Research Article 1525

The structure of versutoxin (d-atracotoxin-Hv1) provides insights into the binding of site 3 neurotoxins to the voltage-gated sodium channel Jamie I Fletcher1, Bogdan E Chapman1, Joel P Mackay1, Merlin EH Howden2 and Glenn F King1*

Background: Versutoxin (δ-ACTX-Hv1) is the major component of the venom of Addresses: 1Department of Biochemistry, the Australian Blue Mountains funnel web , versuta. University of Sydney, Sydney NSW 2006, Australia and 2Department of Pharmacology, University of δ-ACTX-Hv1 produces potentially fatal neurotoxic symptoms in primates by Sydney, Sydney NSW 2006, Australia. slowing the inactivation of voltage-gated sodium channels; δ-ACTX-Hv1 is therefore a useful tool for studying sodium channel function. We have *Corresponding author. determined the three-dimensional structure of δ-ACTX-Hv1 as the first step E-mail: [email protected] towards understanding the molecular basis of its interaction with these channels. Key words: α-scorpion toxin, anthopleurin, gurmarin, sodium channel inactivation, versutoxin Results: The solution structure of δ-ACTX-Hv1, determined using NMR spectroscopy, comprises a core β region containing a triple-stranded Received: 19 May 1997 3 July 1997 antiparallel β sheet, a thumb-like extension protruding from the β region and a Revisions requested: 28 July 1997 β Revisions received: C-terminal 310 helix that is appended to the domain by virtue of a disulphide Accepted: 10 October 1997 bond. The β region contains a cystine knot motif similar to that seen in other neurotoxic polypeptides. The structure shows homology with µ-agatoxin-I, a Structure 15 November 1997, 5:1525–1535 spider toxin that also modifies the inactivation kinetics of vertebrate voltage- http://biomednet.com/elecref/0969212600501525 δ gated sodium channels. More surprisingly, -ACTX-Hv1 shows both sequence © Current Biology Ltd ISSN 0969-2126 and structural homology with gurmarin, a plant polypeptide. This similarity leads us to suggest that the sweet-taste suppression elicited by gurmarin may result from an interaction with one of the downstream ion channels involved in sweet- taste transduction.

Conclusions: δ-ACTX-Hv1 shows no structural homology with either sea anemone or α-scorpion toxins, both of which also modify the inactivation kinetics of voltage-gated sodium channels by interacting with channel recognition site 3. However, we have shown that δ-ACTX-Hv1 contains charged residues that are topologically related to those implicated in the binding of sea anemone and α-scorpion toxins to mammalian voltage-gated sodium channels, suggesting similarities in their mode of interaction with these channels.

Introduction lethal venom [6]. The lack of fatalities arising from H. The venoms of numerous aquatic and terrestrial versuta envenomation probably results from its geograph- have been shown to contain toxins that specifically inter- ical distribution, as it is confined to the sparsely popu- act with and alter the conductance properties of ion chan- lated mountains west of Sydney, whereas A. robustus nels. In addition to being useful pharmacological tools for inhabits the Sydney metropolitan area. There have been exploring the functional diversity of ion channels [1,2], no reported fatalities from funnel web envenomation these toxins can also provide leads for the design of novel since the introduction in 1980 of an antivenom to the A. pharmaceuticals [3] and insecticides [4]. It is in this context robustus venom [7]. that we have begun to examine toxins derived from Aus- tralian funnel web (Hexathelidae: Atracinae), a Intriguingly, both A. robustus and H. versuta venoms group of venomous confined to the south eastern produce severe neurotoxic symptoms in primates and coast of Australia [5]. newborn mice but not in other vertebrates [8–10]. Robus- toxin (δ-atracotoxin-Ar1; δ-ACTX-Ar1) and versutoxin (δ- Since 1927, funnel web envenomation has resulted in at atracotoxin-Hv1; δ-ACTX-Hv1) are 42-residue peptide least 14 human fatalities, all of which have been attrib- toxins that are responsible for the primate-specific effects uted to the male Sydney funnel web spider, robus- of A. robustus and H. versuta venom, respectively. These tus. Several other species, however, including the Blue peptides contain four intramolecular disulphide bonds and Mountains funnel web spider, Hadronyche versuta, produce share 83% sequence identity [11,12] (Figure 1). The 1526 Structure 1997, Vol 5 No 11

Figure 1

β1 β2 β3 II VIa I 1 10203040 δ-Atracotoxin-Hv1 CAKKRNWC- GKTEDCCCPMKCVYAWYNEQGS CQS TI S ALWKKC δ-Atracotoxin-Ar1 CAKKRNWC- GKNEDCCCPMKCI YAWYNQQGS CQTTI TGLFKKC Gurmarin XQCVKKDELCI PYYLDCCEPLECKKVNWWD-HLCIG µ-Agatoxin-I E C V PENGHCRDWYDECCEGFYCS CRQ- PPKCI CRNNN 1 ------15 8 ------20 14 ------31 16 ------42 Structure

Comparison of the primary structures of δ-atracotoxin-Hv1 (δ-ACTX- bonding pattern determined for δ-ACTX-Hv1 is shown below the Hv1), δ-atracotoxin-Ar1 (δ-ACTX-Ar1), gurmarin and µ-agatoxin-I sequences (blue dotted lines); the first three disulphide bonds, which (µ-Aga-I). An X within the sequence indicates pyroglutamate. form a cystine knot motif (see text), are conserved in all four toxins. The Homologies are shown relative to δ-ACTX-Hv1; identities are boxed in secondary structure of δ-ACTX-Hv1, as determined in the current study, β β yellow, and conservative substitutions are shown in red. The numbers is shown above the sequences ( strands in blue, turns in green, 310 above and below the sequences refer to δ-ACTX-Hv1. The disulphide- helix in red). (The diagram was drawn using ALSCRIPT [61].)

sequences are unusual in that they contain three consecu- (rpHPLC) (Figure 2). The fractionation of 180 mg of crude tive cysteine residues at positions 14–16 (Figure 1). Recent venom yielded ~5 mg of δ-ACTX-Hv1 that was judged to electrophysiological studies have shown that δ-ACTX- be >95% pure using rpHPLC and sodium dodecyl sul- Hv1 slows the inactivation of tetrodotoxin-sensitive sodium phate polyacrylamide gel electrophoresis (SDS PAGE). channels, resulting in repetitive firing in autonomic and The molecular weight determined using electrospray motor nerve fibres [10]. Receptor site 3 of the sodium mass spectrometry (4847.2 Da) corresponded to the theo- channel is defined as the site that binds the α-scorpion retical (average mass) value for fully oxidised δ-ACTX- and sea anemone toxins. The effects of δ-ACTX-Hv1 on Hv1 (4847.6 Da). sodium-channel gating are similar to those elicited by these neurotoxins [13], leading to the suggestion that δ- Figure 2 ACTX-Hv1 also binds to receptor site 3 [14]. This pro- posal has been confirmed by recent experiments showing (a) that δ-ACTX-Hv1 inhibits binding of the mammalian- ω-ACTX-Hv1 selective α-scorpion toxin from Androctonus australis Hector II (AaHII) and the insect-selective α-scorpion toxin from Leiurus quinquestriatus hebraeus (LqhαIT) to rat brain and cockroach synaptosomes, respectively (GM Nicholson, personal communication). (b) Here we describe the three-dimensional solution struc- ture of δ-ACTX-Hv1 and show that it has striking δ -ACTX-Hv1 homology with µ-agatoxin-I (µ-Aga-I), a sodium channel modifier from the unrelated North American funnel web spider Agelenopsis aperta [15] and, more surprisingly, with gurmarin [16], a sweet-taste suppressing polypeptide from the Indian plant Gymnema sylvestre [17]. Comparison of δ-ACTX-Hv1 with structurally unrelated site 3 sodium 0 5 10 15 20 25 Retention time (min) channel toxins has allowed us to construct a plausible Structure molecular model to explain how these toxins interact with, and alter the conductance properties of, vertebrate The isolation of δ-ACTX-Hv1 from H. versuta venom. (a) Reverse phase voltage-gated sodium channels. HPLC trace of whole H. versuta venom; the only components so far identified from this venom are the insect-specific calcium channel Results and discussion antagonist ω-ACTX-Hv1 [4] (marked with an arrow) and δ-ACTX-Hv1. d- (b) Reverse phase HPLC chromatogram, run under the same conditions Purification of ACTX-Hv1 δ δ as in (a), of the purified -ACTX-Hv1 used in NMR studies. The purified -ACTX-Hv1 was purified from H. versuta venom using toxin elutes with the same retention time as in the crude venom. reverse phase high pressure liquid chromatography Research Article Versutoxin (d-atracotoxin-Hv1) Fletcher et al. 1527

