Miniaturization of Scorpion β-Toxins Uncovers a Putative Ancestral Surface of Interaction with Voltage-gated Sodium Channels*♦

The bioactive surface of scorpion β-toxins that interact with receptor site-4 at voltage-gated sodium channels is constituted of residues of the conserved βαββ core and the C-tail. In an attempt to evaluate the extent by which residues of the toxin core contribute to bioactivity, the anti-insect and anti-mammalian β-toxins Bj-xtrIT and Css4 were truncated at their N and C termini, resulting in miniature peptides composed essentially of the core secondary structure motives. The truncated β-toxins (ΔΔBj-xtrIT and ΔΔCss4) were non-toxic and did not compete with the parental toxins on binding at receptor site-4. Surprisingly, ΔΔBj-xtrIT and ΔΔCss4 were capable of modulating in an allosteric manner the binding and effects of site-3 scorpion α-toxins in a way reminiscent of that of brevetoxins, which bind at receptor site-5. While reducing the binding and effect of the scorpion α-toxin Lqh2 at mammalian sodium channels, they enhanced the binding and effect of LqhαIT at insect sodium channels. Co-application of ΔΔBj-xtrIT or ΔΔCss4 with brevetoxin abolished the brevetoxin effect, although they did not compete in binding. These results denote a novel surface at ΔΔBj-xtrIT and ΔΔCss4 capable of interaction with sodium channels at a site other than sites 3, 4, or 5, which prior to the truncation was masked by the bioactive surface that interacts with receptor site-4. The disclosure of this hidden surface at both β-toxins may be viewed as an exercise in “reverse evolution,” providing a clue as to their evolution from a smaller ancestor of similar scaffold.

Voltage-gated sodium channels (Na v s) 3 are critical for generation and propagation of action potentials in excitable cells and are targeted by a large variety of ligands that bind at distinct receptor sites on the pore-forming ␣-subunit (1). Most lipid-soluble Na v activators, including pyrethroid insecticides, toxic alkaloids (e.g. veratridine, from the plant family of Liliaceae, and batrachotoxin, from the skin of the Colombian frog Phyllobates aurotaenia), and marine cyclic polyether toxins (e.g. brevetoxins produced by "red tide" dinoflagellates), are effective in invertebrates as well as vertebrates. In addition, a wide array of proteinaceous Na v modifiers exists in the venom of scorpions, spiders, cone snails, and sea anemones, which are utilized for prey and defense (1,2).
Na v gating modifiers from scorpion venom are divided into two classes, ␣ and ␤, according to their mode of action and binding properties (3)(4)(5). ␣-Toxins prolong the action potential by inhibiting the fast inactivation of the Na v upon binding to receptor site-3 (4,6), which involves extracellular loops in domains 1 and 4 (6). These toxins are divided into three groups (7): (i) anti-mammalian ␣-toxins (e.g. Lqh2 from Leiurus quinquestriatus hebraeus), which are highly toxic in mammalian brain and bind with high affinity to rat brain Na v s; (ii) antiinsect ␣-toxins (e.g. Lqh␣IT), which are highly toxic to insects and show weak activity in mammalian brain; and (iii) ␣-like toxins (e.g. Lqh3), which are toxic to insects and mice (by intracerebroventricular and peripheral injections) but bind weakly to rat brain synaptosomes (3). Alterations in the binding of ␣-toxins to receptor site-3 on rat brain and insect Na v s are allosterically modulated by a variety of Na v lipophilic activators such as site-2, site-5, and site-7 toxins (e.g. veratridine, brevetoxin, and pyrethroids, respectively; Refs. 3, 7-10).
A number of reports have described in detail the bioactive surfaces of a few scorpion ␤-toxins, the anti-insect excitatory and depressant toxins, Bj-xtrIT and LqhIT2 (22)(23)(24), and the anti-mammalian ␤-toxin, Css4 (25). These studies highlighted a conserved pharmacophore composed of a key negatively charged Glu at the ␣-helix motif, flanked by hydrophobic aromatic residues that may isolate the point of interaction from bulk solvent. The residues that comprise the pharmacophore of ␤-toxins belong to the protein core ( Fig. 1), which is common to both long chain scorpion ␣ and ␤-toxins and most short chain scorpion toxins affecting voltage-gated potassium channels (20,26). An additional hydrophobic cluster of bioactive amino acids at the C-tail of Bj-xtrIT and on the loop connecting the second and the third ␤-strands of Css4 was suggested to be involved in determining toxin selectivity (5,25).
