Structure-Function Map of the Receptor Site for β-Scorpion Toxins in Domain II of Voltage-gated Sodium Channels*

Voltage-gated sodium (Nav) channels are the molecular targets of β-scorpion toxins, which shift the voltage dependence of activation to more negative membrane potentials by a voltage sensor-trapping mechanism. Molecular determinants of β-scorpion toxin (CssIV) binding and action on rat brain sodium channels are located in the S1-S2 (IIS1-S2) and S3-S4 (IIS3-S4) extracellular linkers of the voltage-sensing module in domain II. In IIS1-S2, mutations of two amino acid residues (Glu779 and Pro782) significantly altered the toxin effect by reducing binding affinity. In IIS3-S4, six positions surrounding the key binding determinant, Gly845, define a hot spot of high-impact residues. Two of these substitutions (A841N and L846A) reduced voltage sensor trapping. The other three substitutions (N842R, V843A, and E844N) increased voltage sensor trapping. These bidirectional effects suggest that the IIS3-S4 loop plays a primary role in determining both toxin affinity and efficacy. A high resolution molecular model constructed with the Rosetta-Membrane modeling system reveals interactions of amino acid residues in sodium channels that are crucial for toxin action with residues in CssIV that are required for its effects. In this model, the wedge-shaped CssIV inserts between the IIS1-S2 and IIS3-S4 loops of the voltage sensor, placing key amino acid residues in position to interact with binding partners in these extracellular loops. These results provide new molecular insights into the voltage sensor-trapping model of toxin action and further define the molecular requirements for the development of antagonists that can prevent or reverse toxicity of scorpion toxins.

which contain 4 -7 conserved positively charged arginine or lysine residues separated by 2 hydrophobic residues, have been postulated to be the voltage sensors of the channels (2). In response to depolarization, the S4 segments are thought to move outward across the cell membrane, and this movement exerts a force on the S4-S5 linker, pulls S5 segments away from each other, and bends the S6 segment to open the pore (3). Structural modeling suggests that in the resting state, the extracellular S1-S2 and S3-S4 loops are close to each other, whereas in the activated state these two extracellular loops move away from each other (3). These voltage-driven conformational changes in voltage sensors are the molecular targets for modulation of Na v channel function by scorpion toxins (4).
␤-Scorpion toxins enhance activation of Na v channels via binding to neurotoxin receptor site 4 (4 -7). A ␤-scorpion toxin, CssIV from the venom of the scorpion Centruroides suffusus suffusus, shifts the voltage dependence of rat brain Na V 1.2 channel activation to more negative membrane potentials (8). However, a strong priming depolarizing prepulse is required for the toxin to have its functional effect (8). A voltage sensortrapping mechanism was proposed to explain the prepulse-dependent enhancement of activation by ␤-scorpion toxins (8). According to the model, the toxin binds to its receptor site in the extracellular loops connecting transmembrane segments S1-S2 and S3-S4 in homologous domain II (IIS1-S2 and IIS3-S4) in the resting state of the voltage sensor in a concentrationdependent, bimolecular reaction to form the initial toxin-channel complex. Upon strong depolarization, the IIS4 segment moves outward and rotates, and the toxin binds to newly accessible amino acid residues on the IIS1-S2 and IIS3-S4 loops and the extracellular end of the IIS4 segment in a unimolecular, concentration-independent reaction. In this activated position, the voltage sensor is tightly bound to the rigid toxin and is trapped in its activated conformation. Upon repolarization of the cell membrane, the trapped IIS4 voltage sensors remain activated; therefore, subsequent depolarizations activate the channels at more negative membrane potentials because one of the four voltage sensors is already activated (8 -10).
Previous studies revealed amino acid residues in IIS1-S2 and IIS3-S4 of rNa v 1.2a channels whose substitution altered the binding and action of CssIV (8 -11). In the present experiments, we have refined the molecular map of the amino acid residues in these two extracellular linkers that are required for CssIV binding and voltage sensor trapping, and we have discovered a hot spot for toxin action where single mutations in adjacent amino acid residues can either enhance or impair toxin action substantially. Our results enable a detailed structural and mechanistic model for CssIV interaction with its receptor site in the voltage sensor of domain II of rNa v 1.2a channels and provide further support for the voltage sensor-trapping model of scorpion toxin action and the sliding helix model of voltage sensor function.

EXPERIMENTAL PROCEDURES
PCR-directed Mutagenesis-Mutations were introduced by site-directed mutagenesis using a PCR-based strategy as described previously (9). Hydrophobic amino acid residues, serine, cysteine, and threonine were substituted with alanine to prevent toxin-binding interactions with their side chains. Glutamine was substituted for the larger charged residues arginine, lysine, and glutamic acid, whereas asparagine was substituted for aspartic acid residues. Single amino acid chimeras were generated by aligning amino acid sequences of Na v channels subtypes and substituting nonconserved amino acid residues. All the WT and mutant cDNAs were subcloned into pCDM8 plasmid.
cDNA Transfection-The methods of cDNA transfection have been described in detail previously (10). Briefly, cDNAs encoding wild-type (WT) and mutant rNa v 1.2a channels were transiently expressed in tsA-201 cells by calcium-phosphate transfection. pEBO-pCD8-leu2 was co-transfected into tsA-201 cells, and the cell surface CD8 marker was identified by incubation with polystyrene microspheres pre-coated with anti-CD8 antibody. The expressed CD8 protein was used to identify cells that express rNa v 1.2a channels. Transfected cells were subcloned 12-18 h after transfection. Electrophysiological recordings were performed 18 -72 h after transfection. WT and mutant rNa v 1.2a cDNAs were always transfected and studied in parallel to ensure that differences from WT were truly representative of the phenotype of the mutation.