NMR resonance assignments from hydrogen-deuterium exchange experiments. The Complete sequence-specific 1H resonance assignments 493 distance restraints were comprised of 93 intraresidue, (see Supplementary material available with the Internet 184 sequential, 90 medium-range (1<|i–j| 5), and 126 version of this manuscript) were obtained for all residues, long-range (|i–j|>5) restraints. The total number of 573 except Gly9–Glu12, using standard two-dimensional (2D) NMR-derived restraints corresponds to an average of 13.6 homonuclear NMR experiments [18]. No proton reso- restraints per residue. nances were observable for residues 9–12 over the pH range 2.6–5.1 and a temperature range of 288–315K. This Structures were calculated using a hybrid distance geom- complete lack of signals is indicative of a chemical etry simulated-annealing protocol [19] in the programs exchange process in which these residues are exchanging DIANA [20] and X-PLOR [21] (see Materials and methods between two (or possibly more) distinct conformations at section). A family of 20 X-PLOR structures with the a rate which results in broadening of the 1H resonances lowest residual restraint violations were used to represent beyond detection. the solution structure of δ-ACTX-Hv1 (Figure 3), and the structural statistics for this family are summarised in Table In support of this suggestion is the observation that the 1. These structures have good nonbonded contacts, as evi- amide-proton signals of several residues flanking this region denced by the low values of the mean Lennard–Jones (Cys8 and Asp13–Cys15) are very broad, although they potential, and good covalent geometry, as evidenced by sharpen upon increasing the temperature in the range the small deviations from ideal bond lengths and bond 303–313K. This is consistent with their peripheral involve- angles (Table 1). There are no violations of distance or ment in the exchange process, which would result in dihedral angle restraints greater than 0.13 Å and 3.8°, smaller chemical shift changes than for residues 9–12, respectively. PROCHECK analysis [22] reveals that if causing them to lie on the fast exchange side of coales- residues Gly9–Asp13 (for which there were no NMR cence, as observed. If we assume that the exchange process restraints) are excluded, all but two of the nonglycine/ results in chemical shift changes of 300 Hz (0.5 ppm at nonproline residues (i.e., Ala23 and Trp24) lie in the 600 MHz) for the amide protons of residues 9–12, then the ‘most favoured’ and ‘additionally allowed’ regions of the rate of this exchange process can be estimated to be in Ramachandran plot. the order of ~700 s–1. It is important to note that the lack of observable amide-proton resonances for residues 9–12 Excluding residues 9–13, which are not defined by the cannot be the result of complete conformational disorder, as NMR data (see Figure 3a), the atomic root mean square this would give rise to very sharp signals for these residues. (rms) differences for the final 20 structures with respect to the mean coordinate positions are 0.36 ± 0.07 Å for the Structure determination and disulphide-bonding pattern backbone atoms and 0.68 ± 0.08 Å for all heavy atoms. In The disulphide-bonding pattern of δ-ACTX-Hv1 was not addition, the φ and ψ angular order parameters (Sφ and Sψ, known prior to commencing NMR structural studies, respectively; [23]) are >0.95 for all regions except however, the positions of two of the disulphide bonds Cys8–Cys14 and Ser33–Thr34. present in δ-ACTX-Ar1 (Cys1–Cys15 and Cys8–Cys20) had been previously determined by chemical methods Description of the structure of d-ACTX-Hv1 (PF Alewood, personal communication). Preliminary dis- The three-dimensional structure of δ-ACTX-Hv1 com- tance geometry calculations, using only these disulphide- prises a central β region (coloured blue in Figure 3), bond restraints in combination with a limited nuclear consisting of residues Cys1–Cys8, Cys14–Val21 and Overhauser effect (NOE) data set, revealed that the Ser30–Ser33, with residues Tyr22–Gly29 (red in Figure 3) Cys31 Sγ atom was separated from the Sγ atoms of Cys14, forming a thumb-like extension projecting from this region. Cys16, and Cys42 by 2.5–5.0 Å, 9.4–11.4 Å, and >12 Å, The β region contains several well-defined secondary respectively, indicative of a Cys14–Cys31 disulphide bond. structure elements (Figure 4), including a triple-stranded Thus, by default, the remaining disulphide bond is antiparallel β sheet with +2x, –1 topology [24] comprising Cys16–Cys42. Significant structural homology between δ- residues Asn6–Trp7 (β strand 1), Met18–Val21 (β strand ACTX-Hv1 and two peptides from unrelated organisms 2), and Ser30–Ser33 (β strand 3). The C-terminal end of (see below) provides further evidence that these disul- the short β strand 1 is held in place by a bifurcated hydro- phide-bond assignments are correct. gen bond between the Cys8 amide proton and the car- bonyl oxygens of the two residues preceding β strand 3 Final structure calculations were based on a total of 493 (Gln28 and Gly29). The β region also contains β turns at nonredundant interproton distance restraints derived from Lys3–Asn6 (type II) and Cys15–Met18 (a rare type VIa1 φ χ 2D NOE spectroscopy (NOESY) spectra, 30 and 20 1 turn encompassing a cis peptide bond at Cys16–Met 17). dihedral angle restraints derived from coupling constant The thumb-like extension, which comprises a well- and NOE measurements, and 30 upper-limit hydrogen- defined eight-residue loop between β strands 2 and 3, bond restraints (defining 15 hydrogen bonds) derived contains a type I β turn at Asn26–Gly29. 1528 Structure 1997, Vol 5 No 11

Figure 3

Solution structure of δ-ACTX-Hv1. (a) Stereo (a) view of the ensemble of 20 δ-ACTX-Hv1 structures superimposed for best fit over the backbone atoms of residues 1–8 and 14–42 of the mean coordinate structure. Only the backbone N, C and Cα atoms are displayed. The core β region is shown in purple, except for the conformationally disordered loop (Gly9–Asp13) which is shown in cyan. The C- terminal 310 helix is shown in green, and the thumb-like extension of the β region is coloured red. (b) As for (a) except that the molecules have been rotated ~180° about an imaginary vertical line dividing the two stereo C N C N images. The four disulphide bonds are displayed in magenta.

