Structure of Membrane-active Toxin from Crab Spider Heriaeus melloteei Suggests Parallel Evolution of Sodium Channel Gating Modifiers in Araneomorphae and Mygalomorphae*

Background: Several toxins from mygalomorph spiders are known to inhibit sodium channel activation. Results: Hm-3 toxin from the araneomorph spider Heriaeus melloteei inhibits mammalian and insect sodium channel activation and possesses membrane activity. Conclusion: Hm-3 binds to sodium channel voltage sensors through membrane access. Significance: The first sodium channel activation inhibitor from Araneomorphae points to parallel evolution of H. melloteei and its distant mygalomorph relatives. We present a structural and functional study of a sodium channel activation inhibitor from crab spider venom. Hm-3 is an insecticidal peptide toxin consisting of 35 amino acid residues from the spider Heriaeus melloteei (Thomisidae). We produced Hm-3 recombinantly in Escherichia coli and determined its structure by NMR spectroscopy. Typical for spider toxins, Hm-3 was found to adopt the so-called “inhibitor cystine knot” or “knottin” fold stabilized by three disulfide bonds. Its molecule is amphiphilic with a hydrophobic ridge on the surface enriched in aromatic residues and surrounded by positive charges. Correspondingly, Hm-3 binds to both neutral and negatively charged lipid vesicles. Electrophysiological studies showed that at a concentration of 1 μm Hm-3 effectively inhibited a number of mammalian and insect sodium channels. Importantly, Hm-3 shifted the dependence of channel activation to more positive voltages. Moreover, the inhibition was voltage-dependent, and strong depolarizing prepulses attenuated Hm-3 activity. The toxin is therefore concluded to represent the first sodium channel gating modifier from an araneomorph spider and features a “membrane access” mechanism of action. Its amino acid sequence and position of the hydrophobic cluster are notably different from other known gating modifiers from spider venom, all of which are described from mygalomorph species. We hypothesize parallel evolution of inhibitor cystine knot toxins from Araneomorphae and Mygalomorphae suborders.

Spider venoms are multicomponent mixtures of biologically active molecules, the main functions of which are to subdue prey and deter predators (1,2). These venoms usually contain low molecular weight compounds (3,4), proteins (5,6), and peptides (7,8) with the latter group most often prevailing over the other two in terms of content and/or functional importance. Based on the target and structural features, most venom peptides can be classified into two large groups: linear cytolytic peptides (2,9) and disulfide-containing neurotoxins (1,2,7,8). The latter usually adopt the inhibitor cystine knot (ICK) 5 or "knottin" fold and target ion channels and receptors of the prey or predator nervous system. Numerous studies have shown that currents through voltage-gated ion channels can be affected either by pore block or modification of the gating mechanisms.
In the case of neurotoxins acting on voltage-gated sodium channels (Na V s), a number of peptide pore blockers were identified. They were shown to bind to the so-called receptor site 1 (as well as the classic low molecular weight blockers tetrodotoxin and saxitoxin) and inhibit ion conductivity due to the direct blockage of the pore. The majority of well described peptide pore blockers were isolated from cone snail venoms (the so-called -conotoxins) (10). So far there is only one known spider venom peptide that acts as a sodium channel pore blocker: PnTx1 from Phoneutria nigriventer. This peptide toxin was shown to compete with -conotoxins but not with tetrodotoxin for channel binding (11).
To modify sodium channel gating, neurotoxins can use different strategies. They can retard or inhibit inactivation and either facilitate or inhibit activation of a channel (12). These types of action are usually associated with binding of the peptides to so-called receptor sites 3 and 4 of the channels. The major parts of the channels contributing to these sites are extracellular loops connecting the S3 and S4 segments in domains IV and II correspondingly. The mode of action of the gating modifier toxins is usually explained by a voltage sensor trapping mechanism (13).
A delay of inactivation is characteristic of such classic site 3 ligands as scorpion ␣-toxins (14) and certain sea anemone toxins (15). Moreover, a similar effect is characteristic of some spider toxins binding to site 3 (16,17) and, what is worthy of note, site 4 (18). Promotion of sodium channel activation is also quite widespread among toxins: scorpion ␤-toxins (19) and some spider toxins (20) binding to site 4 show this kind of effect. Importantly, the channel voltage dependence of activation is shifted to more hyperpolarized voltages under the action of these toxins. Conversely, several spider peptide toxins are known to inhibit sodium channel activation: ProTx-I and ProTx-II from Thrixopelma pruriens (21), JZTX-III from Chilobrachys jingzhao (22), and HWTX-IV from Haplopelma schmidti, all produced by mygalomorph species. The data on the exact receptor site for these toxins presented in the literature are sometimes ambiguous. For instance, HWTX-IV was first misclassified as a site 1 toxin (23), which was further corrected to site 4 when new data were acquired (24). In part, this is due to the apparent lack of a shift in the channel voltage dependence of activation, which is more often the case, as with ProTx-I, ProTx-II, and JZTX-III that are also believed to bind to site 4 (25,26).
