Molecular Requirements for Recognition of Brain Voltage-gated Sodium Channels by Scorpion α-Toxins*

The scorpion α-toxin Lqh2 (from Leiurus quinquestriatus hebraeus) is active at various mammalian voltage-gated sodium channels (Navs) and is inactive at insect Navs. To resolve the molecular basis of this preference we used the following strategy: 1) Lqh2 was expressed in recombinant form and key residues important for activity at the rat brain channel rNav1.2a were identified by mutagenesis. These residues form a bipartite functional surface made of a conserved “core domain” (residues of the loops connecting the secondary structure elements of the molecule core), and a variable “NC domain” (five-residue turn and the C-tail) as was reported for other scorpion α-toxins. 2) The functional role of the two domains was validated by their stepwise construction on the similar scaffold of the anti-insect toxin LqhαIT. Analysis of the activity of the intermediate constructs highlighted the critical role of Phe15 of the core domain in toxin potency at rNav1.2a, and has suggested that the shape of the NC-domain is important for toxin efficacy. 3) Based on these findings and by comparison with other scorpion α-toxins we were able to eliminate the activity of Lqh2 at rNav1.4 (skeletal muscle), hNav1.5 (cardiac), and rNav1.6 channels, with no hindrance of its activity at Nav1.1–1.3. These results suggest that by employing a similar approach the design of further target-selective sodium channel modifiers is imminent.

The pivotal role of voltage-gated sodium channels (Na v s) 4 in excitability mark them as major targets for a large variety of toxins that bind at distinct receptor sites and modify their gating (1). These channels are large membrane proteins made of a pore-forming ␣-subunit of ϳ260 kDa and auxiliary ␤-subunits of ϳ30 kDa. The ␣-subunit is composed of four homologous domains (D1-D4), each consisting of six ␣-helical transmembrane segments (S1-S6) connected by intracellular and extracellular loops. A key feature in Na v s function is their ability to rapidly activate and inactivate, leading to transient increase in Na ϩ conductance through the cell membrane. This mechanism is attributed to the ability of the positively charged S4 voltage sensors to move across the membrane in response to changes in membrane potential (1,2).
In mammals, at least nine genes encode a variety of Na v subtypes (1,3), whose expression varies greatly in different tissues (Na v 1.1-1.3 mainly in the central nervous system; Na v 1.6 in both central and peripheral neurons; Na v 1.7 in the peripheral nervous system; Na v 1.8 and Na v 1.9 in sensory neurons; Na v 1. 4 and Na v 1.5 in skeletal and cardiac muscles, respectively). Na v subtypes are distributed heterogeneously in the human brain and their expression is regulated under developmental and pathological conditions (1,(3)(4)(5). In addition, many disorders in humans result from abnormal function due to mutations in various Na v genes (6 -8). Thus, ligands that show specificity for Na v subtypes may be used for their identification at various tissues and as leads for design of specific drugs. This requires that the bioactive surfaces of these ligands be resolved along with molecular details that determine their specificity.
Among the wide range of Na v modifiers, those derived from scorpion venoms play an important role in studying channel activation (␤-toxins) and inactivation (␣-toxins) (9 -11). The channel site of interaction with scorpion ␣-toxins, named neurotoxin receptor site-3 (12), is shared also by structurally unrelated toxins from sea anemone and spider venoms (13,14), which raises questions as to its architecture and boundaries. Based on the findings that site-3 toxins eliminate a gating charge component associated with the movement of D4/S4 (15,16), and that this segment plays a critical role in coupling channel inactivation to activation (17), scorpion ␣-toxins were postulated to inhibit channel inactivation by hindering the outward movement of this segment during depolarization (9).
Scorpion ␣-toxins constitute a class of structurally and functionally related 61-67-residue long polypeptides reticulated by four conserved disulfide bridges. Despite a common ␤␣␤␤ core (10,18,19) these toxins are highly diverse in sequence and preference for insect and mammalian Na v s. Indeed, the ␣-toxin class is divided to pharmacological groups according to their toxicity in insects and mice brain and ability to compete on binding at insect and mammalian Na v s (10) (supplemental Fig.  S1): (i) classical anti-mammalian toxins, such as Aah2 (from Androctonus australis hector) and Lqh2 (from Leiurus quinquestriatus hebraeus), which bind with high affinity to Na v s at rat brain synaptosomes and are practically non-toxic to insects; (ii) ␣-toxins, such as Lqh␣IT, which strongly affect insect Na v s and are weak in mammalian brain; and (iii) ␣-like toxins, such as Lqh3 and BmKM1 (from Buthus martensii Karsch), which are active in both mammalian brain and insects.