Electrophysiological Recording and Analysis-The electrophysiological recording procedures have been described previously (9). Briefly, the whole-cell patch clamp configuration was utilized for Na ϩ current recording with extracellular recording solution containing (in mM) NaCl (150), Cs-HEPES (10), MgCl 2 (1), KCl (2), CaCl 2 (1.5), and 0.1% BSA, pH 7.4, and intracellular recording solution containing (in mM) N-methyl-D-glucamine (190), HEPES (10), MgCl 2 (10), NaCl (10), and EGTA (5), pH 7.4. Linear leak and residual capacitance currents were subtracted using an online P/Ϫ4 subtraction paradigm. To assess the extent of negative shift of the voltage dependence of activation caused by CssIV E15A , the tsA-201 cells were held at Ϫ100 mV and test depolarizations were applied to potentials from Ϫ100 to ϩ20 mV in 5-mV increments. Current-voltage (I-V) plots were generated from peak currents elicited by each depolarization. Each test depolarization was either applied alone or preceded by a 1-ms prepulse to ϩ50 mV followed by a 60-ms interval at the holding potential. We tested the functional properties of each mutant rNa v 1.2a construct in the absence of toxin to examine the effect of the mutant residue, followed by recordings in the presence of CssIV E15A . The voltage dependence and kinetics of each mutant channel were initially screened with 500 nM CssIV E15A to detect differences from WT. If a mutant channel did have significant differences in function from WT, further tests were performed to investigate the phenotype of mutant channels in detail. All the data were analyzed with Igor Pro (WaveMetrics, Lake Oswego, OR). Normalized I-V curves were fit with a function including a single Boltzmann component of the form: where V 1 ⁄2 is the half-activation voltage in mV, and k is a slope factor in mV. In analyses of the rates of development and reversal of voltage sensor trapping, currents were normalized to the maximal peak current of a preceding I-V plot. All data are presented as mean Ϯ S.E.
Structural Modeling-Homology and de novo modeling of the voltage sensor in domain II of rat Na v 1.2a channels was performed using the Rosetta-Membrane method (3,12). rNa v 1.2a sequence (residues Leu 754 -Gly 875 ) was aligned with the K v 1.2-K v 2.1 chimera channel sequence (residues Ser 158 -Gln 315 (13)) using ClustalX software (14). The alignment of the IIS4 segments was manually adjusted to align the first arginine in the IIS4 of rNa v 1.2 with Gln 290 in K v 1.2-K v 2.1 chimera channels, which corresponds to the conserved R0 position in most K v channels (supplemental Fig. S1). 5,000 models of the voltage sensor were generated, and the lowest scoring model was selected as the best model. Modeling of CssIV toxin was performed using the Rosetta method for modeling of soluble proteins (15). ␤-Scorpion toxin sequence was aligned with the neurotoxin 2 sequence (X8, Protein Data Bank code 1JZA) using ClustalX software (14) (see supplemental Fig. S1). 5,000 models of the toxin were generated and the lowest scoring model was chosen as the best model. Docking simulations of CssIV binding to the voltage sensor in domain II of the rNa v 1.2a channel were performed using the Rosetta docking method (16,17). Backbone flexibility of the extracellular part of the voltage sensor (residues Cys 768 -Gly 800 and Ser 832 -Arg 856 ) was allowed during simulations and Tyr 40 and Tyr 42 residues of the ␤-scorpion CssIV toxin were required to be at the voltage sensor-toxin interface. 10,000 models were generated and the best model was chosen among 20 lowest scoring models as the model that fit the majority of available experimental data on key residues contributing to interaction of the CssIV toxin with voltagesensing domain II of rat Na v 1.2a channels (8,9,18).

