Elucidation of the Molecular Basis of Selective Recognition Uncovers the Interaction Site for the Core Domain of Scorpion α-Toxins on Sodium Channels*

Neurotoxin receptor site-3 at voltage-gated Na+ channels is recognized by various peptide toxin inhibitors of channel inactivation. Despite extensive studies of the effects of these toxins, their mode of interaction with the channel remained to be described at the molecular level. To identify channel constituents that interact with the toxins, we exploited the opposing preferences of LqhαIT and Lqh2 scorpion α-toxins for insect and mammalian brain Na+ channels. Construction of the DIV/S1-S2, DIV/S3-S4, DI/S5-SS1, and DI/SS2-S6 external loops of the rat brain rNav1.2a channel (highly sensitive to Lqh2) in the background of the Drosophila DmNav1 channel (highly sensitive to LqhαIT), and examination of toxin activity on the channel chimera expressed in Xenopus oocytes revealed a substantial decrease in LqhαIT effect, whereas Lqh2 was as effective as at rNav1.2a. Further substitutions of individual loops and specific residues followed by examination of gain or loss in Lqh2 and LqhαIT activities highlighted the importance of DI/S5-S6 (pore module) and the C-terminal region of DIV/S3 (gating module) of rNav1.2a for Lqh2 action and selectivity. In contrast, a single substitution of Glu-1613 to Asp at DIV/S3-S4 converted rNav1.2a to high sensitivity toward LqhαIT. Comparison of depolarization-driven dissociation of Lqh2 and mutant derivatives off their binding site at rNav1.2a mutant channels has suggested that the toxin core domain interacts with the gating module of DIV. These results constitute the first step in better understanding of the way scorpion α-toxins interact with voltage-gated Na+-channels at the molecular level.

as the channel site of interaction with peptide toxins from scorpions, sea anemones, and spiders capable of inhibiting the channel inactivation process. Despite their similar effect and ability to compete for binding (3)(4)(5), these toxins differ tremendously in structure (6,7), thus raising questions as to commonalities and differences in the molecular structure of receptor site-3.
Na v s are composed of a pore-forming ␣-subunit (ϳ260 kDa) that in mammals is associated with one or two ␤-subunits and in insects with the TipE accessory subunit (1, 8 -12). The ␣-subunit consists of four homologous domains (DI-DIV), each composed of six transmembrane segments (S1-S6) connected by intra-and extracellular loops. A key feature of Na v s is the voltage-dependent activation enabled by the gating module formed by transmembrane segments S1-S4 in each of the four domains. The positively-charged S4 segments respond to changes in membrane potential and move outwards across the membrane electric field, leading to opening of the channel pore and transient increase in sodium conductance that is followed by fast inactivation (1,13). The fast inactivation is coupled to the movement of the voltage sensor in DIV, which triggers the occlusion of the inner side of the channel pore by the intracellular loop that connects DIII and DIV (inactivation loop; 5, 14 -16). Site-3 toxins have been shown to impede the movement of the voltage sensor in DIV, thereby inhibiting fast inactivation (17,18).
Receptor site-3 has been localized at low resolution to the extracellular loops in domains I and IV using a photoaffinitylabeled scorpion ␣-toxin (Lqq5 from Leiurus quinquestriatus quinquestriatus) and antibodies directed to specific regions of the external loops in domains I (S5-S6) and IV (S3-S4 and S5-S6) of the rat brain channel rNa v 1.2 (19,20). Channel chimeras between rNa v 1.2 and the cardiac channel subtype rNa v 1.5 (21,22) suggested a role for DIV/S3-S4 loop in the interaction of rNa v 1.2a brain channel with Lqq5. Of particular interest was the result of charge inversion at position 1613 (E1613R), which decreased the affinity for Lqq5 by 62-fold (21). Mutagenesis of nearby residues in the rNa v 1.2a channel and the equivalent loop of the skeletal muscle Na v (rNa v 1.4) suggested a putative role for other residues of DIV/S3-S4 (Asp-1428, Lys-1432, Tyr-1433, Phe-1434, and Val-1435) in the interaction of the sodium channel with scorpion ␣-toxins (21, 23, 24; Fig. 1).
Despite these results, the channel components that constitute receptor site-3 have not been fully determined.