Efforts to identify ␣-toxin residues involved in the interaction with the Na v receptor site-3 revealed a generally common bioactive surface divided to two topologically distinct domains: a conserved "core domain" formed by residues of the loops connecting the secondary structure elements of the molecule core, and a variable "NC domain" formed by the five-residue turn (residues 8 -12) and the C-tail (20 -23). These analyses raised the hypothesis that a protruding conformation of the NC domain correlates with high activity at insect Na v s, whereas a flat conformation of this domain appears in ␣-toxins active at the brain channel rNa v 1.2a (21). The correlation of this structural difference with toxin preference for Na v subtypes was corroborated by constructing the bioactive surface of Lqh␣IT on the scaffold of the anti-mammalian ␣-toxin Aah2 ending up with a chimera (Aah2 Lqh␣IT(face) ) active on insects, whose NC domain is in the protruding conformation (21). Despite this result, the molecular requirements that enable high affinity binding of classical ␣-toxins to mammalian Na v s have not been clarified, and only initial data about the channel region that constitutes receptor site-3 is available (Refs. 24 -26; also see Ref. 10 for review).
Lqh2 is a 64-residue long toxin from L. quinquestriatus hebraeus (Israeli yellow scorpion) (27) that is almost identical in sequence (96% identity) to the most active anti-mammalian toxin, Aah2, whose structure and action are documented (18,28,29). By functional expression and mutagenesis we uncovered residues on the Lqh2 exterior that are putatively involved in bioactivity. By construction of these residues on the scaffold of the anti-insect toxin Lqh␣IT we confirmed their bioactive role and differentiated those that determine toxin potency from those contributing to toxin efficacy. Comparison to other ␣-toxins was then instrumental for the design of an Lqh2 mutant that exhibits high specificity for the neuronal channels hNa v 1.1, rNa v 1.2a, and rNa v 1.3.

EXPERIMENTAL PROCEDURES
Materials, Bacterial Strains, and Animals-Native Lqh2 was purchased from Latoxan (Valance, France). Escherichia coli strain DH5␣ was used for plasmid constructions, and the BL21 strain (DE3, pLys) was used for toxin expression using the vector pET-14b (Novagen) in a protocol similar to that described previously (30). cDNAs encoding hNa v 1.1 (human, Ref. 31), rNa v 1.2 (rat, Ref. 32), and rNa v 1.3 (33) were subcloned into the pCDM8 vector, the cDNA sequence was determined, and any amino acid sequence errors were corrected to the sequences recorded in GenBank. The pAlter vectors encoding rNa v 1.4 and hNa v 1.5 were a gift from Dr. R. G. Kallen (University of Pennsylvania, Philadelphia, PA). The pNa200 vector encoding for rNa v 1.6 was a gift from Dr. A. Goldin (University of California, Irvine, CA). Wistar rats for the preparation of brain synaptosomes were purchased from the Animal Housing at Tel Aviv University.
Expression and Production of Recombinant Lqh2-Lqh2-cDNA was isolated from a cDNA library of the Israeli yellow scorpion (34), cloned at the NdeI-BamHI restriction sites of pET-14B expression vector, and expressed in fusion behind a His 6 tag and a thrombin cleavage sequence. Overexpression, in vitro folding, and purification by reverse phase-high pressure liquid chromatography followed a previously described protocol (30). Mutagenesis was performed via PCR and each toxin mutant derivative was produced and purified as described for the unmodified toxin. Quantification of the purified recombinant toxins was performed by amino acid analysis (21). The activity of His Lqh2 was comparable with that of the native toxin in binding and electrophysiological assays and therefore it was used throughout this study (supplemental Fig. S2).