To examine the commonality of the pharmacophore region in ␤-toxins and the extent of its contribution to recognition of receptor site-4, we deleted the N-and C-terminal regions of two ␤-toxins, Bj-xtrIT and Css4, and analyzed the binding properties of the resulting truncated derivatives (⌬⌬␤-toxins). Unpredictably the ⌬⌬␤-toxins were non-toxic and did not bind at receptor site-4; nonetheless they were able to allosterically modulate ␣-toxin activity on insect and mammalian Na v s and to antagonize the effects of brevetoxins. These results indicated that the seemingly inactive ⌬⌬␤-toxins were able to interact with the Na v .
Binding Experiments-Locust neuronal membranes were prepared from dissected brains and ventral nerve cords of adult Locusta migratoria using an established method (7). Rat brain synaptosomes were prepared from adult (ϳ300 g) albino Wistar strain (laboratory-bred) as was described previously (27). Membrane protein concentration was determined by a Bio-Rad protein assay, using bovine serum albumin as standard. Css4 and Bj-xtrIT were radioiodinated by lactoperoxidase (catalog number L8257; 7 units/60 l of reaction mix; Sigma) using 10 g of toxin and 0.5 mCi of carrier-free Na 125 I (Amersham Biosciences,) and the monoiodotoxin was purified as was described previously (25).
The composition of media used in binding assays and termination of the reactions was previously described (27,28). Nonspecific binding was determined in the presence of 1-10 M of the unlabeled toxin and typically consisted of 10 -30% of total binding (for scorpion toxins). Equilibrium competition binding assays were performed and analyzed as was described previously (25). For [ 3 H]PbTx-3, filters were dried 1 h at 80°C, mixed with 5 ml of ULTIMUMA Gold (PerkinElmer Life Sciences), and the radioactivity was monitored. Nonspecific binding was determined in the presence of 2 M PbTx-2. Each experiment was performed in duplicate and repeated at least three times as indicated (n). Data are presented as mean Ϯ S.E. of the number of independent experiments.
Two-electrode Voltage Clamp Experiments-The genes encoding for the Drosophila melanogaster sodium channel ␣-subunit (DmNa v 1) and the auxiliary TipE subunit were kindly provided by J. Warmke, Merck, and M. S. Williamson, Division of Plant and Invertebrate Ecology-Rothamsted, UK, respectively. The genes encoding for the rat brain and skeletal muscle sodium channels (rNa v 1.2 and rNa v 1.4) were a gift from Dr. R. G. Kallen, University of Pennsylvania, Philadelphia, PA. These genes and that for the auxiliary subunit h␤1 were transcribed in vitro using T7 RNA-polymerase and the mMESSAGE mMACHINE TM system (Ambion, Austin, TX), and the cRNAs were injected into Xenopus laevis oocytes as was described previously (13). Currents were measured 1-2 days after injection using a two-electrode voltage clamp and a Gene Clamp 500 amplifier (Axon Instruments, Union City, CA). Data were sampled at 10 kHz and filtered at 5 kHz. Data acquisition was controlled by a Macintosh PPC 7100/80 computer, equipped with ITC-16 analog/digital converter (Instrutech Corp., Port Washington, NY), utilizing Synapse (Synergistic Systems, Sweden). Capacitance transients and leak currents were removed by subtracting a scaled control trace utilizing a P/6 protocol (29). Bath solution contained (in mM): 96 NaCl, 2 KCl, 1 MgCl 2 , 2 CaCl 2 , and 5 HEPES, pH 7.85. Oocytes were washed with the bath solution using a BPS-8 perfusion system (ALA Scientific Instruments, Westbury, NY) with a positive pressure of 4 p.s.i.
Toxins were diluted with bath solution and applied directly to the bath to the final desired concentration.
GV Analysis-Mean conductance (G) was calculated from peak current/voltage relationship using the equation G ϭ I/(V Ϫ V rev ), where I is the peak current, V is the membrane potential, and V rev is the reversal potential. Normalized conductance voltage relations were fit with one Boltzman distribution according to the following equation G/G max ϭ 1 ϩ exp[(V1 ⁄ 2 Ϫ V)/k], where V1 ⁄ 2 is the respective membrane potential for which the mean conductance is half-maximal; k is the respective slope.
