Synthesis, Solution Structure, and Phylum Selectivity of a Spider δ-Toxin That Slows Inactivation of Specific Voltage-gated Sodium Channel Subtypes*

Magi 4, now renamed δ-hexatoxin-Mg1a, is a 43-residue neurotoxic peptide from the venom of the hexathelid Japanese funnel-web spider (Macrothele gigas) with homology to δ-hexatoxins from Australian funnel-web spiders. It binds with high affinity to receptor site 3 on insect voltage-gated sodium (NaV) channels but, unlike δ-hexatoxins, does not compete for the related site 3 in rat brain despite being previously shown to be lethal by intracranial injection. To elucidate differences in NaV channel selectivity, we have undertaken the first characterization of a peptide toxin on a broad range of mammalian and insect NaV channel subtypes showing that δ-hexatoxin-Mg1a selectively slows channel inactivation of mammalian NaV1.1, NaV1.3, and NaV1.6 but more importantly shows higher affinity for insect NaV1 (para) channels. Consequently, δ-hexatoxin-Mg1a induces tonic repetitive firing of nerve impulses in insect neurons accompanied by plateau potentials. In addition, we have chemically synthesized and folded δ-hexatoxin-Mg1a, ascertained the bonding pattern of the four disulfides, and determined its three-dimensional solution structure using NMR spectroscopy. Despite modest sequence homology, we show that key residues important for the activity of scorpion α-toxins and δ-hexatoxins are distributed in a topologically similar manner in δ-hexatoxin-Mg1a. However, subtle differences in the toxin surfaces are important for the novel selectivity of δ-hexatoxin-Mg1a for certain mammalian and insect NaV channel subtypes. As such, δ-hexatoxin-Mg1a provides us with a specific tool with which to study channel structure and function and determinants for phylum- and tissue-specific activity.

the generation and propagation of electrical signals in excitable cells. At least nine different genes encoding distinct Na V channels isoforms have been identified, and functionally expressed, in mammals (1). They are characterized by their sensitivity to TTX, with Na V 1.5, Na V 1.8, and Na V 1.9 being TTX-insensitive or TTX-resistant, and the remaining subtypes being sensitive to nanomolar concentrations of TTX. In addition, localization of the subtypes also varies, with Na V 1.1-1.3 mostly distributed in the central nervous system, Na V 1.6 -1.9 principally located in the peripheral nervous system, and Na V 1.4 and Na V 1.5 found in skeletal and cardiac muscle, respectively. The structural diversity of Na V channels also coincides with variations in physiological and pharmacological properties (2). In contrast, insects express only one gene (para) that undergoes extensive alternative splicing and RNA editing (3). The para-encoded Na V channel is exceptionally well conserved across diverse orders of insects, with the level of identity ranging from 87 to 98% (3). This is one reason why insecticides that target insect Na V channels have broad activity across many insect orders. In contrast, para-type Na V channels have significantly lower levels of identity with the various types of mammalian Na V channels with the level of identity typically around 50 -60% (3). This explains why a high degree of phylogenetic specificity can be achieved with both Na V channel toxins and insecticides that target the Na V channel.
At least seven distinct toxin-binding sites have been identified by radioligand binding and electrophysiological studies on vertebrate and insect Na V channels (4,5). Toxins interacting with these neurotoxin receptor sites have been instrumental in the study of Na V channel topology, function, and pharmacology (6). In particular, a wide range of scorpion ␣-toxins, sea anemone toxins, and spider ␦-hexatoxins (formerly ␦-atracotoxins (7)) compete for binding to receptor site-3 on the extracellular surface of Na V channels. These polypeptide toxins all inhibit the fast inactivation of Na V channels to prolong Na ϩ currents (I Na ), despite huge diversity in primary and tertiary structures (8,9). Nevertheless, receptor site- 3 has not yet been fully characterized but is believed to involve domains DI/S5-S6, DIV/S5-S6, as well as DIV/S3-S4 (9). Most importantly, however, toxin characterization is often limited to studies using whole-cell I Na or binding studies on neuronal membranes where there are mixed populations of Na V channel subtypes. For all of these toxins, the precise pattern of Na V channel subtype selectivity is either unknown or at best is incomplete.
The aim of this study was to first determine the solution structure of Magi 4 and second to investigate the ability of Magi 4 to discriminate between different Na V channels subtypes. Here we report the tertiary structure of Magi 4 by 1 H NMR and show its disulfide bonding pattern and three-dimensional structure are homologous to ␦-HXTX-1 toxins. We highlight the key residues in Magi 4 that appear to be topologically similar to those residues known to be part of the pharmacophore for site-3 scorpion ␣-toxins, despite Magi 4 having a different overall structure to scorpion ␣-toxins (11). In addition, we provide a detailed characterization of the selectivity and mode of action of Magi 4 on nine cloned mammalian and insect Na V channel subtypes, including a detailed characterization on insect neurotransmission. Given that the toxin potently slows the inactivation of Na V channels, it should be renamed ␦-hexatoxin-Mg1a (␦-HXTX-Mg1a) in accordance with the rational nomenclature recently proposed for naming spider peptide toxins (7) (see ArachnoServer spider toxin data base).
The N-terminal Cys(17)-(1-19)-thioester and C-terminal Cys-(20 -43) peptides were then allowed to ligate by solubilizing them in freshly degassed 0.1 M sodium phosphate buffer (10 ml), pH 8.4, containing 6 M guanidine hydrochloride (GdnHCl). Thiophenol (0.40 ml) was then added, and the whole mixture was stirred for 24 h at room temperature and then treated with dithiothreitol (1.6 mmol). After 20 min, the pH was acidified to less than 2 using 1 M HCl. The mixture was washed with ether, and Cys(17)1-43 peptide was purified from the aqueous layers by preparative RP-HPLC using a linear gradient (20 -40% CH 3 CN in 0.1% trifluoroacetic acid for 80 min) to obtain 51 mg (41%) of purified peptide.
To remove the Acm group, a 9.6 mM solution of Cys-(17)-(1-43) (50 mg) was made up in 5 ml of 95% trifluoroacetic acid containing anisole (50 l), and 0.77 mmol of silver trifluoroacetate was added and stirred for 2 h at room temperature. The product was precipitated as the silver salt, by adding ether to the reaction mixture, and was dissolved in 0.1 M phosphate buffer, pH 8.4, containing 6 M GdnHCl. Dithiothreitol (3.8 mmol) was added to the solution and stirred for 0.5 h at room temperature. After the addition of 1 M HCl, the resulting AgCl was removed by filtration, and the filtrate was applied to preparative RP-HPLC using a linear gradient (20 -40% CH 3 CN in 0.1% trifluoroacetic acid for 80 min) to obtain 23 mg (46% yield) of the reduced peptide (8 SH-peptide-(1-43)).