Yet another sodium channel activation inhibitor is the peptide Hm-3 from the spider Heriaeus melloteei (crab spider, Thomisidae, Araneomorphae), which was isolated from the venom by our group earlier (27). The goal of the present study was to characterize Hm-3 in detail by investigating its threedimensional structure and running electrophysiological measurements and thus gain some insight into its mechanism of action.

EXPERIMENTAL PROCEDURES
Hm-3 Isolation-Native Hm-3 was isolated from the venom of H. melloteei following a technique described previously (27).
Recombinant Peptide Production-To produce recombinant Hm-3, a procedure similar to the one developed earlier was used (28). Hm-3-encoding gene was assembled from a number of synthetic oligonucleotides (Table 1) using a combination of PCR and ligation techniques. It was then amplified using a forward primer containing a BamHI restriction site and a methionine codon and a reverse primer containing a SalI restriction site and a stop codon. The PCR fragment was cloned into the expression vector pET-32b (Novagen) using restriction with BamHI and SalI enzymes followed by ligation. As a result, the plasmid pET-32b-Hm-3 was produced and was then used to transform Escherichia coli BL21(DE3) and Origami B cells.
The bacteria were cultured at 37°C in Luria-Bertani medium with the appropriate antibiotics as a selective factor to the midlog phase. Expression was then induced by 0.2 mM isopropyl ␤-D-thiogalactopyranoside, and bacteria were cultured at room temperature (24°C) overnight (16 h). Then the cells were harvested by centrifugation, and the pellet was resuspended in 300 mM NaCl, 50 mM Tris-HCl buffer (pH 6.8), and 5% glycerol and ultrasonicated.
The hybrid protein Trx-Hm-3 with thioredoxin (Trx) as a carrier was purified by affinity chromatography using TALON Superflow resin (Clontech) following the manufacturer's protocol. Finally, the chimeric protein was cleaved at methionine residues by cyanogen bromide (CNBr) using a procedure described previously (29). HCl was added to the protein solution (ϳ1 mg/ml) to the concentration of 0.5 M, and CNBr was then added to the concentration of 100 mM. The probe was then incubated at room temperature (24°C) overnight (16 h) in the dark. Recombinant Hm-3 was purified by reversed-phase HPLC on a Jupiter C 5 column (250 ϫ 10 mm; Phenomenex) using a linear gradient of acetonitrile concentration (0 -60% in 60 min) in the presence of 0.1% trifluoroacetic acid. The purity of the target peptide was checked by MS, N-terminal sequencing, and analytical chromatography on a Vydac 218TP54 C 18 column (4.6 ϫ 250 mm; Separations Group) in a shallow acetonitrile gradient (25-50% in 50 min).
NMR Experiments and Spatial Structure Calculation-NMR experiments were performed using a 0.5 mM Hm-3 solution in 0.5% D 2 O or 100% D 2 O at pH 5.2. All NMR spectra were acquired on a Bruker Avance 800 spectrometer equipped with a cryoprobe at 35°C. 1 H and 13 C resonance assignments were obtained by a standard procedure using a combination of twodimensional total correlation spectroscopy, NOESY, and 13 C heteronuclear single quantum correlation spectra at natural 13 C abundance (30). 1 (31) in the two-dimensional double quantum-filtered COSY spectrum acquired in 100% D 2 O. Temperature coefficients of amide protons (⌬␦ 1 H N /⌬T) were measured in a temperature range from 25 to 45°C using two-dimensional total correlation spectroscopy spectra. To identify the slowly exchanging amide protons, the Hm-3 sample was lyophilized and redissolved in 100% D 2 O. The hydrogen-deuterium exchange kinetics was measured using one-dimensional 1 H spectra.
Spatial structure calculation was performed in the CYANA 3.0 program (32). Upper interproton distance constraints were derived from the intensities of cross-peaks in two-dimensional NOESY spectra ( m ϭ 100 ms) via a "1/r 6 " calibration. Torsion angle restraints and stereospecific assignments were obtained from J coupling constants and NOE intensities. Hydrogen bonds were introduced based on temperature gradient and deuterium exchange rates of H N protons. The disulfide bond connectivity pattern was established on the basis of the observed NOE contacts and verified on the preliminary stages of the spatial structure calculation.
Hm-3 Binding to Lipid Vesicles-Small unilamellar vesicles (SUVs) were prepared by sonication using POPC or a POPC/ DOPG (3:1) mixture (Avanti Polar Lipids) in 10 mM Tris acetate buffer (pH 7.0) with or without 150 mM NaCl. The final lipid concentrations were measured by one-dimensional 1 H NMR spectroscopy by dissolving small fractions of the SUV preparation in a CDCl 3 /CD 3 OD/D 2 O (15:10:3) mixture. Titration of an Hm-3 sample (20 M; 5% D 2 O; same buffer) with SUVs was performed at 35°C. At each lipid concentration, a one-dimensional 1 H NMR spectrum was measured, and the equilibrium concentration of free peptide in solution (C f ) was determined using the integral intensity of the amide region of the spectrum. The binding isotherms were analyzed using the partition equilibrium Equation 1 and Langmuir Equation 2.