Scorpion ␣-toxins constitute a family of structurally related polypeptides of 61-67 amino acids reticulated by four conserved disulfide bridges. Despite the similarity in sequence and three-dimensional structure, these toxins exhibit substantial differences in preference for mammalian and insect Na v s, and on this basis are subdivided into subgroups ( The pharmacology, electrophysiological effects, structures, and bioactive surfaces (see Fig. 2) of scorpion ␣-toxins have been studied extensively (3). Their functional surface is bipartite and is divided between two domains: a core domain that involves short loops that connect the conserved secondary structure elements of the molecule core and an NC domain composed of the five-residue turn (residues 8 -12) and the C-terminal segment (see Fig. 2A; Refs. 26 -29). The difference in amino acid composition and spatial arrangement of the NC domain was suggested to dictate the variations in preference among ␣-toxins for distinct Na v s.
Here, we used two ␣-toxins, Lqh2 and Lqh␣IT, which vary greatly in preference for insect and mammalian brain Na v , to analyze channel constituents involved in this selective recognition. We chose Na v 1.2 for construction of chimeric Na v chan-nels because it is more sensitive to Lqh2 and much less sensitive to Lqh␣IT than cardiac or skeletal muscle Na V channels. Construction of rNa v 1.2a external loops in the background of DmNa v 1 as well as point mutagenesis revealed the role of DI/S5-S6 and DIV/S3 in the selective recognition by Lqh2. Comparison of depolarization-induced dissociation of Lqh2 and its mutant derivatives from various channel mutants suggested that the core domain of Lqh2 interacts with the voltagesensing module in DIV of the channel.

EXPERIMENTAL PROCEDURES
Toxin Production and Modification-Lqh2 was produced in Escherichia coli strain BL21 as described (29). The toxin derivatives bear a His 6 tag and a thrombin cleavage site at their N termini that does not hamper their activity (25).
Channel Modification-The cDNA encoding the DmNa v 1 sodium channel of Drosophila melanogaster cloned in pAlter expression vector (Promega) was digested by XbaI, ApaI, and NotI, dividing the entire sequence to two fragments, one encoding domains I and II (3650 bp) and the other encoding domains III and IV (2890 bp). The two fragments were cloned into pBluescript (Stratagene), and the resulting plasmids were used in all further steps of PCR-driven mutagenesis and construction of channel mutants. Mutagenized DNA fragments were back inserted to the original plasmid and the DNA sequence was verified prior to RNA production for injection into Xenopus laevis oocytes. The cDNA encoding the rNa v 1.2a rat brain sodium channel, cloned in the expression vector pCDM8 (Invitrogen), was used in a similar way for mutagenesis, and BstEII and BspMI restriction sites were used for back insertion to the original plasmid.
Expression of Na v s in Oocytes and Two-electrode Voltage Clamp Experiments-cRNAs encoding the ␣-subunit of each channel and the auxiliary ␤1 and TipE subunits were transcribed in vitro using T7 RNA polymerase and the mMESSAGE mMACHINE TM system (Ambion, Austin, TX) and injected into Xenopus oocytes as described previously (30). Currents were measured 1-3 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). Capacitance transients and leak currents were removed by subtracting a scaled control trace utilizing a P/6 protocol (31). Bath solution contained the following: 96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 1.8 mM CaCl 2 , 5 mM HEPES, pH 7.5. Toxins were diluted with bath solution containing 1 mg/ml bovine serum albumin and applied directly to the bath in the final desired concentration. To avoid application artifacts, 1 mg/ml bovine serum albumin solution was applied prior to toxin addition. For the G-V analysis, mean conductance (G) was calculated from the peak current-voltage relations 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. The normalized conductance-voltage relations were fit with either a one-or two-component Boltzmann distribution according to Equation 1, Cylinders represent transmembrane ␣-helical segments. The gating module (segments S1-S4) of each domain resides near the pore module region (segments S5-S6) of the next domain in a clockwise orientation. Boldface lines represent the external and internal loops that connect the transmembrane segments. The pore loop (SS1-SS2) is colored red, and the inactivation ball is indicated by the IFM motif. Site-3 has been thus far assigned to the extracellular loops S5-S6 of DI and DIV and DIV/S3-S4 (19 -21, 23, 24). The position of Glu-1613 is highlighted by a blue circle. Bottom, sequence alignment of DIV/S3-S4 extracellular loop (in italics) of the mammalian brain channel (rNa v 1.2a), mammalian skeletal muscle channel (rNa v 1.4), heart channel (hNa v 1.5) and Drosophila channel (DmNa v 1). Residue substitutions that affected the activity of scorpion ␣-toxins are in blue (Glu-1613 at rNa v 1.2a DIV/S3-S4 (21)) and in red (Lys-1432, Tyr-1433, Phe-1434, and Val-1435 of rNa v 1.4 (23,24)).