To obtain a dose-response curve for ␣-toxins, currents were elicited by depolarization to Ϫ10 mV from a holding potential of Ϫ80 mV in the presence of various toxin concentrations. At each toxin concentration, currents were allowed to reach a steady-state level prior to the final measurement, and the steady-state current at Ϫ10 mV was normalized relatively to the current in control.

RESULTS
The N and C termini of ␤-toxins vary in sequence, structure, and disulfide bond organization, but their core is conserved ( Fig. 1) (Ref. 17). To examine the contribution of the pharmacophore of scorpion ␤-toxins (25) to their binding affinity, we deleted ( Fig. 1) the first 10 N-terminal residues as well as the C-tail stretch beyond the penultimate Cys of Bj-xtrIT (anti-insect; 27 residues) and Css4 (anti mammalian; 18 residues). Since the ⌬⌬␤-toxins ended with a Cys residue, we added the triplet Thr-Asn-Lys, which appears at the C-tail of various potassium channel blockers (26). The Cys residues that remained unpaired in the two toxins (position 12 in Css4 and 43 in Bj-xtrIT) were substituted by Ser to prevent incorrect folding or dimerization. ⌬⌬Css4 and ⌬⌬Bj-xtrIT were produced recombinantly using the E. coli expression system as was described for the unmodified parental toxins.
Functional Analysis of the ⌬⌬␤-Toxins-⌬⌬Bj-xtrIT and ⌬⌬Css4 were non-toxic to blowfly larvae and mice by injection even at very high concentrations (Ͼ20 nmol/g of animal) and were unable to displace 125 I-Bj-xtrIT and 125 I-Css4 at locust and rat brain neuronal membranes in binding competition assays (up to 20 M; Fig. 2). These results indicated that the two ⌬⌬␤-toxins lost their ability to bind at receptor site-4 of Na v s. Moreover, ⌬⌬Bj-xtrIT and ⌬⌬Css4 (up to 20 M) had no effect on the conductance and gating properties of the mammalian channels rNa v 1.2 (brain) and rNa v 1.4 (skeletal muscle), and the insect channel DmNa v 1, all expressed in X. laevis oocytes (Table 1). Thus, at first glance, these results have suggested that the truncated toxins lost any bioactive competence. At this stage, we used another sensitive measure to assess whether the truncated toxins could interact  with the Na v . It has been shown that the interaction of scorpion ␣-toxins with receptor site-3 is modulated by a variety of ligands that bind at other receptor sites on the Na v , such as veratridine (site-2; Refs. 8,30), scorpion ␤-toxins (site-4; Ref. 19), brevetoxins (site-5; Refs. 7,9), and pyrethroids (site-7; Ref. 10). We therefore tested whether ⌬⌬Bj-xtrIT and ⌬⌬Css4 were capable of modulating the activity of the scorpion ␣-toxin Lqh␣IT on insects. Surprisingly, ⌬⌬Bj-xtrIT and ⌬⌬Css4 increased 6-fold the toxicity of Lqh␣IT to blowfly larvae, with half-maximal doses (EC 50 ) of 5.2 and 9.6 ng/100 mg, respectively (Fig. 3). This unexpected synergism prompted us to examine the binding capability of the ⌬⌬␤-toxins to Na v s.
⌬⌬Bj-xtrIT and ⌬⌬Css4 were examined in binding studies using iodinated scorpion ␣-toxins that interact well with insect and rat brain neuronal membranes. In locust neuronal membranes, ⌬⌬Bj-xtrIT and ⌬⌬Css4 increased 1.7-fold 125 I-Lqh␣IT binding (EC 50 2.9 Ϯ 0.5 and 3.8 Ϯ 1.4 nM, respectively; Fig. 4A). This positive cooperativity indicated that the ⌬⌬␤-toxins interact with the insect Na v at a site other than receptor site-3. Notably, in rat brain synaptosomes, ⌬⌬Bj-xtrIT and ⌬⌬Css4 inhibited 125 I-Lqh2 binding in a dose-dependent manner with an IC 50 of 76 Ϯ 18 and 105 Ϯ 29 nM, respectively, whereas the parental ␤-toxins had no effect (Fig. 4B). It is noteworthy that 125 I-⌬⌬Bj-xtrIT and 125 I-⌬⌬Css4 did not exhibit significant specific binding to insect or rat brain neuronal preparations, and therefore, direct assessment of their binding interaction could not be performed.