where C b is the bound peptide concentration (C 0 ϭ C b ϩ C f ), K p is the partition coefficient, K N is the affinity constant of the peptide to the site on the vesicle surface formed by N lipid molecules, and L* is the lipid concentration in the outer leaflet of the vesicles (60% of total lipid; L* ϭ 0.6 ϫ L). The effect of dilution was taken into account. Electrophysiology-Genes of the following channels were used to prepare mRNA and inject Xenopus oocytes as described previously (33,34): mammalian Na V 1.1-Na V 1.6 and Na V 1.8; DmNa V 1 from Drosophila; mammalian voltage-gated potassium channels (K V s) K V 1.1, K V 1.3, K V 2.1, K V 4.2, and K V 10.1; and mammalian voltage-gated calcium channel (Ca V ) Ca V 3.3. Two-electrode voltage clamp recordings were performed at room temperature (18 -22°C) using a Geneclamp 500 amplifier (Molecular Devices) controlled by a pClamp data acquisition system (Axon Instruments). Whole-cell currents from oocytes were recorded 1-4 days after injection. Bath solution composition was: 96 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 2 mM MgCl 2 , and 5 mM HEPES (pH 7.4). Voltage and current electrodes were filled with 3 M KCl. Resistances of both electrodes were kept between 0.7 and 1.5 megaohms. The elicited currents were filtered at 1 kHz and sampled at 20 kHz using a four-pole low pass Bessel filter. Leak subtraction was performed using a ϪP/4 protocol. To avoid overestimation of a potential toxin-induced shift in the current-voltage relationships of inadequate voltage control when measuring large sodium currents in oocytes, only data obtained from cells exhibiting currents with peak amplitude below 2 A were considered for analysis.
For the electrophysiological analysis of toxins, a number of protocols were applied from a holding potential of Ϫ90 or Ϫ100 mV with a start-to-start interval of 0.2 Hz. Sodium current traces were evoked by 50-ms or 100-ms depolarizations to V max (the voltage corresponding to maximal sodium current in control conditions).
The current-voltage (I-V) relationships were determined by 100-ms step depolarizations between Ϫ90 and 70 mV using 5-mV increments. Na ϩ conductance (g Na ) was calculated from the currents according to Ohm's law: g Na ϭ I Na /(V Ϫ V rev ) where I Na represents the sodium current peak amplitude at a given test potential V and V rev is the reversal potential. The values of g Na were plotted as a function of voltage and fitted using the Boltzmann equation: Ϫ1 where g max represents maximal g Na , V 1/2 is the voltage corresponding to half-maximal conductance, and k is the slope factor.
To assess the toxin-induced effects on the steady-state inactivation, a standard two-step protocol was applied. In this protocol, 100-ms conditioning 5-mV step prepulses ranging from Ϫ90 to 60 mV were followed by a 50-ms test pulse to V max . Data were normalized to the maximal sodium current amplitude, plotted against prepulse potential, and fitted using the Boltzmann equation: I Na /I max ϭ 1/(1 ϩ exp((V Ϫ V 1/2 )/k)) where V 1/2 is the voltage corresponding to half-maximal inactivation and k is the slope factor.
To assess the concentration-response relationships of the toxin-induced inhibitory effects, dose-response curves were constructed. Current traces were evoked as described above in control and in the presence of a range of toxin concentrations. The percentage of toxin-induced inhibition of the sodium current peak amplitudes was plotted against the logarithm of applied concentrations and fitted with the Hill equation to obtain the EC 50 value (i.e. the toxin concentration that produces 50% of the maximal effect) of the effects of Hm-3. From the Hill equation, the Hill coefficient (i.e. the degree of cooperativity and eventually the number of molecules necessary to modulate one channel) was obtained.
To investigate the voltage-dependent reversal of Hm-3 inhibition, a depolarizing pulse of increasing amplitude (up to ϩ100 mV) and/or duration (up to 800 ms) was followed by a 20-ms repolarization to the holding potential of Ϫ100 mV and then a test pulse to 0 mV. The exponential time course of relief of toxin inhibition was determined by plotting the normalized current as a function of time. The normalization was performed as follows. The steady-state level of I Na peak current obtained after application of 1 M Hm-3 was set to 0. The level of I Na peak current obtained after a 700-ms prepulse duration was set to 1. These normalized current values were plotted as a function of time: f(t) ϭ (I t Ϫ I 0 )/(I 700 ms Ϫ I 0 ). The time values of depolarizing prepulse required to recover 60% of the current from inhibition () were calculated to express the different kinetics of reversal of inhibition for different channels.
Comparison of two sample means was made using a paired Student's t test (p Ͻ 0.05). All data are presented as means Ϯ S.E. of at least five independent experiments (n Ն 5). All data were analyzed using pClamp Clampfit 10.0 (Molecular Devices) and Origin 7.5 software (Originlab).

RESULTS
Hm-3 is a peptide toxin isolated previously from H. melloteei venom (27) with the following amino acid sequence (UniProt accession number C0HJK5): GCIAKNKECAWFSGEWCCG-ALSCKYSIKRNLKICV. Because only minute amounts of toxin may be purified from the natural source, we decided to produce Hm-3 recombinantly.