The eight free cysteines were allowed to oxidize in air for 4 days at 4°C in 0.1 M aqueous ammonium acetate, pH 7.6, containing 2 M GdnHCl, 0.1 mM GSSG, and 1 mM GSH. The peptide/GSSG/GSH ratio was 1:10:100. The solution was then diluted 2-fold with cold water (4°C) to a final peptide concentration of 5 M. After 3 days at 4°C, the mixture was acidified to pH 2 by adding trifluoroacetic acid, and the folded peptide was desalted and purified by preparative RP-HPLC using a linear gradient (17-37% CH 3 CN in 0.1% trifluoroacetic acid for 80 min) to obtain 5.3 mg from the previous 23 mg (23% yield).
Capillary Electrophoresis-Capillary zone electrophoresis analyses were performed on a Jasco CE800 system (Jasco, Japan) equipped with a UV detector connected to a Shimadzu CR-4A recorder and a capillary (0.1 m inner diameter, 70 cm length, 50 cm to detector). 20 mM sodium citrate buffer, pH 2.5, was used for the analyses. Samples dissolved in migration buffer were applied hydrodynamically to the capillary (height 20 cm, 15 s), and analyses were performed with a 20-kV constant voltage drop and monitored at 210 nm.
CD Measurements-CD spectra were obtained on a Jasco J-725 spectropolarimeter (Jasco, Japan). The spectra were measured from 260 to 180 nm in 60% trifluoroethanol, pH 7.1, at room temperature, with a 1-mm path length cell. Data were collected at 0.1 nm with a scan rate of 100 nm/min and a time constant of 1 s. The concentration of the toxins was 30 M, as determined by amino acid analysis. Data from 10 separate recordings were averaged and analyzed by the method of Böhm et al. (18).
NMR Experiments-Synthetic ␦-HXTX-Mg1a (4.7 mg) was dissolved to a final concentration of 3.7 mM in 250 l of H 2 O/ D 2 O (90/10; v/v) containing 50 mM NaN 3 in a susceptibility matched microcell (Shigemi, Japan). The pH was adjusted to 2.4. NMR spectra were recorded on a Bruker DMX-750 spectrometer. The temperature was set to 298 K. Chemical shifts were referenced to internal 3-(trimethylsilyl)[2,2,3,3-2 H 4 ]propionate. Two-dimensional DQF-COSY, NOESY, and TOCSY experiments were performed using standard pulse sequences and phase cycling. The NOESY spectra were acquired with mixing times of 50 and 200 ms. The TOCSY spectrum was recorded with a spin lock time of 71 ms. The two-dimensional spectra were recorded using time-proportional phase incrementation for quadrature detection in the F1 dimension. For water suppression, the NOESY experiments include the WATERGATE sequence, and the DQF-COSY and TOCSY experiments included selective low power irradiation during the relaxation delay. The TOCSY and NOESY spectra were recorded with 512 (t 1 ) ϫ 2000 (t 2 ). The DQF-COSY spectrum was recorded with 512 (t 1 ) ϫ 8000 (t 2 ) for an estimation of the coupling constants. The spectra were processed using the XWIN-NMR 2.5 program (Bruker Biospin) running on an O 2 work station (Silicon Graphics). Chemical shift assignments have been deposited in the BioMagResBank (BMRB), accession code 11044.
Structural Calculations-Proton signal assignments were achieved using the standard strategy described by Wüthrich (27) with the graphic software ANSIG.3.3. The DQF-COSY and TOCSY spectra gave the spin system fingerprint of the peptide. The spin systems were then sequentially connected using the NOESY spectra. Interproton distance restraints were obtained from the NOESY spectra acquired with a mixing time of 200 ms. The NOE volumes were converted into four ranges of distance restraints classified as strong (1.8 -2.7 Å), medium (1.8 -3.3 Å), weak (1.8 -5.0 Å), and very weak (1.8 -6.0 Å). The backbone 3 J NH-H␣ coupling constants were estimated or directly measured from a one-dimensional spectrum or the high digital resolution DQF-COSY spectrum using the DECO program (Bruker Biospin). Hydrogen bonds were identified using amideproton temperature coefficients from NOESY spectra obtained at 288, 298, 308, 318, and 328 K.
The structural calculations were performed using the X-PLOR-NIH 2.9.1 program with 576 NOE-based distance restraints, which contain 207 intraresidue, 181 sequential, 60 medium range, and 128 long range restraints. In the first stage, the starting extended strand structure was subjected to 10 ps and 1,000 steps of torsion angle molecular dynamics at 50,000 K. The structures were then subjected to 15 ps and 1,500 steps of a slow cooling torsion angle molecular dynamics stage in which the temperature was reduced from 50,000 to 298 K over 250 steps. Finally, the structures were subjected to 200 steps of conjugated gradient minimization. The initial runs for structure calculations were performed without hydrogen bond and disulfide bond restraints, and the obtained structure was examined.
The structures were checked for violations of geometric and experimental restraints and atom overlapping, using the AQUA3.2 and PROCHECK-NMR3.4 programs. Finally, a set of 20 conformers was selected based on the lowest X-PLOR energy. Structures were analyzed and visualized with the MOL-MOL 2k.1 program.
Neuronal Isolation Procedures and Na V Channel Expression-Characterization of the actions of ␦-HXTX-Mg1a on native I Na were performed using acutely dissociated newborn rat DRG neurons and cockroach DUM neurons from the terminal abdominal ganglia of adult male American cockroaches (Periplaneta americana) as described previously (20,21). Following isolation, recordings were made within 24 h.
Stage V-VI oocytes were harvested from the ovarian lobes of anesthetized female X. laevis frogs as described previously (22). The oocytes were injected with up to 50 nl of cRNA at a concentration of 1 ng nl Ϫ1 using a Drummond microinjector (Ambion). The ND96 solution used for incubating the oocytes containing the following (in mM): NaCl 96, KCl 2, CaCl 2 1.8, MgCl 2 2, HEPES-acid 5, pH 7.4, supplemented with 50 mg liter Ϫ1 gentamycin sulfate. Whole-cell currents from oocytes were recorded 2-5 days after injection.
Electrophysiological Studies-Voltage or current clamp recordings from single DRG and DUM neurons were made using the whole-cell patch clamp technique of Hamill et al. (23). Recordings from X. laevis oocytes were performed using the two-electrode voltage clamp method as described by Liman et al. (22).
