Structural and Functional Elucidation of Peptide Ts11 Shows Evidence of a Novel Subfamily of Scorpion Venom Toxins

To date, several families of peptide toxins specifically interacting with ion channels in scorpion venom have been described. One of these families comprise peptide toxins (called KTxs), known to modulate potassium channels. Thus far, 202 KTxs have been reported, belonging to several subfamilies of KTxs (called α, β, γ, κ, δ, and λ-KTxs). Here we report on a previously described orphan toxin from Tityus serrulatus venom, named Ts11. We carried out an in-depth structure-function analysis combining 3D structure elucidation of Ts11 and electrophysiological characterization of the toxin. The Ts11 structure is highlighted by an Inhibitor Cystine Knot (ICK) type scaffold, completely devoid of the classical secondary structure elements (α-helix and/or β-strand). This has, to the best of our knowledge, never been described before for scorpion toxins and therefore represents a novel, 6th type of structural fold for these scorpion peptides. On the basis of their preferred interaction with voltage-gated K channels, as compared to all the other targets tested, it can be postulated that Ts11 is the first member of a new subfamily, designated as ε-KTx.


Introduction
Scorpion venom is a rich source of potassium channel blocking toxins (KTxs), which have been used in the structural and functional characterization of various voltage-gated potassium (Kv) channels [1]. Kv channels have received much attention because they are widespread in almost all tissue, and also due to the high diversity of Kv channels expressed in mammalian cells. They play key roles in the regulation of many physiological processes, including neurotransmitter release, immune response, heart rate, insulin secretion, neuronal excitability, epithelial electrolyte transport, smooth muscle contraction and cell proliferation [2].
Interestingly, scorpions have only developed one structural fold arsenal of toxins targeting Na + and Cl − channels, while repeatedly evolving several folds to capture diverse prey via Kv channels [2].
To date, scorpion venom peptides are known to adopt five different structural folds. Most of them contain a common core topology comprised of one or two short α-helices connected to a triple-stranded antiparallel β-sheet stabilized by three or four disulfide bonds [3]. For both classes of toxins-those acting on potassium channels (KTxs) and those acting on sodium channels (NaTxs)-the range of different folds is merely the variability of CSα/β and CSα/α topology [1].
Pimenta et al. [4] reported the primary sequences of three new short peptides from Tityus serrulatus venom (Tsv) previously known as TsPep1 (KPKCGLCRYRCCSGGCSSGKCVNGACDCS), TsPep2 (TVKCGGCNRKCCAGGCRSGKCINGKCQCY), and TsPep3, and henceforth named Ts11, Ts12 and Ts13 (following the nomenclature suggested by Cologna et al. [5]). These peptides are 29 amino-acid residues long, ranging from approximately 2900 to 3000 Da. They are highly reticulated by four disulfide-bridges, which make these peptides the most constrained structures of scorpion venom-derived peptides known to date, and a unique group of neurotoxins found in Tityus serrulatus venom containing the vicinal cysteines [6].
Based on alignment and size evidence, Ts11, Ts12 and Ts13 were previously classified as KTxs. However, no subfamily was specified since there was no functional study or structure to base the classification on. In spite of some sequence similarities within the C-terminal β-sheet of well-characterized toxins active in K + channels, their biological function has not been clarified until now [4].
We hereby present the sixth structural fold to be adopted by scorpion venom peptides, highlighted by an knotting type fold, stabilized by four disulfide bridges and completely devoid of the classical secondary structure elements such as α-helix and β-strand. To the best of our knowledge, this fold has not been described thus far for scorpion toxins. Based on the functional characterization of both the voltage-gated potassium channel (Kv) and voltage-gated sodium channel (Nav), and in depth structure-function analyses, we propose that Ts11 can be seen as member of a new KTx subfamily, designated as ε-KTx1.1. A highly homologous peptide, Ts12, was also purified and characterized as far as we were able. The results hereof are presented as Supplementary Material.