Recombinant Hm-3 Production-To provide enough material for NMR and electrophysiological studies, an E. coli expression system for Hm-3 was developed. An Hm-3-encoding gene was produced from a number of synthetic oligonucleotides and cloned into the pET-32b expression vector. E. coli BL21(DE3) and Origami B cells were then transformed by the recombinant plasmid pET-32b-Hm-3. The target peptide was purified after cleanup of the chimeric protein Trx-Hm-3 by affinity chroma-tography, its cleavage by CNBr, and separation of the hydrolysate by reversed-phase HPLC. We failed to acquire a correctly folded product in E. coli BL21(DE3), but utilization of the E. coli Origami B strain was successful (Fig. 1A). The average molecular mass of the resulting recombinant Hm-3 was measured by MALDI MS (3907.8 Da) and was found to be equal to the mass of native Hm-3 (3907.7 Da) (27) and the calculated mass (3907.7 Da). To prove recombinant and native Hm-3 equivalency, they were co-eluted in analytical reversed-phase HPLC and shown to have the same retention time (Fig. 1B). The yield of recombinant Hm-3 was 0.7 mg/liter of bacterial culture.
Spatial Structure of Hm-3-Spatial structure of Hm-3 was studied by NMR spectroscopy in aqueous solution at pH 5.2 and 35°C ( Fig. 2A). The set of 20 structures (Fig. 3A) was calculated in CYANA from 200 random starts using the following experimental data: upper NOE-based distance restraints, J coupling-based torsion angle restraints, and hydrogen bond restraints ( Table 2). The Hm-3 structure (Fig. 3B) involves a ␤-hairpin formed by two strands (Cys 23 -Ser 26 and Leu 31 -Cys 34 ) connected by a turn in the 3 10 -helix conformation (Ile 27 -Arg 29 ). The ␤-hairpin protrudes from a "globular core" cross-linked by three disulfide bonds (Cys 2 -Cys 18 , Cys 9 -Cys 23 , and Cys 17 -Cys 34 ) presenting the "cystine knot" arrangement (I-IV, II-V, and III-VI). Other possible disulfide bond connectivity patterns were tested in the course of spatial structure calculations, but they did not correspond to the experimental NMR data. The toxin is therefore concluded to assume the conventional ICK (knottin) fold. The Hm-3 "core" accommodates several tight ␤-turns (Fig. 3A). In addition to disulfide bonds, the peptide structure is stabilized by 11 backbone-backbone hydrogen bonds.
The calculated Hm-3 structure is well defined. At the same time, some heterogeneity was observed in the conformation of the Cys 9 -Cys 23 disulfide bridge (Fig. 3A). A lack of structural convergence in this region is probably not due to an enhanced intramolecular mobility but is rather a consequence of insufficiency of the NMR data. Both Cys residues forming this bridge have degenerate 1 H ␤ chemical shifts, thus making the determination of side-chain conformation impossible. The obtained structural ensemble (Fig. 3A) demonstrates no large distance violations, good Ramachandran statistics, and quite low values of the root mean square deviation for backbone and heavy atoms ( Table 2). Thus, the overall Hm-3 spatial structure is reliably and precisely defined by our NMR data.
The analysis of the Hm-3 surface properties (Fig. 3C) revealed the amphiphilic nature of the peptide. Two pronounced neighboring hydrophobic clusters on the toxin surface are formed by residues belonging to the Ala 10 -Trp 16 loop and the central part of the ␤-hairpin. These clusters (consisting of Trp 11 , Phe 12 , Trp 16 , Tyr 25 , Ile 27 , and Leu 31 ) are surrounded by a ring of charged groups, including two negatively charged (Glu 8 and Glu 15 ) and five positively charged (N terminus, Lys 24 , Lys 28 , Arg 29 , and Lys 32 ) moieties. The positive and negative charges are distributed almost uniformly on the opposite (hydrophilic) side of the Hm-3 surface (net charge, ϩ4).
Interaction of Hm-3 with Lipid Vesicles-The amphiphilic properties of the Hm-3 surface pointed to the ability of the peptide to partition into lipid membranes. To investigate the membrane-binding propensity of Hm-3, SUVs consisting of either zwitterionic (POPC) or a mixture of zwitterionic and anionic (POPC/DOPG, 3:1) lipids were used. Titration of the peptide sample with liposomes led to a gradual decrease of the Hm-3 NMR signal intensity (Fig. 2B). The observed attenuation could be explained by a tight association of the peptide molecules with the vesicle surface. In this case, because of a very slow reorientation of SUVs in solution, the bound peptide molecules become unobservable by high resolution NMR spectroscopy, and the intensity of the NMR signal is directly proportional to the equilibrium concentration of the free peptide in solution (C f ). The measured binding curves (Fig. 3D) revealed an effective partitioning of Hm-3 into the zwitterionic and partially anionic SUVs at close to physiological ionic strength conditions (150 mM NaCl). Comparison of these data with the results obtained in low-salt conditions indicated that addition of 150 mM NaCl into the binding buffer diminished the peptide affinity to the POPC/DOPG SUVs only slightly (Fig. 3D).
