago (29), no commercial insecticides have ever been directed against these channels. The limited phyletic specificity of the -agatoxins may have promoted the misconception that insect VGCCs could not be selectively targeted. However, this view is at odds with the fact that insect and vertebrate VGCCs are pharmacologically distinct, with different susceptibilities to a range of peptide toxins and chemical agents (30). Our recent isolation of two families of highly insect-selective -atracotoxins (4, 10) demonstrates that phyletically specific targeting of insect VGCCs is indeed possible. This is further exemplified in the current study, where we demonstrated that -ACTX-Hv1a is lethal to a range of insect species but inactive against heterologously expressed rat HVA VGCCs even at concentrations as high as 10 M.
The Insect-specific Pharmacophore of -ACTX-Hv1a-The specificity of -ACTX-Hv1a for insect VGCCs makes it an excellent lead for the development of insecticides directed against insect calcium channels. In this study, we mapped the pharmacophore of -ACTX-Hv1a to provide a template for rational design of such insecticides. The cockroach neuronal membrane binding assays and the two sets of insect bioassays all identified three functionally critical residues (Pro 10 , Asn 27 , and Arg 35 ; see Fig. 1, D-F) that we presume correspond to the primary VGCC-binding epitope of the toxin. It is remarkable that these residues, although well separated in the amino acid sequence, form a spatially contiguous epitope when mapped onto the three-dimensional structure of -ACTX-Hv1a (Fig.  4A). Arg 35 is positioned at the center of this binding "hot spot," flanked by Pro 10 and Asn 27 .
Arg is one of the most highly enriched residues at proteinprotein interfaces (31), and Arg 35 seems to be the most critical residue for binding of -ACTX-Hv1a to cockroach VGCCs; mutation of Arg 35 to Ala caused a 280-fold reduction in K i (see Fig.  1F), which corresponds to a change in the free energy of binding (⌬⌬G bind ) of 3.3 kcal/mol. Mutation of the flanking Pro 10 and Asn 27 residues to Ala caused a smaller ϳ180-fold reduction in K i , corresponding to a ⌬⌬G bind of ϳ3.1 kcal/mol. Gln 9 and Tyr 13 seem to be of minor importance for binding of the toxin to orthopteroid VGCCs, with mutation of these residues to Ala giving a ⌬⌬G bind of only ϳ1.6 kcal/mol. Gln 9 is contiguous with the core pharmacophore, whereas the Tyr 13 side chain is situated ϳ12 Å from the binding hot spot (Fig. 4A). These residues do not seem to be important for binding of the toxin to dipteran VGCCs.
Sequencing of venom peptides (1, 3) and analysis of venomgland cDNA libraries (4,21) has revealed that each species of Australian funnel-web spider typically expresses a complement of 3-6 -ACTX-Hv1a homologs in its venom. Despite significant intra-and interspecific variations in the sequences of these homologs, the five pharmacophore residues (Gln 9 , Pro 10 , Tyr 13 , Asn 27 , and Arg 35 ) are identical in all -ACTX-1 orthologs that have been sequenced to date (see Fig. 3A in Ref. 4 for an alignment of 11 representative -ACTX-1 sequences). Thus, the data currently available seem to indicate that the pharmacophore of -ACXT-Hv1a has been universally conserved in this family of spiders. (2)) with core pharmacophore residues (Pro 10 , Asn 27 , and Arg 35 ; ⌬⌬G bind Ͼ 2.5 kcal/mol) shown in red. Residues drawn in orange (Gln 9 and Tyr 13 ; ⌬⌬G bind ϳ1.6 kcal/mol) are of minor importance for binding to orthopteroid VGCCs but seem to be unimportant for interaction of the toxin with dipteran calcium channels. Proposed molecular interactions with orthopteroid calcium channels are summarized above each double-headed arrow. ⌬⌬G bind values, which indicate the loss of binding energy (in kilocalories per mole) resulting from mutation of the indicated residue to Ala, were calculated using the equation ⌬⌬G bind ϭ RTln(wild-type K i / mutant K i ). B, molecular surface of -conotoxin MVIIA (Protein Data Base file 1OMG; (48)) with key pharmacophore residues shown in red (⌬⌬G bind Ͼ 2.5 kcal/ mol) and orange (⌬⌬G bind ϭ 1.8 -2.2 kcal/ mol). ⌬⌬G bind values resulting from mutation of the indicated residues to Ala were calculated from IC 50 values reported in Ref. 36. The illustrated mimetic of -conotoxin MVIIA, which was rationally designed to mimic three key pharmacophore residues (Arg 10 , Leu 11 , and Tyr 13 ), had an IC 50 of ϳ3 M against rat Ca v 2.2 channels (46). Molecular surfaces were drawn using PyMOL (49). Toxins were applied to transfected HEK cells in 20 mM Ba 2ϩ recording solution at a holding potential of Ϫ100 mV. Test currents were elicited by voltage stepping to ϩ20 mV. Data are the average of recordings from six cells.