Experiments were performed at constant temperature 18 -24°C using either an Axopatch 200A patch clamp amplifier or GeneClamp 500 amplifier. Current and voltage pulse protocols were generated using the pClamp software system (Molecular Devices, Sunnyvale, CA). Data were digitized at 10 -25 kHz, and low pass filtered at either 1 kHz (oocytes) or 5 kHz (DRG and DUM neurons) using a 4-or 5-pole Bessel filter (Ϫ3 dB). Leakage and capacitative currents were digitally subtracted with P-P/4 procedures and series resistance compensation set at Ͼ80% for all patch-clamped cells to minimize voltage errors. The extracellular Na ϩ concentration was reduced (see below) to minimize series resistance errors.
The voltage clamp data recorded in this study were rejected if there were large leak currents upon seal formation or currents showed signs of inadequate space clamping. At the commencement of each experiment, oocytes or DUM neurons exhibiting leakage currents at a holding potential of Ϫ90 mV of more than Ϫ200 and/or Ϫ600 pA, respectively, were discarded. Only cells exhibiting stable leakage currents throughout the whole experiment (with a maximal deviation of Ϯ10% of initial value) were considered in the data analysis. To avoid overestimation of a potential toxin-induced shift in the current-voltage relationship as a result of inadequate voltage control when measuring large sodium currents in oocytes, only results from cells with currents lower than 1.5 A were considered. Current clamp data were rejected if the initial resting membrane potential was more depolarized than Ϫ45 mV.
Recording Solutions-DRG neurons were perfused with an external solution containing the following (in mM): sodium acetate 30, MgCl 2 1, CaCl 2 1.8, cesium acetate 5, KCl 5, D-glucose 25, HEPES-acid 5, TEA-Br 100, CdCl 2 0.5, 4-AP 1 with the pH adjusted to 7.4 with 1 M TEA-OH. Recordings were performed using micropipettes filled with a solution containing the following (in mM): CsF 135, sodium acetate 8, HEPES-acid 5, with the pH adjusted to 7.0 with CsOH. The osmolarity of internal and external solutions was adjusted to 300 -305 mosM liter Ϫ1 with sucrose, prior to use, to reduce osmotic stress. Pipettes with resistances of 0.8 -2 megohms were used for recording I Na from DRG neurons.
Oocytes were perfused with ND96 solution. Voltage and current electrodes were filled with 3 M KCl and resistances were Ͻ1 megohms.
Voltage and Current Clamp Recordings-In DRG neurons, the predominant TTX sensitivity of the Na v channels present in each cell was determined using a modified steady-state Na v channel inactivation (h ∞ ) voltage clamp protocol (24,25). This takes advantage of the separation of h ∞ curves for TTX-sensitive and TTX-resistant Na v channels (26). Only those found to have less than 10% TTX-resistant I Na were used for TTX-sensitive experiments. Of the remaining cells, those expressing sufficiently large TTX-resistant I Na were used for experiments by perfusing with external solution containing 200 nM TTX.
To determine the effect of the toxin on fast inactivation, currents were recorded in the absence and presence of a range of toxin concentrations. The action of the toxin on fast inactivation was assayed by measuring the late current remaining at 50 ms (I 50ms ) for DRG and DUM neurons or 30 ms for oocytes, as a fraction of peak current (I pk ). Currents were elicited by test potentials to Ϫ10 mV in DUM and DRG neurons or at depolarizing test potentials corresponding to maximal activation in oocytes (ranging from Ϫ10 to ϩ10 mV). Test potentials were applied at 10-s (DRG and DUM) or 5-s (oocytes) intervals. The normalized late current ratio (e.g. I 50ms /I pk ) gives the fractional probability for Na V channels not to be inactivated at the end of the test pulse. The concentration dependence for removal of inactivation was measured by plotting the normalized late current as a function of toxin concentration according to the following Hill Equation 1, where EC 50 is the concentration at half-maximal inhibition of fast inactivation, [toxin] is the toxin concentration, and n H is the Hill coefficient. The effect of ␦-HXTX-Mg1a on the voltage dependence of Na v channel activation was determined using depolarizing test pulses from Ϫ90 to ϩ70 mV for 50 ms, in 5-mV (oocytes) or 10-mV steps (DRG and DUM). The values for sodium conductance (g Na ) were calculated according to Equation 2, SEPTEMBER 4, 2009 • VOLUME 284 • NUMBER 36

Spider Toxin Structure That Targets Sodium Channel Subtypes
where I Na is the absolute value of the sodium current at a given test potential (V), and V rev is the reversal potential. The values of g Na and V Ϫ V rev were then fitted to Boltzmann Equation 3, where g max is maximal g Na ; V1 ⁄ 2 is the half-maximal conductance, and k m is the slope factor.
To determine the effect of ␦-HXTX-Mg1a on the voltage dependence of steady-state Na V channel fast inactivation (h ∞ ), a two-pulse protocol with a 0.5-ms interpulse interval was applied. This consisted of a 500-ms or 1-s conditioning prepulse (V cond ), in which the holding potential of Ϫ90 mV was stepped to potentials ranging from Ϫ130 up to ϩ20 mV in 5-mV (oocytes) or 10-mV increments (DRG and DUM), followed by a 50-ms test pulse (V test ) to Ϫ10 mV in DRG or DUM neurons, or potentials corresponding to maximal activation in oocytes. Pulses were applied every 10 s. Data were normalized to the maximum peak control I Na and fitted using Boltzmann Equation 4, where V1 ⁄ 2 is the voltage at half-maximal inactivation; k h is the slope factor; V is the test voltage, and C is a constant or noninactivating fraction (usually zero in controls).
The effect of the toxin on the rate of recovery from Na V channel inactivation was examined by applying a standard twopulse protocol with a variable interpulse interval (⌬T). A 50-ms conditioning prepulse (V cond ) was applied from a holding potential of Ϫ90 to Ϫ10 mV, followed by a 50-ms test pulse (V test ), with an interpulse interval ranging between 0.5 ms and 4 s.
Mathematical curve fitting was accomplished using Graph-Pad Prism version 5.00 for Macintosh (GraphPad Software, San Diego). All curve-fitting routines were performed using nonlinear regression analysis employing a least squares method. Comparisons of two sample means were made using a paired Student's t test. Multiple comparisons were assessed by repeated measures analysis of variance with a Bonferroni's multiple comparison post hoc test. A test was considered to be significant with p Ͻ 0.05. All data are presented as means Ϯ S.E. of n independent experiments.

Peptide Synthesis and Characterization-␦-HXTX-Mg1a
was synthesized with its C terminus as the free carboxyl, as observed for the native toxin by mass spectrometry and from the cDNA-encoding gene (10). The structural identity between the synthetic and native toxins was verified by ESI/Q-TOF MS and capillary zone electrophoresis (supplemental Fig. S1 and supplemental Table S1). In a co-injection experiment, native ␦-HXTX-Mg1a and synthetic ␦-HXTX-Mg1a co-eluted in a single sharp peak (supplemental Fig. S1A). Moreover, the CD spectra of synthetic and native ␦-HXTX-Mg1a superimposed indicating that their secondary structures were similar. Both native and synthetic ␦-HXTX-Mg1a were able to displace the binding of 125 I-Lqh␣IT from insect Na V channels in cockroach synaptosomes with similar IC 50 values (supplemental Fig. S1, B and C).