where V11 ⁄ 2 and V21 ⁄ 2 are the respective membrane potentials for two populations of channels for which the mean conductance is half maximal, k 1 and k 2 are their respective slopes, and A defines the proportion of the second population (amplitude) with respect to the total. For fits in which only one population of channels was apparent, A was set to zero. The voltage dependence of steady-state fast inactivation was described using a single Boltzmann distribution as shown in Equation 2, where I is the peak current obtained at the depolarizing test step, I max is the current without a preceding conditioning step, V is the membrane potential of the conditioning step, V h is the membrane potential at which half-maximal inactivation is achieved, k is the slope factor, ␣ 0 is the remaining normalized peak current at highly depolarizing conditioning potentials, and ␣ 1 is the normalized amplitude (32).
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, normalized to the maximal effect set to 1, and fitted 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.
Determination of Voltage-dependent Dissociation of Toxin-Voltage-dependent toxin dissociation was measured with a two-pulse protocol. Conditioning dissociation pulses between Ϫ20 mV to ϩ105 mV were applied from a Ϫ100 mV holding potential, following 50 ms at Ϫ100 mV for channels recovery from fast inactivation. Sodium currents were then elicited with a 50-ms test pulse to Ϫ20 mV. The experiments were conducted at saturating toxin concentrations, and a 30-s interval between test pulses ascertained maximal toxin rebinding. The extent of removal of fast inactivation represented by the ratio I ss /I peak was plotted as a function of the conditioning voltage and was fitted with the Boltzmann distribution described in Equation 4, where V1 ⁄ 2 is the half-maximal dissociation voltage, and k1 ⁄ 2 is the corresponding slope factor.

RESULTS
The extreme difference in potency of Lqh2 and Lqh␣IT on rNa v 1.2a rat brain and DmNa v 1 Drosophila voltage-gated sodium channels is correlated with differences in their bioactive surfaces as well as differences in their channel receptor sites. Whereas the toxins have been thoroughly dissected and their bioactive surfaces documented ( Fig. 2) (26,29), their channel receptor sites are described incompletely. Based on previous reports suggesting that receptor site-3 is associated with channel external loops of domains IV and I (19,21,23,24,33), our approach to identify channel constituents that determine toxin recognition was to first uncover the extracellular loops involved with toxin selectivity and then use this information to characterize toxin interaction with receptor site-3.
Construction of rNa v 1.2a External Loops in DmNa v 1 and Analysis of Sensitivity to Lqh2 and Lqh␣IT-Stepwise construction of the four external loops DIV/S1-S2, DIV/S3-S4, DI/S5-SS1, and DI/SS2-S6 from rNa v 1.2a in the background of DmNa v 1 (see sequences in supplemental Fig. 1) provided channel chimeras, which we named by indicating the parent channel in full type and the substituted segments or amino acid residues as superscripts (supplemental Table 1). The voltage-dependent activation and inactivation properties of these chimeras varied only slightly from those of DmNa v 1 (supplemental Table 1). However, the sensitivity of these chimeric channels to the two toxins varied greatly (Table 1). Most dramatic was the decrease in sensitivity of the final chimera DmNa v 1 rNav1.2a(DI/S5-SS1ϩSS2-S6,DIV/S1-S2ϩS3-S4) to Lqh␣IT by nearly 3 orders of magnitude from an EC 50 of 0.36 Ϯ 0.04 nM for DmNa v 1 to an EC 50 of 241 Ϯ 47 nM for the chimera (Table 1). On the other hand, Lqh2 activity increased from an EC 50 value of 6920 Ϯ 1420 nM for DmNa v 1 to an EC 50 value of 26.7 Ϯ 3.1 nM for the chimera (Table 1 and Fig. 3A). This suggested that the mammalian receptor site-3 of Lqh2 has been nearly fully constructed in the background of the insect channel. Therefore, this chimera was named DmNa v 1 rNav1.2a(site-3 face) .