The allosteric effects of the ⌬⌬␤-toxins were analyzed on the activity of Lqh␣IT and Lqh2 on insect and rat brain Na v s expressed in Xenopus oocytes. Lqh␣IT increases the sodium peak current and inhibits the fast inactivation of the insect sodium channel, DmNa v 1, with an EC 50 of 2.6 Ϯ 0.2 nM (  (Fig. 5B), indicating decrease in activity at this mammalian sodium channel. Once more, these opposite effects have indicated that the ⌬⌬␤toxins, which by themselves had no effect on the sodium currents (Table 1), modulate the effect of site-3 ␣-toxins on mammalian and insect Na v s in a different manner, further supporting their interaction with the sodium channels.
Since the structure of the truncated toxins is composed of the ␤␣␤␤ fold as it appears in most scorpion toxins, we questioned the specificity of their interaction with the Na v by employing another short chain toxin of similar size and fold, charybdotoxin (from the scorpion L. quinquestriatus hebraeus; Fig. 1) (Ref. 26). Charybdotoxin, which is a blocker of voltage-gated potassium channels, had no effect on Lqh␣IT activity at DmNa v 1 and on Lqh2 activity at rNa v 1.2 (not shown), suggesting that the activity of the ⌬⌬␤-toxins is not derived only from their ␤␣␤␤ fold.

Alterations of the mid-voltage of activation (V1 ⁄2 ) derived from conductance-voltage (G-V) curves of DmNa v 1/TipE, rNa v 1.2/h␤ 1 and rNa v 1.4/h␤1 expressed in Xenopus oocytes and induced by PbTx-2 and the ⌬⌬␤-toxins
The data represent the mean Ϯ S.E. of at least six independent experiments. The oocytes were clamped at Ϫ100 mV and then 50-ms depolarization pulses from Ϫ90 to ϩ50 mV were applied at 5-mV increments. The V1 ⁄ 2 , was calculated as described under "Experimental Procedures."
toxins were administered to blowfly larvae subsequent to PbTx-2 injection and after a paralytic effect already developed, even at a 1:100 molar ratio, the larvae were not resuscitated, indicating that the ⌬⌬␤-toxins were unable to abolish the brevetoxin effect. To further clarify this point, we examined the effects of PbTx-2 in the presence of the truncated toxins on rNa v 1.2, rNa v 1.4, and DmNa v 1 expressed in Xenopus oocytes. Although PbTx-2 in concentrations of 0.5-2 M shifted the voltage dependence of activation of the three channels to more hyperpolarizing membrane potentials, pre-or co-application of ⌬⌬Bj-xtrIT or ⌬⌬Css4 with PbTx-2 in a 1:2-3 molar ratio abolished these effects ( Fig. 6 and Table 1). Once more, when the ⌬⌬␤-toxins were applied subsequent to PbTx-2, when its effect had already developed, they had no inhibitory effect (not shown). We further tested the effects of the ⌬⌬␤-toxins on the binding of ␣-toxins in the presence of PbTx-2. 1 M PbTx-2 increased the binding of 125 I-Lqh␣IT to locust neuronal membranes by 30% (19). 100 nM ⌬⌬Bj-xtrIT increased the binding of 125 I-Lqh␣IT to the same membranes by 70%. However, co-application of PbTx-2 and ⌬⌬Bj-xtrIT had no additive effect on Lqh␣IT binding (Fig.  7A). These data suggest that the binding sites of brevetoxin and the ⌬⌬␤toxins are not identical.
To examine whether the ⌬⌬␤toxins were directly or indirectly involved with brevetoxin binding, we analyzed their effect on the binding of [ 3 H]PbTx-3, a specific marker of receptor site-5 at rat brain synaptosomes (32). High concentrations (1-3 M) of either ⌬⌬Bj-xtrIT or ⌬⌬Css4 had no effect on [ 3 H]PbTx-3 binding, not even when incubated for 1 h with the membranes prior to the addition of [ 3 H]PbTx-3 (Fig.  7B). These results have indicated that the ⌬⌬␤-toxins do not compete for the brevetoxin binding site at rat brain Na v s.
