Solution Structure of Kurtoxin: A Gating Modifier Selective for Cav3 Voltage-Gated Ca²⁺ Channels

Abstract

Kurtoxin is a 63-amino acid polypeptide isolated from the venom of the South African scorpion Parabuthus transvaalicus. It is the first and only peptide ligand known to interact with Cav3 (T-type) voltage-gated Ca²⁺ channels with high affinity and to modify the voltage-dependent gating of these channels. Here we describe the nuclear magnetic resonance (NMR) solution structure of kurtoxin determined using two- and three-dimensional NMR spectroscopy with dynamical simulated annealing calculations. The molecular structure of the toxin was highly similar to those of scorpion α-toxins and contained an α-helix, three β-strands, and several turns stabilized by four disulfide bonds. This so-called “cysteine-stabilized α-helix and β-sheet (CSαβ)” motif is found in a number of functionally varied small proteins. A detailed comparison of the backbone structure of kurtoxin with those of the scorpion α-toxins revealed that three regions [first long loop (Asp⁸–Ile¹⁵), β-hairpin loop (Gly³⁹–Leu⁴²), and C-terminal segment (Arg⁵⁷–Ala⁶³)] in kurtoxin significantly differ from the corresponding regions in scorpion α-toxins, suggesting that these regions may be important for interacting with Cav3 (T-type) Ca²⁺ channels. In addition, the surface profile of kurtoxin shows a larger and more focused electropositive patch along with a larger hydrophobic surface compared to those seen on scorpion α-toxins. These distinct surface properties of kurtoxin could explain its binding to Cav3 (T-type) voltage-gated Ca²⁺ channels.

Voltage-gated ion channels are expressed by nearly all cells and play a crucial role in regulating membrane potential and a variety of cellular functions. These channels are comprised of two principle domains: a central pore domain formed by two segments, S5 and S6, and four surrounding voltage-sensing domains, each composed of segments S1–S4.¹⁻⁴ Venomous animals (spiders, scorpions, and cone snails, among others) produce a broad array of polypeptide toxins, many of which bind to voltage-gated Na⁺, K⁺, or Ca²⁺ channels,⁵⁻⁸ and have proven to be valuable pharmacological tools for evaluating specific channel characteristics.

Although the origins of the venomous peptide toxins that interact with voltage-gated ion channels are diverse, their modes of action fall into two major categories, pore blockade and gating modification, based on the domain with which they interact and their mechanisms of action. Pore blockers bind to the external vestibule of the channel and physically obstruct the movement of ions by occluding the ion-conducting pore.⁹ The three-dimensional structures of many pore-blocking toxins, including the μ-conotoxins for the Na⁺ channel, charybdotoxin for the K⁺ channel, and ω-conotoxins for the Ca²⁺ channel, have all been determined, allowing investigation of the structure–function relationships of the pore-forming domains of the channels.¹⁰⁻¹⁵ Gating modifiers, on the other hand, bind to the voltage-sensing domains of voltage-gated ion channels and modify the energetics of either activation or inactivation.¹⁶⁻²¹ Established gating modifiers include the α- and β-scorpion toxins, sea anemone toxins, and δ-conotoxins for Na⁺ channels;²²⁻²⁸ hanatoxin (HaTx), SGTx1, GxTx-1E, and VSTx for K⁺ channels;²⁹⁻³³ and ω-agatoxin IVA (ω-Aga IVA) and ω-grammotoxin SIA (GrTx) for Ca²⁺ channels.³⁴⁻³⁷

Studies of the structure and function of gating modifiers have advanced our understanding of the molecular structures and gating mechanisms of voltage-gated ion channels. Studies employing HaTx, SGTx1, GxTx-1E, VSTx, GrTx, and ω-Aga IVA have revealed that, within voltage-gated K⁺ and Ca²⁺ channels, these proteins bind to structurally conserved motifs composed of hydrophobic and acidic residues within the C-terminal end of S3 and the N-terminal end of S4.^{17−19,38−42} The X-ray structures of voltage-gated K⁺ channels (i.e., KvAP and Kv1.2 channels) show that these regions of S3 and S4 form a helix–turn–helix motif termed the voltage sensor paddle.^4,43,44 Gating modifier toxins that partition into the membrane interact with the voltage sensor paddle at the protein–lipid interface.^42,45

Cav3 (T-type) voltage-gated Ca²⁺ channels can be differentiated from other types of Ca²⁺ channels on the basis of their activation at lower voltages, faster inactivation, slower deactivation, and smaller Ba²⁺ conductances.⁴⁶ Their unique gating properties allow Cav3 (T-type) Ca²⁺ channels to trigger low-threshold spikes that can lead to burst firing and oscillatory behavior and can contribute to standing calcium currents near the resting membrane potential in a variety of cell types.⁴⁷⁻⁵⁰ Although these characteristics imply Cav3 (T-type) channels could play important roles in many tissues,^51,52 progress in understanding their subunit composition and physiological functions has been hindered by a scarcity of ligands that interact with these channels.^53,54

A gating modifier kurtoxin, isolated from the venom of the scorpion Parabuthus transvaalicus, is the first peptide ligand known to act on Cav3 (T-type) voltage-gated Ca²⁺ channels.^55,56 Here, we describe the solution structure of kurtoxin determined using proton two-dimensional (2D) and heteronuclear three-dimensional (3D) NMR spectroscopy with dynamical simulated annealing calculations. The structure of kurtoxin closely resembles those of scorpion α-toxins (Figure 1) targeting Na⁺ channels and shows the unique structural characteristic of gating modifiers, electropositive and hydrophobic patches on the surface of the molecule.^39,57,58 Detailed inspection of the structure of kurtoxin offers the possibility of understanding the molecular basis of its Cav3 (T-type) Ca²⁺ channel selectivity and could facilitate clarification of the gating mechanism of voltage-gated ion channels.