We next analyzed the external loops of DI pore module. The sensitivity of DmNa v 1 rNav1.2(DI/S5-SS1ϩSS2-S6) to Lqh2 increased 56-fold compared with the sensitivity of DmNa v 1 to this toxin, whereas these substitutions had barely an effect on Lqh␣IT activity. However, the sensitivity of DmNa v 1 rNav1.2(DI/S5-SS1) and DmNa v 1 rNav1.2(DI/SS2-S6) to Lqh2 was as poor as that of DmNa v 1, and the Lqh␣IT effect on these channel chimeras hardly changed (Table 1). Thus, single loop substitutions were not very effective suggesting that there may be a cooperative effect of the two extracellular loops in the pore module on the ability of Lqh2 to interact with the brain sodium channel. Overall, these results indicated that both the gating module of DIV and the pore module of DI were required to form a complete Lqh2 receptor site in DmNa v 1.

Substitution of Residues in DIV External
Loop That Differ between rNa v 1.2a and DmNa v 1-The substantial role of DIV/S3-S4 in channel sensitivity to the two ␣-toxins (Table  1) as well as previous reports on changes in activity of ␣-toxins upon substitution of Glu-1613 in rNa v 1.2a (21) and its Asp-1428 and Asp-1701 equivalents in rNa v 1.4 (23,33) and DmNa v 1 (7) suggested that differences in this loop between the two channels might be involved with the varying potency of ␣-toxins. Hence, we examined the effect of reciprocal exchange of Glu-1613 at rNa v 1.2a and Asp-1701 at DmNa v 1 on channel sensitivity to Lqh2 and Lqh␣IT. Whereas the substitution D1701E at DmNa v 1 markedly reduced Lqh␣IT activity (139-fold decrease; Table 1), this channel mutant remained insensitive to Lqh2. In sharp contrast was the effect of the reciprocal substitution in rNa v 1.2a (E1613D), which improved Lqh␣IT affinity by ϳ1000-fold, closely resembling the potency of Lqh2 at rNa v 1.2a (EC 50 ϭ 18.8 Ϯ 3.3 nM; Table 1; Fig. 3B). Following this striking result, we analyzed the Glu-to-Asp substitution in the context of the chimera DmNa v 1 rNav1.2a(site-3 face) . We found that the activity of Lqh␣IT on DmNa v 1 rNav1.2a(site-3 face-E1613D) was restored (EC 50 ϭ 0.37 Ϯ 0.03 nM) and was similar to the activity of the toxin at DmNa v 1, whereas Lqh2 activity persisted (Table 1; Fig. 3A). Thus, Glu-1613 at DIV/S3-S4 is a key factor that hinders Lqh␣IT interaction with receptor site-3 of the rat brain channel rNa v 1.2a.  (26,29,40). Residues of the core domain are colored magenta, and residues of the NC domain are colored blue. Bottom, sequence alignment of the two toxins. The bioactive surface of scorpion ␣-toxins consists of the conserved core domain (residues in magenta) and the diverse NC domain (residues in blue). Lqh2 and Lqh␣IT are similar in structure and share ϳ70% sequence similarity, yet they exhibit opposing preferences for the mammalian brain and insect Na v s (Table 1).

TABLE 1 Effect of Lqh2 and Lqh␣IT on channel mutants
The channels were clamped at a holding potential of Ϫ80 mV, and currents were elicited by depolarization to Ϫ20 mV in the presence of increasing toxin concentrations. At each toxin concentration, the currents were allowed to reach a steadystate level prior to the final measurement. Determination of dose-dependent effect of the toxin (removal of fast inactivation) is described in detail under "Experimental Procedures", and the EC 50 values provided are mean Ϯ S.E., where n stands for the number of cells analyzed. No effect denotes lack of effect on channel inactivation in the presence of 5 M toxin. , whose sensitivity to Lqh2 increased 14-fold, with only a minor effect in its sensitivity to Lqh␣IT (Table 1), compared with DmNa v 1. This suggested that this region in DIV gating module of rNa v 1.2a plays a role in Lqh2 interaction with its receptor.
Analysis of Voltage-dependent Dissociation of Lqh2 Mutants from Na v 1.2a-The swap of the Lqh2 receptor at rNa v 1.2a onto DmNa v 1 indicated an important role of the gating module of DIV and pore module of DI in toxin recognition. On the toxin side, the bioactive surface of Lqh2 has recently been shown to be composed of two domains: the core domain (Phe-15, Arg-18, Trp-38, and Asn-44) and the NC domain (Lys-2, Thr-57, Lys-58; Fig. 2A; Ref. 29). On these grounds and the fact that binding of scorpion ␣-toxins is voltage-dependent (1, 34 -37), which suggests toxin binding at the mobile voltage-sensing region, we analyzed which of the toxin bioactive domains interacts with the DIV gating module of rNa v 1.2a. This analysis was based on the assumption that the dissociation of toxin mutants upon depolarization would vary from that of the unmodified toxin if the substitutions affect a site of interaction with the channel gating module.