RESULTS
Modification of Voltage-dependent Activation of Na V 1.2 Channels by CssIV E15A Toxin-Our previous work showed that CssIV shifted the voltage dependence of activation of rNa v 1.2a channels to more negative membrane potentials (8). These functional effects required a strong depolarizing prepulse to activate the channels (8). In the present study we used a highaffinity recombinant derivative of CssIV, CssIV E15A (19). We first verified the requirement of a prepulse to observe the functional effects of CssIV E15A on WT rNa v 1.2a channels. Without a prepulse, current-voltage (I-V) relationships obtained from rNa v 1.2a channels using 15-ms depolarizations to potentials from Ϫ100 to ϩ20 mV from a holding potential of Ϫ100 mV were similar whether or not 500 nM CssIV E15A was present. Under these conditions, no Na ϩ current was activated at test potentials more negative than Ϫ60 mV (Fig. 1). Even in the presence of CssIV E15A , little inward current was activated by a test pulse to Ϫ60 mV (Fig. 1B). In contrast, when a 1-ms pre-pulse to ϩ50 mV was applied 60 ms before each test pulse, marked Na ϩ current was observed at Ϫ60 mV (Fig. 1B), and a substantial component of the Na ϩ current activated with a negatively shifted voltage dependence (Fig. 1C). These effects were not observed in the absence of CssIV E15A . This negative shift in the voltage dependence of activation reflects voltage sensor trapping by bound CssIV (8). At 500 nM CssIV E15A , a mean of 11.9% of the peak Na ϩ current activated at this negative membrane potential for WT rNa v 1.2a channels ( Fig. 1 and Table 3). We used this negative shift in the voltage dependence of activation to assay the effects of mutations in IIS1-S2 and IIS3-S4 on CssIV binding and action, and we have termed the additional current observed at a test pulse to Ϫ60 mV following a prepulse "I VST ," as an abbreviation for voltage sensor-trapping current. To conserve channel function, site-directed mutations of single amino acid residues were designed to reduce the hydrophobicity or hydrophilicity of amino acid side chains by conversion to Ala, to reduce the charge of amino acid side chains while retaining their hydrophilicity by conversion of Arg, Lys, Glu, and Asp to Asn, Gln, or Cys, or to substitute single amino acid residues from homologous Na V channels having different responses to ␤-scorpion toxins. The functional properties of several representative mutants are described first, followed by presentation of a molecular map and high-resolution structural model of the toxin-receptor interaction.
Loss of Voltage Sensor Trapping with Mutants A841N and L846A-In previous studies, we showed that amino acid residues in IIS1-S2 and IIS3-S4 were crucial for the binding and voltage sensor-trapping activity of ␤-scorpion toxin on rNa v 1.2a channels (8,9). In the present work, we used recombinant CssIV E15A to identify individual amino acid residues that are important for toxin action by mutagenesis and characterization of their functional effects by electrophysiological analysis. We found that two mutants (A841N and L846A) in IIS3-S4 greatly reduced the voltage sensor-trapping activity of CssIV E15A .
In the absence of toxin, mutation A841N did not affect the I-V relationship compared with WT ( Fig. 2A, Table 1), indicating that this mutation does not alter the voltage-dependent activation of sodium channels. As for WT channels, CssIV E15A (500 nM) did not induce significant Na ϩ current when cells expressing A841N channels were depolarized to Ϫ60 mV without a depolarizing prepulse (Fig. 2B, Table 2). However the I-V relationship of A841N was altered only slightly by a ϩ50-mV, 1-ms depolarizing prepulse in the presence of the toxin (Fig.  2B), and I VST was only 2.8% at 500 nM and increased to 5.3% at 1 M (Table 3). These results suggest that mutant A841N has reduced affinity for CssIV and possibly also reduced efficacy in voltage sensor trapping.

TABLE 1 Voltage dependence of activation
The voltage dependence of activation was measured as described under "Experimental Procedures" in the absence of toxin. The voltage of half-activation (V 0.5 ) and the slope factor of each channel were derived from fitting the corresponding voltage-dependent activation curve with a single Boltzmann equation. All data are presented as mean Ϯ S.E.
Ϫ26.6 Ϯ 0.6 Ϫ6.0 Ϯ 0.4 6 V843A Ϫ31.9 Ϯ 0.8 Ϫ40.8 Ϯ 2.7 Ϫ5.6 Ϯ 0.6 4 The nearby mutation L846A also impaired voltage sensor trapping by CssIV E15A , but its effect was even greater. It caused a negative shift in the voltage dependence of activation under control conditions (Fig. 2C, Table 1). However, voltage sensor trapping by CssIV E15A was completely abolished for L846A (Fig. 2D, Table 3). Thus, mutation L846A had two functional effects. First, the mutation itself enhanced activation of the channel, either by stabilizing the IIS4 voltage sensor in the activated conformation, destabilizing the resting conformation, or both. Second, the L846A mutation completely blocked voltage sensor trapping by the toxin, either by complete loss of binding of CssIV E15A to the mutant channel in resting state or by complete loss of voltage sensor trapping efficacy by CssIV E15A in the activated state, or a combination of these effects.
Increased Voltage Sensor Trapping with Mutant N842R-To examine the functional properties of mutant N842R, we first tested its voltage-dependent activation in the absence of toxin. We observed that the I-V relationship of the N842R channel was similar to WT (Fig. 3A, Table 1). However, unlike the WT channel, treatment with 500 nM CssIV E15A induced a signifi-cant Na ϩ current during test pulses to Ϫ60 mV without a depolarizing prepulse (Fig. 3, B and C, Table 3), and the I-V relationship revealed a significant component of I VST (Fig. 3C, Table 3). Global fits to the I-V curves indicated that ϳ9% of the current activated with the negative voltage dependence without a prepulse. Approximately 55% of the Na ϩ current activated with negative voltage dependence after a prepulse, about 6-fold more than without the prepulse (Fig. 3C, Table 3). This negatively shifted component of the Na ϩ current is 4.45-fold greater than that observed for WT (Figs. 1B and 3C, and Table 3). Therefore, the mutation N842R greatly enhanced voltage sensor trapping by CssIV E15A .