Figure 1.

Amino acid sequences and alignment of kurtoxin and five scorpion α-toxins. LqhαIT and Lqq III are highly active in insects. Lqh III and Bmk M1 are α-like toxins. AaH II is highly active in mammals. These sequences were aligned using ClustalX. Highly conserved residues are shaded in black or gray. The secondary structure elements of kurtoxin are shown as arrows (β-strand), bars (α-helix), and lines (connecting loops).

Experimental Procedures

Sample Preparation

Functional kurtoxin was obtained using a bacterial expression system.⁵⁹ The recombinant kurtoxin was expressed as an inclusion body, solubilized in a denaturing solution, and then refolded in a refolding solution. The crude folded kurtoxin was purified by preparative RP-HPLC, after which the purity of the recombinant protein was confirmed by analytical RP-HPLC and MALDI-TOF MS measurements.

CD Spectral Analysis

CD spectra were recorded on a JASCO J-750 spectropolarimeter in a solution containing 0.01 M sodium phosphate in H₂O and 0, 10, 15, 20, 25, or 30% CH₃CN at pH 7.0. Measurements were taken at 20 °C using a quartz cell with a 1 mm path length. The spectra were expressed as molecular ellipticity [θ] in degrees square centimeter per decimole.

NMR Spectroscopy

NMR measurements were taken on a Bruker AVANCE 600 spectrometer equipped with an xyz gradient triple-resonance probe. The samples used for proton 2D NMR experiments were 1 mM kurtoxin dissolved in water containing 25% CD₃CN at pH 4.0 (uncorrected for the isotope effect). All proton 2D NMR spectra were recorded in a phase-sensitive mode using time-proportional phase incrementation (TPPI) for quadrature detection in the t₁ dimension at 278, 288, and 298 K. TOCSY spectra were recorded using a MLEV-17 pulse scheme⁶⁰ with isotropic mixing times of 60 and 90 ms. NOESY spectra⁶⁰⁻⁶² were recorded with mixing times of 60, 100, and 150 ms. Suppression of the solvent resonance in both the NOESY and TOCSY measurements was achieved using the WATERGATE scheme.⁶³ E-COSY⁶⁴ spectra were recorded to obtain the constraints for stereospecific assignments.

The following triple-resonance 3D NMR spectra were recorded using 1 mM ¹³C- and ¹⁵N-labeled or 1 mM ¹⁵N-labeled kurtoxin in 25% CD₃CN at 278 and 288 K. Uniformly ¹³C- and ¹⁵N-enriched kurtoxin was used to record 3D HNCACB, CBCA(CO)NH, HNCA, HN(CO)CA, HNCO, and HN(CA)CO spectra.⁶⁵⁻⁶⁸ Uniformly ¹⁵N-enriched kurtoxin was used to record 2D ¹H–¹⁵N HSQC⁶⁹ and 3D ¹⁵N TOCSY-HSQC spectra with 90 ms mixing times and a 3D ¹⁵N NOESY-HSQC spectrum with a 120 ms mixing time. The ³J_HN–H^α values were obtained from the 3D HNHA⁷⁰ and 2D DQF-COSY spectra. Slowly exchanging backbone amide protons were identified by analysis of TOCSY spectra recorded in 75% D₂O and 25% CD₃CN on time scales of 30 min and 3.5 h and then every 3 h up to 25 h. ¹H chemical shifts were referenced to DSS at 0 ppm, and ¹³C and ¹⁵N chemical shifts were calculated from the ¹H frequency. All spectra were processed using AZARA version 2.5 (provided by W. Boucher) or XWIN-NMR and were analyzed using ANSIG version 3.3⁷¹ on a Silicon Graphics Octane2 workstation or on a Linux workstation.

NMR Experimental Restraints and Structure Calculations

The backbone NH–C^αH coupling constants were converted to backbone torsion angle ϕ constraints according to the following rules: for a ³J_NH–C^αH of <5.5 Hz, the ϕ angle was constrained in the range of −65 ± 25°; for a ³J_NH–C^αH of >8.0 Hz, it was constrained in the range of −120 ± 40°.^72,73 Backbone dihedral constraints were not applied to ³J_NH–C^αH values between 5.5 and 8.0 Hz. The range of the χ¹ side chain torsion angle constraints and the stereospecific assignment of the prochiral β-methylene protons were obtained using the ³J_αβ coupling constants combined with the intraresidue NH–C^βH NOEs.⁷⁴ The ³J_αβ coupling constants were determined from the E-COSY spectrum in D₂O. For the t²g³, g²g³, and g²t³ conformations around the C^α–C^β bonds, the χ¹ side chain torsion angle was constrained in the ranges of −60 ± 30°, 60 ± 30°, and 180 ± 30°, respectively.⁷⁵

Quantitative determination of the cross-peak intensities was based on the counting levels. Observed NOE data were classified into four distance ranges (1.8–2.7, 1.8–3.5, 1.8–5.0, and 1.8–6.0 Å) that corresponded to strong, medium, weak, and very weak NOE values, respectively. Pseudoatoms were used for the methyl protons or the nonstereospecifically assigned methylene protons.⁷⁶ Correcting factors for the use of pseudoatoms were added to the distance constraints, and 0.5 Å was added to the distance constraints involving methyl protons.⁷⁷ For each disulfide bond, we used three distance constraints, S(i)–S(j), S(i)–C^β(j), and S(j)–C^β(i), whose target values were set to 2.02 ± 0.02, 2.99 ± 0.5, and 2.99 ± 0.5 Å, respectively.⁷⁸ The hydrogen bond acceptors for the slowly exchanged amide protons were identified by analyzing the preliminarily calculated structures.^79,80 The distance restraints on the hydrogen bonds were added as target values of 1.8–2.3 Å for NH(i)–O(j) and 2.8–3.3 Å for N(i)–O(j) bonds.