Isolation of Native Ts11 and Biochemical Characterization
Ts11 was isolated using a combination of cation exchange chromatography of T. serrulatus venom [7], followed by reversed-phase fast protein liquid chromatography (RP-FPLC) of fraction XIIA, as described in the experimental procedures section. The fraction XIIA was fractionated on a C8 column and afforded 17 chromatoghaphic peaks ( Figure 1A). The second peak was identified as Ts11 by N-terminal sequencing and MALDI-TOF analysis (m/z 2938.2) (using alpha-Cyano-4-hydroxycinnamic acid (HCCA) matrix and reflectron positive ion mode ( Figure 1B). Ts12 was isolated through reversed-phase high-performance liquid chromatography (RP-HPLC) of T. serrulatus venom on a C18 column (Supplemental Materials Figure S1A). The toxin was eluted in approximately 24 mL and its observed m/z was 2991.3 was confirmed by MALDI-TOF analysis (Supplemental Materials Figure S1B).    Figure  2A) and the minimum-energy closest-to-average structure ( Figure 2B) of scorpion peptide Ts11. The peptide structures are well defined with backbone and heavy atom RMSD of 0.44 and 1.03 respectively over the entire chain, and Figure 1C shows the observed medium and long range NOEs that were used for structure determination, 3 JHNHα, and chemical shift index (CSI) along with the amino acid sequences of toxin peptide Ts11. The CSI and 3 JHNHα coupling values suggest the likely absence of any α-helix and/or β sheet structure in the Ts11 peptide.

Structural Elucidation of Ts11
The coordinates for 15 structures, NMR restraints and chemical shifts have been deposited in the RCSB Protein Data Bank with RCSB ID RCSB 103995, PDB ID 2MSF, and Biological Magnetic Resonance Bank BMRB with accession number: 25122.
NMR spectral analysis shows the formation of a single set of resonances for the Ts11, indicating that it adopts one type of structural form in solution. Resonance assignment was performed according to standard procedures as outlined by Wüthrich [8]. Complete sequence specific proton assignments were achieved by analyzing homo-nuclear two-dimensional (2D) spectra (DQF-COSY, TOCSY, and NOESY). Initially, the NH and Hα resonances of the individual spin systems (except Pro) were identified by analyzing the so-called "fingerprint" region of the DQF-COSY and TOCSY spectra and the remaining resonances of the spin systems were identified by following the so-called "TOCSY-tower". Sequence specific assignments were achieved by linking each individual spin system via sequential inter-residue Hαn-HN (n + 1) cross-peaks in the "fingerprint" region of the NOESY spectrum. Partial carbon assignments were also performed by using 1 H-13 C HSQC spectra, which further reconfirmed most of the homo-nuclear proton assignments and clarified side chain proton assignments of other residues that were not resolved in the homo-nuclear 2D spectra. The geminal methylene protons were not assigned stereo-specifically and the NOE distance restraints involving these protons were used ambiguously during structure calculation in the Xplor-NIH program.
The structural evaluation using PROCHECK demonstrates that none of the resulting structures have bad non-bonded contacts and most of the backbone dihedral angles are within the allowed regions of the Ramachandran plot (94.6% residues fall in the allowed region). Detailed structure determination statistics are provided in Table 1.   Figure 2A) and the minimum-energy closest-to-average structure ( Figure 2B) of scorpion peptide Ts11. The peptide structures are well defined with backbone and heavy atom RMSD of 0.44 and 1.03 respectively over the entire chain, and Figure 1C shows the observed medium and long range NOEs that were used for structure determination, 3 J HNHα , and chemical shift index (CSI) along with the amino acid sequences of toxin peptide Ts11. The CSI and 3 J HNHα coupling values suggest the likely absence of any α-helix and/or β sheet structure in the Ts11 peptide.

Structural Elucidation of Ts11
The coordinates for 15 structures, NMR restraints and chemical shifts have been deposited in the RCSB Protein Data Bank with RCSB ID RCSB 103995, PDB ID 2MSF, and Biological Magnetic Resonance Bank BMRB with accession number: 25122.
NMR spectral analysis shows the formation of a single set of resonances for the Ts11, indicating that it adopts one type of structural form in solution. Resonance assignment was performed according to standard procedures as outlined by Wüthrich [8]. Complete sequence specific proton assignments were achieved by analyzing homo-nuclear two-dimensional (2D) spectra (DQF-COSY, TOCSY, and NOESY). Initially, the NH and Hα resonances of the individual spin systems (except Pro) were identified by analyzing the so-called "fingerprint" region of the DQF-COSY and TOCSY spectra and the remaining resonances of the spin systems were identified by following the so-called "TOCSY-tower". Sequence specific assignments were achieved by linking each individual spin system via sequential inter-residue Hαn-HN (n + 1) cross-peaks in the "fingerprint" region of the NOESY spectrum. Partial carbon assignments were also performed by using 1 H-13 C HSQC spectra, which further reconfirmed most of the homo-nuclear proton assignments and clarified side chain proton assignments of other residues that were not resolved in the homo-nuclear 2D spectra. The geminal methylene protons were not assigned stereo-specifically and the NOE distance restraints involving these protons were used ambiguously during structure calculation in the Xplor-NIH program.
The structural evaluation using PROCHECK demonstrates that none of the resulting structures have bad non-bonded contacts and most of the backbone dihedral angles are within the allowed regions of the Ramachandran plot (94.6% residues fall in the allowed region). Detailed structure determination statistics are provided in Table 1.