NMR Spectroscopy-Sequence-specific 1 H resonance assignments for all residues observed were established using standard methods (27). NOE cross-peaks were converted to distance restraints for structural calculation. From NMR experiments, 582 distance constraints and 22 dihedral angle constraints were used for the structural calculations. The total of 582 distance restraints included 576 NOE constraints and 6 hydrogen bonds (Fig. 1B). In addition, disulfide bond constraints were also used (Cys 1 -Cys 15 , Cys 8 -Cys 20 , Cys 14 -Cys 31 , and Cys 16 -Cys 43 ). Simulated annealing calculations were started from an extended structure, and 20 structures were selected with the lowest XPLOR energy, which had no Ն0.4 Å NOE distance violations and no Ն5.0°angle violations from 100 structures obtained at the final calculation (28). Structural statistics for these 20 converged structures are summarized in Table 1. Fig.  2A shows a stereoview of the best fit superposition of the backbone atoms (N, C, C ␣ , and O) for 20 converged structures. The root-mean-square deviations with respect to the mean coordinate positions were 0.483 Å for backbone atoms and 1.178 Å for all heavy atoms of the disulfide-rich structured region (Gly 2 -Ala 23 and Gln 28 -Glu 32 ), excluding the nonstructured loop (Trp 24 -Gln 27 ) and C-terminal regions (Arg 33 -Cys 43 ). Analysis of the 20 structures using PROCHECK NMR and AQUA reveals that 95.1% of the backbone dihedral angles for the structured region lie in the most favored and additional allowed regions of the Ramachandran plot (29).
Six protons from amide groups were identified to form hydrogen bonds according to the amide proton temperature coefficients (supplemental Fig. S2). If the hydrogen bond was in agreement with amide temperature coefficients (more thanϪ4.6 ppb/K), and a hydrogen bond acceptor and an oxygen atom of the backbone was within 2.6 Å of the amide-proton, the amide-proton was identified as a donor of the hydrogen bond. Hydrogen bond acceptors for all the protons were unambiguously determined from preliminary structure calculations. These restraints were then used in the following stage of structure calculations (1. Twenty two dihedral angles estimated from the 3 J NH-Ha values were used as angle constraints within the range of Ϫ90°a nd Ϫ40°for 3 J NH-H␣ Ͻ5.5 Hz (Ala 6 , Trp 7 , Cys 15 , Trp 24 , Cys 31 , Arg 33 , Lys 36 , and Phe 39 ) and between Ϫ160°and Ϫ80°for 3 J NH-H␣ Ͼ8 Hz (Ser 3 , Cys 8 , Lys 9 , Cys 14 , Tyr 18 , Asn 19 , Cys 20 , Ala 23 , Asn 26 , Gln 28 , Ser 29 , Glu 32 , Trp 35 , and Glu 42 ) following standard parameterization (31) (Fig. 1B).
From the results of initial runs without using the disulfide bridge restraints, the disulfide patterns of Cys 1 -Cys 15 and Cys 16 -Cys 43 were unambiguously determined. Analyzing the distances between S␥ i and S␥ j of the remaining four cysteines, three types of disulfide patterns were possible (pattern I, Cys 8 -Cys 14 and Cys 20 -Cys 31 ; pattern II, Cys 8 -Cys 31 and Cys 14 -Cys 20 ; and pattern III, Cys 8 -Cys 20 and Cys 14 -Cys 31 ). Three sets of structure calculations were performed with these possible restraints. The energy minimization was done in vacuo with the GROMOS96 43B1 parameters set, without reaction field (32), to determine the lowest energy state and thus the most stable disulfide pattern of ␦-HXTX-Mg1a. Energy computations were done with the GROMOS96 (32) implementation of Swiss-Pdb-Viewer (33). Total energy calculations for patterns I, II, and III were Ϫ693.5, Ϫ892.7, and Ϫ1523.4 kJ/mol, respectively, therefore unambiguously determining the remaining disulfide bridge patterns as Cys 8 -Cys 20 and Cys 14 -Cys 31 . These disulfide bridges, along with Cys 1 -Cys 15 , form an ICK motif (Fig.  2D) and a pattern (I-IV, II-VI, III-VII, and V-VIII) identical to other ␦-HXTX-1 family members (14, 15) (Fig. 1A). Fig. 2A shows the best fit superposition of the backbone atoms (N, C ␣ , and C) for the 20 converged structures of ␦-HXTX-Mg1a. The three-dimensional structure of ␦-HXTX-Mg1a includes several well defined regions, consisting of an antiparallel ␤-sheet (strand 1, Asn 19 -Ala 23; , strand 2, Gln 28 -Glu 32 ) with some secondary structural elements ( Fig. 1B and Fig. 2B). We identified turns by the standard definition that the distance between C␣ i and C␣ iϩ3 is less than 7 Å (34, 35). The two turns involve Ser 3 to Ala 6 (turn 1) and Cys 15 to Tyr 18 (turn 2). The average dihedral angles for residues at positions i ϩ 1 and i ϩ 2 are follows: 2 ϭ Ϫ40°, 2 ϭ 122°for Lys 4 ; 3 ϭ 57°, 3 ϭ 84°for Arg 5 ; 2 ϭ Ϫ45°, 2 ϭ 128°for Cys 16 ; and 3 ϭ 98°, 3 ϭ Ϫ2°for Gly 17 . Turn 1 is classified as a miscellaneous type IV ␤-turn. There is no hydrogen bonding in this turn conformation. Arg 5 in turn 1 shows a positive 3 value, because nonglycine residues at position i ϩ 2 are rare in type II ␤-turns, and furthermore arginine shows low turn potential. Turn 2 is assigned as a typical type II ␤-turn (␤ ␥L ). In this turn, the Cys 15 oxygen atom forms hydrogen bonding with the Tyr 18 amide-proton (Fig. 1B). The region Trp 24 to Gln 27 is not well defined by the NMR data.