All calculations were conducted on an SGI Octane2 workstation using the X-PLOR version 3.851.⁸¹ The three-dimensional structures were calculated on the basis of the experimentally derived distance and torsion angle constraints using a dynamically simulated annealing protocol starting from a template structure with randomized backbone ϕ and ψ torsion angles. The final 20 structures with the lowest energy and smallest Lennard-Jones van der Waals energy were chosen. The convergence of the calculated structures was evaluated in terms of the structural parameters. There were root-mean-square deviations (rmsds) from the experimental distances and dihedral constraints, from the energetic statistics (F_NOE, F_tor, F_repel, and E_L-J), and from the idealized geometry. The structures were analyzed using the PROCHECK_NMR,⁸² PROMOTIF,⁸³ MOLMOL,⁸⁴ and MolProbity.^85,86 The distributions of the backbone dihedral angles in the final converged structure were evaluated by representation of the Ramachandran dihedral pattern, which indicated the deviations from the allowed ϕ and ψ angle limits. The degrees of angular variation among the converged structures were further assessed using an angular order parameter.⁸⁷ The solvent-accessible surface areas for the side chains of the amino acid residues were calculated with a solvent radius of 1.4 Å. Structural figures were generated using MOLMOL and INSIGHT II 2000 (Accelrys Inc.).

Results and Discussion

NMR Sample Preparation

Kurtoxin was insoluble in aqueous solution, even at a concentration of <1 mM. We therefore tested whether kurtoxin could be solubilized using 0–30% acetonitrile (CH₃CN) without disrupting the protein’s intrinsic structure. Figure 2 shows that the CD spectra for kurtoxin recorded in the absence or presence of 10, 15, 20, 25, and 30% CH₃CN are nearly identical in the far-UV region (200–250 nm), indicating that the backbone structure of kurtoxin is affected little by the addition of CH₃CN. Moreover, we obtained high-quality NMR spectra with 1 mM kurtoxin in the presence of 25% deuterated acetonitrile (CD₃CN), strongly suggesting that the addition of CD₃CN effectively prevented the aggregation of kurtoxin (Figure 3). In the ¹H–¹⁵N HSQC spectra recorded in pure water, several hydrophobic peaks, including Tyr¹⁷, Tyr¹⁸, Trp³⁸, Gly³⁹, Trp⁴⁰, and Leu⁴², could not be detected, most likely because of the line broadening in the absence of CD₃CN (Figure 3). This appeared to be due to the self-aggregation of kurtoxin molecules. On the other hand, in the ¹H–¹⁵N HSQC spectra recorded in the presence of 25% CD₃CN, all of the backbone ¹H–¹⁵N cross peaks were identifiable, and the intensities of nearly all the peaks were increased without large chemical shift variations (Figure S1 of the Supporting Information).

Figure 2.

CD spectra of 0.05 mM kurtoxin in H₂O containing 0, 10, 15, 20, 25, or 30% CH₃CN [0.01 M sodium phosphate (pH 7.0)] at 20 °C.

Figure 3.

¹H–¹⁵N HSQC spectra of kurtoxin in the presence (black) and absence (green) of 25% CD₃CN at 288 K. The resonance assignments are indicated by the one-letter amino acid codes and residue numbers. The dotted circles indicate additional and unassigned peaks.

Addition of CH₃CN can weaken the hydrophobic interactions between surface hydrophobic regions without altering the three-dimensional structure of the proteins.⁸⁸ Nonetheless, CH₃CN can influence chemical shift variations by changing the polarity of the accessible atom environments; of those, the chemical shifts located in the flexible N- and C-terminal regions (Ala⁶³ of kurtoxin in Figure S1 of the Supporting Information) and loops are most sensitive.⁸⁹ Comparison of the CD and ¹H–¹⁵N HSQC spectra recorded in the presence and absence of 25% CD₃CN indicates that CD₃CN prevented the hydrophobic-related aggregation of kurtoxin without disrupting its overall topology. Linear kurtoxin was also refolded in a redox solution containing 30% CH₃CN,⁵⁹ suggesting the nativelike conformation of kurtoxin is stabilized in an aqueous solution containing CH₃CN.

A small number of additional minor peaks were observed in NMR spectra, including ¹H–¹⁵N HSQC spectra in the presence and absence of CD₃CN (Figure 3). The intensity ratio of the minor peaks is <5–10% compared to that of the major peaks according to relevant NMR signals. Unfortunately, we failed to assign these minor peaks because the signals were very weak and it was hard to sequentially connect these minor peaks. In a recent study, two stable conformations of BmκαTx11, a kurtoxin-homologous scorpion α-toxin, were identified by using NMR spectroscopy.⁹⁰ We assume that the minor peaks of kurtoxin also come from a possible minor conformation similar to BmκαTx11.