(b)

N N

C C Structure

The C-terminal residues Ile35–Lys41 form a well-ordered conformational averaging on the millisecond timescale β 310 helix (green in Figure 3) that is held close to the between two (or more) well-defined conformations (leading domain by virtue of a Cys16–Cys42 disulphide bond. The to exchange-broadened NMR resonances). C-terminal helix lies adjacent to the β turn formed by residues Cys15–Met18 of the β domain, but it makes very Three of the four disulphide bonds in δ-ACTX-Hv1 few van der Waals contacts with any of these residues, (1–15, 8–20 and 14–31) form a classical toxic/inhibitory δ including the outward projecting Pro17 sidechain. The 310 polypeptide cystine knot motif [25]. In the -ACTX- helix is highly nonpolar, comprising residues Ile35, Ala37, Hv1 motif, the Cys14–Cys31 disulphide bond passes Leu38 and Trp39, but it is flanked at the C-terminal end by through a 14-residue ring formed by the polypeptide Lys40 and Lys41, while the sidechain of Lys19 from the β backbone and the Cys1–Cys15 and Cys8–Cys20 disul- domain projects towards the N-terminal end of the helix. In phide bonds. However, the half-cystine spacing defini- ten of the 20 structures, a hydrogen bond is observed tion for this family of peptides, which we recently between the sidechain hydroxyl oxygen of the N-cap modified to CX3–7CX4–6CX0–5CX1–4CX4–13C in order to residue (Thr34) and the backbone amide proton of Ala37. include the structure of ω-ACTX-Hv1 [4], needs to be

further modified to CX3–7CX4–6CX0–5CX0–4CX3–13C to The hydrophobic core of the β domain is formed by the include the structure of δ-ACTX-Hv1. sidechains of Cys8, Cys14, Met18, Cys20 and Cys31, and is flanked on the N-terminal side by the aliphatic portion of d-ACTX-Hv1 contains two unusual b turns the Lys4 sidechain. The region Gly9–Asp13 (cyan in All residues in the Asn26–Gly29 type I β turn (Figure 5a) Figure 3) is undefined by the NMR data. As indicated pre- show high positional turn potentials [26] and the hydrogen- viously, however, this should not be interpreted as indicat- deuterium exchange data were consistent with a strong ing extreme conformational disorder (which would lead to C=Oi–N-Hi+3 hydrogen bond. In contrast, the Lys3-Lys4- very sharp NMR resonances), but rather a reflection of Arg5-Asn6 type II β turn (Figure 5b) is unusual in a number Research Article Versutoxin (d-atracotoxin-Hv1) Fletcher et al. 1529

Table 1 Figure 4

Structural statistics for the family of 20 d-ACTX-Hv1 structures.

Distance restraints intraresidue (i–j = 0) 93 sequential (|i–j| = 1) 184 medium range (|i–j| ≤ 5) 90 long range (|i–j| > 5) 126 hydrogen bonds 30 total 523 Dihedral angle restraints φ 30 χ 1 20 total 50 Mean rmsds from experimental restraints NOE (Å) 0.0138 ± 0.0006 dihedrals (°) 0.24 ± 0.10 Mean rmsds from idealised covalent geometry* bonds (Å) 0.0015 ± 0.0001 angles (°) 0.253 ± 0.008 impropers (°) 0.240 ± 0.007 Restraint violations mean NOE violations/structure > 0.1 Å 1.5 ± 0.5 maximum NOE violation (Å) 0.13 mean angle violations/structure > 0.5° 2.5 ± 0.9 maximum angle violation (°) 3.8 Mean energies (kJ mol–1) † δ ENOE 4.99 ± 0.44 Schematic drawing of -ACTX-Hv1. The figure shows the location of † β Ecdih 0.20 ± 0.20 the strands (cyan) and the 310 helix (red and yellow). (This figure was EvdW 1.93 ± 0.54 drawn using the programme MOLMOL [62].) Ebond 1.46 ± 0.12 Eimproper 3.11 ± 0.20 Eangle 11.57 ± 0.75 Etotal 23.26 ± 1.43 Residues Cys15–Met18 form a rare type VIa1 turn Atomic rms differences (Å)‡ (Figure 5c) which connects the antiparallel β strands 2 backbone atoms (1–8, 14–42) and 3. Although there are too few examples of type VIa1 versus mean/pairwise 0.36 ± 0.07/0.52 ± 0.13 heavy atoms (1–8, 14–42) turns to calculate meaningful turn potentials (they com- versus mean/pairwise 0.68 ± 0.08/0.98 ± 0.14 prise <1% of the ~4000 turns identified in a subset of the PDB [26]), this seems to be the only example of a type *Idealised geometry is defined by the CHARMM force field as VIa turn containing a methionine residue in any position. implemented within X-PLOR. †The final values of the square-well NOE and dihedral angle potentials were calculated with force constants of The Cys15–Met18 turn buries the C = Oi–N-Hi+3 hydro- 50 kcal mol–1 Å–2 and 200 kcal mol–1 Å–2, respectively. ‡Atomic gen bond in the interior of the central β region, partly differences are given as the average root mean square (rms) difference explaining why the Met18 amide proton exchanges with against the mean coordinate structure (mean) and as the average rms deuterons more slowly than any other backbone or difference of all pairwise comparisons of the 20 structures (pairwise). All energies, violations, and rms differences are given as the mean ± sidechain amide proton. standard deviation. Structural homologies The tertiary structure of δ-ACTX-Hv1 was compared with of respects. While the initial two lysine residues show high all other structures in the Brookhaven PDB using the DALI positional turn potentials, nonglycine residues are rare in algorithm [27]. δ-ACTX-Hv1 was found to have significant the i+2 position (arginine occurs in <1% of cases) and structural homology with µ-Aga-I [15] and ω-agatoxin-IVB asparagine is uncommon in the i+3 position of type II turns (ω-Aga-IVB) [28] from the North American funnel web (~2% of cases) [26]; this may indicate some functional spider A. aperta, and with gurmarin [16], a sweet-taste sup- significance for these residues as they are conserved in pressing polypeptide from the leaves of G. sylvestre (Figure δ-ACTX-Ar1 (Figure 1). An unusual hydrogen bond is 6). The four peptides share a very similar cystine knot observed between the Asn6 sidechain HD21 proton and motif, and therefore it appears that their global fold is deter- the Lys4 backbone carbonyl oxygen, such that residues mined primarily by their pattern of disulphide bonds. Lys4–Asn6 form a pseudo-turn in which the asparagine Cβ and Cγ atoms geometrically mimic the backbone C and N The structural homology observed between δ-ACTX- atoms of a classical β turn (Figure 5b). Hv1, µ-Aga-I and gurmarin includes conservation of a 1530 Structure 1997, Vol 5 No 11

Figure 5

The β turns in δ-ACTX-Hv1. Backbone N, Cα (a) Gln28 Gln28 and C atoms are coloured purple, O atoms Glu27 Glu27 are in red, HN atoms are cyan, sidechains are magenta, and hydrogen bonds are Asn26 Asn26 represented as dotted green lines. (a) Classical type I β turn (Asn26–Gly29). (b) Type II β turn (Lys3–Asn6); the asparagine β γ Gly29 Gly29 pseudo-turn, whereby the Asn6 C and C atoms geometrically mimic the backbone C and N atoms of a classical β turn and the (b) Arg5 Arg5 Asn6 HD21 sidechain hydrogen forms a Asn6 Asn6 hydrogen bond with the Lys4 carbonyl oxygen, is apparent. (c) Type VIa1 β turn (Cys15–Met18), resulting from a Cys16–Pro17 cis peptide bond. Lys4 Lys4