Discussion
Ts11 has a unique feature in its structure: it is a short peptide of only 29 residues with two vicinal cysteines at the positions 11 and 12, adopting a highly constricted structure that is stabilized by 4 disulfide bonds.
In an earlier study published by Pimenta et al. [4] it was reported that the enzymatic approach to described disulfide mapping for Ts11 (TsPep1) was unsuccessful due to the existence of two vicinal Cys residues and possibly due to the existence of a very compact TS11 structure. In this study, we have undertaken a solution state NMR study to investigate the four-disulfide linkages and to determine the three-dimensional structure of the Ts11 peptide.
A closer look into the structure (Figure 2A,B) of Ts11 reveals that the peptide backbone adopts a very compact cysteine-knot like structure with three loops and one β-turn (V22-A25) at the C-terminal part. The structure is reinforced by four disulfide bonds (Cys 4 -Cys 12 , Cys 7 -Cys 28 , Cys 11 -Cys 21 , and Cys 16 -Cys 26 ). All five glycine residues, as anticipated, are positioned in the β-turn and in the loops. The N-terminal end residues up to Cys 4 -notably containing two positively charged arginine residues-protrude out from the remaining compactly folded C-terminal head region like a screw tail. The exceptionality of the three-dimensional structure of Ts11 is its unique disulfide connectivity and the lack of any regular secondary structural units like α helices and/or β-strand, contrasting with what is commonly observed in the other scorpion toxins.
Ts11 was considered as a Kv blocker without altering the kinetics of channel gating. Although it is not a potent blocker of Kv channels, Ts11 does not target Nav channels, and therefore, it can be surmised that these toxins are representative blockers of Kv channels. However, it is possible that the Ts11 has other targets not yet identified.
Scorpion venom peptides are known to adopt five different structural folds. Most of them contain a common core topology comprised of one or two short α-helices connected to a triple-stranded antiparallel β-sheet stabilized by three or four disulfide bonds [3]. For scorpion venoms that affect Kv channel peptides, the fold known as the cystine-stabilized α/β (CSα/β) motif is the most abundant (around 208 peptides, according to Scorpktx [9]), and occurs in three distinct subfamilies: α-, β-and γ-KTxs [2,3,10,11]. It is important to note that Ts11 is not the first CSα/β motif-containing scorpion venom peptide that folds into a distinct three dimensional structure. For example, Ts16 and Maurotoxin are two CSα/β motif-containing peptides with a distinct cysteine pairing (C1-C5, C2-C4, C3-C6) and their structure only contains α-helices [12,13].
Another subfamily of KTxs with a different structural arrangement is the κ-KTx. These are purely helicoidal 3D structures named the cystine-stabilized helix-loop-helix (CSα/α) fold that consists of two short α-helices connected by a β-turn, stabilized by two disulfide bonds. To date,

