Although we have previously modeled this structure (Diochot et al. 2004), the model only indicates that the overall fold is conserved. Actually, it is of first importance to experimentally describe the precise conformation of the side chains to understand why APETx2 and APETx1 are specific for very different ion channels (respectively, sodium and potassium channels). The precise location of the side chains is of primary importance in the electrostatic anisotropy analysis, which will be used as a guideline for further mutagenesis experiments.
NMR resonance assignment and secondary structure
The proton resonance of APETx2 was sequentially assigned following the standard method first described by Wüthrich (1986) and successfully applied to various toxins such as ADO1, PaTx1, or Mca, respectively, from the assassin bug Agriosphodrus dohrni, the spider Phrixotrichus auratus or the scorpion Scorpiomaurus (Mosbah et al. 2000; Bernard et al. 2004; Chagot et al. 2004).
The spin systems were identified on the basis of both COSY and TOCSY spectra recorded at 280 K and 290 K. The use of two temperatures allowed us to resolve overlapping signals in the fingerprint region. Once the intraresidual and sequential assignment procedures were achieved, almost all protons were identified and their resonance frequency determined (BMRB code 6467). The distribution of the Hαi/HNi+1, Hβi/HNi+1, and HNi/HNi+1 nOe correlations indicates that the toxin is mainly organized in extended regions, characterized by strong Hαi/HNi+1 correlations together with large 3JHN-Hα coupling constants (Fig. 1A).
The structure of APETx2 was determined by using 631 nOe-based distance restraints, including 369 intraresidue restraints, 128 sequential restraints, 41 medium-range restraints, and 93 long-range restraints. The repartition of these nOe along the sequence is shown in Figure 1B (top). In addition, 38 hydrogen bond restraints derived from proton exchange data and 28 dihedral angle restraints derived from the measurement of coupling constants were included, as well the distance restraints derived from the three disulfide bridges, which were deduced from the observation of the structures obtained in the first calculation run. Altogether, the final experimental set corresponded to 16.8 constraints per residue on average.
The calculation using the whole set of restraints and water solvent minimization led to a single family of 25 solutions (Fig. 2A) consistent with the experimental restraints. Their structural statistics are given in Table 1. All the solutions have good nonbonded contacts, as indicated by the negative value of the van der Waals energy, and good covalent geometry as shown by the low values of CNS energy terms and low root mean square deviation (RMSD) values for bond lengths, valence angles, and improper dihedral angles. Correlation with the experimental data shows no nOe-derived distance violation >0.2 Å.
The RMSD calculated on the whole structure is 0.84 ± 0.16 Å for the backbone and 1.54 ± 0.21 Å for all heavy atoms. If N- and C-terminal residues are omitted, these values drop to 0.79 ± 0.16 Å and 1.52 ± 0.21 Å, respectively. If the loop 15–27 and the N and C termini are excluded, the values become 0.43 ± 0.09 Å and 1.09 ± 0.20 Å, which indicates a rather poor resolution of this loop as well as of the N- and C-terminal residues. This is confirmed by the individual RMSD values (Fig. 1B). The analysis of the Ramachandran plot for the ensemble of the 25 calculated models (in PROCHECK software nomenclature) reveals that 75.8% of the residues are in the most favored regions, 23.2% in the allowed regions, 1.0% in the generously allowed, and none in the disallowed regions (data not shown).
The convergence of the 25 final solutions allowed us to describe the structure of APETx2 (Fig. 2) (Protein Data Bank [PDB] code 1WXN), which consists of a compact disulfide-bonded core composed of a four-stranded β sheet from which a loop (15–27) and the N and C termini emerge. APETx2 can therefore be classified as an all-β toxin. The four strands, well detected by the PROCHECK-NMR software, include residues 3–6 (strand I), residues 9–14 (strand II), residues 28–32 (strand III), and residues 35–39 (strand IV). According to the value of dihedral angles, strands I and II are connected by a type II′- β turn (residues 6–9), strands III and IV are connected by a type I-β turn (residues 32–35), while strands II and III are connected by the 15–27 loop.
Comparison with related toxins
The fold of APETx2 places it in the Defensin family (according to the SCOP database, http://scop.mrc-lmb.cam.ac.uk) (Torres and Kuchel 2004), which includes anti-microbial peptides from humans as well as different toxins from the venoms of snake, Platypus or sea anemones.