NMR Spectroscopy

Sequence-specific assignments for the backbone atoms of kurtoxin were obtained from an analysis of heteronuclear 3D NMR spectra [HNCACB, CBCA(CO)NNH, HNCA, and HN(CO)CA] recorded in 25% CD₃CN at pH 4.0 and 288 and 278 K using a uniformly ¹⁵N- and ¹³C-labeled protein. All backbone C^α atoms were assigned except those of Cys¹², Arg¹⁴, Ile¹⁵, Asp²⁵, Leu²⁶, Ala³², and Trp⁴⁰. The backbone and side chain protons were obtained from 2D NMR spectra (DQF-COSY, TOCSY, and NOESY) and 3D ¹⁵N NOESY-HSQC and 3D ¹⁵N TOCSY-HSQC spectra recorded in 25% CD₃CN at pH 4.0 and 288 K. The assignments of all backbone and side chain protons were complete except for those of the H^ζ atoms of Lys¹³ and Lys²¹, H^δ2 of Asn¹⁹ and Asn²⁰, and H^δ and H^ε of Tyr²². For all Pro residues (Pro⁶ and Pro⁵⁰), strong sequential d_α,δ and no d_α,α were observed in the NOESY spectra, indicating the toxin’s proline residues are all in the trans configuration.

Identification of Secondary Structure Elements

As summarized in Figure S2 of the Supporting Information, the pattern of observed NOEs and chemical shift index (CSI) values for H^α was ultimately interpreted in terms of the secondary structure of the molecule. The weak ³J_HN^α coupling constants, strong d_NN NOE peaks and d_αN(i,i+3) and d_αβ(i,i+3) NOE correlations, and the CSI value of −1 all indicate an α-helical conformation for residues Lys²¹–Leu³⁰. The extent of the β-strands and their relative orientations within the β-sheet structure were determined using standard criteria: large ³J_HN^α coupling constants (Ile², Asp³, Try³⁶, Cys³⁷, Tyr⁴⁵, Cys⁴⁶, and Gln⁴⁷), strong sequential d_αN, interstrand NH–NH and NH–C^αH connectivities, and slowly exchanging amide protons (Ile², Gly⁴, Gly³⁵, Tyr³⁶, Cys⁴⁴, Tyr⁴⁵, Cys⁴⁶, Gln⁴⁷, and Leu⁴⁹). Kurtoxin contains three β-strands comprised of residues Ile²–Gly⁴, Ser³⁴–Cys³⁷, and Cys⁴⁴–Leu⁴⁹, which are arranged in an antiparallel fashion with several turns (Figure S3 of the Supporting Information). Our criteria allowed discrimination of the peripheral and central strands within the β-sheet.

Structure Calculations

The structure of kurtoxin was determined from a total of 894 NMR experimental constraints, including 856 experimental distance constraints and 38 dihedral angle constraints, which correspond to an average of 14.2 constraints per residue. Of the 861 distance constraints, there were 311 intraresidue and 492 interresidue NOE distance constraints, 42 hydrogen bond constraints determined from hydrogen–deuterium exchange-out experiments, and 12 disulfide bond constraints. The 42 distance constraints related to hydrogen bonds were as follows: I2(HN)–L49(CO), G4(HN)–C46(CO), D25(HN)–K21(CO), C27(HN)–C23(CO), K28(HN)–N24(CO), G29(HN)–L26(CO), L30(HN)–C27(CO), K31(HN)–K28(CO), A32(HN)–C27(CO), G35(HN)–D33(CO), Y36(HN)–Y45(CO), W38(HN)–S43(CO), C44(HN)–I15(CO), Y45(HN)–Y36(CO), C46(HN)–G4(CO), Q47(HN)–S34(CO), L49(HN)–I2(CO), A53(HN)–P50(CO), K56(HN)–Y5(CO), R62(HN)–W10(CO), and A63(HN)–C61(CO). The disulfide bond pattern of kurtoxin was determined to be Cys¹²–Cys⁶¹, Cys¹⁶–Cys³⁷, Cys²³–Cys⁴⁴, and Cys²⁷–Cys⁴⁶, based on sequential cleavage with proteases and MALDI-TOF MS measurements.⁵⁹

We conducted the simulated annealing calculations starting with 100 random kurtoxin structures. From those, we selected 20 final structures (Figure 4A) that were in good agreement with the NMR experimental constraints (NOE distance and torsion angle violations of <0.2 Å and <2°, respectively). Statistics for the converged structures were evaluated in terms of the structural parameters (Table 1). The deviations from the idealized covalent geometry were very small, and the Lennard-Jones van der Waals energy was large and negative (−212.55 ± 19.40), indicating there were no distortions or nonbonded bad contacts in the converged structures. The atomic rmsd about the mean coordinate positions was 0.87 ± 0.14 Å for the backbone atoms (N, C^α, and C) and 1.54 ± 0.19 Å for all heavy atoms. Ramachandran analysis showed that 98.7% of all residues fell within allowed regions.

Figure 4.