Binding Experiments-Preparation of rat brain synaptosomes, membrane protein quantification, and Lqh2 radioiodination were carried out as described (29). Purification of the monoiodotoxin and determination of its concentration, as well as the composition of media used in the binding assays and termination of the reactions have been previously described (29,35). Nonspecific toxin binding was determined in the presence of excess (1 M) unlabeled toxin. Equilibrium competition assays were performed using increasing concentrations of the unlabeled toxin in the presence of a constant low concentration of the labeled toxin (30 -50 pmol). The median concentration values for inhibition of toxin binding (IC 50 ) were determined by a non-linear regression analysis using the Hill equation, employing a Hill coefficient of 1. Mathematical curve fitting for IC 50 determination was accomplished using KaleidaGraph (version 3.08, Synergy Software). K i values were calculated by the equation K i ϭ IC 50 /1 ϩ [L*/K d ], in which L* is the concentration of the radioactive ligand and K d is its dissociation constant (29,35).
Expression of Na v s in Oocytes and Two-electrode Voltage Clamp Experiments-cRNAs encoding the ␣-subunit of each channel and the auxiliary ␤1 subunit were transcribed in vitro using T7 RNA polymerase and the mMESSAGE mMACHINE TM system (Ambion, Austin, TX) and injected into Xenopus laevis oocytes as was previously described (36). Currents were measured 2-3 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 an 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 (37). Bath solution contained (in mM): 96 NaCl, 2 KCl, 1 MgCl 2 , 2 CaCl 2 , 5 HEPES, pH 7.85. Toxins were diluted with bath solution containing 1 mg/ml bovine serum albumin, and applied directly to the bath to the final desired concentration. To avoid application artifacts, 1 mg/ml bovine serum albumin solution was applied prior to toxin addition.

Dose-Response Curves of ␣-Toxin Effect on Fast Inactivation-
Currents were elicited by a 50 ms depolarization to Ϫ20 mV from a Ϫ80 mV holding potential in the presence of increasing toxin concentrations. At each toxin concentration, the currents were allowed to reach a steady-state level prior to the final measurement. The dose dependence for toxin-induced removal of fast inactivation is calculated by plotting the ratio of the steady-state current remaining 50 ms after depolarization (I ss ) to the peak current (I peak ) as a function of toxin concentration and fitting with the Hill equation, where H is the Hill coefficient, [Toxin] is the toxin concentration, and a 0 is the offset measured prior to toxin application. The a 1 Ϫ a 0 amplitude provides the maximal effect obtained at saturating toxin concentrations. EC 50 is the toxin concentration at which half-maximal inhibition of fast inactivation is obtained. To reduce variability, H was set to 1 in all calculations.
Mutant toxin efficacy was determined as the ratio between its maximal effect (I ss /I peak ) under saturating concentrations relative to that of Lqh2, which was normalized to 1.

RESULTS
The scorpion ␣-toxins Lqh2 and Lqh␣IT are very similar in structure and yet they differ greatly in preference for the mamma-lian brain versus insect sodium channels (rNa v 1.2a and DmNa v 1, respectively) (10,21,27). To identify the Lqh2 face of interaction with the rat brain Na v we produced the toxin in recombinant form and analyzed the contribution of surface amino acids to bioactivity using mutagenesis. Because the activity of the recombinant Lqh2 fused to a His 6 tag and a thrombin cleavage site was similar to that of the native toxin (supplemental Fig. S2), we used in all further experiments the toxin and mutants in their fused forms.
Mutagenic Dissection of Lqh2-Substitution of residues at the Lqh2 surface relied primarily on the results of mutagenic analyses of ␣-toxins Lqh␣IT, Lqh3, and BmKM1, which highlighted the bioactive role of residues at the core and NC domains (21)(22)(23). Each recombinant toxin derivative was assayed at rNa v 1.2a expressed in Xenopus oocytes in the presence of the ␤1 subunit. CD spectroscopy was used to discern effects that were due to structural perturbations from those directly associated with bioactivity. From a total of 29 mutants, the CD spectrum of only Y21A changed, and because a conservative substitution, Y21F, had negligible effect on Lqh2 activity ( Fig. 1 and supplemental Table S1), it is conceivable that Tyr 21 has a structural role. Substitution to Ala of Lys 2 , Phe 15 , Arg 18 , Trp 38 , Asn 44 , Thr 57 , and Lys 58 decreased the potency of Lqh2 more than 5-fold (in EC 50 values; Fig. 1; supplemental Table S1 and Fig. S3), whereas substitution of all other residues had little or no effect ( Fig. 1 and supplemental Table S1). Because residues equivalent to all these amino acids but Lys 2 have been implicated in activity of ␣-toxins, we analyzed in parallel how  supplemental Table S1.