Residues Important for ⌬⌬Css4 Effect-The inability of ⌬⌬Bj-xtrIT and ⌬⌬Css4 to compete on binding at receptor sites 3 (Figs. 4 and 5), 4 (Fig. 2), and 5 (Fig. 7B) have suggested that they recognize another receptor site on the sodium channel. Since the two ⌬⌬␤-toxins exhibited similar activities, we focused on ⌬⌬Css4 and substituted 16 putative solvent-exposed residues to Ala with the objective to identify those residues important for interaction with the channel. The ⌬⌬Css4 mutants were tested for their ability to influence Lqh␣IT toxicity by mixing each mutant in a 5:1 (w/w) ratio with Lqh␣IT, and the ED 50 of the mixtures was determined on blowfly larvae (Fig. 8). A number of substitutions, which in the context of Css4 had no effect on the binding affinity (K13A, D21A, D23A, Q31A, and Q32A; Ref. 25), decreased the positive cooperative effect of ⌬⌬Css4 on Lqh␣IT, with the utmost effect achieved with K13A. In contrast, substitutions, which in the context of Css4 decreased substantially the binding affinity for rat brain synapto- somes, had no effect (L19A, R27A, and Y42A) or only a minor effect (F17A and N22A) on ⌬⌬Css4 ability to enhance Lqh␣IT toxicity (Fig. 8). Since Css4 and ⌬⌬Css4 do not bind at the same receptor site on the sodium channel, the difference of bioactive residues was not surprising. The only substitutions that affected both Css4 and ⌬⌬Css4 activity were E28A (hot spot) and F14A (numbering follows Css4).

DISCUSSION
"Uncoating" the Exterior of Scorpion ␤-Toxins Exposes a Hidden Bioactive Surface-The C-tail, loops, turns, and unstructured stretches that connect the main secondary structure ele-ments in long-chain scorpion toxins constitute a large portion of their exteriors and bear residues that participate in bioactivity (reviewed in Refs. 3 and 5). Here we show that removal of about 50% of the amino acid residues of Bj-xtrIT and Css4, which changed markedly their exteriors (Fig. 1), abolished any detectable activity and ability to bind at receptor site-4 on the Na v . However, the ability of ⌬⌬Bj-xtrIT and ⌬⌬Css4 to reduce the activity of Lqh2 at the brain channel rNa v 1.2 and increase Lqh␣IT activity at receptor site-3 of insect channels (binding, effect on fast inactivation, and toxicity to blowfly larvae) indicated that they bound to the Na v . The bioactive residues of ⌬⌬Css4 are different from those of Css4 except for Glu-28, which was considered a hot spot in Css4 surface of interaction with the rNa v 1.2 (25). It is likely that in the context of the new functional surface of ⌬⌬Css4, Glu-28 is unable to interact with the site-4 residue as it did in the context of Css4.