We hypothesized that the I VST that we observed without a prepulse might be caused by voltage sensor trapping at the holding potential of Ϫ100 mV and might be reversed by a stronger hyperpolarization. In fact, the I VST observed with this mutant channel was eliminated by a 10-s hyperpolarization to Ϫ150 mV (Fig. 3, D and E). Evidently, CssIV E15A trapped a significant fraction of the IIS4 voltage sensors of N842R channels in an activated state at the resting membrane potential. The voltage dependence of activation was measured as described under "Experimental Procedures" in the presence of 500 nM CssIV E15A but without the prepulse. The voltage of half-activation (V 0.5 ) and slope factor of each channel in the presence of 500 nM CssIV E15A but without the prepulse are derived from fitting the corresponding voltage-dependent activation curve with a single Boltzmann equation. All data are presented as mean Ϯ S.E.
Ϫ38. 5    Ϫ7.9 mV. Without the prepulse, 7.5% of the current activated with the negative voltage dependence (V a2 ). With the prepulse, 55% of the current activated with the negative voltage dependence, V a2 . D, voltage protocol used to observe the effects of hyperpolarization upon voltage sensor-trapping activity by CssIV E15A on N842R channels at resting membrane potential. E, I VST traces recorded at Ϫ60 mV before or after the hyperpolarization pulse to Ϫ150 mV for 10 s. F, concentration-response curves for I VST on N842R mutant channels by CssIV E15A with (filled squares, ϩPre) or without (open squares, -Pre) the prepulse. I VST was normalized to the peak of the I-V in the absence of toxin. Concentration-response data were fit with first-order Hill equations (n Ͼ ϭ 4).
We measured the voltage sensor trapping activity of CssIV E15A with N842R channels at three different concentrations of CssIV E15A (Fig. 3F). Values of I VST as a function of concentration were fit with first-order Hill equations (Fig. 3F, Table 3). I VST reached a maximum of 53 Ϯ 4.5% of total Na ϩ current with an EC 50 of 170 nM. Without a prepulse, I VST increased over a similar concentration range to reach a maximum of 9.0 Ϯ 1.5% (Fig. 3F).
Our previous work demonstrated that the rate of voltage sensor trapping by CssIV during the strong depolarizing prepulse is concentration-independent, suggesting that this process does not involve binding of toxin (9). To compare the rate and extent of voltage sensor trapping upon depolarization by CssIV E15A in WT and N842R channels, we applied depolarizations of variable duration (0 -5 ms) to ϩ50 mV from a holding potential of Ϫ100 mV. Trapping was assayed 60 ms later by a test pulse to Ϫ60 mV (Fig. 4A, inset). The peak test pulse current at Ϫ60 mV increased as a function of prepulse duration (Fig. 4, A and B). The time courses were fit with single exponential functions to yield the maximum amount of trapping (Fig.  4C, top) and the rate of trapping (Fig. 4C, bottom) as a function of concentration. For WT, higher concentrations of CssIV E15A resulted in greater I VST after a 5-ms depolarization, whereas I VST was maximal for N842R at all three concentrations of CssIV E15A (Fig. 4C, top). These results indicate that voltage sensor trapping by CssIV E15A in N842R channels is saturated at 200 nM. In contrast to our results for the extent of voltage sensor trapping, the time constants for the development of voltage sensor trapping were concentration-independent for both WT and N842R (Fig. 4C, bottom). These results are consistent with the three-step model for voltage sensor trapping: concentration-dependent binding, voltage-dependent channel activation, and finally concentration-independent voltage sensor trapping (8,9).
Repolarization causes reversal of voltage sensor trapping. To measure the rates of reversal, cells were depolarized to ϩ50 mV for 1 ms to activate and trap a population of channels. They were then repolarized to Ϫ100 mV for variable times followed by a test depolarization to Ϫ60 mV to assay the loss of trapping (Fig. 4D, inset). I VST decayed exponentially with recovery time for both WT and N842R channels (Fig. 4D). However, the rate of decay was ϳ6.8-fold slower for N842R compared with WT channels (Fig. 4, D and E), as quantified by the time constants of single exponential fits to these data (Fig. 4F).