substitution of Arg 2 in Lqh␣IT affects the activity at DmNa v 1. We found that substitution R2A decreased Lqh␣IT potency at DmNa v 1 by 8-fold (not shown) with no change in the CD spectrum, and therefore conclude that the second amino acid in scorpion ␣-toxins is an integral part of the NC domain as was also suggested for BmKM1 (23).
As residues of the five-residue turn in the ␣-toxins Lqh␣IT and BmKM1 have previously been shown to be involved in bioactivity (20,21,38), we have substituted one by one their equivalents (Asp 8 , Asp 9 , and Val 10 ) in Lqh2. Surprisingly, these substitutions had no effect on activity at rNa v 1.2a ( Fig. 1 and supplemental Table S1), which led us to examine the possibility that it is the general fold of the five-residue turn that affects the spatial arrangement of the entire NC domain and hence the activity. We constructed a triple mutant, in which these three residues were simultaneously exchanged by their Lqh␣IT equivalents (D8K, D9N, V10Y). Surprisingly, the potency of Lqh2 D8K,D9N,V10Y for rNa v 1.2a was reduced only 4-fold compared with that of the unmodified toxin (supplemental Table  S1), suggesting that by itself the five-residue turn of Lqh2 contributes only slightly to toxin activity.
A conservative substitution of Phe 15 by Trp had no effect, whereas substitution to Ala decreased the potency of Lqh2 for rNa v 1.2a by 5.5-fold ( Fig. 1 and supplemental Table S1 and Fig.  3). However, substitution of Phe 15 to Glu, which appears at this position in Lqh␣IT, decreased the toxin potency for rNa v 1.2a by 60-fold. These results have suggested that Phe 15 is involved in hydrophobic interactions with the sodium channel.
Unlike in BmKM1 (23,39), Lqh␣IT (21), and Lqh3 (22), only two residues at the C-tail of Lqh2 were important for activity. Although substitution of Lys 58 by Ala decreased the activity nearly 100-fold, a conservative substitution to Arg had no effect on toxin function. In contrast, even a conservative substitution of Thr 57 to Ser strongly reduced the activity ( Fig. 1 and supplemental Table S1 and Fig. 3).

Construction of Lqh2 Bioactive Surface on the Scaffold of Lqh␣IT-
Despite the structural similarity and general conservation of residues at the bioactive domains, the potency of Lqh2 at rNa v 1.2a is 3 orders of magnitude higher than that of Lqh␣IT (Fig. 2). Comparison of the two toxins at rNa v 1.2a has indicated that the efficacy of Lqh␣IT was 40% lower than that of Lqh2 and required 3 orders of magnitude higher toxin concentration (100 M; Fig. 2).