Discussion
Ts11 has a unique feature in its structure: it is a short peptide of only 29 residues with two vicinal cysteines at the positions 11 and 12, adopting a highly constricted structure that is stabilized by 4 disulfide bonds.
In an earlier study published by Pimenta et al. [4] it was reported that the enzymatic approach to described disulfide mapping for Ts11 (TsPep1) was unsuccessful due to the existence of two vicinal Cys residues and possibly due to the existence of a very compact TS11 structure. In this study, we have undertaken a solution state NMR study to investigate the four-disulfide linkages and to determine the three-dimensional structure of the Ts11 peptide.
A closer look into the structure (Figure 2A,B) of Ts11 reveals that the peptide backbone adopts a very compact cysteine-knot like structure with three loops and one β-turn (V22-A25) at the C-terminal part. The structure is reinforced by four disulfide bonds (Cys 4 -Cys 12 , Cys 7 -Cys 28 , Cys 11 -Cys 21 , and Cys 16 -Cys 26 ). All five glycine residues, as anticipated, are positioned in the β-turn and in the loops. The N-terminal end residues up to Cys 4 -notably containing two positively charged arginine residues-protrude out from the remaining compactly folded C-terminal head region like a screw tail. The exceptionality of the three-dimensional structure of Ts11 is its unique disulfide connectivity and the lack of any regular secondary structural units like α helices and/or β-strand, contrasting with what is commonly observed in the other scorpion toxins.
Ts11 was considered as a Kv blocker without altering the kinetics of channel gating. Although it is not a potent blocker of Kv channels, Ts11 does not target Nav channels, and therefore, it can be surmised that these toxins are representative blockers of Kv channels. However, it is possible that the Ts11 has other targets not yet identified.
Scorpion venom peptides are known to adopt five different structural folds. Most of them contain a common core topology comprised of one or two short α-helices connected to a triple-stranded antiparallel β-sheet stabilized by three or four disulfide bonds [3]. For scorpion venoms that affect Kv channel peptides, the fold known as the cystine-stabilized α/β (CSα/β) motif is the most abundant (around 208 peptides, according to Scorpktx [9]), and occurs in three distinct subfamilies: α-, β-and γ-KTxs [2,3,10,11]. It is important to note that Ts11 is not the first CSα/β motif-containing scorpion venom peptide that folds into a distinct three dimensional structure. For example, Ts16 and Maurotoxin are two CSα/β motif-containing peptides with a distinct cysteine pairing (C1-C5, C2-C4, C3-C6) and their structure only contains α-helices [12,13].
Another subfamily of KTxs with a different structural arrangement is the κ-KTx. These are purely helicoidal 3D structures named the cystine-stabilized helix-loop-helix (CSα/α) fold that consists of two short α-helices connected by a β-turn, stabilized by two disulfide bonds. To date, only 18 scorpion peptides adopting this fold have been discovered [11,[14][15][16].
The δ-KTxs subfamily comprises all the Kunitz-type serine protease inhibitor scorpion toxins. These peptides exert both protease and potassium channel inhibiting properties. Conserved in this subfamily is the Kunitz-type structural fold with a double stranded antiparallel β-sheet flanked by an α-helix in both C-terminal and N-terminal segments. This fold is stabilized by three disulfide bridges: the first one connecting the C-terminal α-helix to one of the β-strands, and a second and third linking the C-terminal α-helix with the C-terminal and the N-terminal tail respectively [17,18].
The fourth subfamily of scorpion-venom peptides adopts the Inhibitor Cystine Knot (ICK) motif, with a triple stranded antiparallel β-sheet stabilized by 3 cystine linkages, and are predominantly found in cone snails and spider venoms [19,20]. The ICK is a structural motif that is shared with a large group of polypeptides having diverse primary structures and bioactivities. It is also found in peptides from evolutionarily distant organisms, such as fungi, plants, humans, marine mollusks and insects [19]. In scorpions, ICK peptides only represent minor venom components and target a limited number of receptors, such as Kv channels or ryanodine receptors [2].
The fifth fold is the disulfide-directed β-hairpin (DDH). It was previously suggested that the three disulfide bridge ICK fold is an elaboration of a simpler, ancestral two-disulfide fold coined the disulfide-directed β-hairpin (DDH) [2,20,21]. The ϕ-liotoxin, Lw1a is an example of a native peptide that adopts the previously hypothetical DDH fold. However, a recent critical revision by Undheim and colleagues noted that although both folds are related, it remains unclear whether or not the DDH fold really is the evolutionary precursor of the ICK motif. The fact that no other single-domain DDH peptides have been found in other organisms besides scorpions, suggests that the DDH is likely a derived ICK [20][21][22] As in Ts11, vicinal cysteins also occur in γ-KTxs, δ-KTxs, λ-KTxs, λ-KTxs/calcines [2], and chlorotoxins. The ICK scaffold which we determined for Ts11 is in common with with λ-KTxs and λ-KTxs/calcine, although they possess only 3 disulfide bridges instead of 4.
Further analyses highlight also the unique pattern of disulfide bridges of Ts11 when compared with other ICK type scorpion toxins (λ-KTx and λ-KTx/calcine) and with DDH-fold toxin (ϕ-liotoxins-Lw1a). The new pattern of disulfide connectivity in Ts11 can be regarded as novel organization of ICK type scorpion toxins, conserving elements from the typical ICK-fold scorpion toxins and from the DDH-fold, the evolutionary precursor of ICK motif that are stabilized by 3 and 2 disulfide bonds respectively [21] (Figure 6).