Lys3 Lys3

(c)

Cys15 Cys15 Met18 Met18

Cys16 Cys16

Pro17 Pro17

Structure

conformationally disordered loop in the β domain (residues thought to involve voltage-gated sodium and calcium Gly9–Asp13 in δ-ACTX-Hv1, Pro12–Asp16 in gurmarin channels downstream of sweet-taste receptors [31]. We and Arg10–Glu15 in µ-Aga-I). In all cases, the loop is ter- propose, therefore, that the structural and sequence minated by negatively charged residues (two in the case homology between gurmarin and δ-ACTX-Hv1 supports of µ-Aga-I and δ-ACTX-Hv1, and one in gurmarin; the alternative hypothesis that gurmarin elicits its taste- Figure 1). The structural homology between δ-ACTX- suppressing effect by interacting with an ion channel Hv1 and the agatoxins, despite their lack of sequence involved in taste transduction [30]. We have chemically homology, is perhaps not surprising as they are all ion synthesised gurmarin in order to test this hypothesis. channel toxins. µ-Aga-I, like δ-ACTX-Hv1, modifies the kinetics of voltage-gated sodium channels [15], while ω- Binding of site 3 neurotoxins to mammalian voltage-gated Aga-IVB is an antagonist of Purkinje-type (P-type) voltage- sodium channels gated channels [29]. There are at least five distinct sites on voltage-gated sodium channels that bind different neurotoxins [13]. The comparison between δ-ACTX-Hv1 and gurmarin δ-ACTX-Hv1, α-scorpion, and sea anemone toxins bind to extends beyond structural homology, however, with a sig- a common or overlapping region of the channel referred to nificant degree of sequence homology (44% over the as site 3 [32,33]. Various studies have implicated several aligned region, comprising 29% identities and 15% conserv- loops within the sodium channel as important regions ative substitutions; see Figure 1), a common cis-peptide for recognition of site 3 neurotoxins: the S5–SS1 loop bond (Cys16-Pro17 in δ-ACTX-Hv1 and Glu19-Pro20 in in domain IV and the S5–SS1 and SS2–S6 loops in gurmarin), and similarly located charged residues. It has domain I [13,32]. Recent mutagenesis studies, however, generally been conjectured that gurmarin suppresses the have revealed that the short extracellular S3–S4 loop in response of rats (but not humans) to sweet-tasting sub- domain IV of the sodium channel α subunit (hereafter stances by binding to a sweet-taste receptor [16,17,30]. The referred to as IVS3–S4) contains a number of key determi- transduction of the response to sweet tastants, however, is nants for binding of α-scorpion and sea anemone toxins Research Article Versutoxin (d-atracotoxin-Hv1) Fletcher et al. 1531

Figure 6

Structural comparison of δ-ACTX-Hv1 (PDB accession code 1vtx), gurmarin (1gur) and µ- Aga-I (1eit). Stereo view of an overlay of the backbone N, Cα and C atoms of gurmarin (green), δ-ACTX-Hv1 (red) and µ-Aga-I (blue). Residues 1–8, 12–27 and 29–33 of δ-ACTX- Hv1 superimpose on residues 3–10 and 15–35 of gurmarin with a backbone root mean square difference (rmsd) of 2.0 Å. Residues 1–8, 12–24 and 26–35 of δ-ACTX-Hv1 superimpose on residues 2–9, 14–26 and 27–36 of µ-Aga-I with an rmsd of 2.3 Å. The 1–15, 8–20 and 14–31 disulphide bonds in δ-ACTX-Hv1 overlay with the 3–18, 10–23 and 17–33 disulphide bonds in gurmarin and the 2–17, 9–22 and 16–32 disulphide bonds in µ-Aga-I (not displayed for reasons of clarity). δ -ACTX-Hv1 has a C-terminal 310 helical extension relative to the other two structures.

[32]. The residues identified as critical for toxin binding with the structures of anthopleurin-B [41] and AaHII [42] (Glu1613, Glu1616 and Lys1617) are all charged, consis- such that three cationic and two anionic groups superim- tent with the identification of several anionic and cationic pose very closely. A stereo view of the superimposed residues in the α-scorpion and sea anemone toxins that are charged residues is shown in Figure 7a, while the disposi- crucial for receptor binding. tion of these residues with respect to the overall three- dimensional fold of these molecules is shown in Figure 7b. Given that δ-ACTX-Hv1 contains several distinct clus- It is clear that the charged residues overlay well despite a ters of charged residues (Lys3-Lys4-Arg5, Glu12-Asp13 completely different three-dimensional fold for each of and Lys40-Lys41) and binds to the same site as the α-scor- the toxins. Furthermore, the overlaid residues include pion and sea anemone toxins, we investigated whether, many of those identified as crucial for binding of α-scor- despite their lack of structural homology, these toxins pion and sea anemone toxins to the sodium channel, might share topologically related charged residues that including Asp9, Lys48 and Lys49 (in the case of antho- could interact with charged residues in the IVS3–S4 loop. pleurin-B) and Lys2 and Lys58 (in the case of AaHII). For the purpose of this analysis, we confined our search These overlaid residues comprise a roughly contiguous largely to charged residues in the sea anemone and α- surface on one face of the molecule, suitable for inter- scorpion toxins that have been identified as important for action with the short IVS3–S4 loop. While Lys58 is buried channel binding. Chemical modification of α-scorpion in the crystal structure of AaHII, it is thought to be toxins has revealed that the cationic residues Lys2, exposed in solution as it is the most chemically reactive Lys28, Lys58, Arg62 (using the residue numbering for lysine residue in the toxin [42]. AaHII) and the N-terminal α-amino group are important for toxin binding to the sodium channel, whereas none of The uppermost cationic and anionic clusters in Figure 7 the histidine residues are critical [34,35]. These cationic (Lys10 and Glu12 in δ-ACTX-Hv1) do not include any residues are all highly conserved across the family of residues thus far identified as being important for α-scorpion toxins [33]. binding of site 3 toxins to the sodium channel. Thus, we propose that the lower three clusters, involving two While chemical modification studies of sea anemone toxins cationic residues and one anionic residue, comprise a have been somewhat inconsistent [36,37], an heterologous surface that would be complementary to the two anionic expression system has been developed for producing wild and one cationic residue identified in the IVS3–S4 loop type and mutant versions of the sea anemone toxin antho- as important for binding of sea anemone and α-scorpion pleurin-B [38]. Consequently, for the sea anemone toxins, toxins; this electrostatic complementarity of binding we have restricted our topological search to critical charged surfaces is shown schematically in Figure 8. In support residues identified using this mutagenesis system, namely of this hypothesis, modelling of the IVS3–S4 loop in Asp9, Arg12, Lys37, Lys48 and Lys49 [36,39,40]. an extended conformation places the charged residues of the loop in an excellent position to interact with The results of the topological comparison are summarised the complementary charged groups on the toxins (data in Figure 7. The structure of δ-ACTX-Hv1 can be overlaid not shown). 1532 Structure 1997, Vol 5 No 11