To correlate these differences in effect with structural features at the bioactive surface of the two toxins, and on the basis of a previous successful construction of Lqh␣IT bioactive surface on the scaffold of Aah2 (21), we substituted in a stepwise manner residues on the exterior of Lqh␣IT by their Lqh2 equivalents. The activity of the resulting constructs was examined at rNa v 1.2a expressed in Xenopus oocytes. This stepwise construction was useful in determining the contribution of each of the bioactive domains to Lqh2 function. The core domain of the two toxins is nearly identical in amino acid composition, except for Phe 17 in Lqh␣IT whose Lqh2 equivalent is Gly, and Glu 15 in Lqh␣IT whose Lqh2 equivalent is Phe (Fig. 2C and supplemental Fig. S1). Substitution of Gly 17 to Phe had only a minute effect on the potency of Lqh2, whereas substitution of Phe 15 by nonaromatic residues had a significant effect at rNa v 1.2a ( Fig. 1 and supplemental Table S1). This result led us to examine whether substitution of Glu 15 to Phe in Lqh␣IT would increase its potency for rNa v 1.2a. Indeed, the potency of mutant Lqh␣IT E15F , in which the core domain was aligned in principle with that of Lqh2, increased 30-fold with no change in efficacy, which was similar to that of Lqh␣IT (Fig. 2). In sharp contrast was the result obtained when the NC domain (five-residue turn and C-tail) of Lqh2 was constructed on the scaffold of Lqh␣IT. Whereas the potency of mutant Lqh␣IT Lqh2(8 -10,56 -64) was similar to that of the unmodified Lqh␣IT, its efficacy increased substantially and was comparable with that of Lqh2 (Fig. 2). When both constructs were combined, the potency and efficacy of the chimera, Lqh␣IT Lqh2 (8 -10,15,56 -64) , at rNa v 1.2a were almost indistinguishable from those of Lqh2, and therefore this chimera was named Lqh␣IT Lqh2(face) (Fig. 2). These results suggest that the full effect at rNa v 1.2a requires the cooperative interaction of the two functional domains at the toxin exterior.
Abolishment of Lqh2 Activity at rNa v 1.4, hNa v 1.5, and rNa v 1.6-Because the bioactive surfaces of scorpion ␣-toxins show general resemblance, their diverse activities at various Na v subtypes are likely conferred by subtle differences either in structure or amino acid composition at their face of interaction with receptor site-3 (40). Any attempt to design a toxin specific for a particular Na v requires the identification of these fine differences. Although Lqh2 is highly active at a variety of mammalian Na v s such as Na v 1.2a (EC 50 ϭ 13.4 Ϯ 1.5 nM; n ϭ 6), Na v 1.4 (EC 50 ϭ 42 Ϯ 1.2 nM; n ϭ 3), Na v 1.5, Na v 1.6, and Na v 1.7 (Refs. 35, 41, and 42 and insets in Fig. 3), the scorpion ␣-toxin Amm8 (from Androctonus mauretanicus mauretanicus), which exhibits 90% sequence identity with Lqh2 (supplemental Fig. S1), shows clear preference (14.3-fold) for rNa v 1.2a over rNa v 1.4 (43). To identify the amino acid residues that determine this differential effect we mutagenized Lqh2 in a sequential manner focusing on residues that vary in Amm8, particularly at the NC domain, and examined the activity of the mutants at rNa v 1.2a versus rNa v 1.4. Interestingly, substitutions D8/A/K/N (where D8 was replaced by A, K, or N), V10/A/Y/I (where V10 was replaced by A, Y, or I), or Arg 64 by Asn and addition of Asp (R64N-D) (a single substitution where two residues replace one), as appears in Amm8, did not alter the activity at rNa v 1.4 (not shown). Therefore, guided by the idea that the general shape of the NC domain is important, we analyzed how combined substitutions at Lqh2 would affect its activity at rNa v 1.4. Complete exchange of the NC domain in Lqh2 by its Amm8 counterpart, achieved through substitutions D8N, V10I, R64N-D, resulted in a toxin mutant with Ͼ10-fold preference for rNa v 1.2a over rNa v 1.4 similar to that of Amm8 (EC 50 ϭ 42 Ϯ 6.5 nM, n ϭ 3, and 438 Ϯ 19 nM, n ϭ 3, respectively). To correlate this change in selectivity with the two entities that constitute the NC domain, we mutagenized separately the C-tail and the five-residue turn. The activity of mutant Lqh2 R64N-D was similar to that of Lqh2. However, mutant Lqh2 D8N,V10I in concentrations up to 5 M hardly slowed the decay of the sodium current at rNa v 1.4, hNa v 1.5, and rNa v 1.6, although the sodium peak current increased especially at hNa v 1.5. It has been suggested that the increase in peak current in the presence of an ␣-toxin varies with sodium channel subtype and the details of experimental conditions and likely results from increased channel open probability due to prevention of inactivation from channel states other than the open state (41,44). Therefore the most direct measure of the effect of an ␣-toxin is on the steady-state current (I ss ) normalized to the peak current (I peak ). In contrast to these results, Lqh2 D8N,V10I was highly active at rNa v 1.2a (only a 3-fold decrease; EC 50 ϭ 37 Ϯ 2.9, n ϭ 4; Figs. 1 and 3), as well as at hNa v 1.1 and rNa v 1.3 (Fig. 3).