Conclusions
Many peptide toxins obtained from animal venoms have proved to be valuable tools for the elucidation of the pharmacological, physiological and structural features of their pharmacological receptors.
Unravelling molecular determinants by which animal toxins are able to recognize a receptor or channel, and a detailed examination of their folds, provides several interesting research avenues in terms of protein engineering and therapeutic potential. Together, these tools can offer potential for Figure 6. Comparison of the Ts11 with DDH-fold and ICK-fold toxins disulfide bond patterns. Disulfide patterns were compared with ϕ-Liotoxin-Lw1a (DDH motif), λ-KTxs and λ-KTx/calcine (three disulfide bridges ICK-type toxins, Imperatoxin A and Maurocalcin). Black lines represent the disulfide connectivity unique for Ts11. Red lines represent the DDH motif on ϕ-Liotoxin-Lw1a. Purple lines represent the disulfide connectivity on ICK-type toxins (λ-KTxs and λ-KTx/calcine). Long dashes: disulfide bond shared between Ts11 and ICK-type toxins. Dotted lines: disulfide connectivity shared between DDH motif and the ICK-type toxins. The green arrow indicates the positive charged residue of a possible dyad.

Conclusions
Many peptide toxins obtained from animal venoms have proved to be valuable tools for the elucidation of the pharmacological, physiological and structural features of their pharmacological receptors.
Unravelling molecular determinants by which animal toxins are able to recognize a receptor or channel, and a detailed examination of their folds, provides several interesting research avenues in terms of protein engineering and therapeutic potential. Together, these tools can offer potential for altering pharmacological selectivity, specificity and potency of these toxins, making them a unique source of lead compounds and templates from which agents of specific therapeutic value may be designed and generated [1].
Ts11 tertiary structure obtained through solution NMR showed an ICK-type scaffold lacking the classical secondary structures, such as α-helix or β-strands, which, to the best of our knowledge has never been described thus far. Ts11 presents itself as a Kv blocker with unique structural features. Based on the novel scaffold of Ts11 and its high similarity with Ts12, we propose that these peptides are the first members of a sixth structural fold adopted by scorpion venom peptides. On the basis of a functional analysis evidencing these toxins as preferential Kv blockers as compared to all the other targets tested, and due to the poor percentage of identity with the other KTxs, we suggest that they can be regarded as the first members of a new subfamily of KTxs, named as ε-KTx. Therefore, Ts11 and Ts12 are named ε-KTX 1.1, and 1.2, respectively.
Ts12 was isolated by injecting whole venom of T. serrulatus in reversed-phase chromatography using the HPLC AKTA Explorer 100 system (GE Healthcare, Uppsala, Sweden), with the analytical column PepMap™ C18 (4.6 mm × 150 mm; Applied Biosystems, Foster City, CA, USA) which was previously equilibrated with 0.1% aqueous trifluoroacetic acid (TFA) (solution A). For each run, 50 mg of the lyophilized venom was dissolved in 1 mL of solution A. Samples were centrifuged, filtered and the supernatant applied to the column. Elution of the components was obtained by the following gradient system: 0-15 min, 0% B (solution B: 0.1% trifluoroacetic acid in acetonitrile); 15-50 min, 0%-30% B; 50-60 min; 30%-60% B. Flow was 1 mL/min and absorbance was monitored at 214 nm and 280 nm. Fractions were collected using an automated fraction collector Frac920 (GE Healthcare) on 96 deep well plates. Samples of interest were lyophilized and stored at −20 • C until required.
The K V 1.1-K V 1.6, K V 2.1, K V 3.1 and K V 4.2 and Shaker IR currents were evoked by 250 ms depolarizations to 0 mV followed by a 250 ms pulse to −50 mV, from a holding potential of −90 mV. Current traces of hERG channels were elicited by applying a pulse from −90 mV to +40 mV for 2.5 s followed by a step to −120 mV for 2.5 s. The Kv 10.1 currents were evoked by 1 s depolarization to 0 mV, from a holding potential of −90 mV. Sodium current traces were evoked, from a holding potential of −90 mV, by 100 ms depolarization to 0 mV. In order to investigate the current-voltage relationship, current traces were evoked by 10 mV depolarization steps from a holding potential of −90 mV.
To assess the concentration-response relationship of Ts11 on Kv1.3, data were fitted with the Hill equation: y = 100/[1 + (IC50/[toxin]) h ], where y is the amplitude of the toxin-induced effect, IC 50 is the toxin concentration at half maximal efficacy, [toxin] is the toxin concentration and h is the Hill coefficient (Hill coefficient: 0.8). Current-voltage relationship was determined by 100 ms step depolarization between −90 and +70 mV, using 10 mV increments. All data represent at least 3 independent experiments (n = 3) and are presented as mean ± standard error.
In the one-dimensional and two-dimensional spectra, the water signal was suppressed by using excitation sculpting with gradients [29]. The two-dimensional NOESY (mixing time 200 and 300 ms) was recorded with a sweep width of 7210 Hz in both dimensions, 64 scans, 2048 data points in t 2 , and 1024 free induction decays (FIDs) in t 1 .
A two-dimensional total correlation spectroscopy (2D-TOCSY) [30] was recorded with DIPSI2 sequence for mixing (mixing time 80 ms). A double quantum-filtered correlation spectrum (DQF-COSY) [31] was acquired using excitation sculpting with gradients for water suppression with a sweep width of 7210 Hz in both dimensions, 64 scans, 2048 data points in t 2 , and 1024 FIDs in t 1 .
In the processing of two-dimensional spectra the data were apodized with a shifted sine-bell square function in both dimensions. Proton and carbon chemical shifts were calibrated by using external DSS signal as reference (0.000 ppm).
Natural abundance 1 H, 13 C heteronuclear single quantum correlation ( 1 H-13 C HSQC) spectrum was recorded on the natural abundance sample with sensitivity enhancement and gradient coherence selection optimized for selection of aliphatic CH groups (J CH = 135 Hz) using 64 scans, 1024/2048 complex data points, and 12,072/7210 Hz spectral widths in t 1 and t 2 respectively. For the selection of aromatic CH groups 170 Hz was used for J CH along with 32 scans, and 64/2048 complex data points.