Solution structure of kurtoxin. (A) Stereopairs of backbone heavy atoms (N, C^α, and C) for the 20 converged structures of kurtoxin. These are the results of the best-fit superposition of the backbone heavy atoms of the molecule. N and C indicate N- and C-terminal positions, respectively. (B) Ribbon structure of kurtoxin. Schematic diagram of kurtoxin illustrating the location of the β-strands (cyan), α-helix (red and yellow), and disulfide bonds (numbered ball and stick). Long loop I and long loop II correspond to residues Tyr⁵–Asn²⁰ and Pro⁵⁰–Ala⁶³, respectively. (C) Tubular representation of kurtoxin illustrating its motional properties. The diameter of the tube is proportional to the atomic rmsds of the backbone atoms: red for residues whose backbone rmsds are >1.0, cyan for those with β-strands, and yellow for those with an α-helix. (D) Surface profile of kurtoxin. The surface hydrophobic patches and charged residues are indicated. The molecular surface of kurtoxin is shown in color according to the electrostatic potential: red for negatively charged amino acids, blue for positively charged amino acids, and white for uncharged or hydrophobic amino acids. The left and right figures are rotated 180° with respect to one another about a vertical axis. These figures were generated using MOLMOL.

Table 1. Structural Statistics for the 20 Lowest-Energy Structures of Kurtoxin^a.

rmsd from experimental distance constraints (Å)b (856)	0.0115 ± 0.0012
rmsd fom experimental dihedral constraints (deg)b (38)	0.1555 ± 0.0012
energetic statistics (kcal/mol)c
FNOE	5.7261 ± 1.2582
Ftor	0.9845 ± 0.0964
Frepel	5.3839 ± 1.3992
EL-J	–212.55 ± 19.4000
rmsd from idealized geometry
bonds (Å)	0.0017 ± 0.0001
angles (deg)	0.5010 ± 0.0065
impropers (deg)	0.3391 ± 0.0060
Ramachandran analysisd (%)
most favored regions	56.0
additionally allowed regions	36.1
generously allowed regions	6.5
disallowed regions	1.3
average rmsd (Å)
backbone (N, Cα, C)	0.87 ± 0.14
all heavy atoms	1.54 ± 0.19
MolProbity analysise
Clash score	94.39 ± 9.94
MolProbity score	4.37 ± 0.08

^aNone of these 20 structures exhibited distance violations of >0.2 Å or dihedral angle violations of >2°.

^bThe number of each experimental constraint used in the calculations is given in parentheses.

^cF_NOE, F_tor, and F_repel are the energies related to the NOE violations, the torsion angle violations, and the van der Waals repulsion term, respectively. The values of the force constants used for these terms are the standard values as depicted in the X-PLOR 3.1 manual. E_L-J is the Lennard-Jones van der Waals energy calculated with the CHARMm empirical energy function.¹⁰²E_L-J was not used in the dynamical simulated annealing calculations.

^dPROCHECK_NMR was used to assess the stereochemical quality of the structures.

^eThe MolProbity webserver was used to evaluate the determined kurtoxin ensemble structures.

Structure Description

The molecular structure of kurtoxin has a compact core consisting of an α-helix and a triple-stranded antiparallel β-sheet stabilized by four disulfide bridges. The 2.5-turn α-helix is composed of residues extending from Lys²¹ to Leu³⁰ and is linked to the central strand of the β-sheet by two disulfide bridges (Cys²³–Cys⁴⁴ and Cys²⁷–Cys⁴⁶) (Figure 4B). The three β-strands are formed by residues Ile²–Gly⁴ (β-strand I), Ser³⁴–Cys³⁷ (β-strand II), and Cys⁴⁴–Leu⁴⁹ (β-strand III), with residues Ile², Gly⁴⁸, and Leu⁴⁹ involved in a β-bulge conformation. The last two residues of β-strand III (Gly⁴⁸ and Leu⁴⁹) form a classical β-bulge and interact with the first residue of β-strand I (Ile²). As a result, the ϕ angle of Gly⁴⁸ is positive (105.6°), which causes a distortion in the β-sheet. Two long loops (loop I, Tyr⁵–Asn²⁰; loop II, Pro⁵⁰–Ala⁶³) extend from the core (Figure 4B). Long loop I includes two type IV β-turns (Asp⁸–Asn¹¹ and Asn¹¹–Arg¹⁴) and adopts the positive ϕ angles (58.9° and 89.9°, respectively) of Asn¹¹ and Arg¹⁴ in the average structure. Long loop I is connected to β-strand II by a disulfide bond (Cys¹⁶–Cys³⁷) and to the C-terminus by a disulfide bond (Cys¹²–Cys⁶¹) and a hydrogen bond [R62(HN)–W10(CO)]. Long loop II starts with a well-defined type I β-turn structure (residues 50–53) and extends to the C-terminus. It is stabilized through hydrogen bonding [N11(HD21)–S58(CO), K56(HN)–Y5(CO), and R62(HN)–W10(CO)] and formation of a disulfide bridge (Cys¹²–Cys⁶¹) to long loop I. The short loop regions (Lys¹³–Ile¹⁵ and Trp⁴⁰–Thr⁴¹) and the six C-terminal residues (Ser⁵⁸–Ala⁶³) are less defined in the final 20 structures than the other regions in kurtoxin (Figure 4C). This may reflect a lack of medium- and long-range NOE constraints due to the inherent flexibility of these regions.