Figure 7

Structural comparison of δ-ACTX-Hv1, α-scorpion and sea anemone toxins. (a) Stereo view of an overlay of topologically similar residues in δ-ACTX-Hv1 (thick rendered tubes; representative structure from the ensemble 1vtx), AaHII (thin rendered tubes; 1ptx) and anthopleurin-B (fine rendered tubes; representative structure from the ensemble 1apf). Cationic and anionic residues are coloured blue and red, respectively, and the residues from δ-ACTX-Hv1 are labelled. Only the sidechain heavy atoms of the overlaid residues are shown. (b) Structures of anthopleurin-B, δ-ACTX-Hv1 and AaHII showing the disposition of the topologically similar charged residues with respect to each of the individual structures. The orientation is the same as in (a) except that the molecules have been separated by translation in the plane of the page for easy comparison. The sidechain heavy atoms of the cationic and anionic residues are coloured blue and red, respectively, while the backbones are drawn as grey rendered tubes. The three molecules have distinctly different three-dimensional folds. Asp9 in anthopleurin-B, and Lys10, Glu12 and Asp13 in δ-ACTX-Hv1 are part of disordered loops and hence their orientation in solution is not fixed.

Figure 8 Further circumstantial evidence in favour of this model comes from structural studies of the toxins. The anionic Lys2 (AaHII) Lys58 (AaHII) residues Asp9 in anthopleurin-B [41] and Asp13 in Lys49 (AP-B) Lys48 (AP-B) δ-ACTX-Hv1 are both part of disordered loops that could Arg5 (δ-ACTX) Lys3 (δ-ACTX) be reconfigured to complement a binding site, while there is disorder in the region of Asp8 and Lys58 in the crystal Asp8 (AaHII) structure of AaHII [42]. Furthermore, the model impli- Asp9 (AP-B) cates Arg5 of δ-ACTX-Hv1 in channel binding, perhaps Asp13 (δ-ACTX) explaining why it has been conserved in the disfavoured i+2 position of a type II β turn.

Clearly, other regions of the sodium channel are also impli- cated in the binding of site 3 neurotoxins, as exemplified by the fact that photoreactive α-scorpion toxin V from L. quin- Glu1613 Leu1614 Ile1615 Glu1616 Lys1617 Structure questriatus (LqTx) covalently attaches to the S5–SS1 loop of domain I [32]. In this respect, it is worth mentioning that each of the toxins has a rich cluster of aromatic residues, Summary of proposed electrostatic interactions between site 3 and these clusters roughly superimpose when the toxins are neurotoxins and residues in the IVS3–S4 loop of the mammalian voltage-gated sodium channel. The sequence of the relevant portion of overlaid as in Figure 7. This hydrophobic surface has been the IVS3–S4 loop is shown at the bottom of the diagram; residues implicated in the binding of site 3 toxins to the sodium identified from mutagenesis studies to be important for binding site 3 channel [35], and Trp33 in anthopleurin-B [43] and Trp38 neurotoxins are coloured (cationic residues are in blue, anionic are in in Aah II [35], which are both situated in this hydrophobic α red). Residues from -ACTX-Hv1, AaHII and anthopleurin-B (AP-B) region, are known to be crucial for channel binding. Three that are proposed to interact with these residues in the IVS3–S4 loop are indicated at the top of the diagram. of the five aromatic residues in δ-ACTX-Hv1 (Trp7, Trp23 and Tyr24) are present in this hydrophobic cluster, and Research Article Versutoxin (d-atracotoxin-Hv1) Fletcher et al. 1533

these residues are strictly conserved in δ-ACTX-Ar1. This Materials and methods region might therefore be important for the interaction of Nomenclature all of these toxins with other sites on the sodium channel, We recently introduced a rational nomenclature for toxins derived from Australian funnel web spiders [4]. Based on this nomenclature, we such as the S5–SS1 loop of domain I. suggest that the trivial names robustoxin and versutoxin be replaced with δ-atracotoxin-Ar1 (δ-ACTX-Ar1) and δ-atracotoxin-Hv1 (δ-ACTX- Biological implications Hv1), respectively, where the δ indicates inhibition of sodium channel Australian funnel web spiders produce a neurotoxic inactivation as the predominant pharmacological activity. venom that has caused at least 14 human fatalities. The Purification of δ-ACTX-Hv1 human envenomation syndrome includes symptoms Approximately 180 mg of lyophilised H. versuta venom was dissolved µ µ such as lachrymation, salivation, generalised skeletal in 1350 l H2O and centrifuged through a 0.2 m filter. Samples muscle fasciculation, sweating, nausea, vomiting, diar- (50 µl) were applied to a Pharmacia semi-preparative rpHPLC column × µ rhoea, pulmonary oedema, disturbances in respiration, (Pep-S C2/C18, 9.3 250 mm, 15 m) and the venom components were eluted at a flow rate of 5 ml min–1 using a linear gradient of blood pressure and heart rate, followed by severe 10–28% acetonitrile over 8 min followed by a gradient of 28–33% hypotension and respiratory or circulatory failure. acetonitrile over 12 min. Fractions containing δ-ACTX-Hv1 (which eluted Intriguingly, this syndrome is manifest in primates and at 30% acetonitrile) were pooled and lyophilised. This yielded ~10 mg of crude δ-ACTX-Hv1 which was dissolved in 1800 µl H O, cen- newborn mice but not other vertebrates. 