Description of the Three-dimensional Structure-
Electrophysiological Studies on Mammalian Na V Channels-The effects of ␦-HXTX-Mg1a were initially examined on neonatal rat DRG neurons as these were employed in the original characterization of ␦-HXTX-1 and ␦-AOTX-Mb1a toxins (36 -38). Rat DRG neurons express two types of Na V channel currents, TTX-sensitive (mainly Na V 1.1, Na V 1.6, and Na V 1.7) and TTX-resistant (Na V 1.8 and Na V 1.9) (26,39). Like ␦-HXTX-1 toxins, ␦-HXTX-Mg1a slowed inactivation of TTX-sensitive I Na in DRG neurons (Fig. 3A). This was observed as a sustained current in the presence of the toxin during depolarizing test   (Fig. 3E). This steady-state current was completely blocked by 100 nM TTX indicating that it is mediated exclusively via Na V channels (data not shown). In parallel, the peak I Na amplitude recorded at Ϫ10 mV transiently increased, but after 5 min perfusion was slightly reduced by 6.6 Ϯ 11.0% (n ϭ 7). Similar to ␦-HXTX-1 toxins (37,38), this effect was accompanied by a modest 10-mV hyperpolarizing shift in the threshold, but not the midpoint (V1 ⁄ 2 ), of Na V channel activation (Fig.  3C). Furthermore, ␦-HXTX-Mg1a, at concentrations up to 1 M, failed to modulate TTX-resistant I Na (Fig. 3, B and D).
Unlike ␦-HXTX-1 toxins, however, the effect to slow Na V channel inactivation in TTX-sensitive DRG neurons was almost completely reversible after washing with toxin-free solution with a on of 84.9 Ϯ 18.4 s and off of 81.5 Ϯ 9.2 s (n ϭ 5).
Given the potent action of ␦-HXTX-Mg1a on DRG neurons, we then assayed the effect of the toxin on eight mammalian Na V channel clones (Na V 1.1/␤ 1 to Na V 1.8/␤ 1 ) by analyzing g Na /V relationships and steady-state inactivation. Effects of the toxin on Na V 1.9 channels were not investigated as this channel subtype currently fails to express in standard heterologous systems (40). Addition of ␦-HXTX-Mg1a (up to 5 M) to the bath medium produced a marked slowing of I Na inactivation with a rank order in magnitude of Na V 1.6/␤ 1 Ͼ Na V 1.1/␤ 1 Ͼ Na V 1.3/ ␤ 1 . However, with Na V 1.2/␤ 1 and Na V 1.7/␤ 1 , there were only weak effects to slow I Na inactivation (Fig. 4A, left-hand panels, and supplemental Table 2S). These effects were accompanied by ϳ20% increase in peak I Na amplitude with Na V 1.6/␤ 1 and Na V 1.1/␤ 1 channels but less significant changes with the remaining channel clones (supplemental Table 2S).
The above effects were accompanied by only weak hyperpolarizing shifts in the voltage dependence of activation (g Na ) to a maximum ⌬V1 ⁄ 2 of Ϫ8.9 mV for Na V 1.3/␤ 1 channels (Fig. 4A, right-hand panels). In contrast, the toxin failed to slow I Na inactivation of Na V 1.4/␤ 1 , Na V 1.5/␤ 1 , or Na V 1.8/␤ 1 channels at concentrations up to 5 M (Fig. 4B). This lack of activity on Na V 1.8/␤ 1 channels is consistent with the lack of effects on TTX-resistant I Na in DRG neurons that express both Na V 1.8 and Na V 1.9.
In the mammalian Na V channel clones, steady-state inactivation (h ∞ ) in the absence of toxin was best described by a single Boltzmann function. ␦-HXTX-Mg1a (5 M) caused a strong Ϫ19.7 mV hyperpolarizing shift in the voltage at halfmaximal inactivation (V1 ⁄ 2 ) of Na V 1.3/␤ 1 but only weak nonsignificant shifts in V1 ⁄ 2 in Na V 1.1/ ␤ 1 , Na V 1.2/␤ 1 , Na V 1.6/␤ 1 , and Na V 1.7/␤ 1 channels of less than Ϫ4 mV (Fig. 4A, right-hand panels). Moreover, in the presence of ␦-HXTX-Mg1a, steady-state inactivation became incomplete, causing the appearance of a noninactivating component at prepulse test potentials more depolarized than Ϫ40 mV, an effect previously noted with ␦-HXTX-1 toxins (37,38). This noninactivating component was up to 27 Ϯ 10% (n ϭ 5) of peak current in the case of the Na V 1.1/␤ 1 channel (determined from C in Equation 4) with a rank order in magnitude of (Fig. 4A, right-hand panels). This was sufficient to induce significant changes in the slope factor, k h , in all Na V channel clones except Na V 1.2/ ␤ 1 . Although time-dependent shifts in h ∞ have been demonstrated in patch clamp configurations (41), these are unlikely to account for the observed changes given the magnitude of the shift in Na V 1.3/␤ 1 and the absence of a noninactivating component in controls.
Electrophysiological Studies on Insect Na V Channels-In contrast to the clear effects, but modest potency, of the toxin on mammalian Na V channel gating and kinetics, ␦-HXTX-Mg1a showed high affinity for the Drosophila Na V channel clone DmNa V 1/TipE. This was evident first by a concentration-dependent increase in control peak I Na amplitude up to a maximum of 290 Ϯ 37% at 1 M (n ϭ 5; Fig. 5C). This was most likely the result of the marked concentration-dependent slowing of fast inactivation (Fig. 5, A and C) with an EC 50 of 22.8 nM (Fig. 5H). Importantly ␦-HXTX-Mg1a completely removed fast inactivation at concentrations of 1 M (Fig. 5C). This was accompanied by a 21.8-mV hyperpolarizing shift in DmNa V 1/TipE channel activation at 1 M (n ϭ 3, Fig. 5D). At a concentration of 15 nM, ␦-HXTX-Mg1a also caused steady-state DmNa V 1/TipE channel inactivation to become incomplete at prepulse test potentials more depolar- ized than Ϫ50 mV (noninactivating component C ϭ 49.2 Ϯ 0.3%, see Fig. 5B). At higher concentrations, the toxin completely prevented channel inactivation (Fig. 5D).
To further probe the mode of action and insect selectivity of ␦-HXTX-Mg1a, we assayed the toxin on cockroach DUM neurons, previously employed in the characterization of ␦-HXTX-Hv1a (42). Like ␦-HXTX-Hv1a, ␦-HXTX-Mg1a produced a potent time-and concentration-dependent slowing of I Na inactivation (Fig. 5, E and H). Maximum slowing after 50 ms was 46.1 Ϯ 4.5% (n ϭ 10) of control peak I Na at 300 nM ␦-HXTX-Mg1a (Fig. 5E). The EC 50 for the slowing of inactivation in DUM neurons was 823 pM, ϳ56-fold lower than in DRG neurons (Fig. 5H). Hence, although this toxin is not insect-specific, it shows a marked selectivity for insect Na V channels. The toxin also caused 14.3 and 13.5 mV hyperpolarizing shifts in the V1 ⁄ 2 of Na V channel activation and steady-state inactivation, respectively (Fig. 5F).