Two surface hydrophobic patches were observed in the kurtoxin structure. The major patch is composed of the solvent-exposed side chains of Tyr⁵, Tyr³⁶, Trp³⁸, Trp⁴⁰, Leu⁴², and Tyr⁴⁵ (Figure 4D), which, except for Tyr⁵, are located in the hairpin structure (strands II and III). Their side chains are well stacked, creating a compact hydrophobic patch on the surface of the molecule. The minor hydrophobic patch is situated on the opposite side of the protein and is centered around Trp¹⁷, Tyr¹⁸, Tyr²², and Leu²⁶ (Figure 4D). Interestingly, all line-broadened residues (including Tyr¹⁷, Tyr¹⁸, Trp³⁸, Gly³⁹, Trp⁴⁰, and Leu⁴²) in pure water are located within either the major or minor surface hydrophobic patch in kurtoxin, strongly suggesting that the aggregation of kurtoxin results from these hydrophobic surface properties of kurtoxin. All of the charged residues except Asp⁸ are highly exposed on the surface of kurtoxin. Several charged residues, including Asp³, Lys²¹, and Lys⁵⁶, surround the major surface hydrophobic patch, while Lys¹³, Arg¹⁴, and Asp²⁵ are situated near the minor hydrophobic patch (Figure 4D).

Structural Comparison of Kurtoxin with Scorpion α-Toxins

The amino acid sequence of kurtoxin is homologous with those of the scorpion α-toxins targeting Na⁺ channels (Figure 1), which is consistent with the finding that kurtoxin can also interact with voltage-gated Na⁺ channels.⁵⁵ The three-dimensional structures of the scorpion α-toxins have been determined by ¹H 2D NMR spectroscopy and X-ray crystallography.⁹¹⁻¹⁰¹ Their overall folds are remarkably similar to that of kurtoxin and consist of an α-helix and three β-strands stabilized by four disulfide bridges, which has been termed the CSαβ (cysteine-stabilized α-helix and β-sheet) motif.¹⁰² In addition, like kurtoxin, the α-helices of the scorpion α-toxins contain 2.5–3 helical turns and are connected to the central β-stand by two disulfide bridges.

Figure 5 shows the geometric average backbone structure from the 20 NMR kurtoxin models superimposed on the backbones of AaH II (antimammal α-toxin), LqhαIT (anti-insect α-toxin), and BmK M1 (α-like toxin). Comparison of the kurtoxin backbone with those of the scorpion α-toxins clearly highlights three regions of structural difference: the first long loop region (Asp⁸–Ile¹⁵), the β-hairpin loop (Gly³⁹–Leu⁴²), and the C-terminal segment (Arg⁵⁷–Ala⁶³). The structural differences among these regions are strongly correlated with a marked difference in sequence (Figure 1). Among the eight residues extending from Asp⁸ to Ile¹⁵, only two (Ans¹¹ and Cys¹²) are conserved in the amino acid sequences of these toxins. The 9th and 10th residues are variably polar or hydrophobic in the scorpion α-toxins, with a non-proline cis peptide bond in BMK M1.⁹⁹ On the other hand, they are bulky hydrophobic residues (Tyr⁹ and Trp¹⁰) with a common trans peptide bond in kurtoxin. In addition, the structurally well-defined hydrophobic residues Val¹³ and Tyr¹⁴ found in LqhαIT and BmK M1 are replaced with disordered positively charged residues Lys¹³ and Arg¹⁴, respectively, in kurtoxin.

Figure 5.

Stereopairs showing the superposition of the kurtoxin structure on scorpion α-toxin structures. The backbone (C, C^α, and N) atoms of kurtoxin are superimposed on those of the scorpion α-toxins. The top panel shows the superposition of the backbone of kurtoxin on that of the anti-mammal α-toxin AaH II (PDB entry 1PTX). The middle panel shows the superposition on the backbone of the anti-insect α-toxin LqhαIT (PDB entry 1LQH). The bottom panel shows the superposition on the backbone of the α-like toxin Bmk M1 (PDB entry 1DJT). The backbone structures of kurtoxin and scorpion α-toxins are shown as thick and thin lines, respectively. N and C indicate the N- and C-terminal positions, respectively. Labeling shows the kurtoxin residues in regions of structural difference between the two backbones. The backbone rmsd values are 3.13, 2.98, and 2.62 Å for AaH II, LqhαIT, and Bmk M1, respectively.

The sequence of the hairpin loop (Gly³⁹–Leu⁴²) also differs between kurtoxin and the scorpion α-toxins. The length of the loop in kurtoxin (four residues) is shorter than in other toxins (approximately seven residues). In addition, whereas the loop is disordered and involved in the formation of the major hydrophobic patch in kurtoxin, it protrudes from the CSαβ core and turns toward the C-terminal segments in the scorpion α-toxins (Figure 5). Because the C-terminal segments are disordered in both kurtoxin and scorpion α-toxins, it is difficult to assess structural differences in that region. However, site-directed mutagenesis studies and functional assays of scorpion α-toxins have shown that there is a functional site composed of the five-residue reverse turn (Asp⁸–Cys¹²) and the C-terminal segment, and that the conserved hydrophobic surface may be involved in maintaining the stability of the protein and its biological activity.^103−107 Taken together, these findings indicate that the core region of kurtoxin (i.e., the CSαβ motif) is well-defined and superimposes well on those of the scorpion α-toxins, but the Asp⁸–Ile¹⁵, Gly³⁹–Leu⁴², and C-terminal segments of kurtoxin are structurally different from those of the scorpion α-toxins, suggesting it is these regions that are responsible for the functional differences between kurtoxin and scorpion α-toxins.