The general fold is characterized by a short α-helix or turn followed by two or three anti-parallel β-strands. β-Defensin type peptides are stabilized around a three disulfide-bonded hydrophobic core with a 1–5, 2–4, 3–6 cysteine pairing. Many variations on the same motif lead to molecules with a variety of biological properties, as observed in other short peptide families.
Toxins with a similar fold and disulfide arrangement have been characterized from other sea anemone venoms, and Figure 3A shows the three-dimensional structures of the related sea anemone toxins anthopleurin A (PDB code 1AHL) from the giant green anemone Anthopleura xanthogrammica, and ATX Ia from Anemonia sulcata, which both inhibit sodium channels (PDB code 1ATX), ShI from Stichodactyla helianthus (PDB code 1SH1), and BDS1 from Anemonia sulcata (PDB code 1BDS) first described as an anti-hypertensive and anti-viral protein, and later shown to be an inhibitor of the potassium channel KV3.4 (Diochot et al. 1998). APETX2 and APETx1 (PDB code 1WQK) (Diochot et al. 2003; Chagot et al. 2005), both from Anthopleura elegantissima, are shown in Figure 3D. All these toxins are organized around the same overall fold.
APETx2 possesses some sequence similarities with BDS1 and APETx1 (respectively, 32% and 69% sequence similarity, excluding cysteine residues) and little sequence similarity with other sea anemone toxins (Diochot et al. 2004) (Fig. 3B). While organized around the same fold, these three toxins, however, show different pharmacological properties: BDS1 and APETx1 act respectively on the potassium channels KV3.4 and HERG, and APETx2 inhibits ASIC3, a proton-gated channel primarily permeant for sodium. The dissimilarity of these target proteins would thus suggest different modes of toxin-channel interaction, and therefore a different organization of toxin surfaces.
The phylogenic tree (Fig. 3C) shows that the sea anemones toxins probably have a common ancestor and have diverged during evolution. An interesting fact is that the toxins closest to APETx2 are potassium channel inhibitors; therefore, APETx2 may have diverged to become a sodium channel effector.
As APETx1 and APETx2 share very similar primary sequences, they are good models for the study of the influence of toxin surfaces on receptor selectivity. To gain insight into the structural features leading to such different pharmacological properties, we first compared the three-dimensional structures of APETx1 and APETx2 (Fig. 3D). With a RMSD of 1.65 Å for all the backbone atoms of APETx1 and APETx2, and a poor fit of the 6–12, 15–20, 33–35, and 39–42 region, these toxins are clearly structurally related. The two toxins share a similar cysteine pattern and among the 15 amino acid differences, three are conservative (T9, T19, and S22 in APETx1 are replaced by S9, S19, and T22 in APETx2) and the 12 others are significantly different in the physicochemical nature of the residues involved. Residues T3, Y5, K8, I10, G16, T17, K18, N23, G31, I36, Y39, and V41 in APETx1 are replaced respectively by A3, S5, N8, K10, Y16, R17, P18, D23 R31, T36, T39, and A41 in APETx2.
The spatial locations of the residues which differ in APETx2 and APETx1 are shown in Figure 4A. Meaningful substitutions can be linked to two different spatial clusters: residues 3, 5, 8, 10 in the first and second strand, together with residues 39 and 41 in the C terminus of the protein form the first cluster. Residues 16, 17, 18, 31, and 36 located after the first strand and in the third and fourth strand form the second. The first cluster appears to be formed of mostly uncharged and hydrophobic residues, surrounding a central Lysine (K10). Conversely, the second cluster is biased toward positive charges, with two basic residues close to a hydrophobic one (Y16) and a nonpolar core. The basic-aromatic residue combination is a feature of toxin interaction surfaces in other peptide toxins, and has been termed the characteristic “basic-aromatic dyad” (Dauplais et al. 1997). A similar dyad is also found in the other known ASIC toxin, PcTx1 (see Fig. 4B and text below). Additionally, residue 23, located on the loop between the two clusters is also changed from an uncharged residue (N23) to a negatively charged one (D23), a significant mutation in APETX2. Taken together, these features strongly suggest that at least some of these residues and perhaps the two clusters considered as toxin “patches” are important for the interaction of APETx2 with the ASIC3 channel subunit and also for the different specificity of APETx1 and APETx2.