2 trifuged through a 0.2 µm filter, and further purified using a Vydac analytical rpHPLC column (C18, 4.6 × 250 mm, 5 µm). The toxin was The neurotoxic effects of funnel web venom in primates eluted at a flow rate of 1 ml min–1 using a linear gradient of 25– are caused by the binding of a 42-residue polypeptide 32% acetonitrile over 11 min. δ-ACTX-Hv1 fractions were pooled and δ named δ-atracotoxin (δ-ACTX) to neuronal voltage- lyophilised to give ~5 mg of -ACTX-Hv1. δ gated sodium channels. The effects of -ACTX on NMR spectroscopy sodium channel gating are very similar to those elicited An NMR sample was prepared by dissolving ~5 mg of δ-ACTX-Hv1 in α µ µ µ by the so-called site 3 neurotoxins, such as the -scor- a solution of 250 l H2O, 15 l D2O, and 5 l chloramphenicol (1 mM) pion and sea anemone toxins, which bind to site 3 on the in a 5 mm o.d. susceptibility-matched microcell (Shigemi). The pH was adjusted to either 2.6, 3.6, or 5.1. For experiments in D2O, the sample sodium channel. These similarities have led to the pro- µ was lyophilised and reconstituted in a microcell in 280 l 99.96% D2O posal, and subsequent experimental confirmation, that (Sigma). To resolve spectral ambiguities, spectra were collected under δ-ACTX also binds to site 3 on the sodium channel. three separate pH/temperature conditions: pH 2.6 and 300K, pH 3.6 Comparison of the solution structure of δ-ACTX-Hv1 and 313K, pH 5.1 and 293K. The following homonuclear 2D-NMR with the structures of sea anemone and α-scorpion spectra were recorded for each set of pH/temperature conditions: double-quantum-filtered correlated spectroscopy (DQF COSY) [44] toxins reveals that, even though the three toxins have with phase cycling modified for fast recycle times [45]; total correlation very dissimilar three-dimensional folds, they share a spectroscopy (TOCSY) [46] with MLEV isotropic mixing periods of 70 number of topologically related anionic and cationic and 100 ms; ECOSY [47]; and NOESY [48] with mixing times of 250 residues. We propose that these charged residues and 300 ms. provide a binding surface that facilitates a precise elec- All spectra of the pH 3.6 sample were recorded with a 5 mm inverse trostatic interaction between these toxins and a comple- broadband gradient probe on a Bruker DRX 500 spectrometer using mentary charged surface on the S3–S4 loop of domain the WATERGATE gradient module [49] for water suppression. All 1 IV of the voltage-gated sodium channel. remaining spectra were acquired using a 5 mm H probe on a 600 MHz Bruker AMX-600 spectrometer that was not equipped with gradients. Thus, water suppression was achieved by selective irradia- Surprisingly, the solution structure of δ-ACTX-Hv1 is tion during the relaxation delay between transients (1.2–1.4 s) and very similar to that of gurmarin, a plant polypeptide. It during the mixing period in NOESY experiments; all pulse sequences has generally been conjectured that gurmarin sup- were modified by the insertion of a SCUBA sequence [50] prior to the first 90° pulse in order to facilitate the recovery of bleached presses the response of rats to sweet-tasting substances Hα resonances. by binding to a sweet-taste receptor. The structural and sequence homology between gurmarin and δ- Spectra were processed using XWIN-NMR software (Bruker). Dis- ACTX-Hv1, however, supports the alternative hypothe- tance restraints were derived from the intensities of cross-peaks in NOESY spectra recorded at pH 2.62 (300K) and pH 3.5 (313K) in sis that gurmarin elicits its taste-suppressing effect by H2O and D2O with mixing times of 250 and 300 ms. Hydrogen-deu- interacting with an ion channel involved in taste trans- terium exchange experiments were carried out by lyophilising the H2O µ duction. It is striking that such structurally similar pep- sample, reconstituting it in 280 l of 99.96% D2O, and running 12 consecutive one-dimensional (1D) spectra of ~5 min duration fol- tides have evolved (presumably independently) in such τ lowed by three TOCSY experiments ( m = 40 ms) of ~2 h duration. unrelated phyla. The occurrence of very similar cystine Slowly exchanging amides were interpreted as those that still gave knot motifs in polypeptides from plants and both terres- amide resonances in the final TOCSY experiment. trial and aquatic animals [25] argues that this protein fold has been particularly favoured during the course of Structure calculations evolution. It may be that the toxic/inhibitory polypeptide Analysis of NOESY spectra yielded 493 nonredundant interproton distance restraints which were assigned upper-distance bounds of cystine knot motif represents the most economical way 2.80 Å (strong), 3.50 Å (medium), or 5.00 Å (weak) based on the cor- to construct a very small globular protein. responding cross-peak intensity. Lower distance bounds were taken as 1534 Structure 1997, Vol 5 No 11