In DUM neurons, the slowing of inactivation described above was accompanied initially by a slight increase in the peak I Na . This effect would reach a maximum after ϳ70 s followed immediately by a decrease in peak and late I Na amplitude (Fig.  5E). If depolarizing test pulses were abolished for 2 min after reaching steady-state toxin effects (e.g. after 4 -5 min perfusion), peak and late I Na amplitude increased markedly (Fig. 5G, trace marked 120 s). If depolarizing test pulses (⌬t ϭ 10 s) were then reapplied, peak and late I Na amplitude would decline back to steady-state levels within ϳ1 min (Fig. 5G, trace marked 180  s). This effect may result from a slowed rate of recovery from fast inactivation or a voltage-dependent dissociation of the toxin from the channel. To test these possibilities, we first assessed the effect of ␦-HXTX-Mg1a (30 nM) on the rate of recovery from inactivation. Although the toxin markedly increased repriming kinetics at interpulse intervals less than 3 ms, presumably reflecting a more rapid transition from open to closed states, the major effect of the toxin was to prolong the rate of recovery from fast inactivation at interpulse intervals greater than 3 ms, reflecting a slowed transition between closed-inactivated and closed states (Fig. 6A). This action resulted in a use-dependent decrease in peak I Na amplitude during repetitive stimulation at 30 Hz (Fig. 6B).
To determine whether the effect of the toxin was dependent on the holding potential, the amplitude of the toxin-modified I Na measured at the end of the test pulse was compared with peak I Na at hyperpolarized holding potentials. In comparison with currents recorded at Ϫ90 mV, the fraction of the sustained I Na , measured at the end of the 50-ms depolarizing test pulse (I 50ms ), compared with the peak I Na amplitude (I pk ) was not significantly increased except at Ϫ150 mV (p Ͻ 0.05, n ϭ 7; Fig.  6C). In addition, to test if depolarizing pulses could cause dissociation of the toxin, post-pulses up to ϩ200 mV were applied immediately following a test pulse to Ϫ10 mV from Ϫ90 mV every 2 s. At this stimulation frequency there was some initial rundown in peak and late I Na amplitude, but the depolarizing post-pulses failed to cause any significant decrease in fractional I 50ms /I pk compared with data recorded in the absence of the post-pulse (p Ͼ 0.05, n ϭ 5; Fig. 6D).
To investigate the effects of the toxin on membrane excitability, DUM neurons were held under current clamp conditions, and action potentials were evoked by depolarizing current pulses. ␦-HXTX-Mg1a (100 nM) initially produced a prolongation of the repolarizing phase of the action potential (Fig. 7A). The falling phase of the action potential developed a broad shoulder in the last two-thirds of the repolarization phase (Fig. 7A). This caused a suppression of the after hyperpolarization and an increase in spike duration resulting in "plateau" action potentials (Fig. 7, A-C). Although a small depolarization of 3.1 Ϯ 0.6 mV (n ϭ 5) was observed, neither resting membrane potential (RMP) nor spike amplitude was significantly altered. Applying artificial hyperpolarization to a level 20 -40 mV more negative than the RMP facilitated the appearance of these plateau potentials. In addition, the duration of plateau potentials were heavily influenced by the stimulation frequency such that the duration increased from ϳ70 ms at a RMP of Ϫ60 and stimulus interval of 10 s (Fig. 7A) to ϳ300 ms at stimulus intervals of 1 min at an RMP held at Ϫ90 mV (Fig.  7C). Interestingly, 100 nM ␦-HXTX-Mg1a co-applied with 500 M 3-4 diaminopyridine (3,4-DAP), a K V channel blocker, markedly prolonged the action potential duration to ϳ150 ms at Ϫ60 mV (Fig. 7D) and ϳ800 ms at Ϫ90 mV (Fig. 7E). In this case, the action potential duration was more than 80-fold longer than with 3,4-DAP applied alone.
As described previously, DUM neurons are spontaneously active (43). At resting membrane potentials, most DUM neurones were capable of generating repetitive action potentials with firing frequencies of 31 Ϯ 4.6 Hz (n ϭ 4; Fig. 7F). In the presence of 100 nM ␦-HXTX-Mg1a, repetitive plateau action potentials of ϳ70 ms duration would also occur spontaneously, albeit with a much reduced firing frequency of 3.2 Ϯ 0.6 Hz (n ϭ 4; Fig. 7G). However, when the membrane potential was hyperpolarized to Ϫ80 mV, spontaneous plateau action potential duration increased up to ϳ500 ms, but firing frequency was still low at 0.7 Ϯ 0.2 Hz (n ϭ 4; Fig. 7H). Using a ramp current from Ϫ80 mV at 0.2-0.4 nA/s, there was also a Ϫ26 Ϯ 2.0 mV shift in the threshold of spontaneous firing from Ϫ44.6 Ϯ 4.0 mV in controls to Ϫ70.6 Ϯ 3.5 mV (n ϭ 7) in the presence of 100 nM ␦-HXTX-Mg1a.

DISCUSSION
Actions on Na V Channel Gating-In all cases, ␦-HXTX-Mg1a modulated channel gating and kinetics in a similar fashion to other ␦-HXTX-1 toxins by causing the following: (i) a slowing/removal of Na V channel inactivation typically associ- FIGURE 4. Differential effects of ␦-HXTX-Mg1a on mammalian Na V channels expressed in Xenopus oocytes. A, left-hand panels show superimposed current traces illustrating typical effects on mammalian Na V 1.1-Na V 1.3/␤ 1 , and Na V 1.6 -1.7/␤ 1 following a 5-min perfusion with 5 M ␦-HXTX-Mg1a. Currents were elicited every 5 s by 100-ms test pulses from Ϫ90 mV to the voltage of maximum activation of the Na V channel subtype under control conditions. Corresponding right-hand panels show normalized g Na /V (squares) and h ϱ /V (circles) relationships in the presence (open symbols) and absence (closed symbols) of 5 M ␦-HXTX-Mg1a (n ϭ 3-6). The g Na /V curves were fitted using Equation 1 under "Experimental Procedures" although h ϱ /V curves were fitted according to Equation 4. B, superimposed current traces showing typical lack of effect on mammalian Na V 1.4/␤ 1 , Na V 1.5/␤ 1 , or Na V 1.8/␤ 1 channel currents following a 5-min perfusion with 5 M ␦-HXTX-␤Mg1a. ated with an increase in peak Na V channel currents; (ii) a hyperpolarizing shift in the voltage dependence of channel activation; and (iii) a hyperpolarizing shift in the voltage dependence of steady-state inactivation. In insect neurons, further characterization revealed that the toxin (iv) slowed Na V channel repriming kinetics and (v) caused a use-dependent reduction in I Na amplitude.