Comparison of the Surface Profiles of Kurtoxin and Scorpion α-Toxins

Kurtoxin contains five negatively charged and 11 positively charged residues in its amino acid sequence (Figure 1), and all of these charged residues except Asp⁸ are highly exposed on the water-accessible surface of the molecule. The side chain oxygen of Asp⁸ in AaH II forms a hydrogen bond with the amide proton of Val¹⁰, and the side chain of Gln⁸ in Lqh III forms a hydrogen bond with the oxygen of Val¹³.¹⁰⁰ Asp⁸ of kurtoxin is directed toward Lys¹³ and Arg¹⁴ (Figure 6A). Although there are no experimental data for the hydrogen bond interactions between Asp⁸ and any other residues in kurtoxin, some side chain oxygens of Asp⁸ in 20 ensemble structures are close enough to form hydrogen bonds with Lys¹³ and/or Arg¹⁴ in the determined kurtoxin structures. Along the α-helix, the positive and negative charges align toward the solvent-accessible region of the molecule in both scorpion α-toxins and kurtoxin (Figure 6), suggesting that this feature may be involved in ion channel binding and determining selectivity.¹⁰⁰ Kurtoxin is highly basic, as compared to the α-scorpion toxins; the net charge of kurtoxin is +6, while the others have net charges ranging from −2 to +3. As shown in Figure 1, the two hydrophobic residues (Val¹³ and Tyr¹⁴) conserved in all scorpion α-toxins except AaH II are replaced with two positively charged residues (Lys¹³ and Arg¹⁴, respectively) in kurtoxin. Figure 6B shows that the side chains of Val¹³ and Tyr¹⁴ in Lqq III (an anti-insect α-toxin) are largely buried in the molecular core. By contrast, Lys¹³ and Arg¹⁴ in kurtoxin are exposed to solvent (Figure 6A) and form a local electropositive surface (Figure 4D). In addition, a large electropositive patch (surface area of 660 Ų) is formed by the five positively charged C-terminal residues (Arg⁵⁴, Lys⁵⁶, Arg⁵⁷, Arg⁶⁰, and Arg⁶²). This is in contrast to the C-terminal structure of Lqq III, which contains only two positively charged residues (Arg⁵⁸ and Lys⁶²). The water-exposed, positively charged residues of kurtoxin form a distinctive large electropositive surface, which is located around the five-residue reverse turn and C-terminal segment and is the proposed Na⁺ channel binding site in scorpion α-toxins.¹⁰⁴

Figure 6.

Ribbon diagrams and heavy atom side chains of kurtoxin (A) and Lqq III (B). The surface hydrophobic patches and charged residues are indicated: red for the negatively charged amino acids are colored, blue for the positively charged amino acids, purple for the major surface hydrophobic amino acids, and orange for the minor surface hydrophobic amino acids. The left and right figures are rotated 180° relative to one another about a vertical axis.

A surface hydrophobic patch is a conserved feature of all scorpion α-toxins and is involved in mediating their interaction with Na⁺ channels.^108,109 The orthogonal arrangement of the aromatic side chains in the surface hydrophobic patch, termed a “herringbone” structure, is found in all scorpion α-toxins and has been identified as the lowest-energy configuration of solvent-exposed aromatic rings.¹¹⁰ There are two hydrophobic patches in scorpion α-toxins, a major patch commonly composed of five residues (Tyr⁵, Tyr³⁵, Tyr⁴², Trp⁴⁷, and Tyr⁴⁹) (Figure 6B) with a surface area of ∼280 Ų and a minor one with a surface area of ∼210 Ų. Kurtoxin exhibits a larger hydrophobic surface than the scorpion α-toxins. The major hydrophobic patch of kurtoxin consists of the six solvent-exposed side chains of Tyr⁵, Tyr³⁶, Trp³⁸, Trp⁴⁰, Leu⁴², and Tyr⁴⁵ and has a surface area of 600 Ų (Figure 6A). The side chains are well-packed on each other, creating a compact hydrophobic patch on the protein surface. The minor hydrophobic patch (surface area of 500 Ų) is centered on Trp¹⁷, Tyr¹⁸, and Tyr²² and also includes the δ-methyls of Leu²⁶ (Figure 6A). Overall, it appears that kurtoxin shows a distinct surface profile, composed of both positive and hydrophobic residues, compared to other scorpion α-toxins.