Increased Voltage Sensor Trapping with Mutants V843A and E844N-Mutations V843A and E844N also increased voltage sensor trapping following a prepulse (Fig. 5), but the effects were less complex than for mutation N842R. In the absence of CssIV E15A , mutation V843A shifted the voltage dependence of activation ϳ7 mV to more negative potentials (Fig. 5A, Table 1), suggesting that the mutation stabilized the activated state of the domain II voltage sensor. In the presence of 500 nM CssIV E15A but without a prepulse, the voltage-dependent activation of V843A was shifted 11.4 mV in the positive direction (Fig. 5B,  Tables 1 and 2). This is consistent with the idea that toxin binding stabilizes the voltage sensor in its resting conformation. No I VST was detectable at Ϫ60 mV with V843A channels (Fig. 5B,  inset). However, following a prepulse to ϩ50 mV in the pres-ence of 500 nM CssIV E15A , I VST was increased to 56 Ϯ 5% of the maximal peak current for V843A (Fig. 5B), which was 4.7-fold greater than WT. In contrast to V843A, the voltage dependence of activation of E844N was similar to WT (Fig. 5C, Table 1). However, following a prepulse to ϩ50 mV in the presence of 500 nM CssIV E15A , I VST was increased to 50.3 Ϯ 6.5% of the maximal peak current for E844N (Fig. 5D)  response curves for voltage sensor trapping gave EC 50 values of 187 Ϯ 18 nM for V843A and 210 Ϯ 25 nM for E844N (Fig. 5E).
We measured the rates of development of voltage sensor trapping for mutants V843A and E844N (Fig. 6, A and B). The level of I VST increased with the duration of the prepulse and reached maximal effect at ϳ1 ms for E844N and 3 ms for V843A (Fig. 6B). The extent of voltage sensor trapping for both mutants was substantially greater than that of WT (Fig. 6B). The extent of voltage sensor trapping also increased with toxin concentration for both mutants (Fig. 6C, top), whereas the rate constants for the development of voltage sensor trapping were concentration-independent for both mutants (Fig. 6C,  bottom), consistent with our three-step model for voltage sen-sor trapping in which the final trapping step is both concentration-and voltage-independent (8,9).
We also measured the rate of reversal of voltage sensor trapping for mutants V843A and E844N (Fig. 6D). The maximal level of voltage sensor trapping was greater for both V843A and E844N than for WT (Fig. 6, E and F, top). The rate of reversal of voltage sensor trapping at Ϫ100 mV for E844N was comparable with that of WT but was accelerated for V843A (Fig. 6F, bottom). The striking differences in the kinetics of the development and reversal of voltage sensor trapping for mutants N842R, V843A, and E844N are considered under "Discussion."

A Molecular Map of the ␤-Scorpion Toxin Receptor Site-
Positively charged and hydrophobic amino acid residues in CssIV are important for its binding (7). Our previous studies of channel chimeras implicated the extracellular IIS1-S2 and IIS3-S4 loops of Na v channels in formation of the receptor site for ␤-scorpion toxins and provided evidence that the IIS4 segment itself is not directly involved in toxin binding (8 -10). To complete the mapping of the receptor site for ␤-scorpion toxin on Na v channels, we constructed and analyzed 23 mutations in IIS1-S2 and IIS3-S4. A linear map of the functional effects of all of the mutants that have been characterized in this study and previous work is illustrated in Fig. 7 in terms of the K d ratio (mutant/WT) measured in radioligand binding studies or the I VST ratio (WT/mutant) measured in this work. This complete scan of amino acid residues in the IIS1-S2, IIS3-S4, and IIS4 segments reveals two areas of interest (Fig. 7). Eight residues in the IIS3-S4 loop (Glu 837 , Leu 840 , Ala 841 , Asn 842 , Val 843 , Glu 844 , Gly 845 , and Leu 846 ) are important in CssIV binding and action on rNa v 1.2a channels, and seven of them form a hot spot of contiguous amino acid residues for toxin action. Mutations of these amino acid residues can either increase or decrease voltage sensor trapping by CssIV. In addition, two residues in IIS1-S2 (Glu 779 and Pro 782 ) also contribute significantly to CssIV binding and action (Fig. 7). These results indicate that the IIS3-S4 loop plays the primary role in binding of CssIV toxin and controls the functional effects of the toxin, whereas two amino acid residues in the IIS1-S2 loop play a secondary role.
Structural Model of ␤-Scorpion Toxin Binding to the Na v 1.2 Channel-We previously developed a structural model of the CssIV-rNa v 1.2a complex (9) using the Rosetta rigid-body docking method (16) and the structure of the bacterial K v AP voltagesensing domain (20) as a template. Here we present a new structural model of the CssIV-rNa v 1.2a complex that was generated using the Rosetta flexible-backbone docking method (17) and the high-resolution structure of the mammalian K v 1.2-K v 2.1 chimera channel (13) as a template (see "Experimental Procedures"). Our new structural model shows that the CssIV toxin has an extensive surface interacting with the amino acid residues that line the extracellular water-accessible cavity of the rNa v 1.2a voltage sensor in domain II (Fig. 8, A and  B). The overall orientation of the CssIV toxin relative to the voltage sensor in our new model (Fig. 8) is very similar to our original model (9). The majority of residues in the voltage sensor that have significant effects on CssIV toxin binding and voltage sensor trapping are at the interface with the toxin in our model (Fig. 8, A and C). Specifically, Glu 779 in S1 interacts with Phe 44 of CssIV, Glu 837 in S3 interacts with Phe 44 of CssIV, and Ala 841 , Asn 842 , Glu 844 , and Leu 846 in the IIS3-S4 loop are in direct contact with the CssIV toxin (Fig.  8C), consistent with the significant effects of mutations in these residues on toxin binding and action (Fig. 7). Although Leu 840 , Val 843 , and Gly 845 seem not to be in direct contact with the CssIV toxin in our model, mutations L840C and V843A may alter the local conformation of the IIS3-S4 loop  (8,9) or the I VST ratio for mutant channels studied functionally here (filled bars). For mutant channels that enhance voltage sensor trapping, the I VST ratio was defined as Ϫ1[I VST (WT)/I VST (mut)]. For mutant channels that reduce voltage sensor trapping, the I VST ratio was defined as I VST (WT)/I VST (mut). At some loci multiple mutants were studied and the data corresponding to each mutant and its corresponding label are color coded. For instance, E837Q, E837R, and E837C are plotted in black, green, and red, respectively, in the bar graph. Empty bars represent previously studied amino acid residues (8,9), whereas the filled bars represent new residues whose function was analyzed in this paper.