We first analyzed the voltage-dependent dissociation of Lqh2 at a saturating concentration from rNa v 1.2a expressed in Xenopus oocytes by applying a series of depolarizing prepulses to ϩ105 mV of variable durations (see "Experimental Procedures"). However, under this protocol with prepulse duration of up to 200 ms, the toxin had barely dissociated, as indicated by the persistent inhibition of inactivation following the test pulse (Fig. 4A). Considerable dissociation of Lqh2 was obtained following a prepulse duration of at least 500 ms, and the apparent toxin effect was nearly abolished after 2 s (not shown). Because depolarizing prepulses exceeding 200 ms promote channel entrance into slow inactivation (38), we limited the assays with saturating toxin concentrations to a 200-ms depolarizing pre-pulse (Fig. 4A, inset). The same protocol was used to analyze Lqh2 mutants K2A, F15A, N44A, and T57A. This analysis revealed that the voltage-dependent dissociation of Lqh2 substituted at the core domain was markedly enhanced, especially for F15A and N44A. The V1 ⁄ 2 for dissociation of mutants Lqh2 F15A and Lqh2 N44A was 82.5 Ϯ 0.6 and 53 Ϯ 2.7 mV, respectively, compared with 129 Ϯ 13 mV for Lqh2 (Table 2), and a complete dissociation was observed following a depolarizing prepulse to ϩ105 mV (Fig. 4A). On the other hand, the voltage-dependent dissociation of toxin mutants Lqh2 K2A and Lqh2 T57A of the NC domain ( Fig. 2; Ref. 29) was similar to that of the unmodified toxin, despite a large increase in their EC 50 values ( Fig. 4B and Table 2). These results have suggested that the core domain of Lqh2 interacts with a channel region undergoing modification upon depolarization.
Effects of Channel Substitutions on Voltage-dependent Dissociation of Lqh2-Because an E1613R substitution in rNa v 1.2a has been shown to enhance the dissociation of the ␣-toxin Lqq5 (21), we examined the effects of charge neutralization and inversion at this position (E1613N and E1613R) on Lqh2 voltage-dependent dissociation (Tables 2 and 3). Notably, Lqh2 voltage-dependent dissociation off rNa v 1.2a E1613N and rNa v 1.2a E1613R channel mutants was markedly facilitated compared with its dissociation off the unmodified channel. The dissociation off rNa v 1.2a E1613R was more prominent as it began at ϩ40 mV and was complete following a prepulse to ϩ105 mV (Fig. 4, A and C, and Table 2). However, despite the substantial shift in V1 ⁄ 2 for both channel mutants, the EC 50 values of Lqh2 barely changed (Tables 2 and 3). This result suggested that substitution of Glu-1613 at DIV/S3-S4 impairs an interaction of Lqh2 with an activated and/or fast-inactivated channel state.