The toxin-induced increase in peak I Na seen with DmNa V 1 and to a lesser extent in DUM neurons, rNa V 1.1, rNa V 1.3, and Na V 1.6, was in general well correlated with the magnitude of the sustained current at the end of the depolarizing test pulse. This increase in peak I Na can be explained by a delay in Na V channel fast inactivation from open to open-inactivated states. In contrast, ␦-HXTX-Mg1a did not slow I Na inactivation as markedly in Na V 1.2 or Na V 1.7 and therefore did not appreciably increase peak I Na amplitude. This slowing of the transition from open to open-inactivated states also generated a noninactivating component in the steady-state inactivation curve. This occurred at prepulse test potentials more depolarized than Ϫ40 mV, an effect previously noted with ␦-HXTX-1 toxins (37,38). This is the result of a more rapid rate of recovery from fast inactivation when test pulses are separated by less than 3 ms. This reflects a higher proportion of transitions between open and closed states than open-inactivated to closed states via a closed-inactivated intermediate.
In cockroach DUM neurons, the slowing of Na V channel inactivation by ␦-HXTX-Mg1a induced the development of pro- longed plateau action potentials. This was accompanied by spontaneous repetitive firing as a result of a hyperpolarizing shift in the threshold of action potential generation because of a negative shift in the voltage dependence of Na V channel activation. The reduction in firing frequency observed in tonically active neurons probably reflects the reduced rate of recovery from fast inactivation at interpulse intervals greater than 3 ms, presumably resulting from a slowing of the frequency of transition between the closed-inactivated and closed states. This is supported by the use-dependent decrease in I Na amplitude during repetitive pulses. Interestingly, in DUM neurons, ␦-HXTX-Mg1a produced greater inhibition of I Na inactivation (Fig. 7I), with resultant prolongation of action potential duration, when the membrane potential was held at more hyperpolarized potentials (Fig. 8H). This suggests that the toxin partially dissociates from the channel at more depolarized potentials. Voltage-dependent dissociation of bound toxins from vertebrate Na V channels has already been demonstrated with scorpion and sea anemone toxins (9,11,45,46) and has also been observed with ␦-HXTX-Hv1a on insect DUM neurons (42). However, there was no indication of any depolarization-dependent dissociation of toxin from DUM neurons using hyperpolarized holding potentials or depolarizing post-pulse protocols.
Comparison with Orthologous ␦-HXTX-1 Toxins-These actions on TTX-sensitive Na V channel subtypes in rat DRG and insect DUM neurons resemble the effects of other spider ␦-HXTX-1 toxins, with which ␦-HXTX-Mg1a shares some sequence homology (37,38,42). In contrast to ␦-HXTX-1 toxins, however, ␦-HXTX-Mg1a is more efficacious on insect rather than mammalian Na V channel subtypes, including markedly different actions on repriming kinetics and usedependent activity (this study and see Ref. 10). ␦-HXTX-Mg1a therefore shows greater similarity in its phylum-selective actions with scorpion ␣-like toxins. Like ␦-HXTX-Mg1a, ␣-like toxins are toxic by direct injection into rat brain but fail to compete for site-3 on rat brain synaptosomes (9). Rat brain synaptosomal membranes are rich in mainly rNa V 1.2/1.2a (78%) and rNa V 1.1 (15%) (47). Consistent with a lack on inhibition of 125 I-Lqh2 binding to site-3 in rat brain synaptosomes, ␦-HXTX-Mg1a showed only weak effects on Na V 1.2/␤ 1 at 5 M. In addition, the toxin had even weaker activity on Na V 1.7/␤ 1 and no activity on Na V 1.4/␤ 1 , Na V 1.5/␤ 1 , and Na V 1.8/␤ 1 channels (Fig. 4A). Although there is currently no reliable expression system for Na V 1.9 channels, the lack of effect of ␦-HXTX-Mg1a on TTXresistant I Na in DRG neurons would indicate that ␦-HXTX-Mg1a does not have any appreciable affinity for this channel subtype.
Scorpion ␣-like toxins, similar to ␦-HXTX-Mg1a, specifically target Na V 1.1, Na V 1.3, and Na V 1.6 channels (48). Indeed FIGURE 6. Typical effects of ␦-HXTX-Mg1a on DUM neuron Na V channel gating. A, effects of 30 nM ␦-HXTX-Mg1a on Na V channel repriming kinetics in cockroach DUM neurons. Na V channel repriming rate was determined by normalizing peak I Na elicited during a 50-ms test pulse from Ϫ90 mV to Ϫ10 mV against peak I Na recorded during a 50-ms conditioning pulse from Ϫ90 mV to Ϫ10 mV and plotted as a function of the interpulse interval. With interpulse intervals greater than 2 ms, 30 nM ␦-HXTX-Mg1a (open circles, n ϭ 6) slowed the rate of recovery from inactivation in comparison with controls (closed circles, n ϭ 6). B, use-dependent actions of ␦-HXTX-Mg1a on Na V channels in cockroach DUM neurons. Effects of 30 nM ␦-HXTX-Mg1a on use-dependent decline in I Na during 20 depolarizing test pulses from Ϫ90 to Ϫ10 mV at 30 Hz are shown. Currents were normalized to the peak I Na amplitude of the first pulse in the train in the absence (filled symbols, n ϭ 6) and presence (open symbols, n ϭ 4) of 30 nM ␦-HXTX-Mg1a. C, effects of ␦-HXTX-Mg1a (30 nM) at different hyperpolarized holding potentials. Late current (I 50ms ) amplitude only increased at holding potentials more negative than Ϫ140 mV, resulting in a modest but significant increase in the normalized late current (I 50ms /I pk ) (n ϭ 3, *, p Ͻ 0.05, one-way repeated measures analysis of variance). D, effects of depolarizing post-pulses on dissociation of 30 nM ␦-HXTX-Mg1a were assessed using 10-ms post-pulses to ϩ140 or ϩ200 mV applied immediately following a test pulse to Ϫ10 from Ϫ90 mV every 2 s (n ϭ 5). NS, not significant at p Ͻ 0.05. the lethal actions of ␦-HXTX-Mg1a, when injected intracranially, most likely results from its action on these channels and not Na V 1.2. However, in this study ␦-HXTX-Mg1a was not active on Na V 1.4/␤ 1 , which ␣-like, ␣-insect, and anti-mammalian ␣-toxins specifically target (49). Moreover, it was only weakly active on Na V 1.7/␤ 1 , which the novel scorpion toxin OD1 specifically targets (50). ␦-HXTX-Mg1a is therefore unique from scorpion ␣-toxins in that it does not target Na V 1.2, Na V 1.4, or Na V 1.7 channels with high affinity.