during channel activation, and the mutation G845N would create a structural conflict for toxin binding according to our model. On the CssIV toxin side of the interaction, Glu 15 , Asn 22 , Tyr 24 , Glu 28 , and Trp 58 are either in direct contact or within close proximity to the IIS3-S4 loop of the channels, which is in agreement with the significant effects of mutations at these positions on toxin binding to the channel (Fig.  8, B and D).

DISCUSSION
The Receptor Site for ␤-Scorpion Toxins Includes a Hot Spot in the IIS3-S4 Loop-In the experiments described here, we have substantially extended our previous studies of the molecular determinants of voltage sensor trapping (8 -10) and further established that both IIS1-S2 and IIS3-S4 loops are required for normal binding and action of ␤-scorpion toxins, whereas the IIS3-S4 loop plays a dominant role. These data suggest that ␤-scorpion toxins interact with a short segment of IIS1-S2 and a broader region of IIS3-S4. Evidently, these two distant regions of the primary structure of Na v channels are close to each other in the folded structure of the channels and form a single toxinbinding site. Similarly, the CssIV residues that are crucial for toxin binding to Na v channels come from different regions of the primary structure but are brought together in a wedgeshaped interactive surface in the tertiary structure of the toxin (18).
Seven of the eight influential residues in IIS3-S4 are in consecutive positions in the primary structure of the rNa v 1.2a channels. This cluster of amino acid residues in the IIS3-S4 loop defines a hot spot for toxin action. The amino acid sequence of this segment is well conserved between rNa v 1.2a and hNa v 1.4, consistent with the similar effects of CssIV on these two channels (21). The IIS1-S2 and IIS3-S4 loops may play cooperative roles in toxin binding and voltage sensor trapping. In molecular models of K ϩ channels, the IIS1-S2 loop is relatively static (22,23), whereas the IIS3-S4 loop undergoes outward displacement and rotation relative to IIS1-S2 in response to voltage sensor activation, and the cleft between IIS1-S2 and IIS3-S4 widens (3,24). Molecular models of the voltage sensor of a bacterial Na v channel suggest similar movements during activation (25,26). As the cleft formed by the IIS1-S2 and IIS3-S4 linkers widens during the transition from the resting state to the activated state, newly accessible residues in both the IIS3-S4 loop and the N-terminal end of the IIS4 segment may interact with the toxin. CssIV may bind in a state-independent manner to the IIS1-S2 loop and then cause voltage sensor trapping by preferential state-dependent interactions with the IIS3-S4 loop in its activated conformation.
Mutations in the IIS3-S4 Loop Can Either Strengthen or Weaken Voltage Sensor Trapping-In the IIS1-S2 and IIS3-S4 extracellular loops, mutations at seven positions greatly weakened or even abolished binding or voltage sensor trapping by CssIV (Fig. 7). This is in agreement with previous findings that mutations in these extracellular loops generally weaken the binding interaction of toxins to a variety of voltage-gated ion channels (8,9,27). These results point to the S3-S4 loops of voltage-gated ion channels as the primary molecular target for gating modifier toxins. Surprisingly, and in contrast to these previous studies, we found three mutations that greatly enhanced voltage sensor-trapping activity by a ␤-scorpion toxin on rNa v 1.2a channels. The bidirectional effects of mutations in the IIS3-S4 loop on toxin efficacy in voltage sensor trapping suggest that the native residues in these positions make interactions that contribute both positively and negatively to the binding energy of the toxin and that toxin interactions with this extracellular loop determine the efficacy of the voltage sensor-trapping process.