Commonalities and Variability in the Bioactive Surfaces of
Scorpion ␣-Toxins-Comparison of the bioactive surface of Lqh2 with those of the insecticidal ␣-toxin Lqh␣IT (20,21) and the ␣-like toxin Lqh3 (22) highlights considerable commonality. In the three toxins this surface is divided into two distinct domains (Fig. 4). The conserved core domain, composed in Lqh2 of Phe 15 , Arg 18 , Trp 38 , and Asn 44 , resembles that of Lqh␣IT (Phe 17 , Arg 18 , Trp 38 , and Asn 44 ), and Lqh3 (His 15 , Phe 17 , Pro 18 , Phe 39 , and Leu 45 ) (Fig. 4). Although the spatial orientation of amino acid side chains involved in this domain relative to the ␤␣␤␤ toxin core is not identical, its general similarity may explain the ability of scorpion ␣-toxins to compete in binding at various Na v s (reviewed in Ref. 10). Of particular interest in ␣-toxins that show high activity at mammalian Na v s is residue 15 (supplemental Fig. S1). It is evident from substitution of Phe 15 in Lqh2 ( Fig. 1 and supplemental Table S1) and  (22), as well as the 30-fold increase in potency of Lqh␣IT E15F for rNa v 1.2a, that position 15 is critical for the potency at mammalian Na v s (Fig. 2) and not at the insect Na v (21).

His 15 in Lqh3
In the NC domain of Lqh2, substitution of three residues (Lys 2 , Thr 57 , and Lys 58 ) decreased the activity compared with eight such residues including Arg 2 , Ile 57 , and Arg 58 in the NC domain of Lqh␣IT, and a different set of three residues in Lqh3 (Fig. 4). The conserved positively charged residue 58 had attracted much attention, and its bioactive role was shown by chemical modifications (45,46) and mutagenesis (21,39,46,47). It was suggested for the chimera Aah2 Lqh␣IT(face) (21) that aside of the intramolecular hydrogen bond network that residue 58 is involved with, which stabilizes the structure of the NC domain, the side chain of Arg 58 may interact with a negatively charged channel residue. The latter was implicated from the finding of a co-crystallized negatively charged sulfate ion at this positively-charged niche. In the recently solved x-ray structures of Lqh␣IT and a mutant (PDB accessions 2asc and 2atb, respectively), negatively-charged ions (chloride or nitrate, respectively) have been identified at precisely the same position of that of the sulfate ion. This strengthens the suggestion that this is a focal point of Lqh␣IT interaction with a negatively-charged residue of the channel receptor. Because a similar positivelycharged niche has been shown for Aah2 (48), we further assume that residue 58 in Lqh2 interacts with a negatively-charged channel residue.
In previous work it was suggested that the shape of the NC domain in scorpion ␣-toxins determines their selectivity (21,22,38,39,49). It was further proposed that the shape of the NC domain is established by the peptide bond conformation between residues 9 and 10 in the five-residue turn. In the cis conformation, the NC domain adopts a protruding structure capable of high affinity interaction with insect Na v s. In the trans conformation, the NC domain is flat, which probably enables interaction with mammalian brain Na v s (21,50). On these grounds the unchanged activity of Lqh2 when residues at the five-residue turn were substituted (mutants Lqh2 D8K,D9N,V10Y and Lqh2 D8N,V10I ; Fig. 1 and supplemental Table S1) was intriguing, suggesting that the general shape of the NC domain did not change. Close inspection of the x-ray structure of Aah2 (PDB accession 1aho; almost identical with Lqh2 except for the N and C termini; supplemental Fig. S1) reveals that besides a disulfide bond, backbone atoms of the five-residue turn form a network of chemical interactions with C-tail residues, thereby stabilizing the structure of the NC domain. This may explain how substitution of side chains of residues at the five-residue turn probably did not change the overall shape of the NC domain in Lqh2 D8K,D9N,V10Y and Lqh2 D8N,V10I .