The partition equilibrium equation (Equation 1) is usually used for the analysis of peptide interactions with lipid membranes. Interestingly, this simple isotherm containing only one variable parameter (partition coefficient K p ) did not fit satisfactorily the measured Hm-3 binding curves (Fig. 3D, dashed  lines). The obtained K p values are collected in Table 3 for comparison with previous studies. Conversely, the usage of a more complex Langmuir isotherm (Equation 2) having two variable parameters (K N and N) provided a reliable approximation of the measured data (Fig. 3D, solid lines). The results (Table 3) revealed that irrespectively of the lipid composition of SUVs and salt concentration the site of the peptide binding on the vesicle surface is formed by ϳ10 lipid molecules (N). The observed changes in Hm-3 affinity toward charged SUVs upon variation of salt concentration are mostly induced by the changes in the peptide affinity to this site (parameter K N ) ( Table  3). The absence of a pronounced dependence in the efficiency of Hm-3 membrane binding upon variation of the charge of lipid headgroups and salt concentration indicates that hydrophobic interactions play a central role in the formation of the peptide⅐membrane complex.
Electrophysiological Characterization-At a concentration of 1 M, recombinant Hm-3 was tested against a panel of Na V s (Fig. 4A). The toxin was found to inhibit sodium current through mammalian channels Na V 1.2 and Na V 1.4 -Na V 1.6 and the insect channel DmNa V 1. Na V 1.1, Na V 1.3, and Na V 1.8 channels were not affected by 1 M Hm-3. A clear rightward shift of the I-V curve was observed for Na V 1.4, Na V 1.5, and DmNa V 1 channels, whereas this was less pronounced for Na V 1.2 and Na V 1.6 channels (Fig. 4A, right panels). It should be noted that when higher concentrations of the toxin were applied a shift of the activation dependence of the Na V 1.2 and Na V 1.6 channels

in aqueous solution, comparison of Hm-3 spatial structure with other ion channel activation inhibitors from spider venom, and isotherms of Hm-3 binding to lipid vesicles.
A, a set of 20 "best" Hm-3 structures superimposed over the backbone atoms. The disulfide bonds are shown in orange. B, ribbon representation of Hm-3 spatial structure. Hydrophobic, positively charged, and negatively charged residues are colored in green, blue, and red, respectively. (C) Ribbon representation and two-sided view of Hm-3, HWTX-IV, JZTX-III, ProTx-I, and SGTx1 surfaces (Protein Data Bank codes are 2MQU, 2M4X, 2I1T, 2M9L, and 1LA4, respectively). The molecules are superimposed over the heavy atoms (carbon, nitrogen, and sulfur) of six conserved Cys residues. On the surfaces, the hydrophobic (Ala, Met, Ile, Leu, Val, Phe, Trp, Tyr, and Pro), polar (Asn, Gln, Gly, His, Ser, and Thr), positively charged (Arg and Lys), and negatively charged (Asp and Glu) residues are colored in green, magenta, blue, and red, respectively. Cys residues are in yellow. D, the binding curves describing Hm-3 interactions with POPC and POPC/DOPG (3:1) SUVs are approximated by the partition equilibrium equation (Equation 1; dashed lines) and by the Langmuir isotherm (Equation 2; solid lines). Fitted parameters are summarized in Table 3. The dilution curve is shown by a dotted line. did occur (data not shown). The native venom-extracted toxin showed identical activity to the recombinant Hm-3 against Na V 1.4 (data not presented). Because it was established that voltage clamp errors might occur in two-electrode experiments with heterologously expressed channels in oocytes, we verified whether the IC 50 concentration of tetrodotoxin (1 nM for Na V 1.2 and 4 nM for Na V 1.4) alters the voltage dependence of activation and found no alteration of the activation and steadystate inactivation curves (data not shown).
To assess the concentration dependence of the Hm-3-induced inhibitory effects, dose-response curves were constructed in which the percentage of current inhibition was plotted as a function of toxin concentration (Fig. 4B) To investigate whether Hm-3 acts in a voltage-dependent manner, the sodium peak current in toxin conditions was calculated as a fraction of the corresponding peak current in control using the I-V relationship data. The obtained data were fitted with a linear regression (Fig. 4C). It was concluded that the degree of Hm-3-induced inhibition is voltage-dependent.
The characteristics of Hm-3 modulation of Na V 1.4 and DmNa V 1 channel kinetics were investigated further. For DmNa V 1, a significant shift of the midpoint of activation is clearly observed in the normalized activation curves (Fig. 5A): the V 1/2 shifted from Ϫ21.7 Ϯ 0.1 mV in control conditions to Ϫ4.1 Ϯ 0.4 mV after toxin application. For Na V 1.4, on the contrary, no significant shift was seen (Fig. 5C).