From a structural viewpoint, the ability of ⌬⌬Bj-xtrIT and ⌬⌬Css4 to bind to the Na v , as manifested in the modulation of receptor site-3, suggests that a masked functional surface was exposed by miniaturization of the two ␤-toxins. The probability that this hidden surface has been formed by coincidence is very low as it appears in two pharmacologically distinct toxins, Css4 and Bj-xtrIT. One way to rationalize such a finding is by assuming that scorpion toxins affecting sodium channels evolved from smaller peptides bearing the conserved ␤␣␤␤ core that were able to interact with sodium channels (26,33). It is conceivable that along evolution, these peptides gained higher affinity and specificity for distinct Na v targets, whereas increasing in volume. Along this process, the bioactive surface of the ancestral toxin has been masked by new functional exteriors, such as those resolved nowadays by mutagenesis in various ␤-toxins (22,23,25). This hypothesis is in concert with the idea that diversification of long-chain scorpion toxins occurred mainly by mutations and slight structural alterations at the loops, turns, and C-tail rather than at the ␤␣␤␤ core (17). Thus, the truncation of these toxins may be considered an exercise in "reverse evolution" since it was opposite to the precedent natural path of evolution of these molecules.  Table 1. Inset, rNa v 1.2/h␤1 currents induced by the indicated toxin in a test pulse to Ϫ35 mV. Similar results were obtained with ⌬⌬Css4 (Table 1).    The ⌬⌬␤-Toxins as Antagonists of Brevetoxins-The similar allosteric effects of the ⌬⌬␤-toxins and brevetoxins on the activity of ␣-toxins (Figs. 4 -7) (Refs. 7,9,31) have raised the possibility that they both interact with receptor site-5. However, because simultaneous application of ⌬⌬Bj-xtrIT and PbTx-2 had no additive effect on Lqh␣IT binding at insect Na v s and because ⌬⌬Bj-xtrIT and ⌬⌬Css4 did not compete with brevetoxin on binding at receptor site-5 (Fig. 7), it was apparent that the truncated ␤-toxins bind at a different site. Interestingly, this binding induced allosterically a similar conformational alteration at site-3 to that induced by brevetoxins binding. This conclusion raises the question as to why the ⌬⌬␤-toxins are capable of inhibiting PbTx-2 effects on rNa v 1.2, rNa v 1.4, and DmNa v 1 activation only when pre-or co-applied (Fig. 6), whereas having no antagonistic effect when applied after PbTx-2. The 30-Å-long brevetoxin has been suggested to orient its "head" (Lacton ring A) down across the membrane, parallel to the transmembrane hydrophobic ␣ helices S5 and S6 between domains 1 and 4 of the Na v , whereas its "tail" ("rigid region"; rings H-K) points outward (32,34). Chemical modifications at the side chain of the K-ring of PbTx-3 (rigid region) hardly affected its binding but markedly influenced its activity (35,36). In addition, brevetoxins that vary in their rigid region structure exhibit different effects on the binding of ␣-toxins at receptor site-3 (37). It is therefore likely that the rigid region of brevetoxins interacts with an external channel site whose conformational alteration allosterically affects the activity of ␣-toxins at receptor site-3. If the ⌬⌬␤-toxins interact at least partially with this external channel site, it may suggest that their surface of interaction with the channel, which is responsible for the allosteric effects, is involved in the inhibition of brevetoxin activity. This assumption is supported by previous observations that subtle modifications induced by tetrodotoxin binding to the Na v external pore are sufficient to prevent the allosteric inhibition caused by brevetoxin on the binding of a scorpion ␣-toxin to rat brain Na v s (31, 37). Therefore, preapplication of the ⌬⌬␤-toxins may prevent brevetoxin activity without affecting its binding. As brevetoxins are responsible for massive fish kills and marine mammal mortalities, as well as illness in humans who ingest filter-feeding shellfish or inhale toxic aerosols (34), the ⌬⌬␤toxins may serve as leads for development of brevetoxin antagonists.
General Implications-The continuous search for novel pharmacologically valuable compounds in arthropod venoms requires tedious fractionation and purification followed by biological assays to validate that a desired activity still persists at each purification step. Inactive fractions or substances are usually disregarded as they are not expected to yield a valuable end product. These basic considerations have led to the present situation where of 100 -300 distinct polypeptides that can be monitored by mass spectroscopy in the venom of a single scorpion or a spider, the biological activity of only a small portion has been characterized (4,38,39). However, there are numerous reasons to believe that compounds that at first glance seem inactive should not be ignored. We have recently shown that scorpion ␤-toxins that bind to receptor site-4 at the extracellular region of Na v s may influence allosterically the effects produced by ␣-toxins that bind at receptor site-3 (19). Moreover, even mere binding of a ␤-toxin, whose activity was abolished by mutagenesis, was shown to induce this allosteric effect at site-3 (19). We have also shown that the alleged anti-insect selective scorpion depressant toxins affect the mammalian skeletal muscle Na v provided that the channel was allosterically modulated by the binding of an ␣-toxin at receptor site-3 (40).
Here we demonstrate the use of the sensitive measure of allosteric interactions to evaluate the ability of substances that a priori were considered "biologically inactive" to bind and alter the Na v conformation. It is therefore rational to believe that when scorpion toxins that affect Na v s degrade either in the venom glands (39) or in the stung animal (hemolymph or blood), their partially degraded derivatives may "boost" the impact of the entire venom. This phenomenon also provides a conceptual basis to future attempts to decrease doses of drugs by harmless, allosterically acting compounds.