Based on the results of the dissociation assays, we examined the effects of Lqh2 and its bioactive surface mutants on a number of channel mutants modified at the DIV extracellular loop S1-S2, the distal region of S3 and the beginning of S3-S4 (Table  3). These channel determinants were selected on the basis of the swap experiments, which suggested that they determine the selectivity of Lqh2, and on unpublished results of the effects of point mutations at DIV/S1-S2 on Lqh2 activity (39). Of all
Voltage-dependent Dissociation of Toxin Mutants from Channel Mutants-Analysis of the voltage-dependent dissociation of Lqh2 mutants Lqh2 F15A and Lqh2 N44A off rNa v 1.2a E1613R revealed substantial enhancement, as indicated by the prominent shifts in V1 ⁄ 2 to more negative membrane potentials (V1 ⁄ 2 values of 6.0 Ϯ 0.3 mV and Ϫ10.7 Ϯ 2 mV, respectively, compared with 64.4 Ϯ 4.4 mV for Lqh2; Fig. 4C; Table 2). In light of this result, we sought other residues that putatively interact with Lqh2. As replacement of the Leu-Val-Leu-Ser sequence at the C-terminal end of DIV/S3 in DmNa v 1 by its rNa v 1.2a equivalent, Met-Phe-Leu-Ala, resulted in gain of Lqh2 function at the chimeric channel, DmNa v 1 rNav1.2a(DIV/S3-MFLA) ( Table 1), we examined the effect of substitutions M1609A and F1610A at rNa v 1.2a on Lqh2 activity. Whereas substitution M1609A had no effect (Table 3), the substitution F1610A resulted in 5-fold increase in the EC 50 of Lqh2 (Table  2). Accordingly, we analyzed the effect of F1610A substitution

TABLE 2 Effects of Lqh2 mutants and their voltage-induced dissociation off rNa v 1.2a and the F1610A and E1613R channel mutants
The EC 50 values for rNa v 1.2a are from Kahn et al. (29). The voltage at which 50% of the toxin dissociated (V1 ⁄ 2 ) was calculated from the slopes presented at Fig. 4, A, C, and D. The values provided are mean Ϯ S.E. of three to six measurements (n). ND, not determined (when the EC 50 values could not be calculated because a saturating effect could not be reached). on the voltage-dependent dissociation of Lqh2 and its mutant derivatives. As shown in Fig. 4D and Table 2, the dissociation of Lqh2 off rNa v 1.2a F1610A was markedly facilitated and observed at a lower membrane potential compared with the toxin dissociation off the unmodified channel. Furthermore, whereas the dissociation of the NC-domain toxin mutant, Lqh2 T57A , off rNa v 1.2a F1610A resembled that of the unmodified toxin, the dissociation of the core domain toxin mutant Lqh2 F15A off rNa v 1.2a F1610A was markedly facilitated (Fig. 4D; Table 2). These results suggested that upon interaction, the core domain of Lqh2 is in close proximity to the distal region of S3 at DIV in the rat brain sodium channel.

DISCUSSION
Determination of receptor sites of Na v modifiers is complex due to the lack of channel structure and their conformational rearrangements during gating. The experimental approach in the present study was to first identify channel regions involved in toxin selectivity and then dissect the relevant regions in search for specific residues associated with the receptor site. Although receptor site-3 is not necessarily constituted from components that differ between the brain and insect sodium channels, the successful swap and gain of high activity of Lqh2 at the chimeric channel DmNa v 1 rNav1.2a(site-3 face) has indicated that the external loops of the gating module of domain IV and pore module of domain I in rNa v 1.2a play an important role in toxin selectivity and that they are spatially arranged in the chimeric channel as in rNa v 1.2a. The inverse experiment to construct rNa v 1.2a such that it acquires high sensitivity to Lqh␣IT surprisingly did not require swap of external loops from DmNa v 1. Instead, a single conservative substitution, E1613D, converted the brain channel to high sensitivity toward Lqh␣IT. In the skeletal muscle channel Na v 1.4 and cardiac channel Na v 1.5, the position equivalent to Glu-1613 is occupied by an Asp residue, and both channels are sensitive to Lqh␣IT (32,36). Moreover, substitution at this position in rNa v 1.4 (D1428E) decreased the effect of Lqh␣IT (23). These observations are in concert with the gain of function of Lqh␣IT at the E1613D mutant of rNa v 1.2a and indicate that Glu-1613 at DIV/S3-S4 of rNa v 1.2a is in close proximity to the surface of interaction with FIGURE 5. Model of Lqh2 interaction with rNa v 1.2 resting state. The external loops DIV/S1-S2 and S3-S4 of rNa v 1.2a were constructed on the structural model of K v 1.2 in its resting state (43) using the Swiss-PdbViewer. The internal loops and the gating modules of DI, DII, and DIII were removed. Shown in black are the remaining pore modules of DII, DIII, and DIV, whereas the DI pore module is shown in light gray. Due to the substantial difference in size, the DI/S5-SS1 external loop is omitted (indicated by the orange dashes). DIV/S1 is shown in green, DIV/S2 is shown in blue, DIV/S3 is shown in light orange, and DIV/S4 is shown in dark orange. Lqh2 modeling relied on its close resemblance to Aah2 (29; Protein Data Bank code 1AHO). Phe-15 and Asn-44 are bioactive residues of Lqh2 that are in close proximity to Phe-1610 and Glu-1613 of the channel, respectively (colored sticks according to their chemical nature). Docking of Lqh2 core domain at DIV gating module was performed using DockingServer. Although the residues of the toxin NC domain that may interact with residues at DI/S5-SS1 and DIV/S1-S2 have not been clarified yet, further modeling was performed manually to show this proximity, while avoiding side chain clashes. The final figure was drawn using PyMOL.