the sum of the van der Waals radii. Pseudo-atom corrections were Supplementary material added to distance restraints where necessary [51]. A 0.5 Å empirical Supplementary material available with the Internet version of this correction was added to the upper bound for restraints involving manuscript contains details of the sequence-specific 1H resonance methyl protons [52]. assignments.

16 φ dihedral angles restraints were derived directly from 3J cou- NHα Acknowledgements pling constants [53] measured from either 1D NMR spectra or the This work was supported by the Australian Research Council in the form of antiphase cross-peak splitting in a high digital resolution 2D DQF a research grant to GFK and a Research Fellowship to JPM. JIF gratefully COSY spectrum. Following the standard parameterization of the acknowledges the support of a Junior Research Fellowship from the Aus- Karplus relationship [18], the φ angle was restrained to –35 ± 15° for tralian Grains Research and Development Corporation. Peter Barron and 3 3 JNHα < 3.0 Hz (Cys16), –65 ± 15° for JNHα = 3.0–5.8 Hz (Lys4, Ile35, Peter Caha of Bruker Australia are gratefully acknowledged for providing 3 time on their DRX 500 NMR spectrometer, and Lindsey Mackay is thanked Ser36, Ala37, Leu38), –120 ± 30° for JNHα = 8.0–9.5 Hz (Asn6, Cys15, Lys19, Tyr22, Ala23, Asn26, Gln32, Thr34), and –120 ± 15° for electrospray mass spectrometry. Bill Bubb is thanked for expert mainte- 3 φ nance of the Department of Biochemistry NMR spectrometers. for JNHα > 9.5 Hz (Lys3, Val21). 13 additional angle restraints of –100 ± 80° (Ala2, Trp7, Cys8, Met18, Trp24, Tyr25, Ser30, Cys31, Ser33, Trp39, Lys40, Lys41, Cys42) were applied for residues where References the intraresidue Hα-HN NOE was clearly weaker than the NOE between 1. Olivera, B.M., Miljanich, G.P., Ramachandran, J. & Adams, M.E. (1994). Calcium channel diversity and neurotransmitter release: the ω- HN and the Hα of the preceding residue [54]. The intense intraresidue 3 conotoxins and ω-agatoxins. Annu. Rev. Biochem. 63, 823–867. Hα-HN NOE for Arg5, combined with a JNHα value of ~7 Hz, allowed its φ angle to be restrained to 50 ± 40° [15,55]. 2. McIntosh, J.M., et al., & Olivera, B.M. (1995). A new family of conotoxins that blocks voltage-gated sodium channels. J. Biol. Chem. 270, 16796–16802. χ Stereospecific assignment of methylene protons and 1 dihedral 3. Valentino, K., et al., & Ramachandran, J. (1993). A selective N-type angle restraints were derived for 13 residues (Cys1, Lys3, Lys4, calcium channel antagonist protects against neuronal loss after global Arg5, Asn6, Cys15, Met18, Tyr22, Glu27, Cys31, Gln32, Leu38, cerebral ischemia. Proc. Natl. Acad. Sci. USA 90, 7894–7897. 3 3 Cys42) using Jαβ2 and Jαβ3 coupling constants measured from 4. Fletcher, J.I., et al., & King, G.F. (1997). The structure of a novel insecticidal neurotoxin, ω-atracotoxin-HV1, from the venom of an ECOSY spectra in combination with HN-Hβ2, HN-Hβ3, Hα-Hβ2, and χ Australian funnel web spider. Nat. Struct. Biol. 4, 559–566. Hα-Hβ3 NOE intensities [56]. The 1 restraints, but not stereospecific assignments, were obtained for four residues (Cys16, Ser30, Ser33, 5. Gray, M.R. (1988). Aspects of the systematics of the Australian funnel χ web spiders (Araneae: Hexathelidae: Atracinae) based upon Ser36). Additional 1 dihedral angle restraints were derived for morphological and electrophoretic data. In Australian arachnology. 3 Val21, Thr34 and Ile35 from measurement of Jαβ in ECOSY spectra (Austin, A.D. & Heather, N.W., eds), pp. 113–125, The Australian [57]. The two sets of Val21 γ-methyl protons had coincident chemical Entomological Society, Brisbane, Australia. shifts and therefore could not be stereospecifically assigned. Ranges 6. Sheumack, D.D., Baldo, B.A., Carroll, P.R., Hampson, F., Howden, χ β of ±20° were used for all 1 dihedral angle restraints. The -methyl- M.E.H. & Skorulis, A. (1984). A comparative study of properties and ene protons of Pro17 were stereospecifically assigned from NOESY toxic constituents of funnel web spiders (Atrax) venoms. Comp. Biochem. Physiol. 78C, 55–68. spectra on the basis that Hα is always closer to Hβ than Hβ [58]. 3 2 7. Sutherland, S.K. (1980). Antivenom to the venom of the male Sydney The Cys16-Pro17 peptide bond was clearly identified as being in the ω funnel web spider, Atrax robustus. Med. J. Aust. 2, 437–441. cis conformation ( ~0°) on the basis of an intense NOE between 8. Mylecharane, E.J., Spence, I., Scheumack, D.D., Claassens, R. & the Hα protons of these two residues [18]. Howden, M.E.H. (1989). Actions of robustoxin, a neurotoxic polypeptide from the venom of the male funnel-web spider (Atrax Hydrogen-deuterium exchange experiments were used to identify 14 robustus) in anaesthetized monkeys. Toxicon 27, 481–492. amide proton hydrogen-bond donors (Arg5, Cys8, Met18, Lys19, 9. Phillips, C.A., et al., & Howden, M.E.H. (1987). The effects of the Val21, Gly29, Ser30, Cys31, Gln32, Ser33, Thr34, Ala37, Leu38, venom from the Blue Mountains funnel-web spider (Atrax versutus) in Trp39, Lys40, Lys41), the H resonances of which were still visible anaesthetized monkeys. Clin. Exp. Pharmacol. Physiol. Suppl. N 10, 14–15. after ~340 min in D O. Hydrogen-bond acceptors for all of these 2 10. Nicholson, G.M., Willow, M., Howden, M.E.H. & Narahashi, T. (1994). amide protons except Ser33 and Thr34 were unambiguously deter- Modification of sodium channel gating and kinetics by versutoxin from mined from preliminary structure calculations, and the hydrogen bonds the Australian funnel-web spider Hadronyche versuta. Pflügers Arch. (i–j) were subsequently defined in the structure calculations by dis- 428, 400–409. ≤ ≤ ≤ ≤ tance restraints of 1.7 d 2.2 Å for HNi–Oj and 2.7 d 3.2 Å for 11. Sheumack, D.D., Claassens, R., Howden, M.E.H. & Whitley, N.M. (1985). Complete amino acid sequence of a new type of lethal Ni–Oj. Preliminary structure calculations showed that the Cys8 amide proton formed a bifurcated hydrogen bond with the carbonyl oxygens neurotoxin from the venom of the funnel-web spider Atrax robustus. of both Gln28 and Gly29; restraints were included for both hydrogen- FEBS Lett. 181, 154–156. 12. Brown, M.R., Sheumack, D.D., Tyler, M.I. & Howden, M.E.H. (1988). bond acceptors. Amino acid sequence of versutoxin, a lethal neurotoxin from the venom of the funnel-web spider Atrax versutus. Biochem. J. 250, 401–405. The distance geometry programme DIANA [20] was used to calculate 13. Catterall, W.A. (1988). Structure and function of voltage-sensitive ion 1800 structures from random starting conformations, with disulphide channels. Science 242, 50–61. bonds defined as described previously [4]. DIANA calculations incor- 14. Nicholson, G.M., Little, M.J., Tyler, M. & Narahashi, T. (1996). porated a single cycle using redundant dihedral angle constraints Selective alteration of sodium channel gating by Australian funnel-web (REDAC) in order to reduce the computation time required for obtain- spider toxins. Toxicon 34, 1443–1453. 15. Omecinsky, D.O., Holub, K.E., Adams, M.E. & Reily, M.D. (1996). ing a family of acceptable conformers [59]. The best 75 structures Three-dimensional structure analysis of µ-agatoxins: further evidence (selected on the basis of their final penalty function values) were then for common motifs among neurotoxins with diverse ion channel refined in X-PLOR using the dynamical simulated-annealing protocol specificities. Biochemistry 35, 2836–2844. described previously [4]. Elements of secondary structure in the final 16. Arai, K., et al., & Akasaka, K. (1995). Three-dimensional structure of ensemble of 20 structures were identified using the programme PRO- gurmarin, a sweet-taste suppressing polypeptide. J. Biomol. NMR 5, MOTIF [60]. 297–305. 17. Imoto, T., Miyasaka, A., Ishima, R. & Akasaka, K. (1991). A novel peptide Accession numbers isolated from the leaves of Gymnema sylvestre. I. Characterization and δ its suppressive effect on the neural responses to sweet taste stimuli in The NMR restraints file and coordinate file for the family of 20 -atraco- the rat. Comp. Biochem. Physiol. 100A, 309–314. toxin-Hv1 structures have been deposited with the PDB with accession 18. Wüthrich, K. (1986). NMR of Proteins and Nucleic Acids. John Wiley codes r1vtxmr and 1vtx, respectively. & Sons, Inc., New York. Research Article Versutoxin (d-atracotoxin-Hv1) Fletcher et al. 1535