Functional analyses of N842R, V843A, and E844N mutant channels showed that the voltage sensor-trapping action by CssIV was strongly enhanced. Similar enhancement of voltage sensor trapping was observed for mutations that neutralize the two outermost arginine gating charges in the IIS4 segment (10) or the negatively charged residues with which they interact (28). It was proposed that the loss of electrostatic interactions between the negatively charged residues in IIS2 and IIS3 and the arginine gating charges in IIS4 increased the mobility of IIS4 voltage sensor, thereby allowing more rapid and complete volt- Side chains of key residues for the CssIV-rNa v 1.2a interaction are shown in space filling representation and all other side chains shown in stick representation. A probe radius of 1.4 Å was used to scan the molecular surface of each structural model. This figure was generated using Chimera (43).
age sensor trapping (10). This increased mobility might reduce the energy barrier for the voltage sensor to be trapped by ␤-scorpion toxin in the activated state. N842R, V843A, and E844N differ from those mutants studied previously. First, all of the residues studied previously are localized within the cell membrane electrical field and are thus directly involved in voltage sensing (10,28). In contrast, Asn 842 , Val 843 , and Glu 844 are all localized in the IIS3-S4 extracellular loop outside the membrane electrical field and thus cannot contribute directly to voltage sensing. Second, the amino acid residues within the IIS2, IIS3, and IIS4 transmembrane segments do not contribute to the receptor site for ␤-toxins (9), whereas Asn 842 , Val 843 , and Glu 844 evidently do play an important role in toxin binding. The gain-of-function effects of the N842R, V843A, and E844N mutations may include both enhanced binding affinity and enhanced efficacy for voltage sensor trapping (see kinetic model below), whereas the gain-of-function effects of mutations in the IIS4 segment must result from specific effects on voltage sensor trapping.
N842R is a single residue chimera between rNa v 1.2a, on which ␤-toxins have strong voltage sensor-trapping action, and hNa v 1.5, on which ␤-toxins have very weak voltage sensortrapping action (8). Therefore, it is surprising that N842R strongly enhances the voltage sensor-trapping action of ␤-toxins and that CssIV can trap the voltage sensor of N842R at the resting membrane potential without a depolarizing prepulse. The adjacent mutations V843A and E844N also increased voltage sensor trapping, but only following a depolarizing prepulse. The striking differences in CssIV action caused by these three adjacent mutations highlight the extreme sensitivity of this hot spot for toxin action in IIS3-S4 to changes in single amino acid residues.
Kinetics of ␤-Scorpion Toxin Action and Recovery Fit the Voltage Sensor-trapping Model-Previous studies showed that CssIV purified from scorpion venom negatively shifts the voltage dependence of activation following a strong depolarizing prepulse (8,9). A voltage sensor-trapping mechanism was proposed to explain the prepulse-dependent enhancement of activation by ␤-scorpion toxins (8,9). According to the model, before the prepulse, the toxin binds to its receptor site in the resting state of the channels in a bimolecular chemical reaction to form a toxin-channel complex. Upon strong depolarization, the IIS4 segment in the voltage sensor in domain II moves outward, and the toxin binds to newly accessible amino acid residues in the IIS3-S4 loop and the extracellular end of the IIS4 segment. In this activated position, the IIS4 segment is tightly bound to the toxin and trapped in its activated, outward position. Upon repolarization of the cell membrane, the trapped IIS4 voltage sensor remains activated, which reduces the energy required to re-activate the channels because one of the four voltage sensors is already activated. The reduced electrical energy required to activate one fewer voltage sensor causes the negative shift of the voltage dependence of activation. Kinetic analysis showed that voltage sensor trapping by CssIV is well fit by a model that incorporates these three steps in toxin action (9), and kinetic analysis of the effects of a toxin partial agonist (CssIV E15R ) further supported a three-step voltage sensor-trapping process. The voltage sensor-trapping mechanism also explains the actions of tarantula huwena toxin-IV (29), ␤-scorpion toxin Ts1 (30), tarantula protoxin-II (31,32), ␣-scorpion toxins (33,34), and ␦-conotoxins (35).
Our kinetic analysis of the effects of gain-of-function mutations in rNa v 1.2a channels also supports the voltage sensortrapping mechanism. Below saturating concentrations of toxin, the extent of voltage sensor trapping is increased with increasing toxin concentration, whereas the rates of voltage sensor trapping and the reversal of voltage sensor trapping are both independent of toxin concentration. Our evidence that the effects of both a loss-of-function toxin partial agonist, CssIV E15R (19), and the three gain-of-function rNa v 1.2a mutants studied here are all well fit by a kinetically similar voltage sensor-trapping mechanism provides additional strong support for this model of toxin action.
In general, our results do not provide sufficient data to separate the effects of the mutations on toxin binding affinity versus efficacy in voltage sensor trapping because we cannot reach saturating effects for mutants with increased EC 50 values. However, we were able to use this kinetic model to fit the results and derive K d values for the mutants with enhanced voltage sensortrapping activity. The mutations V843A and E844N in the hot spot for CssIV action both reduced the K d value for binding to the resting state by 2.7-4.3-fold (Table 3). Thus, increased binding affinity makes a substantial contribution to increased voltage sensor trapping by these mutants.