TABLE 3 Lqh2 activity and voltage-dependent dissociation off rNa v 1.2a and mutants
Determination of the dose-dependent effect of the toxin (removal of fast inactivation) is described in detail under "Experimental Procedures", and the EC 50 values provided are mean Ϯ S.E. of at least three measurements (n). V1 ⁄ 2 , the voltage at which 50% of the toxin dissociated off the channel. Lqh␣IT. The gain of rNa v 1.2a sensitivity to Lqh␣IT upon a single substitution demonstrates that the brain channel bears a receptor site for Lqh␣IT, and the primary reason for lack of Lqh␣IT activity is the hindrance caused by Glu-1613. Furthermore, these results also suggest that receptor sites 3 at both the brain and insect channels are similar although not identical. The interaction of DIV/S3-S4 external loop with scorpion ␣-toxins has been demonstrated in previous studies (21,23,40,41), and it was proposed that the positively charged S4 segment moves outwards upon depolarization and is capable of removing the bound toxin from its receptor site (1,5,15,16,21,34). Perturbation of this movement by the bound toxin then inhibits the subsequent conformational change required for fast inactivation. However, in contrast to the large detrimental effect of E1613R substitution at rNa v 1.2a on Lqq5 activity (21), this substitution had comparatively little effect on Lqh2 activity except when assayed in the toxin dissociation protocol (Fig. 4A; Table  2). This difference in effect may be attributed to the difference between the toxins but also to the experimental system employed for channel expression, such that DIV/S3-S4 may not be displayed identically when expressed in mammalian cells versus Xenopus oocytes. Indeed, we find a large effect of the E1613R mutation on Lqh2 action for Na v 1.2 channels expressed in the human embryonic kidney cell line tsA-201 (39). Nevertheless, the results obtained with Lqq5 and Lqh2 applied onto rNa v 1.2a expressed in mammalian cells (21,23,39) and the results regarding toxin dissociation obtained in this study corroborate the suggestion that Glu-1613 in rNa v 1.2a is in very close proximity to the interacting surface of scorpion ␣-toxins with this brain sodium channel.

Channel
On the basis of the bipartite bioactive surface of Lqh2 (29) and the successful swap of its receptor, we assume that Lqh2 interacts with the channel such that one of the two functional domains at the toxin surface recognizes the gating module at DIV and the other toxin domain interacts with the pore module of DI. The involvement of the distal part of DIV/S3 in this interaction might take place within a crevice in the membrane-channel interface that enables access of the toxin core domain, a scenario that resembles the interaction of a scorpion ␤-toxin with DII of rNa v 1.2a (42).
Our suggestion that the core domain of Lqh2 interacts with DIV gating module is based on the findings of large changes in the depolarization-induced dissociation of the core domain toxin mutants compared with the unmodified toxin (Fig. 4A), as well as on the enhanced dissociation of Lqh2 from the channel mutants modified at this channel region, rNa v 1.2a E1613R and rNa v 1.2a F1610A (Table 2; Fig. 4, C and D). We used these data to construct an initial model of the putative interaction of Lqh2 with rNa v 1.2a by employing the three-dimensional structure of the potassium channel K v 1.2 (43) and assuming that the intersegmental region of both channel types would be similar (Fig. 5). In this initial model, Phe-15 of the toxin is in close proximity to Phe-1610 in DIV/S3 and to Glu-1613 in DIV/S3-S4, whereas Asn-44 of the toxin is in close proximity to Glu-1613, in agreement with our conclusion that the toxin core domain interacts with the voltage-sensing module of the channel. Although at this stage we are unable to determine which of the toxin domains interacts first with the channel, from the toxin unbinding experiments it seems that, upon depolarization, dissociation of the toxin off the channel is initiated at the core domain.
In summary, this study reveals sodium channel determinants involved in scorpion ␣-toxin selectivity as well as illuminates for the first time the ␣-toxin domain that interacts with the channel voltage sensor, which enables initial modeling of its docking at the channel. Further mutagenesis and double-mutant cycle analysis, including residues that are spatially conserved in sodium channels, are required to identify the individual amino acid residues in the DI/S5-S6 and DIV/S1-S2 loops that participate in toxin binding.