Structural Constraints
Distance restraints were derived from cross-peak volumes of the NOESY spectrum recorded with 200 ms mixing time. Estimated interproton distances were derived using the isolated spin pair approximation, r ij = r ref (v ref /v ij ) 1/6 where r ij is the estimated interproton distance, r ref is the fixed internal reference distance, and v ref and v ij are the NOE cross-peak volumes of the reference and estimated cross-peaks respectively. Average cross-peak volume of the geminal methylene proton pairs was used as reference volume which corresponds to the fixed reference distance of 1.8 Å. Generally an experimental error of ±20% on the calculated interproton distances was used for upper and lower bounds. The 3 J HNHα coupling constants were measured from the one-dimensional proton spectrum recorded in H 2 O and then converted to dihedral restraints as follows: 3 J HNHα > 8 Hz, ϕ = −120 • ± 30 • ; 3 J HNHα < 6 Hz; ϕ = −60 • ± 30 • ; ω = 180 • ± 30 • to define the trans X-Pro conformation as confirmed by the observation of strong NOE interactions between Hα(n) and HD2, HD3(n + 1) Pro.

Structure Calculations
All structure calculations were performed by using Xplor-NIH program, version 2.25 (National Institutes of Health Bethesda, Bethesda, MD, USA) [32]. A set of 100 structures was generated by torsion angle molecular dynamics, starting from an extended strand and by using only NMR-derived restraints, excluding any disulfide restraints. After the torsion angle molecular dynamics round [33], the majority of the structures had converged to very similar structures with similar total energies and with no violations of the NOE and dihedral restraints. In the initially derived structures all the disulfide bonds could be identified unambiguously by the observation of side chain proximity of eight Cys residues. Torsion angle molecular dynamics round was repeated for the second time including both the disulfide bond and NMR-derived restraints. The fifteen lowest energy structures from the second round were used for further refinement during a "gentle molecular dynamics" round in explicit water [34]. A box of water was constructed and optimized around selected structures obtained from the second torsion angle dynamics step. The final stage of refinement commenced with a 20 ps constant temperature molecular dynamics simulation at 300 K (20,000 steps of 0.001 ps) and was followed by a 200-step conjugate gradient energy minimization of the average structure of the last 10 ps of the 20 ps simulation. Structures were analyzed by using PROCHECK [35]. Visual representations were created by using UCSF Chimera software (version 1.9, University of California, San Francisco, CA, USA).