Although voltage sensor trapping by all three gain-of-function mutants in IIS3-S4 of rNa v 1.2a has generally similar dependence on toxin concentration and voltage, the effects of these mutations on the rate of onset and reversal of voltage sensor trapping are quite different and these factors affect toxin efficacy. N842R greatly slows the reversal of voltage sensor trapping upon repolarization, whereas neither V843A nor E844N have this effect (Figs. 4 and 6). Corrected time courses of the onset of voltage-sensor trapping were generated (supplemental Fig. S2) by taking into account the time course of trapping (Figs. 4B and 6, A and B) and the loss of trapping during the 60-ms repolarization intervals (Figs. 4D and 6D). These corrected time courses show that the amount of trapping for V843A is much greater than for N842R, E844N, or WT, even though the toxin receptor was 100% occupied at the 500 nM test concentrations for each of these mutant channels (Figs. 3F and 5E). These differences likely result from more efficient trapping of the depolarized voltage sensor by the toxin bound to the V843A mutant channel and greater stability of the trapped, activated voltage sensor. Despite these differences in detail, the corrected time courses were well fit with a voltage sensor-trapping model assuming 100% occupancy of the toxin receptor (supplemental Fig. S2D) with the parameters given in supplemental Fig. S2E.
The corrected time course of voltage sensor trapping (supplemental Fig. S2C) for V843A reaches a final value of 3.2 or 320% of the original peak conductance in the absence of toxin. This value indicates that the trapped voltage sensor for this mutant channel causes the peak open probability (P O ) for this channel to increase above the maximum level reached in the absence of toxin by at least 3-fold. Based on previous single channel recording studies of Na V channels, this would bring the maximum single channel open probability in the presence of CssIV toxin to nearly the theoretical upper limit of 1.0 for this mutant.
Structural Model for the Toxin-Receptor Complex-The functional surface of ␤-scorpion toxins is composed of two clusters of discontinuous residues on the two sides of the wedge-shaped toxin molecule (18). One is associated with the ␣ helix of CssIV, which includes a hot spot that is conserved among ␤-scorpion toxins and contains their primary pharmacophore. The other cluster of important amino acid residues is hydrophobic and is associated mainly with the ␤2 and ␤3 strands and the C-terminal tail. This cluster of amino acid residues determines the species selectivity of the ␤-scorpion toxins (18). Our structural model indicates that residues in IIS1-S2 interact with the species selectivity cluster on the toxin, but not with the pharmacophore cluster, whereas IIS3-S4 residues interact with both clusters on the toxin surface (Fig. 8). Both sets of interactions contribute to toxin binding affinity, but binding to the IIS3-S4 loop appears to be the primary determinant of toxin efficacy in voltage sensor trapping, as illustrated by the large changes in efficacy of voltage-sensor trapping by mutations of the adjacent residues at positions 842, 843, and 844. Thus, the S3-S4 linker emerges not only as the center for high affinity toxin binding but also as the site of interaction that determines whether ␤-scorpion toxins act as agonists or antagonists of voltage-dependent channel activation.
Voltage Sensor Trapping Supports a Sliding Helix Model of Voltage Sensor Function-␤-Scorpion toxins bind to their receptor site in the resting state of Na v channels and then trap the voltage sensor in its activated state (8,9). Therefore, the receptor site composed of the IIS1-S2 and IIS3-S4 loops must be available for toxin binding at the extracellular surface of the membrane in the resting state of the channels. This characteristic distinguishes the mechanism of ␤-scorpion toxins from hanatoxin and related cysteine-knot toxins that inhibit potassium channels (36) and channel chimeras (27). The receptor site for these toxins can be transferred from one ion channel to another with significant recovery of toxin action by forming chimeras in which the extracellular half of the S3 and IIS4 segments and the connecting IIS3-S4 loop are inserted (27). These studies support the concept that the IIS3-S4 linker is the primary component of the receptor sites for these gating modifier toxins, as demonstrated in earlier work on ␣and ␤-scorpion toxins (8,9,34). However, hanatoxin partitions into the cell membrane and interacts with amino acid residues in the transmembrane part of the S3b-S4 helical hairpin (37,38), but CssIV does not (39). Our data indicate that ␤-scorpion toxins can bind to the resting state of Na v channels via the extracellular surface only. Our finding that CssIV bound to the N842R mutant channel can trap its voltage sensor without a depolarizing prepulse further confirms that the IIS3-S4 loop is at the cell surface in a resting state of the channel. These results agree closely with the predictions of the sliding helix model of voltage-dependent gating (40 -42), which posits that the IIS4 segments are in a transmembrane position in the resting state of the channel and that, in response to depolarization, the IIS4 segments move outward and rotate. As expected from this gating model, our results confirm that the IIS3-S4 loop is on the extracellular surface of the membrane in the resting state of the channels.
The sliding helix mechanism also predicts that the IIS3-S4 linker undergoes substantial conformational changes during activation. Our results show that the majority of the residues forming the receptor site for ␤-toxins in sodium channels are in a cluster of consecutive residues in the IIS3-S4 extracellular loop and that the amino acid residues in this loop control the efficacy of voltage sensor trapping by the toxin. The sensitivity of voltage sensor trapping to mutation of these individual residues is also consistent with the movement and conformational changes of the IIS3-S4 loop that are predicted by the sliding helix model of voltage sensor function (3,42). A major step toward further definition of this mechanism of voltage sensor function, as well as toward understanding the structural basis for ␤-scorpion toxin action, will eventually come from structural analysis of voltage sensors in defined functional states stabilized by binding of specific toxins and other effectors.