Structure-Function Relationships-Given that ␦-HXTX-Mg1a has a similar solution structure to ␦-HXTX-1 toxins (Fig. 2C) and comparable actions on Na V channel gating and kinetics, we posited that ␦-HXTX-Mg1a might also share key residues important for binding to their Na V channel target. Fig. 8 (A-C) shows a comparison of residues that are conserved in spider ␦-toxins targeting site-3 (14,15). A number of residues of these spider toxins are oriented in a similar fashion because of the sequence and structural homology between ␦-HXTX-Mg1a and ␦-HXTX-1 toxins (Fig. 1A and Fig. 2C). A cluster of positively charged residues of ␦-HXTX-Mg1a (Lys 4 , Arg 5 , and Arg 33 ) appears in a similar position on the surface of ␦-HXTX-Hv1a and ␦-HXTX-Ar1a (Lys 3 , Lys 4 , and Arg 5 ). The positively charged Lys 9 of ␦-HXTX-Mg1a also seems to be oriented similarly to Lys 10 of ␦-HXTX-Hv1a and ␦-HXTX-Ar1a, whereas aromatic residues Trp 7 and Tyr 22 are conserved in all these spider toxins. Finally, ␦-HXTX-Mg1a possesses Ser 6 in the same position as the nonpolar Asn 6 of the ␦-HXTX-1 toxins. However, other variations in the sequence of ␦-HXTX-Mg1a are most likely responsible for the dramatic loss in affinity for FIGURE 7. Effect of ␦-HXTX-Mg1a on action potential duration and firing frequency in cockroach DUM neurons. A and B, typical superimposed action potentials generated by a single supramaximal current pulse every 10 s before (black traces) and following (gray traces) a 5-min perfusion with 100 nM ␦-HXTX-Mg1a. Traces were recorded from a resting membrane potential of Ϫ60 (A) and Ϫ90 mV (B), following manual hyperpolarization. C, same as B except stimuli were delivered every minute rather than at 10-s intervals. D, representative effects of a 5-min perfusion with 500 M 3,4-di-aminopyridine (3,4-DAP, dark gray), 100 nM ␦-HXTX-Mg1a (gray), and 100 nM ␦-HXTX-Mg1a in the presence of 500 M 3,4-DAP (light gray), on action potential duration at a resting membrane potential of Ϫ60 mV recorded every 10 s. E, same as D except data were recorded at Ϫ90 mV following manual hyperpolarization. Note the slower time scale. F-H, alterations in spontaneous action potential firing frequency and duration. Typical effects on tonic action potential firing before (F) and following a 5-min perfusion with 100 nM ␦-HXTX-Mg1a recorded at a resting membrane potential of Ϫ60 mV (G) and a holding potential of Ϫ80 mV following manual hyperpolarization (H). rNa V 1.2 observed in 125 I-Lqh2 binding experiments (10) and the present voltage clamp experiments (Fig. 4A). In particular, the addition/removal of charged side chains at positions 3, 10 -12, 19, 34, 36, and 40 -43 are dramatically different to the corresponding residues in ␦-HXTX-1 toxins.
Site-3 toxins are proposed to slow the fast inactivation of Na V channels by preventing the outward movement of the S4 segments (6). ␦-HXTX-Mg1a has been previously shown to compete with the scorpion ␣-insect toxin Lqh␣IT at subpicomolar concentrations for binding to insect site-3 (10). Recently, extensive mutagenesis studies have identified the pharmacophore of Lqh␣IT (51). Fig. 8E shows the critical residues in Lqh␣IT form a functional surface composed of two domains, the NC domain and core domain. The positively charged residues of Lys 4 , Arg 5 , and Arg 33 in ␦-HXTX-Mg1a, are oriented similarly to Lys 8 , Arg 58 , and Lys 62 in the NC domain of Lqh␣IT. The aromatic Trp 7 and Tyr 22 and basic Lys 9 residues of ␦-HXTX-Mg1a occupy similar positions to Phe 17 , Trp 38 , and Arg 18 in the core domain of Lqh␣IT. These aromatic residues in ␦-HXTX-Mg1a are also present in ␦-HXTX-1 toxins. Despite these topologically similar residues, ␦-HXTX-Mg1a and ␦-HXTX-1 toxins are smaller in bulk than scorpion ␣-toxins and lack aliphatic residues corresponding to the bioactive Ile 57 and Val 59 residues of Lqh␣IT. Indeed, ␦-HXTX-Mg1a shows the greatest similarity in phylum selectivity to scorpion ␣-like toxins, e.g. Lqh3. Results from a mutagenesis study suggest that the core domain of Lqh3 plays an important role in interaction with the receptor site and its toxin selectivity (52). Aromatic (Phe 17 and Phe 39 ), positively charged (His 15 ), and aliphatic residues (Pro 18 and Leu 45 ) were assigned to the bioactive surface in Lqh3. Importantly, the aliphatic Val 21 and Ala 23 of ␦-HXTX-Mg1a are oriented similarly to those of Lqh3. The core domain of ␦-HXTX-Mg1a therefore seems to be involved in hydrophobic-aromatic interactions and play a role in its phylum selectivity, although this awaits experimental confirmation.
In conclusion, although peptide neurotoxins binding with site 3 on Na V channels are known to possess phylum selectivity, our present study highlights a novel feature of ␦-HXTX-Mg1a, namely target selectivity for distinct Na V channel subtypes. This discriminating activity of the spider toxin is somewhat unexpected given that, like homologous ␦-HXTX-1 toxins, it interacts with both mammalian and insect Na V channels. The effects of ␦-HXTX-Mg1a on DmNa V 1/ TipE, Na V 1.1/␤ 1 , Na V 1.3/␤ 1 , and Na V 1.6/␤ 1 channels as opposed to the limited or lack of activity on Na V 1.2/␤ 1 , Na V 1.4/␤ 1 , Na V 1.5/ ␤ 1 , Na V 1.7/␤ 1 , Na V 1.8/␤ 1 , and Na V 1.9/␤ 1 channels are remarkable because they represent the first exhaustive characterization of a selective interaction of any peptide neurotoxin across the complete range of Na V channel subtypes. Our findings indicate that specific insect and mammalian Na V channel subtypes can be pharmacologically discriminated by their sensitivity to ␦-HXTX-Mg1a as has only been partially described for scorpion ␣-toxins, sea anemone, and other spider toxins (53,54). This should provide new tools to study the functional role and distribution of various Na V channel subtypes. Despite very low sequence homology with all three scorpion ␣-toxin groups (e.g. Lqh3, Lqh␣IT, and Aah2), the three-dimensional structure of ␦-HXTX-Mg1a reveals an apparently similar functional surface with a number of these site-3 toxins and spider ␦-toxins. Several positively charged and aromatic residues of ␦-HXTX-Mg1a are arranged in a topologically similar manner to those of site-3 toxins. Thus, the structure of ␦-HXTX-Mg1a provides an important lead for understanding phylum and subtype specificity and awaits the determination of the pharmacophore of the toxin.