Various folds can accommodate ASIC channel recognition
To date, the only other known toxin acting on ASIC channels is PcTx1 (PDB code 1LMM). This toxin, folded according to the ICK motif, has been purified from the venom of the tarantula Psalmopoeus cambridgei, and is the first and only high-affinity ligand described for homomeric ASIC1a channels (Escoubas et al. 2000). Its structure was solved by our group (Escoubas et al. 2003), and analysis of its electrostatic anisotropy led us to suggest a possible surface involved in the interaction with the ASIC1a channel. We proposed the surface to be composed of the four basic residues K25, R26, R27, and R28 associated with aromatic residues W7, W24, and F30 surrounding the basic patch (Fig. 4B) (Escoubas et al. 2003).
Comparison of the two toxins, acting on different ASIC subunits, nevertheless reveals some degree of similarity in their exposed side chains, and that despite different folds (Fig. 4B). Both peptides possess a basic/aromatic cluster (K25, R26, R27, R28, W7, W24, and F30 for PcTx1 and R17, R31, F15, Y16, Y32, and F33 for APETx2), and a basic/hydroxyl cluster (T37, T40, and K39 for PcTx1 and S9 and K10 for APETx2) located on opposite sides of the molecule. The presence of a single acidic residue somewhat equidistant to these two clusters is also a constant (E19 or D16 for PcTx1 and D23 for APETx2). Another common characteristic is the orientation of the dipole moment resulting from the electrostatic anisotropy: In both APETx2 and PcTx1, this dipole emerges through the basic/aromatic cluster.
We have previously proposed (Ferrat et al. 2001) that electrostatic anisotropy could play an orientating force within the electrostatic field of the membrane receptor, and therefore that the orientation of the dipole could be construed as an indication of the interaction surface between a toxin and its target. This prediction method has been extensively validated through site-directed mutagenesis studies, for toxins acting by pore occlusion such as the scorpion Csαβ toxins (Sabatier et al. 1994; Inisan et al. 1995; Fremont et al. 1997). Although mutagenesis studies are far fewer in ICK gating modifier toxins, it appears that at least in gating-modifier toxins acting on members of the KV2 and KV4 potassium channel subfamilies, a consensus structural interpretation of the electrostatic anisotropy is beginning to emerge, which nevertheless remains to be fully confirmed for all toxins by additional mutagenesis studies (Bernard et al. 2000; Chagot et al. 2004).
Applied to APETx2, the calculation shows that the dipole moment emerges between the third and fourth strands of the β-sheet near residue R31, defining the cluster composed of Y16, R17, and R31 associated with F33 and F15 as a putative functional surface. A similar result was obtained for PcTx1, in which the dipole moment emerges through a patch of basic residues located at the tip of the cone-shaped peptide. In both cases the dipole moment calculation predicts an involvement of both basic and hydrophobic residues, an interesting feature in that it recalls the role of the basic-hydrophobic dyad demonstrated for scorpion toxins interacting with voltage-dependent KV1 channels (Dauplais et al. 1997). Another element that appears to support our approach is the fact that a single amino acid extension of either the N- or C-terminal part of the PcTx1 toxin has no effect on channel inhibition or toxin binding, showing that the opposite part of the peptide (the base of the “cone”) is not crucial for channel interaction (P. Escoubas, unpubl.).
Despite their similarity in sequence and in fold, APETx1 and APETx2 have specific electrostatic characteristics. This fact is demonstrated by the difference in dipole orientation, which result in a different orientation of the toxins toward their receptor sites. This slight difference results in a drastic modification of the toxin specificity. The dipole orientation clearly distinguish APETx1 and APETx2, and more generally, could distinguish effectors of various ion channel ligands.
As both toxins acting on ASIC channels possess the same pattern of residues and have the same electrostatic anisotropy repartition, we therefore propose that their interaction with the receptor is basically mediated by this basic/aromatic cluster, although a complementary role of the second common cluster or for the negatively charge residues (D23 in APETx2) cannot be excluded, and will need to be explored by mutagenesis studies. Further understanding of structure–activity relationships in the toxin–ASIC interaction model will be useful in the future design of novel therapeutic agent specifically targeting this important ion channel class.