Kurtoxin Binding Site on Cav3 (T-type) Ca²⁺ Channels

Voltage-gated ion channels consist of a central ion conduction pore (segments S5 and S6) surrounded by voltage sensors (segments S1–S4), which form “voltage sensor paddles” that move in response to changes in membrane voltage.^4,43 The overall structure of the voltage sensor paddles includes hydrophobic, cationic, and helix–turn–helix structures formed by the S3b–S4 segment, and it has been suggested that it is the voltage sensor paddles that are recognized by gating modifier toxins.^{17−19,38−40,42,45} Kurtoxin has been identified as the first high-affinity (K_d = 15 nM) gating modifier of Cav3.1 (α1G T-type) Ca²⁺ channels, and also the first to show cross-reactivity with voltage-gated Na⁺ channels.⁵⁵ This is similar to HaTx and GrTx, which exhibit cross-reactivity with the Kv2.1 K⁺ channel and P/Q-type Ca²⁺ channel, modifying the energetics of their gating.³⁸ It has been suggested that hydrophobic and negatively charged residues (Ile²⁷³, Phe²⁷⁴, and Glu²⁷⁷) in the Kv2.1 channel form the binding site for HaTx and GrTx.^38,111 Glu¹⁶¹³ in the rat brain IIA Na⁺ channel is equivalent to Glu²⁷⁷ in the Kv2.1 channel, and mutation of Glu¹⁶¹³ has a large effect on the affinity of scorpion α-toxins for Na⁺ channels.¹⁷ In addition, Glu¹⁶⁵⁸, situated at the end of S3, within repeat IV of the P/Q-type Ca²⁺ channel, contributes to the binding of ω-Aga IVA.¹⁹ These findings prompt us to speculate that kurtoxin may bind to the S3b–S4 motif in domain IV of both Cav3 (T-type) Ca²⁺ channels and Na⁺ channels, a region that has some conservation between the two channels and that corresponds to the HaTx/GrTx binding site on voltage-gated K⁺ and Ca²⁺ channels.⁵⁵ The region conserved in Cav3 (T-type) Ca²⁺ channels contains three glutamate residues [Glu¹⁶⁶¹, Glu¹⁶⁶², and Glu¹⁶⁶⁴ for Cav3.1 (Figure 7A)]. Notably, the region in the Cav3 (T-type) Ca²⁺ channel makes a larger negative patch than in Na⁺, K⁺, or other types of Ca²⁺ channels, and this negative domain is conserved in all Cav3 (T-type) Ca²⁺ channel subtypes. As mentioned, kurtoxin is highly electropositive because of the presence of an electropositive patch formed by Lys¹³, Arg¹⁴, and five positively charged residues in the C-terminal segment (Arg⁵⁴, Lys⁵⁶, Arg⁵⁷, Arg⁶⁰, and Arg⁶²). We therefore speculate that the positively charged surface of kurtoxin is an important determinant of its binding to the conserved negative domain in Cav3 (T-type) Ca²⁺ channels. Figure 7B shows a comparison of the surface profiles of kurtoxin and three scorpion α-toxins. Kurtoxin exhibits surface characteristics arising from a surface hydrophobic patch in combination with a large electropositive patch. By contrast, scorpion α-toxins show a rather small surface hydrophobic patch with a mixed charged surface. These distinct surface profiles may explain why only kurtoxin is able to interact with the binding domain on Cav3 (T-type) Ca²⁺ channels, which is composed of electronegative and hydrophobic residues. We are currently in the process of preparing alanine mutants of kurtoxin and the Cav3.1 channel to examine the molecular basis of the interaction between the toxin and channel.

Figure 7.

Comparison of the amino acid sequences of the indicated voltage-gated ion channels and the surface profiles of kurtoxin and scorpion α-toxins. (A) Comparison of the amino acid sequences of the domain IV S3–S4 linker in different voltage-gated ion channels (Cav, voltage-gated calcium channels; Nav, voltage-gated sodium channels; and Kv2.1, voltage-gated potassium channels). These sequences were aligned using ClustalX. Highly conserved residues are shaded in black or gray. The red rectangular boxes highlight the proposed kurtoxin binding site on Cav3 (T-type) Ca²⁺ channels. The blue rectangular boxes highlight the sequence involved in the binding sites of gating modifiers (Ile²⁷³, Phe²⁷⁴, and Glu²⁷⁷ in K⁺ channels; Glu¹⁶¹³ in Na⁺ channels; and Glu¹⁶⁵⁸ in P/Q-type Ca²⁺ channels). (B) Surface profiles of kurtoxin, AaH II, LqqhαIT, and Bmk M1: yellow for hydrophobic residues (Ala, Cys, Gly, Leu, Ile, Phe, Pro, Trp, Tyr and Val), blue and red for basic (Arg and Lys) and acidic (Asp and Glu) residues, respectively, and white for other residues. The surface hydrophobic patch residues and charged residues are indicated.

Conclusion

We investigated the three-dimensional structure of the first peptide toxin known to inhibit Cav3 (T-type) voltage-gated Ca²⁺ channels and suggest that its unique surface properties are likely responsible for its binding selectivity. Interestingly, kurtoxin can interact with high affinity with native neuronal high-threshold L-type, N-type, and P-type Ca²⁺ channels in central and peripheral neurons, producing complex gating modifications specific to each channel type.⁵⁶ When the channels are expressed in Xenopus oocytes, however, kurtoxin interacts only with the α-subunit of Cav3.1 (α1G T-type) voltage-gated Ca²⁺ channels; it does not interact with any other type of Ca²⁺ channel.⁵⁵ At present, the mechanism by which kurtoxin interacts with Ca²⁺ channels remains unknown. The structural studies of kurtoxin reported here provide clues about the molecular mechanism by which Cav3 (T-type) Ca²⁺ channel activity is regulated by selective ligands and could contribute to the development of highly specific Cav3 (T-type) Ca²⁺ channel inhibitors.

Glossary

Abbreviations

CD

circular dichroism

DSS

4,4-dimethyl-4-silapentane-1-sulfonic acid

DQF-COSY

double-quantum-filtered correlation spectroscopy

E-COSY

exclusive COSY

HSQC

heteronuclear single-quantum coherence

MALDI-TOF MS

matrix-assisted laser desorption ionization time-of-flight mass spectrometry

NMR

nuclear magnetic resonance

NOE

nuclear Overhauser effect

NOESY

nuclear Overhauser effect spectroscopy

PDB

Protein Data Bank

RP-HPLC

reverse phase high-performance liquid chromatography

TOCSY

total correlated spectroscopy.

Supporting Information Available

Figures S1–S3. This material is available free of charge via the Internet at http://pubs.acs.org.

The authors declare no competing financial interest.

Accession Codes

The coordinates for kurtoxin have been deposited in the Protein Data Bank as entry 1T1T.

Funding Statement

National Institutes of Health, United States