19. Nilges, M., Clore, G.M. & Gronenborn, A.M. (1988). Determination of Hector refined at 1.3 Å resolution. J. Mol. Biol. 238, 88–103. the three-dimensional structures of proteins from interproton distance 43. Dias-Kadambi, B.L., Combs, K.A., Drum, C.L., Hanck, D.A. & data by hybrid distance geometry-dynamical simulated annealing Blumenthal, K.M. (1996). The role of exposed tryptophan residues in calculations. FEBS Lett. 229, 317–324. the activity of the cardiotonic polypeptide anthopleurin B. J. Biol. 20. Güntert, P., Braun, W. & Wüthrich, K. (1991). Efficient computation of Chem. 271, 23828–23835. three-dimensional protein structures in solution from nuclear magnetic 44. Rance, M., Sørensen, O.W., Bodenhausen, G., Wagner, G., Ernst, resonance data using the program DIANA and the supporting R.R. & Wüthrich, K. (1983). Improved spectral resolution in COSY 1H programs CALIBA, HABAS and GLOMSA. J. Mol. Biol. NMR spectra of proteins via double quantum filtering. Biochem. 217, 517–530. Biophys. Res. Commun. 117, 479–485. 21. Brünger, A.T. (1992). X-PLOR Version 3.1. A System for 45. Derome, A.E. & Williamson, M.P. (1990). Rapid-pulsing artifacts in Crystallography and NMR. Yale University Press, New Haven, CT, double-quantum-filtered COSY. J. Magn. Reson. 88, 177–185. USA. 46. Bax, A. & Davis, D.G. (1985). MLEV-17 based two-dimensional 22. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M. homonuclear magnetisation transfer spectroscopy. J. Magn. Reson. (1993). PROCHECK: a program to check the stereochemical quality 65, 355–360. of protein structure coordinates. J. Appl. Cryst. 26, 283–291. 47. Griesinger, C., Sørensen, O.W. & Ernst, R.R. (1987). Practical 23. Hyberts, S.V., Goldberg, M.S., Havel, T.F. & Wagner, G. (1992). The aspects of the E.COSY technique. Measurement of scalar spin-spin solution structure of eglin c based on measurements of many NOEs coupling constants in peptides. J. Magn. Reson. 75, 474–492. and coupling constants and its comparison with X-ray structures. 48. Kumar, A., Ernst, R.R. & Wüthrich, K. (1980). A two-dimensional Protein Sci. 1, 736–751. nuclear Overhauser enhancement (2D nOe) experiment for the 24. Richardson, J.S. (1981). The anatomy and of protein elucidation of complete proton-proton cross-relaxation networks in structure. Adv. Protein Chem. 34, 167–339. biological macromolecules. Biochem. Biophys. Res. Comm. 95, 1–6. 25. Pallaghy, P.K., Nielsen, K.J., Craik, D.J. & Norton, R.S. (1994). A 49. Piotto, M., Saudek, V. & Sklenár, V. (1992). Gradient-tailored common structural motif incorporating a cystine knot and a triple- excitation for single-quantum NMR spectroscopy of aqueous stranded β-sheet in toxic and inhibitory polypeptides. Protein Sci. solutions. J. Biomol. NMR 2, 661–665. 3, 1833–1839. 50. Brown, S.C., Weber, P.L. & Mueller, L. (1988). Toward complete 1H 26. Hutchinson, E.G. & Thornton, J.M. (1994). A revised set of potentials NMR spectra in proteins. J. Magn. Reson. 77, 166–169. for β-turn formation in proteins. Protein Sci. 3, 2207—2216. 51. Wüthrich, K., Billeter, M. & Braun, W. (1983). Pseudo-structures for 27. Holm, L. & Sander, C. (1993). Protein structure comparison by the 20 common amino acids for use in studies of protein alignment of distance matrices. J. Mol. Biol. 233, 123–138. conformations by measurements of intramolecular proton-proton 28. Reily, M.D., Thanabal, V. & Adams, M.E. (1995). The solution structure distance constraints with nuclear magnetic resonance. J. Mol. Biol. of ω-Aga-IVB, a P-type calcium channel antagonist from venom of the 169, 949–961. funnel web spider, Agelenopsis aperta. J. Biomol. NMR 5, 122–132. 52. Clore, G.M., Wingfield, P.T. & Gronenborn, A.M. (1991). High- 29. Yu, H., et al., & Schreiber, S.L. (1993). Sequential assignment and resolution three-dimensional structure of interleukin 1β in solution by structure determination of spider toxin ω-Aga-IVB. Biochemistry three- and four-dimensional nuclear magnetic resonance 32, 13123–13129. spectroscopy. Biochemistry 30, 2315–2323. 30. Miyasaka, A. & Imoto, T. (1995). Electrophysiological characterization 53. Pardi, A., Billeter, M. & Wüthrich, K. (1984). Calibration of the angular of the inhibitory effect of a novel peptide gurmarin on the sweet taste dependence of the amide proton-Cα proton coupling constants, 3 3 response in rats. Brain Res. 676, 63–68. JHNα, in a globular protein. Use of JHNα for identification of helical 31. Lindemann, B. (1996). Taste reception. Physiol. Rev. 76, 719–766. secondary structure. J. Mol. Biol. 180, 741–751. 32. Rogers, J.C., Qu, Y., Tanada, T.N., Scheuer, T. & Catterall, W.A. 54. Clubb, R.T., Ferguson, S.B., Walsh, C.T. & Wagner, G. (1994). Three- (1996). Molecular determinants of high affinity binding of α-scorpion dimensional solution structure of Escherichia coli periplasmic toxin and sea anemone toxin in the S3-S4 extracellular loop in domain cyclophilin. Biochemistry 33, 2761–2722. IV of the Na+ channel α subunit. J. Biol. Chem. 271, 15950–15962. 55. Ludvigsen, S. & Poulsen, F.M. (1992). Positive theta-angles in proteins 33. Gordon, D., et al., & Rochat, H. (1996). Scorpion toxins affecting by nuclear magnetic resonance spectroscopy. J. Biomol. NMR sodium channel current inactivation bind to distinct homologous 2, 227–233. receptor sites on rat brain and insect sodium channels. J. Biol. Chem. 56. Wagner, G., Braun, W., Havel, T.F., Schaumann, T., Gõ, N. & 271, 8034–8045. Wüthrich, K. (1987). Protein structures in solution by nuclear 34. Loret, E.P., Mansuelle, P., Rochat, H. & Granier, C. (1990). magnetic resonance and distance geometry. The polypeptide fold of Neurotoxins active on insects: amino acid sequences, chemical the basic pancreatic trypsin inhibitor determined using two different modification and secondary structure estimation by circular dichroism algorithms, DISGEO and DISMAN. J. Mol. Biol. 196, 611–639. of toxins from the scorpion Androctonus australia Hector. 57. Zuiderweg, E.R., Boelens, R. & Kaptein, R. (1985). Stereospecific Biochemistry 29, 1492–1501. assignments of 1H-NMR methyl lines and conformation of valyl 35. Kharrat, R., Darbon, H., Rochat, H. & Granier, C. (1989). residues in the lac repressor headpiece. Biopolymers 24, 601–611. Structure/activity relationships of scorpion α-toxins. Multiple residues 58. Kline, A.D., Braun, W. & Wüthrich, K. (1988). Determination of the contribute to the interaction with receptors. Eur. J. Biochem. complete three-dimensional structure of the alpha-amylase inhibitor 181, 381–390. tendamistat in aqueous solution by nuclear magnetic resonance and 36. Khera, P.K. & Blumenthal, K.M. (1994). Role of the cationic residues distance geometry. J. Mol. Biol. 204, 675–724. arginine 14 and lysine 48 in the function of the cardiotonic 59. Güntert, P. & Wüthrich, K. (1991). Improved efficiency of protein polypeptide anthopleurin B. J. Biol. Chem. 269, 921–925. structure calculations from NMR data using the program DIANA with 37. Norton, R.S. (1991). Structure and structure-function relationships of redundant dihedral angle constraints. J. Biomol. NMR 1, 447–456. sea anemone proteins that interact with the sodium channel. Toxicon 60. Hutchinson, E.G. & Thornton, J.M. (1996). PROMOTIF — A program to 29, 1051–1084. identify and analyze structural motifs in proteins. Protein Sci. 38. Gallagher, M.J. & Blumenthal, K.M. (1992). Cloning and expression of 5, 212–220. wild-type and mutant forms of the cardiotonic polypeptide 61. Barton, G.J. (1993). ALSCRIPT. A tool to format multiple sequence anthopleurin B. J. Biol. Chem. 267, 13958–13963. alignments. Protein Eng. 6, 37–40. 39. Gallagher, M.J. & Blumenthal, K.M. (1994). Importance of the unique 62. Koradi, R., Billeter, M. & Wüthrich, K. (1996). MOLMOL: a program for cationic residues arginine 12 and lysine 49 in the activity of the display and analysis of macromolecular structures. J. Mol. Graph. cardiotonic polypeptide anthopleurin B. J. Biol. Chem. 269, 254–259. 14, 51–55. 40. Khera, P.K. & Blumenthal, K.M. (1996). Importance of highly conserved anionic residues and electrostatic interactions in the activity and structure of the cardiotonic polypeptide anthopleurin B. Biochemistry 35, 3503–3507. 41. Monks, S.A., Pallaghy, P.K., Scanlon, M.J. & Norton, R.S. (1995). Solution structure of the cardiostimulant polypeptide anthopleurin-B and comparison with anthopleurin-A. Structure 3, 791–803. 42. Housset, D., Habersetzer-Rochat, C., Astier, J.-P. & Fontecilla-Camps, J.C. (1994). Crystal structure of toxin II from Androctonus australis