A reversed situation was observed in the steady-state inactivation curves. For DmNa V 1, little change in the voltage dependence of inactivation was seen: the V 1/2 values were Ϫ49.8 Ϯ 2.2 mV in control and Ϫ45.2 Ϯ 1.6 mV in toxin conditions (Fig.  5B). But for Na V 1.4, a shift was observed: V 1/2 values were Ϫ56.0 Ϯ 0.3 mV and Ϫ62.1 Ϯ 0.7 mV in control and toxin situations, respectively (Fig. 5D). Different sensitive channels exhibit strikingly different patterns of changes in the I-V curves. These differences could be attributed to the different binding affinity of the Hm-3 toxin to specific channels. We investigated whether this was due to different toxin binding off-rates among the sensitive channels. A strong depolarizing pulse to ϩ100 mV of increasing duration was followed by 20 ms at the holding potential of Ϫ100 mV, which was subsequently followed by a test pulse to 0 mV (Fig.  6A). It was found for both tested channels that the conditioning pulse caused a reversal of the Hm-3-induced inhibition of the sodium current (Fig. 6B). The strongest conditioning depolarization pulse did not result in a complete reversal of Hm-3induced inhibition. Maximum values yielded 68 (Na V 1.4) and 90% (DmNa V 1) of control at ϩ100 mV and 635-ms pulse duration (p Ͻ 0.05, n ϭ 5). The inhibition was reversed in an exponential time course. However, different kinetics was observed in Na V 1.4 and DmNa V 1. To quantify these differences in the kinetics of reversal of inhibition, the above protocol was applied with conditioning pulses ranging from ϩ40 to ϩ100 mV (Fig.  6C). For both channels, the relief of inhibition was the fastest at ϩ100 mV and slowest at ϩ40 mV (Fig. 6, D and E). However, at both voltages, the off-rate was much higher for DmNa V 1 compared with Na V 1.4. At ϩ100 mV, ϩ100 mV ϭ 227 ms for Na V 1.4, and ϩ100 mV ϭ 114 ms for DmNa V 1. Accordingly, at ϩ40 mV, ϩ40 mV ϭ 910 ms for Na V 1.4, and ϩ40 mV ϭ 370 ms for DmNa V 1. Isochronal activation curves for toxin-bound channels can be obtained by plotting the normalized percent

TABLE 3 Energetic and stoichiometric parameters of Hm-3 interactions with SUVs obtained using partition equilibrium equation (Equation 1) and Langmuir isotherm (Equation 2)
Lipids Partition equilibrium a , K p increase in test pulse peak amplitude after 700-ms conditioning depolarizations against the range of potentials used in the conditioning pulses. A shift toward more positive potentials was observed for toxin-modulated Na V 1.4 and DmNa V 1 channels. However, a more pronounced shift was seen for the Na V 1.4 channels because the V 1/2 shifted 30 mV toward more positive potential values compared with a moderate 10-mV shift of the midpoint of activation for DmNa V channels (Fig. 6, F and G).
Finally, to check the toxin selectivity to sodium channels, Hm-3 was investigated for its activity on K V s and Ca V s. No activity was observed for K V 1.1, K V 1.3, K V 2.1, K V 4.2, K V 10.1, or Ca V 3.3 at 1 M Hm-3 (data not shown). . Hm-3 activity against Na V s. A, whole-cell current traces recorded from oocytes expressing cloned Na V isoforms (left panels) and current-voltage relationships of Na V s (right panels) in control measurements and after application of Hm-3. Arrows mark traces after application of 1 M toxin. Shown are representative traces of at least three independent experiments. Current-voltage relationships of Na V s in control measurements (f) and after application of 1 M Hm-3 (Ⅺ) are normalized; Norm. curr., normalized current. B, dose-dependent curves for Hm-3 inhibitory action against Na V s. Calculated EC 50 values are presented in a box at right. C, voltage-dependent activity of Hm-3. Sodium peak current after toxin application was normalized to control and fitted with a linear regression. Error bars represent S.E.

Hm-3 Is a Sodium Channel Gating Modifier-Electrophysi-
ological studies demonstrate that Hm-3 effectively inhibits both mammalian and insect Na V s (Fig. 4A, left panels). Moreover, the toxin shifts the dependence of channel activation to more depolarized voltages (Fig. 4A, right panels). Several spider toxins were shown to exhibit a similar effect on sodium channels, such as ProTx-I and -II from T. pruriens (21), HWTX-IV from H. schmidti (24), JZTX-III from C. jingzhao (26), CcoTx1 and CcoTx2 from Ceratogyrus marshalli, and PaurTx3 from Paraphysa scrofa (35) (see Fig. 7 for amino acid sequences). All these toxins are believed to represent gating modifiers; i.e. their blocking effect on the channel is associated with the inhibition of the gating mechanism as opposed to a direct pore block.
The following observations justify allocation of Hm-3 to gating modifiers. (i) Our results indicate that the Hm-3-induced inhibition of the sodium peak current is voltage-dependent (Fig. 4C). It was shown in previous studies that voltage-dependent enhancement or reversal of toxin activity can be considered as a hallmark of gating modifier toxins (25,36,37). (ii) Strong depolarizing prepulses provide relief from Hm-3-induced inhibition (Fig. 6, A and B), also characteristic of gating modifiers (24,25). (iii) Furthermore, the reversal of the inhibition is voltage-dependent (Fig. 6, C-G), a phenomenon described previously for other gating modifiers from spider venoms such as ProTx-I and -II (25) and HWTX-IV (38). We therefore conclude that Hm-3 represents the first sodium channel gating modifier from an araneomorph spider.