The close resemblance of the pharmacological properties of Lqh␣IT Lqh2(face) to those of Lqh2 suggest similar orientation of the functional surfaces with contribution of both the core domain and the NC domain to the activity at rNa v 1.2a. Although the efficacy is mainly determined by the configuration of the NC domain (see Lqh␣IT Lqh2(8 -10,56 -64) in Fig. 2), the core domain contributes to potency as indicated by (i) the 30-fold increase at rNa v 1.2a when just Glu 15 of Lqh␣IT was substituted with Phe; (ii) the 1000-fold increase when Glu 15 was substituted by Phe in Lqh␣IT Lqh2(8 -10,56 -64) providing the fully active chimera Lqh␣IT Lqh2(8 -10,15,56 -64) (Fig. 2). The stepwise construction of the bioactive surface demonstrated that cooperative interaction of the two bioactive domains at receptor site-3 is required to obtain the full toxin effect (i.e. high potency and high efficacy). The division of the bioactive surface to two distinct domains suggests that receptor site-3 is also divided to two domains, whose identification may require mutagenesis of rNa v 1.2a and thermodynamic double cycle analysis against various toxin mutants.
Subtle Alterations at the Bioactive Surface Lead to Marked Differences in Toxin Selectivity-As previously mentioned, a rational approach in the design of a target-selective scorpion toxin first requires that amino acid residues involved in bioac-  1fh3 and 2asc, respectively). Modeling of the Lqh2 structure was based on the known structure of Aah2 (PDB code 1aho) employing the SWISS-MODEL protein homology-modeling server (EXPASY). The ribbons indicate the backbone structures covered by a semi-transparent molecular surface of the toxins. Residues composing the bioactive surfaces are space-filled and colored according to their chemical nature (aliphatic, green; aromatic, magenta; polar, yellow; and positive, blue). The bioactive surfaces of Lqh␣IT and Lqh3 have been determined (21,22,51). Note the division of the bioactive surface in the three toxins to a core domain and an NC domain. tivity would be identified, and then examination of the variety of toxin mutants against different Na v subtypes may uncover a toxin derivative with altered selectivity. Whereas the first part in this approach was achieved with relative ease, none of the mutations introduced at the bioactive surface of Lqh2 changed the selectivity toward rNa v 1.2a brain and rNa v 1.4 skeletal muscle channels. Still, comparative analysis of scorpion ␣-toxins demonstrates natural variability in preference for Na v subtypes. In this respect, the high sequence similarity between Amm8 and Lqh2 (supplemental Fig. S1), on the one hand, and the 14.3-fold lower potency of Amm8 at rNa v 1.4 over rNa v 1.2a (43), on the other hand, suggested that minor differences at the toxin exterior might dictate its selectivity. This implication served as the basis of further mutagenesis of Lqh2 with focus on residues associated with the NC domain. By using this approach, the combination of two substitutions at the five-residue turn resulted in a toxin mutant, Lqh2 D8N,V10I , which hardly affected rNa v 1.4, whereas the activity at rNa v 1.2a decreased only to a small extent (Fig. 3). It seems that the double mutation changed the toxin exterior in such a way that had little effect on binding to Na v 1.1, Na v 1.2a, and Na v 1.3, but a strong effect on the binding to Na v 1.4. This suggestion is supported by the inability of high Lqh2 D8N,V10I concentrations (5 M) to inhibit Lqh2 (50 nM) activity at rNa v 1.4 (n ϭ 4; not shown). A possible explanation to the substantial difference in Lqh2 D8N,V10I interaction with these Na v subtypes is that receptor site-3 differs between Na v 1.1-1.3 and Na v 1.4 -1.6. Validation of this suggestion may require comparison of the structure of the toxin-channel complexes, which currently is still an unachievable task.
Overall, Lqh2 D8N,V10I seems to be a valuable toxin derivative in that it may be used as a specific probe for analysis of the distribution and function of rNa v 1.1-1.3 channels in various tissues, in studying their expression through embryonic development, and as a model for design of selective drugs in genetic disorders that involve these channels, such as epileptic seizures where the inactivation of Na v 1.2 is accelerated (8). The design of a selective Lqh2 derivative for rNa v 1.1-1.3 may now facilitate the design of ␣-toxins with high preference for other Na v s.