The observed differences in channel affinity as characterized by the off-rates of toxin binding to Na V 1.4 (Fig. 6, B and D) and DmNa V 1 (Fig. 6, B and E) provide a possible explanation for the striking difference of change in the I-V curves between these Na V s (Fig. 4A). According to the voltage sensor trapping model (24,25), gating modifiers such as Hm-3 bind to a voltage sensor in its inward position and are thrown off by the voltage-driven outward movement of the sensor upon depolarization. The difference in binding affinity may result in different I-V curves. It was found that strong positive conditioning pulses provide the necessary energy to force the outward movement of the sensor into the activated position and thus provide the necessary energy to remove Hm-3 from its receptor site. However, the energy required to remove Hm-3 from Na V 1.4 is much greater than from DmNa V 1. We conclude that both Hm-3-bound channels are activated, but the activation is greatly retarded and positively shifted because of the energy required to dissociate Hm-3.
The observed differential activity of Hm-3 against Na V 1.4 and DmNa V 1 activation and steady-state inactivation (Fig. 5) may be interpreted accordingly. Hm-3 is hardly removed upon depolarization from Na V 1.4, and a preconditioning pulse is required. The Hm-3-bound channels are not activated, and this is observed by the strong reduction in Na ϩ conductance. The activation curve is basically formed by non-bound channels showing control characteristics (Fig. 5C). Moreover, the bound Hm-3 interferes with the inactivation process and hampers the steady-state inactivation (Fig. 5D). On the contrary, Hm-3 is more easily removed upon depolarization from DmNa V 1. The activation curve is therefore a superposition of bound and unbound channels (Fig. 5A). Similarly, the steady-state inactivation curve is less affected due to the easy dissociation of the toxin (Fig. 5B).
It has been demonstrated for ProTx-II (25,38), HWTX-IV (24), and JZTX-III (26) that they bind to receptor site 4. Judging by its activity, we suggest that Hm-3 might also bind to site 4. To unambiguously identify the potential site of interaction, structure-function experiments such as alanine scanning of Hm-3 together with site-directed mutagenesis of the expected site are needed.
One important issue about gating modifiers is their usually low specificity or promiscuity (39). As such, HWTX-IV not only affects Na V activation by binding to site 4 but also influences the inactivation by interacting with site 3 (40). In addition to Na V s, ProTx-I and -II also block T-type calcium channels (41), ProTx-I blocks TRPA1 (42), and JZTX-III was found to block voltage-gated potassium K V 2.1 channels (43). It is there-fore not surprising that Hm-3 shows quite a broad activity spectrum with respect to different Na V s.
Hm-3 Shares Low Homology but Similar Fold with Other Spider Neurotoxins-We previously isolated three peptides from the crab spider H. melloteei called Hm-1, Hm-2, and Hm-3 (27,44). The first two toxins were shown not to affect the activation or inactivation properties of sodium channels, thus presumably acting on receptor site 1. On the contrary, Hm-3, the subject of this study, shifted the voltage dependence of activation of Na V 1.4 to more depolarized voltages (27). A search for potential homologs revealed that a number of toxins show only limited similarity to Hm-3. Among them are Hm-2 (44), ␤/␦agatoxins from Agelena orientalis (45), -agatoxins from Agelenopsis aperta (46), and CSTX-17 from Cupiennius salei (47), and the level of their sequence similarity with Hm-3 is ϳ50 -60%. Interestingly, ␤/␦and -agatoxins also affect sodium channel activation albeit in a reversed manner: they promote channel activation similarly to scorpion ␤-toxins that bind to receptor site 4 (45). Normalized current is plotted as a function of time. Data are presented for Na V 1.4 (‚) and DmNa V 1 (ƒ). D and E, voltage dependence of reversal at ϩ100 (f), ϩ80 (F), and ϩ40 mV (OE) for Na V 1.4 (D) and DmNa V 1 (E); the protocol is shown in C. Normalized current is plotted against conditioning pulse duration. F and G, voltage dependence of activation of toxin-modified Na V 1.4 (F) and DmNa V 1 (G) channels. Peak test pulse currents after 700-ms prepulses to various potentials from the applied protocol as described in C were measured and normalized to the sodium peak current achieved at ϩ100 mV. These normalized peak currents are plotted as a function of prepulse potential (Ⅺ) and fitted with a Boltzmann relationship. Control activation curves (f) are shown for comparison. Voltages of half-maximal activation (V 1/2 ) are shown in the panels. Error bars represent S.E.
Despite low sequence similarity to other known peptides, our NMR studies demonstrate that Hm-3 adopts the ICK (or knottin) fold (Fig. 3A), the most widespread fold in spider toxins (2,48). Other sodium channel activation inhibitors as well as potassium channel activation inhibitors known from spiders (Figs. 3C and 7) also form the ICK fold. As a result of fold conservation, the disulfide-containing core of the Hm-3 molecule could be relatively well superimposed with ICK motifs from other spider toxins (Fig. 3C, upper row; root mean square deviation values calculated over the heavy atoms of six conserved Cys residues are in the range from 0.8 to 1.5 Å). Despite the Hm-3 overall fold being similar to that of other gating modifiers, notable differences can be spotted. Some loops that connect the cysteine residues contributing to the conventional ICK signature are of different length in Hm-3 and other ion channel activation inhibitors from spiders. (i) The most remarkable difference is loop 4 (between the last two cysteine residues, Cys V and Cys VI ), which is at least four residues longer in Hm-3. Consequently, the ␤-hairpin is larger in Hm-3 compared with other gating modifiers. (ii) The C terminus, on the contrary, is very short in Hm-3, containing only one residue following the last cysteine. Moreover, the distribution of physicochemical properties on the surface of Hm-3 is different from that in other toxins (see below).
Hm-3 Molecule Is Amphiphilic-All sodium channel activation inhibitors from spiders with a known three-dimensional structure present a Janus-faced molecular surface (23,42,49) (Fig. 3C), and analysis of Hm-3 spatial structure has shown that it is not an exception. One Hm-3 face contains a ridge of hydrophobic residues formed by Trp 11 , Phe 12 , Trp 16 , Tyr 25 , Ile 27 , and Leu 31 , whereas the opposite face comprises charged residues Lys 5 , Lys 7 , Glu 8 , Glu 15 , Lys 28 , Arg 29 , and Lys 32 (Fig. 3C). The hydrophobic residues forming the ridge are located in loops 2 (Cys II -Cys III ) and 4 (Cys V -Cys VI ) (Fig. 7). Other well studied ion channel activation inhibitors from spiders (Fig. 7) present a conserved pattern containing hydrophobic residues in loops 1 (Cys I -Cys II ) and 4 (Cys V -Cys VI ) and the C terminus that donate to the hydrophobic cluster (Fig. 3C). This pattern is also found in potassium channel activation inhibitors from spiders (50 -52) (Fig. 7) but differs from the distribution of key hydrophobic residues in the sequence of Hm-3. This leads to a completely different localization of the hydrophobic clusters on the Hm-3 surface as compared with other spider toxins. Indeed, superimposition of the toxin molecules by the conserved ICK motif revealed that the hydrophobic face of Hm-3 lies on the opposite side to the hydrophobic faces of "classical" gating modifiers (Fig. 3C).
Hm-3 Action Is Membrane-mediated-Experiments on Hm-3 partitioning into lipid membranes showed the toxin affinity for liposomes, which is characteristic of ion channel activation inhibitors from spiders (Figs. 2B and 3D). Hm-3 bound to both neutral and negatively charged lipid vesicles, and the affinity was high (Table 3). To compare, (i) VsTx1 does not bind to POPC in 150 mM NaCl (K p Ͻ 0.002 ϫ 10 3 M Ϫ1 ) (52), although it binds to 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine/POPG (3:1) in 150 mM KCl (K p ϳ2 ϫ 10 3 M M Ϫ1 ), and POPC in 150 mM NaCl (K p ϳ9.5 ϫ 10 3 M Ϫ1 ). The fact that Hm-3 binding to charged liposomes moderately depends on salt concentration indicates that electrostatic interactions are involved in the formation of the Hm-3⅐membrane complex only partially. At the same time, a comparatively high Hm-3 affinity to both charged and neutral liposomes under physiologically relevant conditions (in salt solution) means that hydrophobic interactions play the major role in the binding.
It is commonly hypothesized that the observed affinity of ion channel activation inhibitors to lipids underlies their mode of action, the so-called "membrane access" mechanism (53). According to this hypothesis, the toxin first binds to the cell membrane with its hydrophobic patch, then drifts to its target (voltage-gated ion channel), and finally binds to a voltage sensor, thereby freezing the channel in the resting state. For a long time, the S3-S4 linker of domain II (site 4) was thought to represent the sole binding site for activation inhibitors, but new data demonstrate that the situation is more complicated and that voltage sensors from other domains could also be involved (56,57). Taking all the presented data on Hm-3 activity into consideration, we believe that it adds to the diversity of spider venom peptides inhibiting sodium channels and featuring the membrane access mode of action.
Evolutionary Considerations-We note that whereas H. melloteei belongs to the Araneomorphae suborder all other spider toxins with a known spatial structure that inhibit sodium channel activation (see Fig. 7) were purified from species belonging to the Mygalomorphae suborder. The following possibilities may be considered. (i) "Pure" divergence may have occurred, i.e. mutation of a primordial gating modifier toxin gene that originated in a common ancestor of mygalo-and araneomorphs. In this case, it is hard to explain the marked difference in the amino acid sequence and the location of the hydrophobic cluster in Hm-3 and gating modifiers from mygalomorphs. (ii) "Pure" convergence may have occurred, i.e. independent evolution of ICK toxins targeting Na V s in mygalo-and araneomorphs. However, spider ICK toxins are currently believed to have a common ancestor. (iii) We therefore propose that parallel evolution took place. Based on our current grasp of spider toxin diversity, we assume the following evolutionary scenario: common ancestor of ICK toxins 3 divergence of ICK toxins in Mygalomorphae and Araneomorphae 3 parallel evolution of ion channel gating modifiers featuring the membrane access mechanism.